Evangelium secundum Matthaeum

Anselmus Laudunensis et schola, in Migne's nineteenth-century abridged recension, attributed by him to Walafrid Strabo. 42 chunks, 45,922 Latin words, eight translation stints, each of which read its own plates on two witnesses as it went.

This file is the merge required by translation-runbook.md step 4. §1 is the work-wide conventions section: what one voice this work speaks in, what the eight stints were reconciled on, and — where they ruled differently on the same class — which ruling stands and why. §2–§9 are the eight stints' own files, preserved unedited, because the lemma divergence lists, the declined-oddity lists and the arguments for the declines are the most valuable thing they contain. Nothing in them has been summarized away.

⛔ A later sweep should read §1 before changing anything here. Every variation §1 records as licensed was measured across all eight ranges before it was licensed, and normalizing it would undo work, not tidy it.

Merged 2026-09-05. Method: the eight files' stated policies were diffed against each other and against the English chunks themselves, mechanically, on every axis GLOSSA-NT-AGENT-BRIEF.md names. The instrument checks the brief demands were run before any drift was reported (\bturba catches turbare; natio catches damnatio-, generatione-, incarnatione-; \bdaemoni misses daemones/daemonum; a raw usque ad grep overcounts the formula). Where a claim in a stint's file could be checked against the chunks, it was.

1.1 The settled Glossa conventions, as this work applies them

1. VERS. n.-- passes through VERBATIM, untranslated. Checked mechanically: the sequence of VERS. addresses is identical in the Latin and the English of all 42 chunks, character for character, including the addresses Migne prints without the -- dash.
2. The lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for (convention 3). The line between "fragment, expand it" and "complete one-word lemma, render as it stands" was drawn by the 0012–0015 stint — can the printed lemma stand as an English clause or phrase on its own? — and adopted by name by the 0027–0031 stint. It is the work's rule. ⛔ And its limit, stated by the same stint and binding here: an expansion that restores a word Migne does not print is 7a″ conformation, not convention 3. Honora patrem. is "Honour father.", not "Honour thy father"; Magna est fides. is "Great is the faith.", not "thy faith"; Hic est Filius. is "This is the Son.", not "my beloved Son"; Propter incredulitatem. is "Because of unbelief.", not "your unbelief". The slight oddness in English is the information.
3. The attribution sigla ride as ordinary [n: …] notes, verbatim, unanglicized, uncorrected. 285 of them. Migne's attributions in this recension are unreliable and passing one through is not endorsing it.
4. ## heads: numerals mirror the plate. CAPUT PRIMUM. → CHAPTER ONE., CAPUT XIII. → CHAPTER XIII. Checked chunk by chunk across all 42: 28 CAPUT heads — one spelled (PRIMUM/ONE) and 27 in numerals (II–XXVIII) — and not one stint spelled a numeral out. PROLOGUS. → PROLOGUE.
5. Italic spans are 1:1 with the Latin, in the same place. Verified mechanically at the merge, after markers are stripped: exact parity in all 42 chunks. Two supplied spans were found and removed — see §1.4.
6. Greek is reproduced verbatim, not transliterated (Τέλος @0116A, Κωφὸς @0117C).
7. There is not one guillemet in the work. Matthew is an italic-lemma book, like Luke and Mark and unlike Genesis; 0 in 45,922 Latin words. « » therefore never arises, and every quotation mark in the English is ours, per Pattern 6.

1.2 The vocabulary axes — measured across all eight ranges, not asserted

Every figure below is a count over the whole work, Latin against English, made at the merge.

Axis 1 — the commentator's own voice is NOT archaised. dicit/ait/dixit → says/said throughout; no saith. Third-person -eth forms outside an italic span number four in 74,630 English words (taketh @0097A, asketh @0097C, believeth @0135C, knoweth @0123B) and all four sit inside quoted scripture carried in roman — Mt 5:42, Lc 6:29, Rom 10:4, Mt 11:27. The -est forms are second-person singular and are Pattern 17's, not archaism. ⚑ The recorded test is italic membership; the practice is slightly wider, and deliberately so: a scripture quotation Migne sets in roman still takes its received English form. A sweep that strips -eth on the italic test alone would damage four faithful quotations.

Axis 2 — turba → crowd, multitudo → multitude, never one word for both. 33 turba forms, 2 multitudo. The pair stands inside one range twice (0024, 0034) and inside three columns of each other at 0157A/0152C, and the two words are kept apart at every site. ⚑ Two apparent excesses were checked and are not drift: "throng" at 0006 renders frequentia hominum, not turba; the third "crowd" in 0019 is convention 3's expansion of the clipped lemma Et cum ejecta esset (Mt 9:25, turba), so the supplied word is the axis's own.

Axis 3 — daemonium → demon · diabolus → devil · satanas → Satan. 19 · 55 · 1, and the three are never crossed in 42 chunks. The adjective diabolicus → "diabolical" at all four of its occurrences (@0105A, @0126D, @0127B ×2) and daemoniacus → "demoniac" at both (@0113D, @0114C) — which is why a naive noun-count reads four devils and two demons short, and why the check has to be run on the lemma and not on the string. ⚑ One "demon" is supplied and is the axis's own: @0114C Duo habentes. is convention 3's expansion of Mt 8:28 duo habentes daemonia. ⛔ The Douay's "devils" for daemonia never once pulled the first word into the second.

Axis 4 — the gentes family. Matthew is the Judaei/gentes gospel and this is the axis that earns its keep here. Measured site by site over the whole work:

Zero crossings. Every stint ran the natio instrument check and said so: the raw grep returns nine false hits in 0032–0037 alone (destinatione, praedestinati, generatione ×4, damnatur, indignati, contaminatione), three in 0027–0031, four in 0016–0021, eight in 0038–0041 of which seven are false. This is the axis 9004 got wrong three ways across four stints. Matthew got it right eight times out of eight. ⛔ One drift was found and repaired — a ninth word entering the family from outside it. See §1.4(a).

Axis 5 — mysterium → mystery, sacramentum → sacrament, 1:1. 19 mysteri- → 19 "mystery/mysteries"; 35 sacrament- → 33 "sacrament" and 2 "oath". One drift repaired (§1.4(b)) and one declared departure, which stands (§1.3(a)).

Axis 6 — usque ad, by the etc. splice test. The work prints it 159 times: 136 are the formula (each immediately follows an etc., and names where the abridged father's sentence resumes) and 23 are ordinary Latin. Every one of the 136 is "as far as"; not one of the 23 is. The per-stint split — 8/4 · 29/5 · 12/0 · 53/2 · 10/1 · 20/0 · 2/6 · 2/4 — was re-derived at the merge and agrees with every stint's own reported figure.

⚑ The 23 ordinary ones are rendered by SENSE, and the variety is licensed, not drift: "up to" (0006:21, 0040:21), "unto" (0010:23, 0010:65), "even unto" (0018:35, the lemma Usque ad mortem @0170A), "even to" (0038:31), "until" (0022:37, 0038:23), "from … to" (0002, 0005, 0011, 0032, 0035, 0036), "as far as" by sense at three sites of physical extent (0001:33 usque ad Deum ascendit, 0005:31, 0021:55 usque ad naves prosequerentur). ⛔ Do not normalize these. Three of them are the verse's own words — Is 49:6 @0091B, Mt 26:38 @0170A, Mt 28:20 @0167C/@0178C — and reading them as the formula would destroy the note that glosses them. ⚑ Two formula sites are not wrapped usque ad because Migne's italic run continues past the formula into the resumed quotation: @0086C usque ad, Hoc fecit initium signorum Jesus. and @0089B usque ad, Omnis qui audit …. The English mirrors the Latin exactly at both. That is why the English carries 134 as far as and not 136.

1.3 The declared departures — argued in public, and they survive the merge

(a) @0135B — sacramentum is Herod's OATH, and the axis is departed from openly. Isidore's sentence is Non est conservandum sacramentum quo malum incaute promittitur … non implere sacramentum, quam manere in stupro, and the passage's own words are jurasse, perjurio, promittitur, stupro. Rendered "oath" at both occurrences, the rival ("hold the axis, print sacrament") named and rejected in the 0022–0026 stint's §(d) 1, and every other sacramentum in that range kept to the axis (0130A, 0133C, 0134B). ⛔ This is not to be normalized. "Sacrament is not to be kept whereby an evil is incautiously promised" would not be awkward, it would be false — it would put the Church's sacraments where Isidore has put a perjury. A departure the Latin forced, argued where a reader meets it, is exactly what the axis rules license. ⚑ Verified at the merge: these are the only two sacramentum in the work not rendered "sacrament".

(b) @0125A — the printed sed makes the proverb command the opposite, and it rides. Sir 4:36 joins its members with et, so the single negative governs both; Migne prints sed, so it does not. Rendered as printed, [var:] fired. Pattern 7a, working.

(c) @0170B — a [cj:] on an intruded negation Migne prints on both witnesses. non occidant kept in the English, the conjecture occidant standing beside it in its marker. The negation is not repaired inside the translation, which is the one error class nobody downstream can find.

1.4 ⛔ THE DRIFT FOUND AT THE MERGE — four English edits, named per site

These are the only changes made to the English chunks. Both gates were re-run after: plate-gate OK, verify-english OK (42 chunks), and italic parity is now exact in all 42.

(a) @0080D (chunk 0006) — pagani was rendered "the Gentiles", eleven words from gentiles. Migne writes alii sunt aridi, id est, pagani incendio apti … Omnis ergo arbor, ut gentiles, hypocritae. Both words had come out "the Gentiles" in one sentence. That is axis 4's exact failure mode — one English word for two Latin words, in the gospel whose argument is that opposition — and it is the gentes family's ninth member entering from outside the axis's own list. paganus occurs once in the whole work, which is why no stint's own count could show it. Changed to "the pagans", which collides with nothing: "Gentile", "heathen", "nation" and "the Gentile world" are all taken, "pagan" was unused in 74,630 words.

(b) @0120C (chunk 0021) — sacramenta was rendered "mysteries". An alium ad haec sacramenta missurus es? had come out "art thou about to send another for these mysteries?" ⚑ The stint's own file names this site and states the opposite rule: cruces-0016 §(a) 6 reads "sacramentum → sacrament (0108C, 0113B, 0120B)". So this is an execution slip against a declared policy, not a policy divergence, and the fix is the stint's own. Changed to "sacraments". With it, axis 5 is 1:1 everywhere in the work except the two declared @0135B sites.

(c) @0080D (chunk 0006) — a supplied italic span, raise up. Migne prints Suscitare signanter dicit with no italic. The English had italicized the mentioned word. ⚑ The same stint's own file forbids this (cruces-0006 §(d), on words-as-words: "rendered with our own quotation marks (Pattern 6: « » is Migne's, " " is ours) rather than by adding italic spans the Latin does not"). Changed to "raise up" — the stint's own remedy, which keeps the mention visible and restores the span count.

(d) @0092B (chunk 0010) — a supplied italic span, They do not light,. Migne prints Non accendunt, id est, in roman: it is the gloss picking its lemma back up, not a lemma he sets. Italics removed. With (c), the work's italic spans are now 1:1 with the Latin in every one of the 42 chunks, which is the sacred-marker rule and was the one place it had slipped.

⚑ Nothing else in the English was touched. In particular no [var:], [sic:] or [cj:] was added, withdrawn or reworded at the merge, and no plate claim rests on anything but the two witnesses the stint that made it read.

1.5 ⚖️ WHERE TWO STINTS RULED DIFFERENTLY — the rulings, stated so they can be argued with

(a) THE CONNECTIVE SUBSTITUTION — @0165B declined, @0156A fired, and @0174B fired on a clip. This is the live one the coordinator named, and checking it across all eight ranges sharpens the rule rather than overturning either stint.

The 0032–0037 stint stated its line in advance and named the evidence that weakens it: it fired autem for enim at @0156A because Multi enim sunt vocati opens Mt 22:14, and declined autem for et at @0165B because Migne's lemma is clipped out of the middle of Mt 25:19 (venit dominus … et posuit rationem cum eis), where et cannot stand at the head of a lifted clause and a postpositive autem sits exactly where a clip needs a connective. It then recorded, against itself, that Migne keeps the verse's own connective in four other clipped lemmata in that same range (@0155A, @0155D, @0165D, @0166C). That is model practice and the decline stands.

⛔ But as literally worded — "where the lemma starts" — that rule is broken by the 0038–0041 stint, which fired @0174B: Migne prints Jesum vero flagellatum where Mt 27:26 reads Jesum autem flagellatum, and that lemma is also clipped from mid-verse.

The ruling, and it reconciles all four sites without overruling any stint: the question is not where the lemma starts, it is WHETHER THE RECEIVED CONNECTIVE COULD HAVE STOOD IN THE CLIPPED POSITION.

Two more fire on the same test and were checked: @0106B Si enim foenum agri for Mt 6:30's Si autem (opens the verse) and @0136A Animalis enim homo for 1 Cor 2:14's Animalis autem (opens the verse). Six sites, one rule, no exceptions. The @0165B decline remains the work's one connective decline, and the 0032–0037 stint's §(d) 6 is still the paragraph to argue from if a later reader wants it fired — the plates are already read.

(b) "ye" — 0012–0015 declared "no ye", 0022–0026 declared "ye inside a quoted scripture lemma". Measured at the merge: 26 occurrences, and every single one is inside quoted scripture (24 in an italic lemma, 2 inside a [var:]'s English of the Vulgate). ⛔ So the two declarations do not actually conflict about the gloss's voice, which is what matters: "ye" never once appears in the commentator's own prose, in any of the eight ranges.

Where they do differ is finer, and it is real: a bare plural imperative lemma is rendered without "ye" at every site but three (Petite. "Ask." · Attendite. "Beware." ×3 · Nolite judicare. "Judge not." · Sinite. "Suffer." · Orate. "Pray." · Diligite. "Love." · Benefacite. "Do good." · Nuntiate. "Tell." · Docete. "Teach."), and with it at exactly three (Intrate. "Enter ye." @0110A · Et alligate. "And bind ye." @0132D · Vigilate. "Watch ye." @0164D). ⚑ All three of those stints are internally split on it — 0016–0021 prints "Ask." and "Beware." beside "Enter ye."; 0022–0026 prints "Suffer." beside "And bind ye."; 0032–0037 prints "Give us of your oil," and "And give it to him," beside "Watch ye." So this is not one stint's house style against another's; it is the same hand choosing per verse.

RULING: this is not swept, and the reason is a rule of this edition and not fatigue. Both forms are plural and both satisfy Pattern 17, so neither is a defect; and the two available sweeps are each worse than the variation. Adding "ye" everywhere would import the Douay's phrasing into lemmata that do not carry it — 7a″ conformation, the defect this edition exists to catch. Removing it everywhere would strip the received English from twelve verses that plainly have it (Come ye after me · Go ye not into the way of the Gentiles · Teach ye all nations · Be ye merciful · Receive ye the Holy Spirit · Have ye not read · It is I, fear ye not · Hear ye him · Tell ye the daughter of Sion · Give ye them · Come, ye blessed · Watch ye therefore). The work's rate — 26 in 74,630 words — sits inside the corpus band (Mark 28 in 41,831; 9004 57 in 36,250; John 8; Luke 9). The inventory above is the record that this was measured, not missed.

(c) The [sic:] marker's own formatting — italic in two, roman in two. @0097A [sic: nomo] and @0158D [sic: in est] italicize the quoted word; @0131A [sic: Excaevavit] and @0135B [sic: Constristatus] do not. This is not drift and no edit was made. Checked at the merge: both roman ones sit inside an italic run and both italic ones sit in roman prose. Nesting … inside … is impossible in this dialect and would break the span parity the sacred-marker rule requires. The convention this work follows, stated here so it is not "fixed" later: italicize the quoted word wherever the marker stands in roman prose; leave it plain where the marker stands inside a lemma. The corpus majority form is the italic one (1,537 to 366), and this work uses it wherever it can.

(d) The [var:] threshold — four wordings, one bar. The 0000–0005 stint stated it (substitution fires; ellipsis, and what Migne's own gloss already discusses, decline); 0016–0021 adopted it and extended it to inflectional change (tense, mood, stem) on the ground that the English word on the page changes, citing 0000's own @0069A as precedent; 0022–0026 adopted 0006–0011's worked application of it; 0027–0031 and 0032–0037 say in terms that they inherited the bar and did not loosen it; 0038–0041 restated it in one line (substitution or addition fires; elision or transposition does not). Read together these are one bar, and the 73 firings are consistent with it. Recorded here because four different wordings of one rule is exactly what a later sweep would mistake for four rules.

(e) Orthography and proper names never fire, and all eight ranges agree. Esrom/Esron · Josaphat/Josophat · Eliachim/Eliacim · Sadoch/Sadoc · Nazareus/Nazaraeus · Gerazenorum · Scariothis · Osanna · saba for Sabaoth; Nunquid/Numquid · quandiu/quamdiu · Quicunque/Quicumque · Abscinde/abscide · Dextera/dextra · circumibat/circuibat · Iota/jota · Inexstinguibili/inextinguibili · Nummulariis/numulariis · patrem familias/patremfamilias. Firing on these would bury the 73 real divergences in noise, which is Pattern 9's own argument.

1.6 Pattern 17 — thou is the singular, you the plural

Applied by number and never by reverence, in all eight ranges, and every stint itemized its own switches (they are preserved below). The switch is visible inside single glosses at @0152B and @0160A. you is the plural throughout the gloss's own voice; ye is §1.5(b)'s.

1.7 What the English translates: the plate where read, and the whole work was read

Every column of cols 0063A–0178D was read on two witnesses by the stint that translated it — the archive PDF (raw/scans/pl114/patrologiaecurs06saingoog.pdf, PDF page = (col+11)/2) and Gallica (ark:/12148/bpt6k54939667, f = (col+1)/2), corner numbers read first on every page. Where the plate and our Corpus Corporum twin disagree, the English renders the plate and the site is filed as a pending TEI patch (§1.8). This is translation-runbook.md 4a″ applied to a whole work for the first time, and it is why the apparatus could fire at all.

⭐ The archive copy of PL 114 is the worse witness, and the eight independent measurements are now the strongest statement of that in the corpus: 12 of 13 refuted · 20 of 21 · 11 of 12 · 9 changed answers · 14 of 16 · 17 of 17 · ~34 · 19 of 19. ⛔ And it does not follow that Gallica is right: @0131B the archive prints a terminal stop that Gallica and our twin lack, and Gallica's own sni for sui was caught elsewhere. The method that found those is the 0022–0026 stint's — read the second witness across the whole range, not only at your candidates — and it is the method this work should be cited for.

1.8 THE TALLIES

⚠ These counts are AS OF THE MERGE, and the step-4a polarity pass is running concurrently and is adding to them. It has already landed one [sic:] this file's own stint files do not have — @0077D justitiam for tristitiam (Jn 16:21, mulier cum parit tristitiam habet, which is the verse the clause quotes and which its own next words answer). ⭐ That site is worth naming for a reason bigger than the count: four earlier readers walked past it, and it is the class runbook 4a exists for — the plate hands a lemma-and-gloss commentary its own contradicting neighbour. ⛔ Re-derive these numbers before deploy; do not read them off this table.

Markers, by stint (chunks · [var:] · [sic:] · [cj:] · [d:] · [ed:] · [n:]):

Zero [d:], zero [ed:], zero [nt:], zero [cn:]. plate-gate.mjs 9002 passes on all of them — 78 markers standing on 64 recorded reads at the time of writing.

The [sic:] — each on two witnesses and each quoting a word still verbatim in our Latin: @0097A nomo · @0131A Excaevavit · @0135B Constristatus · @0158D in est, and from the polarity pass @0077D justitiam. ⚑ Four or five in 45,922 words is what this class looks like when it is measured. Four of the eight stints report zero, and each says why: every defect they found was the digitization's, and at those sites the marker cannot even be written — it would fail the verbatim-substring guard and would accuse Migne of Corpus Corporum's error.

The three [cj:]: @0107B Dei for diei · @0146B Jesus for laesus · @0170B non occidant for occidant. All three keep Migne's word in the English and stand the conjecture beside it.

TEI patches — 26 sites, and every one of the 26 is Corpus Corporum's. None is Migne's.

⭐ One of the 26 is the load-bearing one: @0142A Etiam for Eliam. With Eliam restored the lemma matches Mt 16:14 word for word — so until the patch lands the phrase search fails, the lemma looks divergent, and every instinct points at Migne's plate. That is launch-brief §7's fifth cause of a failed ✓, and it is the only one that reads as evidence against Migne while being evidence against us.

Press variants left UNMARKED — 20 sites, no [sic:], no patch, crux only. The two copies are two printings, not two photographs, so where both are clean there is no one to accuse: @0073A Bethehem/Bethlehem · @0121D and @0122A terminal stops · @0125A accipendum/accipiendum · @0131B a terminal stop · @0153A aliqui/alii · @0154C patrisfamilias/patris familias · @0155D restitum/vestitum · @0157C Cujus/cujus · @0158B proselytum/Proselytum and tollendo. Exemplo/tollendo, exemplo · @0164B quinquepertita/quinque pertita · @0172A rerurrectionis/resurrectionis · @0172C the VERS. 4. dash · and six further VERS. n.-- dash disagreements in cols 0124A–0137C, in both directions. ⛔ The dash is not stable type and is evidence of nothing — never mark it, never patch it, and never let its presence or absence decide whether a span is a verse lemma. The VERS. number and the CAPUT head decide that.

Plate reads — 64 read records in data/plate-reads.json, covering the whole work on both witnesses, corner numbers checked first on every page, and the negatives recorded: the feet of every page in cols 0096–0108 and of the last page of the work were inspected on both copies and found empty. ⛔ That check exists because of @0103C, where the archive copy's raised asterisk and empty page-foot fabricated an apparatus layer in Migne's own editorial voice and Gallica prints a plain colon. Never recover a [cn:] or an [nt:] from one witness.

Lemma spans — 1,826 across the work, and all eight counts check. 204 · 241 · 161 · 335 · 211 · 217 · 268 · 189 = 1,826, which is data/briefs/9002-lemmata.counts.json's master total, and every stint counted its own Latin against its own file's header and reported the number. The 9001 mis-split does not recur here.

The eight ⚑ MATCHES ONLY ELSEWHERE candidates, all adjudicated, none left open: @0080B, @0093A, @0095B, @0106B, @0161D real divergences ([var:] fired) · @0157D a clipped lemma (convention 3, not a divergence) · @0169D a word division (Etsi/Et si, not a divergence and not a [sic:]) · @0165B a declined connective (§1.5(a)).

1.9 What could NOT be reconciled

Nothing was left unreconciled that bears on the English. Two things are recorded as open rather than settled, and both are open at the source, not at the merge:

1. @0097D fiili for filii is filed by the 0012–0015 stint as PROPOSED, NOT CERTAIN — read it again before applying. It is the only one of the 26 patches not marked certain, and it must not be applied on this file's authority.
2. The 26 patches are pending. Until they land, our Latin column prints in fifteen places a word Migne did not print, and the English renders the plate at those places. That divergence between the two columns is deliberate and documented per site in data/briefs/9002-PENDING-TEI-PATCHES-*.md; it is not a defect in the English.

The eight stint files follow, preserved unedited. Headings are demoted one level; nothing else is changed. Where a stint file's ruling has been sharpened or corrected above, §1 says so by column and the stint's own argument is still here to be argued with.

Preserved from cruces-0000.md, unedited; headings demoted one level. Companion files: data/briefs/9002-PLATE-READS-0000.json, data/briefs/9002-PENDING-TEI-PATCHES-0000.md, data/briefs/9002-lemmata-0000.txt.

9002 — cruces, stint 0000–0005 (cols 0063A–0078D)

Glossa ordinaria, Evangelium secundum Matthaeum, PL 114. Chunks 0000–0005: PROLOGUS. through CAPUT II. (cont.), 5,831 Latin words, the opening of the work.

Companion files from this stint: data/briefs/9002-PLATE-READS-0000.json (every page opened, both witnesses) and data/briefs/9002-PENDING-TEI-PATCHES-0000.md (three sites where our Latin has departed from Migne).

(a) THE POLICY I APPLIED — stated before the findings, so it can be argued with

1. Markers. Zero [sic:], zero [cj:], zero [d:], six [var:].

2. When a [var:] fires. Migne's quoted scripture takes a [var:] where his words assert something different from the received text — a substitution. It does not fire where he merely shortens a verse (an ellipsis, which is what an abridged recension does constantly and which asserts nothing new), nor where Migne's own gloss already discusses the divergence.

That second decline class is Pattern 18a's ❌ "anything the EDITION already glosses" carried across to Pattern 14, and it does real work here: this book twice tells the reader in as many words that it is not quoting the Vulgate (@0072A, @0078D). A marker there would tell him twice and blur whose observation it was — exactly the argument Pattern 12 makes about Migne's own ( sic ). I would rather be argued out of this than have applied it silently, so both sites are itemized below with what they would have said.

I have not applied the ellipsis decline to the reverse case: where Migne shortens a verse and the abbreviation itself is what the gloss argues from, that is a substitution in effect and it fires (@0075B Puer for Parvulus).

3. When a single-word [var:] fires. Launch brief §2 requires a witness of its own, because a one-word divergence has two possible owners. Two of my six are single words (@0069A invocavit, @0069B adjuvit), and for both, the second owner is eliminated: I read the word on both witnesses. Ownership is settled by the plate, not by the argument, which is the whole point of reading inline. Both also satisfy the other limb — the gloss leans on the form.

4. Proper-name orthography is not a divergence. Migne's Esrom / Josaphat / Eliachim / Sadoch / Nazareus against the Clementine's Esron / Josophat / Eliacim / Sadoc / Nazaraeus are spellings of transliterated Hebrew names, not different readings, and none changes what any sentence says. All five are listed below as checked and declined. Firing on these would bury the four real divergences in noise, which is Pattern 9's argument, and it would also misrepresent the class: these are not the Douay-conformation defect 7a″ exists to catch, since nothing in the gloss turns on the letter.

5. Every claim about the type rests on two witnesses. I read all eight pages of my range on the archive PDF and all eight leaves on Gallica, corner numbers first on each. This changed the answer twelve times — see §(e). Nothing in this file rests on one copy.

6. Where the plate and our twin disagree, the English renders the plate and the site is filed as a pending TEI patch. Three sites, all of them Corpus Corporum's error, none of them Migne's.

7. The vocabulary axes bind, including where they read oddly. sacramentum → sacrament even where the sense is "sacred sign" (see §(d) 1); gentes → Gentiles in the Judaei/gentes typology and nations in a generic quoted verse; singular gens → nation. The gloss's own voice is not archaised (dicit/ait → says); thou is singular and you plural throughout.

(b) SPAN COUNT — 204 against 204, and the count is paragraph-bounded

data/briefs/9002-lemmata-0000.txt declares 204 spans for chunks 0000–0005.

Counting my own Latin paragraph by paragraph (per the file's own warning that a bare total can agree falsely when two errors cancel):

204 = 204, exact. No paragraph in my range carries an odd number of asterisks and no paragraph carries an unmatched guillemet, so the two failure modes that make a bare total lie are both excluded. Matthew is an italic-lemma book, like Luke and Mark and unlike Genesis: there is not one guillemet in 5,831 words.

⚑ One note on the file's own header, which says "bands 0065A–0078D" while my range is 0063A–0078D. That is not a mis-split: the Prologus (0063A–0065A) carries exactly one marked span, the rubric Incipit Evangelium secundum Matthaeum., and it is listed, at 0065A, which is where it stands. The inventory is complete for my range.

(c) THE LEMMA DIVERGENCE LIST — all 204 spans, collated against sources/vulgate/clementine-flat.txt

Migne's citation sigla ((Ose. XI) for what is Mt 12:18, (Psal. CXXI) forms, (Ibid. …)) pass through verbatim under convention 4 and are not treated as divergences; a wrong siglum is Migne's attribution practice, and refDisplay keeps it while refKey resolves it.

C1. The six that FIRED — substitutions the gloss builds on

C2. DECLINED — Migne shortens the verse (ellipsis), asserting nothing new

Each was collated word by word; the words Migne does print agree with the received text.

• 0067B quia nihil ad perfectum adduxit lex — Hbr 7:19 has nihil enim. quia is the splice into his own sentence; both are causal and the clause is unchanged.
• 0067B Egredietur virga de radice Jesse — Is 11:1 opens Et egredietur.
• 0068B Non videbo Dominum Deum, … non aspiciam hominem ultra, etc. — Is 38:11 continues in terra viventium; Migne's own etc. marks the cut.
• 0068C Omnium iniquitatum ejus non recordabor — Ez 18:22 has quas operatus est between.
• 0068C Dirigatur oratio mea sicut incensum, etc. — Ps 140:2 continues in conspectu tuo.
• 0069A Pater sancte, serva eos quos dedisti mihi — Jo 17:11 has in nomine tuo after eos.
• 0069B Justus Dominus et justitias dilexit — Ps 10:8 opens Quoniam justus.
• 0069B Omnis qui videt Filium … habebit vitam aeternam, et ego resuscitabo eum, etc. — Jo 6:40 has the purpose-clause ut … habeat and continues in novissimo die. With Migne's ut gone, the indicative habebit is what the free-standing sentence requires; that is quoting practice, not a reading.
• 0069C Sic Deus dilexit mundum, ut Filium suum daret — Jo 3:16 has Sic enim and unigenitum before daret. ⚠ The nearest thing in this range to a decline I could be argued out of, since unigenitum is a substantive word to lose from a famous verse. It stays declined because the clause Migne prints says what it says: the gloss adduces it for daret (Mathan. Donans vel donatus … qui dedit bona hominibus), and nothing in the argument touches the only-begottenness.
• 0069B Deus meus ut quid dereliquisti me — Mt 27:46 doubles Deus meus. The single occurrence is deliberate: the lemma being glossed is Eliud, id est, Deus meus, so one is the point. Declined on the sense, not merely on the ellipsis.
• 0070B Unxit te Deus tuus. — Ps 44:8 / Hbr 1:9 read unxit te Deus, Deus tuus. A contraction, introduced only as de quo propheta, and the argument is on unxit.
• 0072C Donec ponam inimicos tuos scabellum pedum tuorum — Ps 109:1 verbatim. No divergence.
• 0075C Ego sum via, veritas, et vita. — Jo 14:6 has via, et veritas, et vita.
• 0076B ingredieturque Aegyptum — Is 19:1 has et ingredietur; a -que for an et-.
• 0077D Vos plorabitis, etc. — Jo 16:20 has quia plorabitis, et flebitis vos; the pronoun is fronted, the person is unchanged.
• 0074D si Moysen et prophetas non audiunt: neque si quis ex mortuis, etc. — Lc 16:31 verbatim as far as Migne's etc.
• 0076A In Aegyptum, et erat ibi, — the lemma splices across the Mt 2:14/2:15 verse break (secessit in Aegyptum : et erat ibi). That is why the pre-scan marked it ⚠; it is not a divergence.
• 0068A Ego ero illi in Patrem, quia tanquam Dominus omnia quaecunque voluit fecit — a conflation, not a divergence: the first half is 2 Rg 7:14 (Ego ero ei in patrem), the second Ps 113:11 (omnia quaecumque voluit fecit), with quia tanquam Dominus Migne's own splice. Declined because a [var:] names one received reading and there is no single verse here to name; the crux is the right home. Recorded so a later reader does not read the ⚠ in the pre-scan as an unexamined site.

C3. DECLINED — Migne's own text already glosses the divergence

• 0072A Ecce virgo in utero habebit [n: (Isa. VII)]. The words are Mt 1:23's; Is 7:14 reads Ecce virgo concipiet. A [var:] would have said exactly what Migne says three lines later in his own voice: "Sensum Isaiae sequitur, non verba. Pro concipiet, ponit: In utero habebit: et vocabunt, pro vocabis, vel vocabitur." He names the divergence, both halves of it, and gives his reason. Declined.
• 0078D Nazaraeus de radice ejus ascendet [n: (Isa. XI)]. Is 11:1 reads flos de radice ejus ascendet. This is the sharpest single divergence in the range and it is declined for the same reason: Migne introduces it as "eisdem verbis in Isaia juxta Hebraicam veritatem"*, i.e. he tells the reader in advance that he is quoting the Hebrew and not the Vulgate. ⚑ Note the gloss twelve lines above expounds flos campi and unde flos interpretatur, so an English reader meets flower and Nazarene* in the same column with Migne's own explanation between them.

C4. CHECKED AND AGREEING WITH THE CLEMENTINE EXACTLY

Verse lemmata, verbatim: Mt 1:1 Liber generationis · Filii David · Filii David filii Abraham (Migne prints no comma where the Vulgate has one; punctuation, carried as printed) · 1:2 Abraham genuit Isaac: Isaac autem genuit Jacob (Migne's colon for the Vulgate's stop) · Judam et fratres · 1:3 Judas autem · 1:4 Aram autem · 1:5 Salmon autem genuit Booz · Booz autem · Ex Ruth · Jesse autem · Regem · 1:6 David autem rex · Uriae · 1:7 Salomon autem · Roboam · Abias · 1:8 Josaphat · Joram autem genuit · 1:9 Ozias autem · Ezechias · 1:10 Manasses · Amon · 1:11 Josias · 1:12 Jechonias · Salathiel · 1:13 Zorobabel autem genuit Abiud · Abiud · Eliachim · 1:14 Azor · Sadoch · Achim · Eliud · 1:15 Eleazar · Mathan · 1:16 Joseph · Virum Mariae · 1:17 Omnes itaque · 1:18 Christi autem generatio · Sic erat · Cum esset desponsata · Antequam convenirent · Inventa est in utero habens · 1:19 Justus · Et nollet eam · 1:20 Haec autem eo cogitante · Ecce angelus Domini · Fili David · Noli timere · Conjugem · In ea natum est · De Spiritu sancto est · 1:21 Pariet (×2) · Et vocabis nomen ejus Jesum · Ipse enim · 1:22 Hoc autem · Ut adimpleretur · 1:23 Ecce virgo (×2) · In utero habebit · Emmanuel · 1:24 Exsurgens autem Joseph · Fecit sicut praecepit · Angelus Domini · Et accepit conjugem · 1:25 Donec peperit · Primogenitum · 2:1 Cum ergo natus · Cum ergo · In diebus · Venerunt · 2:2 Ubi est, qui natus est rex · Vidimus enim stellam · 2:3 Audiens autem Herodes (×2) · Turbatus est · Cum illo · 2:4 Et congregans omnes principes sacerdotum · Ubi Christus nasceretur · 2:5 In Bethlehem Judae · 2:6 Et tu, Bethlehem · 2:7 Clam vocatis Magis · 2:8 Adorem eum · 2:9 Et ecce stella quam viderant · 2:10 Gavisi sunt gaudio magno · 2:11 Et intrantes domum · Puerum cum Maria matre ejus · Adoraverunt eum · Et apertis · 2:12 Et responso accepto · Per aliam viam · 2:13 Accipe puerum · 2:14 Qui consurgens (×2, at 0075D and 0078A) · Nocte · 2:15 Ex Aegypto vocavi · 2:16 Tunc Herodes videns · Occidit omnes pueros · Et in omnibus finibus ejus · A bimatu · 2:17 Tunc adimpletum est · 2:18 Vox in Rama (×2) · Audita · 2:19 Defuncto autem Herode, ecce · 2:20 In terram · Defuncti · 2:22 Timuit illo ire · Secessit in partes · 2:23 Nazareth · Per prophetas.

Inline scripture, verbatim: et benedicentur in semine tuo omnes gentes terrae (Gn 22:18) · Ego hodie genui te (Ps 2:7) · Ecce puer meus quem elegi (Mt 12:18, under Migne's (Ose. XI)) · Excelsus super omnes gentes Dominus (Ps 112:4) · Voluntarie sacrificabo tibi (Ps 53:8) · in novissima tuba (1 Cor 15:52) · Ecce Dominus Deus in fortitudine veniet (Is 40:10) · Emitte agnum, Domine, dominatorem terrae (Is 16:1) · Ecce odor filii mei sicut odor agri pleni (Gn 27:27) · Dominus fortis et potens (Ps 23:8) · Speciosus forma prae filiis hominum (Ps 44:3) · Similis ero Altissimo (Is 14:14) · Nemo ascendit in coelum… (Jo 3:13) · Dispone domui tuae, quia morieris tu et non vives (Is 38:1) · resurrectio et vita (Jo 11:25) · Nuntiabo nomen tuum fratribus meis (Hbr 2:12 verbatim, under Migne's Psalm siglum — the wording is Hebrews', not Ps 21:23's Narrabo; his citation, his practice, carried) · Deus meus sperabo in eum (Ps 90:2) · Ecce Dominus ascendet super nubem levem (Is 19:1) · Tristitia vestra vertetur in gaudium (Jo 16:20).

⚑ Two ✓ in the pre-scan that a careless reader would have taken as clearance, checked and sound: Ecce puer meus quem elegi scores off Mt 12:18 inside a commentary on Mt 1:4 — it is Migne's proof-text, correctly quoted, under a wrong siglum; and in novissima tuba scores off 1 Cor 15:52, which is right, in a gloss on Rahab. Neither is a harmony parallel answering for the wrong evangelist, which is the failure the brief warns of.

C5. NOT SCRIPTURE — name-etymologies and technical tags, not collatable

The single-word spans the pre-scan could not search are, with two exceptions, Jerome's onomastic glosses, not quotations, and no Vulgate collation applies to them: Genuit · genuit · Qui fuit (Luke's genealogical formula) · Christus electus · excelsus · sensibilis · fames · beatitudo · impetus · videns · festinans · lux mea Deus · pacificus · Judicans · judicium Dei · fidelis · nutritius · salus Domini · incensum · resurrectio Domini · praeparatio Domini · pater meus · frater meus iste · Deus meus adjutor · germen novellum · flos campi · flos. The two exceptions are Incipit Evangelium secundum Matthaeum. (0065A), a rubric, and usque ad ×8 (below).

⚠ Deus meus adjutor. scores ✓ Ps 17:3 in the pre-scan. It is not a quotation of Ps 17:3 — it is the etymology of Eleazar, and Migne quotes Ps 90:2 immediately after it as the proof-text. Recorded because the ✓ invites the opposite reading.

C6. usque ad — the etc. splice test, run per occurrence

12 occurrences. 8 are the formula, 4 are ordinary Latin. Every one of the eight formula occurrences is preceded, immediately, by etc., and stands inside an italic span; all eight are rendered "as far as":

@0065A (ut liber Genesis, etc.,) · @0070C (et ut per Joseph, etc.,) · @0072D (locus et tempus, etc.,) · @0073A (videatur impletum, etc.,) · @0073A (praevalet namque, etc.,) · @0074A (ad agnitionem veritatis, etc.,) · @0075B (munera aperta, etc.,) · @0075D (Dominum majestatis, etc.,).

The four ordinary uses carry no etc. and stand outside every italic span; each is rendered by sense and is named here because the brief requires a departure from the axis to name its Latin:

• @0065D A baptismo enim incipiens usque ad Deum ascendit — Luke's genealogy ascending; rendered "ascends as far as God". ⚑ The English words coincide with the formula's, by accident of sense and not by application of the axis.
• @0068D Omnes qui a David usque ad transmigrationem Babylonis numerantur — "from David to the transmigration".
• @0076D a filio unius noctis usque ad filium duorum annorum — temporal, "from a son of one night to a son of two years".
• @0077A gementis usque ad solium superni ascendit judicis — spatial, "ascends to the throne".

A fifth, @0077B auditus est usque Rama, is bare usque and is rendered "as far as Rama".

(d) CRUCES OF THIS RANGE, BY COLUMN

1. @0063B, @0064A, @0064B, @0065A, @0066A, @0076A — sacramentum under the axis, and it reads oddly

The mysterium/sacramentum 1:1 axis is settled and I applied it without exception. It is worth saying plainly what that costs in this work, because the next Matthew stint will meet the same thing and should not silently re-open it. Matthew's Prologus and CAPUT I use sacramentum in its patristic sense of a sacred sign or hidden significance, not of a rite:

• @0064A propter virtutem sacramenti → "on account of the power of the sacrament" (why the gospels are four and not more);
• @0066A ne ultra numerum sacramenti → "beyond the number of the sacrament" (the twelve);
• @0064B scribens sacramenta Divinitatis / @0065A exprimitur Divinitatis sacramentum;
• @0065D formam sacramenti aperit (Luke's ascending genealogy);
• @0076A Quod fugit, sacramenti fuit non timoris → "was a matter of sacrament, not of fear".

Set beside them, @0064A jucundi mysterii sibi consciae and @0074A tantae visionis mysterium take "mystery". The distinction the axis preserves is real and visible in the Latin column; I record the awkwardness rather than smooth it, and I have not invented a third word.

2. @0064A — the unexpressed subject of nolebant, and it has no antecedent in the sentence

unde et aliorum evangelia deciderunt, nec recepta sunt: quia nolebant praefinitum numerum cessari. Rendered "they were unwilling". Naming the "they", as the brief requires: it cannot be the gospels (neuter plural, and they are the object of the unwillingness), and it cannot be the authors of the other gospels, which would make the sentence say the apocryphists enforced the canon. It must be those who fixed the canon — the Fathers, the Church. There is no antecedent anywhere in the paragraph, so this is supplied from the argument alone, and I say so here rather than let a bare "they" pass as though the Latin had named someone.

3. @0065B, @0065C, @0067B, @0070B, @0072B — the evangelist as the unexpressed subject

Repeatedly a third-singular verb has no subject and the subject is Matthew, carried from the head of the chapter: aperiret (0065B), Amplectitur (0065C), ponit/reservat (0070B), astruit (0072B), ponit (0074C). Each is secure — the paragraphs around them name Matthew explicitly (Matthaeus generationem incipit, Matthaeus ponit genuit) — and I record the class rather than each instance.

⚠ One in the class is NOT secure and I have not hidden it. @0067B Hic autem ideo dicitur Jesse: ut sciret eum venisse. Rendered "that he might know". Two readings construe: the Jewish reader/hearer (the argument two paragraphs above is about disclosing Christ Judaeis), or Isaac in the following sentence. I took the first, on the run of the argument; the second cannot be excluded from the words.

4. @0065D — genuit against generavit, a distinction English has no exponent for

Matthaeus ponit genuit, et non generavit. Both are "begot". Rendering both would delete the sentence's whole point. Migne sets neither in italics here (though genuit is italic two clauses earlier), so italicising them in the English would break span parity and put markup on the page that the plate does not have. I used our own double quotes (Pattern 6: " " is ours, « » is his) and kept the Latin forms: Matthew sets down "genuit," and not "generavit,". The same device carries @0072A's concipiet / vocabunt / vocabis / vocabitur — there rendered in English inside our quotes, because Migne's point at that place is about sense, not about the Latin forms as forms.

5. @0067B — a quae clause with no verb, and a bracketed supply

Ruth significat Ecclesiam quae est de Moab … quae ex patre diabolo et petra deserti propter idololatriam de qua Deus suscitavit filios Abrahae. The second quae clause has no predicate. Read on both witnesses; Migne prints it so. I rendered "which [was] of her father the devil", bracketing the supplied verb, which is the corpus's convention for our own supplied words. The supply is licensed by Migne's own parallel eleven words earlier: ex patre diabolo erat. No marker: nothing here is defective type, and an anacoluthon is not a false assertion.

6. @0067C — the negation sits on solus, not on the verb (test 2a)

Non quod solus in hac genealogia rex, sed quia primus rex de Juda. The trigger condition is exactly the launch brief's: a negative, a verb-shaped clause, and a quantifier beside it. The sed clause proves the host: David is king in this genealogy; what is denied is that he is the only one. Rendered "Not that he alone is king in this genealogy, but because he is the first king of Judah." Recorded as handled, not as a defect — this is the 8990 @0614B class caught rather than missed.

The same test was run on, and passed by: @0066A non feminas, sed mares interimebat (host feminas); @0068B flevit non propter longiorem vitam … sed quia dubitabat (host is the cause-phrase — he did weep, which is the 8960 @0740B model); @0069D non Elimelech … sed Obeth vocavit (host Elimelech); @0070A non per Mariam, sed per Joseph; @0071B non concubitu, sed affectu; @0074C non ex ignorantia, sed ex cordis duritia; @0076A sacramenti fuit non timoris (host timoris); @0076D non in Judaea tantum, sed ubique; @0077C non tam morte translatos, quam supplicio peremptos.

7. @0068B — dubitabat ne is an indirect question, not a negation

sed quia dubitabat ne promissio Dei impleretur. Read as a purpose clause this says Hezekiah feared the promise would be fulfilled, which is the reverse of the argument (he is childless and fears it will fail). No negative is at stake and none was dropped: dubito ne is the ordinary "doubt whether" construction, and I rendered "because he was in doubt whether the promise of God would be fulfilled". Recorded explicitly because a negation-counting check will see a ne here and a reader may wonder where it went. It is in "whether".

8. @0068D — in fide praecedentis where the parallel wants in fine

unus et idem est in fide praecedentis et in principio sequentis, three lines after unus … pater sit in fine praecedentis, alter in principio sequentis. Read on both witnesses: both print fide. So it is Migne's, not ours, and Pattern 7 renders it: "one and the same is in the faith of the preceding and at the beginning of the following."

⛔ A [cj:] was weighed and declined, and here is why the rival lost. fine is near-certain on the parallelism, and the site has the shape Pattern 18 was written for — a real word whose faithful English is wrong. But 18a's bar is that the English must assert something false, and 18a's explicit ❌ is "a merely awkward or elliptical sentence." "In the faith of the preceding" asserts nothing the author is arguing against; it is opaque, not false. The ✅ extension for negations and correlatives does not reach it either: in principio is not a correlative whose answering member is in fine in the way 11535 @1231B's quam demanded a comparative. The conjecture lives here, and in the text stands the printed word.

9. @0068D — the full stop our twin lost (patch, not marker)

Both witnesses print sequentis. Sed mystice; our TEI runs the two sentences together. The English carries the stop, under the plate-where-read rule. Filed as patch 1. Nothing here is Migne's defect, so no [sic:] — and none could be written, since the marker must quote a word that is in our Latin and the defect is an absence.

10. @0069C — Heriles and libri Dierum

viri studiosi ex Judaeis dicti Heriles. Read on both witnesses. A real, well-formed word, and by the launch brief's own prior a real word that reads oddly is Migne's until a plate says otherwise — the plate says it is his. (The people usually called Desposyni.) Rendered as printed; no marker. Likewise ex libris Dierum rendered "from the books of Days", the Hebrew title of Chronicles, not silently modernised.

11. @0069D — qui per Mathan descendit de David

Mathat is said to descend from David per Mathan, four words after Mathan has been said to descend per Salomonem; the genealogy wants per Nathan. Both witnesses print Mathan. A real name, rendered as printed. No [cj:]: the sentence is a tangle either way and the reader has the two lines of descent in front of him.

⚑ The archive copy alone reads "Mathan, qui per Solomonen descendit" here — a non-word, refuted by Gallica, and a good illustration of why a one-witness read of this volume is worthless.

12. @0069B — a question Migne closes with a full stop

Quis alius quam ille qui homo fieri voluit … de quibus diceret: Nuntiabo nomen tuum fratribus meis. Migne prints a stop, not a question mark, on a clause that opens Quis alius quam*. Pattern 8: never supply a mark the plate lacks. The English ends with a stop. Read on both witnesses. This is the mirror of the class Pattern 8 protects and it costs nothing to obey.

13. @0070C — Non quod postea convenerint printed twice, and it is NOT a dittography

Non quod postea convenerint: ut si diceremus antequam poenituerit, morte praeventus est. Non quod postea convenerint; sed ostendit proximum tempus nuptiarum…

Read on both witnesses; both print both copies. No [d: …] fired, and the reason matters because Pattern 11a asks which copy is the intrusion. Neither is. The two runs are punctuated differently and each carries its own distinct continuation — the first an analogy (ut si diceremus…), the second the positive explanation (sed ostendit…). An eye-skip duplicates a run and its tail; this is the abridgment splicing two glosses on one lemma, which is what the recension does everywhere. Marking one would quarantine a clause Migne meant.

14. @0071D — oerum for eorum (patch, not marker), and why the plate decided it

Our twin's oerum is a non-word; the same sentence is restated four lines later under a HIER. siglum with eorum spelt correctly in our own twin. That internal corroboration is textbook, and by reference_plate-read-triage it proves an error exists and says nothing about whose. Both witnesses print eorum. Ours. Filed as patch 2, English renders eorum, no [sic:].

15. @0074D — an anchor that falls inside a word, and where I put it

Migne's column break at 0074D falls inside promittit (Herodes devotionem pro | mittit). Splitting an English word is not readable, so the anchor stands one word later: "Herod promises [0074D] devotion." Recorded because verify-english checks anchor presence and order but not position (2b), so a displaced anchor passes silently and only a note like this one makes it visible. The displacement is one word.

16. @0073A — Bethehem Judae is a press variant, and the two copies are two printings

Gallica f37 prints the lemma Bethehem; the archive p.42 prints Bethlehem. Both clean, both normally spaced, and the very next line of both copies sets Bethlehem correctly three times. Our twin follows Gallica.

No [sic:]. A [sic:] needs two witnesses and the second witness refutes it. This is 8950 @0171D's Mititte/Mittite exactly — a difference no ink produces, i.e. two printings, which this volume turns out to have: the line breaks of the two copies diverge from the first page of my range onward. The English renders the place-name; the fact is here and in the plate-reads file.

17. @0075B — quem actum esse hominem intelligunt

Read on both witnesses; both print actum. A real word, so Pattern 7 renders it and Pattern 12 has nothing to wrap: "whom they understand to have been brought to be man". The evident reading is factum, and I record the conjecture here rather than in the text. A [cj:] was weighed and declined: the rendering is opaque, not false — nobody reads "brought to be man" as a claim the author is arguing against — and 18a's ❌ covers it. Had I written "made man" I would have made the repair invisible, which is the 7a class this edition exists to expose.

18. @0076A — two elliptical clauses where I supplied an object

• nocte ignorantiae his a quibus ipse recessit reliquit incredulis — reliquit has no expressed object. Rendered "he left them to those unbelieving ones from whom he himself withdrew."
• Omnigenum Deum monstrat venerantem — monstrat has no expressed object. Rendered "He shows [Egypt] venerating a God of every kind", bracketing the supply. Egypt is the subject of the whole paragraph (Fugit in Aegyptum), and the point is Egypt's idolatry against Israel's refusal. An alternative parse taking Deum as the object and omnigenum … venerantem predicatively yields nothing the argument uses. Both witnesses print the clause as we have it.

19. @0077D — Josephus for Joseph

ideo Josephus intellexit Judaeam. Both witnesses print Josephus, in a paragraph that names Joseph twice in the preceding sentence. A real, well-formed name; rendered as printed; no marker.

20. @0077D — the Pattern 17 payoff, visible without apparatus

unde dicitur: Vos plorabitis, etc., sed additur consolatio: Tristitia vestra vertetur in gaudium. Both are plural (Jo 16:20). Apart from the plural imperative Nolite timere at @0068B, addressed to the apostles, these are the only second-person plurals in 5,831 words — vos once and vestra once, both inside this one quotation. Everything else in this range that addresses one person takes thou — the angel to Joseph (Recognosce … de qua tu es, Noli timere, Et vocabis nomen ejus Jesum, eris necessarius), the commentator to his reader (ut eum cognosceres, ut eum intelligas, subaudis), Isaiah to Hezekiah (morieris tu et non vives). So a reader can see, with no apparatus, that the Church's weeping is addressed to a body and everything else here to an individual.

21. @0075C — pius taken two ways, disclosed

isti pio affectu interrogabant → "with devout affection". But at @0070D–0071A sed et pius dum nollet propalare … ex conscientia castitatis justus, ex timore pius, and nec pietas sine justitia, nec sine pietate justitia, I rendered pius/pietas merciful / mercy, because there the word is half of the technical justitia/pietas pair on which Joseph's whole characterisation turns. One Latin word, two English words, in one stint. It is not on any settled axis; I record the split rather than let it pass, since that is how the axes got made.

22. @0066C — sensibilis rendered "perceiving"

Christus est Salmon, id est, sensibilis, quia omnia sentit. "Sensible" is the traditional English of the etymology but is now ambiguous in exactly the wrong direction. I rendered "perceiving", which is what Migne's own quia omnia sentit says the name means.

23. Punctuation, band by band — the mechanical test, and what it leaves

Column anchors are verified 1:1, so they partition both twins identically. Comparing marks band by band across all 66 bands of my six chunks (60 column anchors plus each chunk's head band):

• question marks: zero mismatched bands (7 in the Latin, 7 in the English, each in its own band);
• guillemets: zero, in both — this book prints none;
• colons: zero mismatched bands after I removed four I had supplied (@0065A, @0069C, @0072A, and the head band of chunk 0004);
• semicolons: 10 bands where the English has one more than the Latin, and none where it has fewer. Every one is a comma in Migne that English cannot hold as a comma without splicing. Nothing Migne prints was destroyed anywhere in the range, which is the direction the test exists to catch; and the brief's standing ruling is that comma→semicolon is a rate question, not an absence. I have not swept the other way either.

24. Sentence-type marks Migne prints, all carried

Seven question marks, all his, none supplied and none deleted: @0068D (Sed quare praetermissus est Joachim…?), @0069B (Deus meus ut quid dereliquisti me?), @0070A (Quid ad Christum generatio … cum Christus non ex semine Joseph?), @0072A (Sed quomodo tunc est signum?), @0075B (Cur non et Joseph cum Maria inventus est a Magis?), @0078B (cur non et Galilaeam…?), @0078B (Quaeritur: quare non timuit Joseph ire in Galilaeam…?).

⚑ Pattern 8b was not needed once. The one Quaeritur in the range is already set by Migne himself as a colon followed by a direct question, so the mark rides as printed with no recasting.

25. Verse addresses — one irregularity, carried

@0075A prints VERS. 11. with no .--, where every other address in the range prints VERS. n.--. Our twin has it so. The archive copy prints a dash there and Gallica does not; a .-- is precisely what fading removes, so by the brief's own rule a verse address may not be altered on one witness — and here the witnesses disagree, which is weaker still. Left exactly as our twin has it. Convention 1 makes the address pass through verbatim in any case, and nothing machine-readable consumes it.

26. The ⚑ MATCHES ONLY ELSEWHERE candidates — none in this range

The launch brief lists eight, the first at @0080B. My range is clean of them, as the brief says. I confirm it rather than assume it: no span in cols 0063A–0078D carries that flag in 9002-lemmata-0000.txt.

(e) DECLINED-ODDITY LIST — every suspect word I did NOT mark, with its column

Fourteen of the sixteen entries below are archive-only readings that Gallica refutes. They are listed in full because the count is the finding: had I read one witness, this file would carry a dozen confident claims about Migne's type and twelve of them would be false.

And the suspicions that were NOT about the archive's ink

Suspiciously TIDY forms — brief §4's other direction, and a checked negative

The class that leaves nothing to notice is the one where the digitization has corrected Migne. I looked for it deliberately at the places where our Latin reads smoothly in rough surroundings, and I read whole lines rather than target words. In this range the answer is: three found, and all three run the other way — patches 1–3 are the digitization corrupting, not tidying.

⭐ The negative worth recording (brief §4 asks for these explicitly): the passage most likely to have been tidied is @0067B's verbless quae ex patre diabolo et petra deserti clause, which is exactly the shape a transcriber "fixes" — and both witnesses print it defective, as our twin has it. Our transcription is faithful there. Likewise @0068D's in fide, @0069D's per Mathan and @0075B's actum: four places where our Latin is rough, four places where the plate is rough in the same way. This volume's digitization drops letters; in my range it did not add any.

Preserved from cruces-0006.md, unedited; headings demoted one level. Companion files: data/briefs/9002-PLATE-READS-0006.json, data/briefs/9002-PENDING-TEI-PATCHES-0006.md, data/briefs/9002-lemmata-0006.txt.

9002 — cruces, stint 0006–0011 (cols 0079A–0096B)

Glossa ordinaria, Evangelium secundum Matthaeum, PL 114. Chunks 0006–0011: CAPUT III (the Baptist), CAPUT IV (the Temptation and the call of the first disciples), and CAPUT V as far as Mt 5:37 — 7,153 Latin words, 15 columns, both plates read.

(a) The policy applied, stated before the findings so it can be argued with

1. Plates first, and both of them. Every one of the nine leaves in this range was rendered from the archive copy at 300 dpi and read column by column with the printed corner numbers checked first, and every candidate raised there was then checked at Gallica. Twenty-one candidates, twenty refuted, one confirmed. No [sic:] is fireable in this range and none is fired. The two divergences that survived are Corpus Corporum's and are filed as TEI patches (data/briefs/9002-PENDING-TEI-PATCHES-0006.md); the per-page ledger is data/briefs/9002-PLATE-READS-0006.json.

2. Where the plate was read and our twin diverges, the English renders the plate (translation-style.md §"What the English translates"). Two sites: @0090D maledicta (our twin prints the non-word maledieta) and @0081D the colon after Ac si dicat. Both are patched-pending, and the second is the single deliberate band-level punctuation mismatch in this stint.

3. [var:] threshold — stated, so the declines can be judged against it. Launch brief §2 rules that a single-word [var:] stands only where the divergence has a witness of its own: either the gloss argues from the word, or the form is one that letter-level corruption cannot manufacture. I applied the second limb mechanically:

• fires — a different word (Habitantibus/sedentibus, illi/ei, autem/enim), a changed number carried across two words (fructus dignos/fructum dignum), an added or restructured clause. None of these can be produced by a broken sort.
• declines — a one-letter difference in a form that means the same thing (Abscinde/abscide, Dextera/dextra, circumibat/circuibat, supra/super, Inexstinguibili/inextinguibili, Iota/jota), and a fragment cut at a particle (enim, autem, ergo absent because the quotation starts after them).

13 fired, and every declined span is listed in §c with the reason.

4. Vocabulary axes, applied from the first chunk. turba → crowd (0088D Quadripartitae turbae "fourfold crowds"; 0089A in turba remanere "in the crowd"); no multitudo in the range. daemonium → demon (0095B incenderent daemoniis); diabolus → devil throughout CAPUT IV; no satanas. gentes → Gentiles everywhere in this range, because every occurrence sits in the Judaei/gentes typology (0080C excaecatio Judaeorum et gentium; 0086C the two Galilees; 0086D in medio gentium, vocandi … de gentibus); gentiles → Gentiles (0080D, 0081B); the one singular gens is the Jewish people and takes "nation" — 0092B terminis unius gentis inclusit, "within the bounds of a single nation". No natio, no ethnicus, no gentilitas. mysterium → mystery, sacramentum → sacrament, 1:1 and never crossed (0081C incarnationis mysterium; 0082A Evangelii sacramenta; 0088C mysterium justitiae; 0089B profunditatem … sacramenti; 0092C sacramentis spiritualibus).

5. usque ad — the etc. splice test, applied per occurrence. 34 in the range: 29 formula, 5 ordinary. The five ordinary ones, each rendered by sense and named here as the axis requires:

6. Second person by number, not by reverence (Pattern 17). The range moves between the two repeatedly and the Latin decides every time: plural you for the Pharisees at 0080B (deponitis), for the crowd baptized at 0081B (baptizo vos), and for the apostles through 0090D–0092B (Vos estis sal terrae, regite, constituatis); singular thou for the individual worshipper from 0093C to the end (munus tuum, frater tuus, tu laesisti, reddas, jurabis, potes), for the Son at 0083D (ut tu … placeas) and 0082D (a te), and inside every singular scripture address (Dominum Deum tuum, Erue eum, angelis suis mandavit de te).

7. Nothing supplied. Where Migne's sentence lacks a word English cannot do without, the English is left as broken as the Latin and the site is logged in §d — three of them, at 0079B, 0083D-side 0084A, and 0094B. No bracketed conjecture stands in the running text anywhere in this stint.

(b) Span count against the brief's total

data/briefs/9002-lemmata-0006.txt heads its inventory 241 spans, 0 open-ended.

Counted in my own Latin, chunks 0006–0011, both whole-file and paragraph-bounded (the brief requires both numbers, because an unmatched delimiter defeats a counter in either direction):

241 = 241, on both counts, with 482 asterisks and zero guillemet characters in the range — so the two figures cannot have cancelled a pair of errors. Matthew is an italic-lemma book, like Luke and unlike Genesis. The cut is clean: the file opens at [0078D] VERS. 1.-- In diebus autem., which is chunk 0006's first span, and closes at [0096B] VERS. 37.-- Est, est; non, non., which is chunk 0011's last.

(c) The lemma divergence list — all 241 spans, in order, declines and reasons included

The ✓ / ⚠ / ⚑ column is the brief's mechanical pre-scan; the last column is my verdict after reading the line. ⛔ A ✓ was never taken as clearance: the launch brief's warning that in a gospel the harmony answers for the wrong verse is the reason §c reads the printed references rather than the ticks, and it caught @0087A Habitantibus in regione, which carries no VERS. address and therefore never showed as ⚑ at all.

(d) This range's cruces, by column

⚑ THE THREE MATCHES ONLY ELSEWHERE CANDIDATES — each adjudicated, with the losers named

The launch brief §3 lists three of its eight live candidates in this range and requires me to say which of three explanations each is — a real divergence, a harmony parallel Migne has reached for, or our own file's corruption — and why the other two lost.

1. @0080B VERS. 8.-- Facite ergo fructus dignos. (matched only Lc 3:8; the lemma is Mt 3)

It is a real divergence, and the plural is the Lucan form. [var:] fired.

• Our file's corruption loses first, and decisively. Both witnesses print fructus dignos — archive p.45 and Gallica f40, at 300 dpi and native resolution respectively, in the italic lemma immediately after VERS. 8. —. There is nothing for the digitization to have done.
• "Harmony parallel Migne reached for" and "real divergence" are not rivals here, and the second is what the apparatus records. Mt 3:8 reads Facite ergo fructum dignum poenitentiae, Lc 3:8 fructus dignos poenitentiae. Migne prints the plural under a Mt 3 address, so the lemma diverges from the verse it addresses — which is precisely what Pattern 14 marks. Whether the plural got there because a compositor's eye slipped to Luke, or because the glossator's copy of Matthew read so, is a transmission question, and Pattern 14 forbids speculating about it in the marker. The marker therefore states both readings and names Luke as the place the plural is found, without asserting a route.
• ⭐ The internal evidence pushes the same way and is worth recording: the work's own prose uses the plural twice more in the same two columns, outside any lemma — @0079B Poenitentiam agere est dignos fructus poenitentiae facere, and @0081B Et si moneo dignos fructus poenitentiae facere. Migne's Matthew glossator says the phrase in the plural as a matter of course. That is corroboration, not proof of ownership, and the plates are what settled it.

2. @0093A VERS. 20.-- Dico autem vobis. (matched only Lc 9:27, Lc 12:4, Lc 12:8 …; lemma is Mt 5)

It is a real divergence in a single connective. [var:] fired.

• Our file's corruption loses. Both witnesses print Dico autem vobis — archive p.52, Gallica f47, in the italic lemma after VERS. 20. —.
• Harmony parallel loses, and this is the one where the pre-scan's own list is misleading. The Lucan verses it matched (Lc 9:27, 12:4, 12:8) are not parallels of Mt 5:20 at all — they are simply other places in the corpus where the three words dico autem vobis happen to stand together. There is no Lucan verse corresponding to Mt 5:20's nisi abundaverit justitia vestra; Luke has no Sermon-on-the-Mount antithesis here. So nothing was reached for.
• Real divergence wins. Mt 5:20 reads Dico enim vobis. The lemma is unmistakably Mt 5:20: the very next lemma in the gloss is Nisi, etc. and the one after it Justitia vestra, which are that verse's own following words. And it passes the single-word test's second limb: autem cannot be manufactured out of enim by a broken sort — they share no letter in any position. So the marker stands on the form, not on my confidence.

3. @0095B VERS. 31.-- Det illi. (matched only 2Tim 1:18; the lemma is Mt 5)

It is a real divergence in the pronoun. [var:] fired.

• Our file's corruption loses. Both witnesses print Det illi. — archive p.53, Gallica f48. (The archive appears to drop the final period; Gallica has it, as our twin does. The period is archive ink, and that is in the plate ledger, not here.)
• Harmony parallel loses, and here the pre-scan's single match is a coincidence rather than a parallel. 2 Tim 1:18 is det illi Dominus invenire misericordiam — Paul on Onesiphorus. It has no relation whatever to the bill of divorce, it is not in a gospel, and Migne could not be "reaching for" it: the gloss underneath is entirely about Deuteronomy's libellus repudii and Moses' permission. Two words coinciding is not a citation.
• Real divergence wins. Mt 5:31 reads det ei libellum repudii. illi for ei is a different pronoun, not a letter-level slip, so it clears the single-word bar.

The rest of the range's cruces, by column

@0079B — Migne's sentence has no verb, and both plates agree. The plate prints qui vere poenitet, in se praeteritos errores, deinde erigit animum ad coeleste regnum — there is no verb governing in se praeteritos errores, and the sense wants one (damnat, plangit, condemnat). Read on archive p.45 and Gallica f40, at the same line and with the same wording on both. ⛔ Nothing is supplied: the English reads "he who truly repents, in himself his past errors, then lifts up his mind to the heavenly kingdom", which is as broken as the Latin. No marker is available — nothing here is a non-word, so [sic:] cannot reach it, and the missing word is not a printed word that misleads, so [cj:] cannot either. This is what Pattern 7 asks for and it is logged here because a blind reader will stop at it.

@0084A — the lemma is in the wrong case, and the italic boundary proves it is the lemma. Migne prints Spiritui Sancto (italic on the first word only, roman on Sancto), glossing Mt 4:1's ablative ductus est … a Spiritu. Confirmed on both witnesses. Rendered "To the Spirit, the Holy, who sends strong to the battle those whom he fills" — the dative kept, the italic boundary kept where Migne set it. No marker: a case-ending is the one thing letter-level corruption manufactures most easily, and nothing in the gloss turns on the case.

@0087B — in Jordane organo, where the argument wants in Joanne. Migne prints Christus enim verbum quod clamat in Jordane organo et in omnibus qui ab initio aliquid divinitus dixerunt. The whole paragraph is the voice/word argument about John (Joannes vox, Christus verbum, qui clamat in Joanne at 0079C says it in those words), and organo — "by his instrument" — has no antecedent under Jordane. Both witnesses print Jordane, at normal spacing, so this is Migne's, not ours. Rendered as printed: "which cries in the Jordan by his instrument". Logged, not marked: Jordane is a real, well-formed word in a construction that parses, so [sic:] is closed to it and [cj:] would be a conjecture with no plate behind it.

@0090D — our twin's maledieta, and the plates' maledicta. The only confirmed defect in the range and it is the digitization's. The English renders the plate ("since revilings come from hatred of the heart"); patch filed. No [sic:]: the marker would accuse Migne of an error he did not make, and would in any case be quoting a word about to disappear from our Latin.

@0094B — Migne contradicts himself two clauses apart, on both witnesses. He prints Sic temperavit, ut poenitentiae locum non afferret and then, four lines later, Sed et locum poenitentiae reliquit. Carried as printed — "He so tempered it, that it should not bring in a place of repentance… But he both left a place of repentance" — because 7a is explicit that a printed negation rides including when carrying it makes the sentence contradict itself, and that is exactly what this edition exists to show. afferret (not afferet) confirmed at Gallica f47 after the archive suggested the shorter form.

@0094B — Christus ministro has no verb. Christus ministro, id est angelo, qui colliget zizania — the verb (tradit, carried over from Judex tradat te ministro in the sentence before) is unexpressed. Rendered "Christ to the officer, that is, to the angel, who shall gather the cockle." Nothing supplied.

@0090A — an unexpressed antecedent, named here because two readings construe. Pro suis vel aliorum peccatis, qui est ab irriguo inferiori — the masculine relative has no expressed antecedent in its own sentence. I take it as luctus, "mourning", which the gloss supplies explicitly six lines later (Luctus pro peccatis vel pro desiderio coelesti consolationem meretur), and rendered "which mourning is from the lower watering". The rival is qui = "he who [mourns]", which also construes; it loses because the following sed qui fit ab irriguo superiori est ex desiderio has a neuter/abstract predicate that only luctus takes.

@0093D — offeres, and why the English follows our twin and not the archive. The archive prints si praesens est, offeres. Hoc ex spiritu — a comma and a lower-case lemma, which would make offeres the tail of the previous sentence rather than the lemma of the next. Gallica prints est. Offeres., exactly as our twin has it, and the archive's comma is a broken period. Recorded because it is the one refuted candidate in this range that would have changed the English.

@0093D — ei prosternas with the reflexive unexpressed. Non pedibus, sed animo humiliter ei prosternas — prosternas is transitive and its object (te) is not printed. Rendered "mayest thou humbly prostrate thyself before him"; the reflexive is English grammar, not a supplied Latin word.

@0094A — pro consentiens ex Graeco habetur benignus vel benevolus. Migne discusses three words as words, none of them italicised on the plate. Rendered with our own quotation marks (Pattern 6: « » is Migne's, " " is ours) rather than by adding italic spans the Latin does not carry, so that the span inventory stays 1:1.

@0093A — sed facit quod docet where the sense wants nisi. Qui non solverit, et sic docet, non minimus, non tamen magnus: sed facit quod docet. Carried as "but he does what he teaches"; the connective is Migne's and Pattern 7's third mechanism (a smoothed connective) is exactly what changing it would be.

@0096A — one Non governing two purpose clauses, and no second negative supplied. Non juremus per creaturas ut veneremur eas ultra quam debemus per eas, ut viles existimemus… An English ear wants "nor by them so as to esteem them cheap"; Migne prints no nec. Both ut clauses are left under the single printed Non.

Migne's own citations, all confirmed on both witnesses and all riding verbatim — they are his, not ours, and correcting one would be an emendation: (Matth. X) for Mt 22:14 @0088A; (Psal. CLXI) for Ps 141:6 @0089D; (Luc. IV) for Lc 14:34 @0091C; (Marc. XI) for Mt 5:17 @0094C; (Joan. I) for Mt 4:19 @0087D.

Verse addresses. The sequence in this range runs 1–13, 15, 16, 18–21, 23–27, 29–33, 36, 37 — i.e. it skips 14, 17, 22, 28, 34, 35, and 6–7 fall inside CAPUT IV as 6 and 8. Nothing is reordered and nothing is touched. Per the withdrawn-verse-address rule I checked no number against the plate as a target and altered none.

(e) Declined-oddity list — every suspect word I did NOT mark, in one place

⚑ Twenty-one of these were raised from the archive copy and killed by Gallica. They are listed in full in data/briefs/9002-PENDING-TEI-PATCHES-0006.md under "NOT patches" rather than repeated here; the summary is that the archive copy of pp. 45–53 is over-inked badly enough to manufacture turned letters in both directions, lost letters, and false word-splits, and that it lost 20 of 21 to the second witness. What follows is the list of oddities that survive after the plates — things that are really on Migne's page and that I chose not to mark.

⭐ And the negatives worth recording, because a checked zero is a result. Three things I went looking for in this range and did not find: (i) no [sic:] is fireable anywhere in 0079A–0096B — twenty-one attempts, twenty refuted at the second witness and the twenty-first belonging to Corpus Corporum; (ii) no anchor is displaced by more than two lines — all 70 column anchors in the range were checked against the word they stand beside on the plate, and against Migne's marginal band letter wherever the margin was legible in the render; the largest divergence is a two-line drift at [0083D], [0087B] and [0091C], and the systematic one is that Corpus Corporum sets the <pb> after a word broken across the column break where Migne's column begins with its second half (bapti-|zando at 0079A, praedicatio-|nem at 0082A, nesci-|vit at 0083A, bapti-|zatos at 0084A, su-|perbiam at 0086A). That is the granularity at which the digitization places an anchor, not a defect, and it is uniform across the range; (iii) no Forte or foot-of-page conjecture appears on any of the nine leaves — the bottoms of pp. 45–53 and f40–f48 carry no apparatus layer at all, numbered or asterisked, which is consistent with PL 114's recorded sample and is recorded here as a checked zero for these nine pages.

⚠ One bookkeeping note for the orchestrator, not a crux: chunk 0010's frontmatter declares noteCount: 16 and the chunk body carries 17 [n: …] markers. The frontmatter is copied verbatim into the English as the brief requires, and the marker sequence itself is 17:17 across the pair, so nothing is wrong with the text — but the declared count is off by one and something upstream generated it.

Preserved from cruces-0012.md, unedited; headings demoted one level. Companion files: data/briefs/9002-PLATE-READS-0012.json, data/briefs/9002-PENDING-TEI-PATCHES-0012.md, data/briefs/9002-lemmata-0012.txt.

Glossa ordinaria, Evangelium secundum Matthaeum (PL 114) — cruces, chunks 0012–0015

Stint 4 of 8. Cols 0096C–0107B, chunks 0012–0015, 4,593 Latin words (frontmatter counts) → ~6,900 English words (ratio 1.49–1.54, on the register's expected ~1.5×). verify-english.mjs is clean on all four; column anchors (44), [n:] notes (30), italic spans (161), VERS. addresses and paragraph counts are all 1:1 with the Latin twins, and question marks, colons and semicolons match paragraph by paragraph — a finer partition than the column bands (Test 3).

Apparatus fired: 1 [sic:] · 12 [var:] · 1 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:]. Guillemets: zero in either language, either direction.

Plates read: 7 pages / 12 columns on BOTH witnesses — archive patrologiaecurs06saingoog pp. 53–59 at 300 dpi (600 dpi where a letter was in doubt) and Gallica ark:/12148/bpt6k54939667 ff. 48–54 at native resolution. Corner numbers read first on every page; both maps confirmed against them ((col+11)/2 and (col+1)/2). Details in data/briefs/9002-PLATE-READS-0012.json.

§0 · THE POLICY I APPLIED, STATED BEFORE THE FINDINGS

1. No claim about Migne's type on one witness, in either direction. Every letter-level candidate below was checked on Gallica before it was written down, and the archive copy lost eleven times out of twelve in this range. (See §1 — this independently reproduces the 12:1 figure CLAUDE.md now records for Matthew's opening range.)
2. Where the plate is read and our TEI differs, the English renders what Migne printed (translation-style.md §"What the English translates"), and the divergence is filed in data/briefs/9002-PENDING-TEI-PATCHES-0012.md. One site in this range: eleborata @0105C.
3. A [sic:] accuses Migne's type and needs two witnesses. One fired (nomo @0097A).
4. A [var:] states BOTH readings — what Migne prints and what the Clementine has — so it can be audited. Fired only on quoted scripture whose divergence changes the sense, and only after the column was read on both plates.
5. A real word that reads oddly is Migne's — rendered as printed, logged, no marker. The declined list is §4 and it is longer than the fired list.
6. usque ad: all twelve occurrences in this range are the etc. splice formula (each follows an etc., and names where the father's sentence resumes) and each is rendered "as far as". There is no ordinary temporal or spatial usque ad in these four chunks.
7. Vocabulary axes: gentiles/gentilis → Gentile(s) throughout (0099D gentilium, 0106D gentili twice, 0106D lemma Gentes) — Matthew's Judaei/gentes opposition is explicit here (facit similem gentili, id est infidelem), so "nations" was never in question; daemonis @0105B → demon, diabolus → devil (they stand two lines apart at 0105A–B); turbae phantasmatum @0099C → crowds; no natio, no multitudo, no mysterium/sacramentum pair in the range. Commentator's voice not archaised (dicit, ait, inquiunt → says, say); -eth used only inside italic lemmata (He seeth in secret, For the Father knoweth, It sufficeth).
8. Pattern 17: thou for every singular address (0096C cogeris, 0097A repercutias, 0099D intuearis, 0103C unge, refice, lava, 0104D facis), you for every plural (0098C oraveritis, 0104A ponite, 0105C sitis, 0105D committite). No "ye": the corpus renders Audistis "You have heard" (9001 @0002, 9004 @0004), and Pattern 17 makes the archaism a number marker, not a register marker.

⭐ The lemma-expansion line I drew, stated so it can be argued with

Convention 3 expands a fragmentary lemma; the "control's class" rule says a complete one-word lemma is rendered as it stands. My test: can the printed lemma stand as an English clause or phrase on its own? If yes, render it as it stands; if it is grammatically headless, expand it from the verse.

• Rendered as they stand (complete): Audistis. "You have heard." · Diligite. "Love." · Benefacite. "Do good." · Orate. "Pray." · Odio habebis. "Thou shalt hate." · Perfecti. "Perfect." · Nesciat. "Let it not know." · Exterminant. · Lucerna. · Panem. · Nostrum. · Sicut. · Amen, · Dies. · Sufficit. "It sufficeth."
• Expanded (headless): Qui solem. → "Who maketh his sun to rise" (Mt 5:45 — the accusative has no verb in the lemma, and Hunc visibilem glosses the sun) · Et volenti. → "And to him that would borrow" (Mt 5:42 — the participle has no governing infinitive, and the gloss is entirely about lending) · Aut enim unum odio. → "For either he will hate the one" (Mt 6:24 — odio with no habebit).
• ⚑ One deliberate exception, and the reason: Praebe illi. is rendered "Offer to him.", NOT expanded to "offer him the other also" — because the elided word is itself the next lemma Migne prints (Alteram., glossed separately three lines down). Expanding here would make the English print the same word twice and would swallow a lemma that has its own gloss. This is not conformation and it is not the stump class; it is a collision the general rule does not foresee.

§1 · ⛔⛔ THE ARCHIVE COPY LOST 11 TIMES OUT OF 12 IN THIS RANGE

Every candidate I raised from the archive copy was taken to Gallica before it was recorded. Eleven were refuted, with our TEI right every time; one was confirmed. This is the same result CLAUDE.md records for cols 0063–0080, arrived at independently, on a different seven pages, by a reader who had not seen that entry when the reads were made.

⚑ The line is worth its own sentence. At 300 dpi the archive shows what looks exactly like Migne's raised asterisk key (Pattern 19's second layer) after diebus abstinentiae, and the foot of that page carries nothing — which is the shape of a lost note. Gallica prints a bare colon. A [cn: | …] and its [nt:] twin were within one HTTP request of being invented. The feet of pp. 53–59 were also inspected on both copies: no foot-of-page apparatus of either layer anywhere in cols 0096–0108. The only mark below the text is a printer's signature 4 under col 0106.

§2 · THE HANDED-OVER LEAD @0099B — SETTLED, AND IT IS NEITHER OF THE TWO OFFERED ANSWERS

The launch brief (§4) handed this stint coram homnibus in the Mt 6:1 citation inside the gloss on VERS. 5, read at 150 dpi on the archive copy alone, and asked whether it is Migne's broken sort (→ [sic:], two witnesses) or our file's corruption (→ TEI patch).

Our Latin is not corrupt: it reads coram hominibus, correctly — so the second answer was never live. And the first fails its own test:

• Archive p.55, 300 dpi: the reading is real. homnibus — an m followed directly by an n, with no dotted i between them, against dotted is in stitiam and faciatis on the same line. It is not a 150 dpi artefact.
• Gallica f50 (corners 99/100), native resolution: hominibus, the i plainly there.
• ⚑ And the two copies differ TWICE on these two lines, in a way ink cannot produce. Two lines above, the archive prints haec sententia, supra dixerat : (comma, lower-case s) and Gallica prints haec sententia. Supra dixerat : (full stop, capital S). Over-inking can close a counter; it cannot turn S into s. These are two printings of PL 114, not two photographs of one — the same finding as 8950's Mititte/Mittite, and the same finding CLAUDE.md now records for this volume.

Verdict: no marker, no patch, crux only. A [sic:] needs two witnesses and has one; a patch needs our file to be wrong and it is right. The honest statement is that one printing of PL 114 carries homnibus at 0099B and the other carries hominibus, and our text follows the second. Recorded here because it is a fact about the volume that a later reader of the archive copy will otherwise re-discover as a defect.

⚑ The general lesson, and it cost this stint nothing because the rule was obeyed: the launch brief says a non-word "is the class that has been right every time." In this range it was right once in three — nomo @0097A stood, homnibus @0099B and fiilius @0100C both fell. A non-word is a reason to look, not a reason to fire.

§3 · THE APPARATUS FIRED, WITH THE ARGUMENT FOR EACH

3a. The single [sic:] — @0097A nomo (chunk 0012)

Alteram. (CHRYSOST.) Primum indigna res est ut nomo fidelis stet in judicio ante conspectum judicis infidelis.

Sense requires homo. Both witnesses print a word that is not homo: the archive at 300 dpi reads nomo (initial n, unambiguous, no ascender); Gallica's initial letter is blotted into a closed bowl and reads n or o, but carries no ascender and cannot be h. Our TEI carries nomo verbatim, so the marker's substring guard is satisfied and the accusation is against type that both copies show to be defective.

Rendered per the 2026-09-05 amendment to Pattern 7 — the marker discloses, so the sense is rendered beside it: "that a faithful man [sic: nomo] should stand in judgment". 7a⁗ check, all four questions: the clause keeps its verb (stet → "should stand"), its negation (none present), its conjunction (ut → "that"), and its subject — which is what the marker would otherwise have taken.

⚠ A corpus control I record because it argues the other way and lost. nomo occurs three times in src/latin/: here, at 9000 @0314C inside oeconomo (a false hit), and at 8991 @0599D, ut nomo homini conditionis vinculo subderetur, where the word wanted is nemo. Two Glossa books printing the same non-word for two different targets looks like a Corpus Corporum OCR habit, and on the frequency argument alone I would have filed a patch. The plates overrule it: this is exactly reference_plate-read-triage's rule — frequency proves an error exists, never whose it is. 8991 @0599D is not in my range and I make no claim about it, but it is worth a plate read by whoever holds that work.

3b. The single [cj:] — @0107B malitia Dei (chunk 0015)

Sufficit. Et quia cura praesenti mentes obruitur, non debet germinari malitia Dei.

Dei is a perfectly good word, so Pattern 12 has nothing to wrap and Pattern 7 renders it — and the page then says "the malice of God ought not to be made to sprout." The lemma being glossed is Sufficit, i.e. Mt 6:34 sufficit diei malitia sua, and the preceding lemma is Dies. The conjecture is diei, and it is not ours to invent: it is printed on the same page, twice, as the lemma this sentence exists to gloss.

the malice of God [cj: Dei; read diei, "of the day"]

Read on both witnesses (archive p.59, Gallica f54): malitia Dei on each. So this is Migne's own text, not the digitization's, and the marker is additive — nothing is removed.

⚑ Why this clears Pattern 18's bar and the rival reading loses. The bar is that the printed word makes the sentence assert something the author is not arguing. "The malice of God" is not merely awkward: it predicates malitia of God in a gloss whose whole subject is the anxiety of the human mind. The rival treatment — leave it to the crux, per 18a's decline classes — loses because 18a's declines are mood, voice, number and person anomalies that merely fail to construe, and this is none of those: it is a noun substitution that construes perfectly and says something false, which is the ✅ side of the list (10703 @1126B, Oro for Pro).

3c. The twelve [var:], each with the printed reading and the Clementine's

Every one is a quoted scripture (the class that stands 120 times in 121 across four Glossa books), and every one was read on both plates before firing.

§4 · THE FULL LEMMA WALK — 161 spans, and the count checks

Count check first (7a″ addendum). data/briefs/9002-lemmata-0012.txt declares 161 spans for chunks 0012–0015. Counting the marked spans in my own Latin, paragraph-bounded, gives 38 + 55 + 45 + 23 = 161, with *zero paragraphs carrying an odd number of and zero guillemets in any of the four chunks, so the paragraph-bounded and whole-file counts are the same number and cannot have cancelled. The file's first entry ([0096C] A malo est.) and last ([0107B] Sufficit.) are the first and last spans of my Latin. The cut is right.**

Of the 161: 12 diverge from the Clementine materially and are marked [var:] (§3c) — one of them, 0104A, was not flagged by the pre-scan. The rest are set out below by class. ⛔ Nothing on this list is a report of a search; every span was read against sources/vulgate/clementine-flat.txt by hand.

4a. The eight ⚠ NOT in Clementine verbatim flags that are INNOCENT, with reasons

4b. The spans that agree with the Clementine EXACTLY — checked, not assumed

⛔ These are the ones a report of findings alone would hide, and in a gospel a bare ✓ is the weakest mark in the corpus, so each address below was read, not trusted.

Mt 5:37 A malo est. · Mt 5:39 Non resistere. ×2, Praebe illi., Alteram. · Mt 11:29 Discite a me, quia mitis sum, et humilis corde · Ps 7:5 Si reddidi retribuentibus mihi mala · Mt 5:40 Et ei qui,, Tunicam tuam., Et pallium; · Mt 5:41 Mille passus., Alia duo. · Mt 5:42 Da ei. ×2, Et volenti., Ne avertaris, · 2Cor 9:7 Hilarem enim datorem diligit Deus · Mt 5:43 Audistis., Odio habebis. · Mt 5:44 Diligite., Benefacite., Orate. · Mt 5:45 Filii., Patris., Qui solem. · Mt 5:46 Publicani. · Mt 5:47 Et si salutaveritis. · Mt 5:48 Perfecti. · Mt 6:1 Attendite,, Apud Patrem vestrum., Attendite ne justitiam vestram faciatis coram hominibus; · Mt 6:2 Cum ergo,, Tuba canere., Ante te., Receperunt. ×2 · Mt 6:3 Nesciat. · Mt 6:4 In abscondito., Videt in abscondito. · Mt 6:5 Et cum oratis. ×2, In angulis. · Mt 6:6 Tu autem., Clauso ostio., Patrem. · Ps 41:9 Apud me oratio Deo vitae meae · Mt 6:7 Orantes autem., Putant enim. · Mt 6:8 Nolite ergo assimilari eis., Scit enim pater. · Mt 6:9 Pater noster. ×3, noster,, In coelis., Sanctificetur nomen tuum., Sanctificetur, ×2 · Mt 6:10 Adveniat regnum tuum., Fiat voluntas tua., Sicut in coelo. · Ps 83:3 Cor meum et caro mea exsultaverunt in Deum vivum. · Mt 6:11 Panem., Nostrum., Supersubstantialem. · Sir 24:29 qui edunt me adhuc esurient · Mt 6:12 Et dimitte,, Dimitte nobis debita nostra:, Sicut. · Mt 6:13 Et ne nos inducas., Et ne nos,, Sed libera nos a malo., Sed libera nos., A malo, · Mt 6:14 Si enim dimiseritis. · Mt 6:15 Si autem non dimiseritis. · Mt 6:16 Nolite fieri., Exterminant., Facies suas. · Ps 50:19 Cor contritum et humiliatum, Deus, non despicies · Mt 6:17 Unge caput tuum. ×2 · Mt 6:18 Et Pater tuus. · Mt 6:19 Nolite thesaurizare. · Mt 6:20 In coelo. ×2 · Lc 21:33 / Mc 13:31 Coelum et terra transibunt: verba autem mea non transibunt. · Mt 6:21 Ubi enim est thesaurus. · Mt 6:22 Lucerna. · Mt 6:23 Si autem oculus., Tenebrosum erit., Si ergo lumen. · Mt 6:24 Nemo potest,, Aut enim unum odio., Aut unum sustinebit., Alterum contemnet., Et mammonae. · Ps 73:23 Superbia eorum qui te oderunt · Mt 6:25 Ideo dico vobis,, Animae vestrae., Nonne anima? · Jo 12:25 Qui amat animam suam, perdet eam · Mt 6:26 Respicite., Et Pater vester. · Mt 6:27 Quis autem vestrum, · Mt 6:28 Considerate lilia agri. · Mt 6:29 Quoniam nec Salomon, · Mt 6:30 In clibanum mittitur., Quanto magis vos,, Modicae fidei. · Mt 6:31 Nolite ergo solliciti,, Nolite,, Solliciti esse. · Mt 6:32 Haec enim omnia., Gentes., Scit enim., Pater vester. · Mt 6:33 Quaerite ergo., Et haec omnia. · Mt 6:34 Solliciti., In crastinum., Dies., Sufficit. — and the twelve usque ad spans, which are Migne's formula and not scripture at all.

⭐ Two of these deserve a line, because the gospel trap runs both ways.

• 0096D oculum pro oculo, dentem pro dente: — Mt 5:38 prints Oculum pro oculo, et dentem pro dente, so against Matthew this looks like a dropped et. It is not a divergence: Lv 24:20 and Dt 19:21 print exactly Migne's form, and the gloss's own frame is Lex modum ultionis instituit — he is quoting the law, not the evangelist. A [var:] here would have accused Migne of mis-citing a verse he is not citing.
• 0104B Coelum et terra transibunt: verba autem mea non transibunt. — printed with no reference. Mt 24:35 has non praeteribunt; Lc 21:33 and Mc 13:31 have non transibunt, exactly as printed. Inside a commentary on Matthew, the Matthaean parallel is the one that diverges and the non-Matthaean ones agree. No marker.

§5 · ⚑ THE MATCHES ONLY ELSEWHERE CANDIDATE IN MY RANGE — @0106B, ADJUDICATED

0106B VERS. 28.-- Et de vestimentis. only Tob 4:17 (lemma is Mt 6)

Verdict: (1) a real divergence. [var: Vulg. Mt 6:28 Et de vestimento, singular].

• (3) our file's corruption — refuted at the plate, and this had to go first. Archive p.58 and Gallica f53 both print Et de vestimentis. in the italic lemma of VERS. 28. A -o→-is ending is well within OCR's reach, so this was the live rival, and only the two plates could kill it. They did.
• (2) a harmony parallel Migne has reached for — loses on the evidence, not on plausibility. This is the explanation the brief warns is the commonest in a gospel, so I looked for the parallel and there is none: Lc 12:22–23 reads neque corpori quid vestiamini / corpus plus quam vestimentum; no gospel verse anywhere prints de vestimentis. The Tob 4:17 hit (et de vestimentis tuis nudos tege) is a three-word phrase collision in a work about almsgiving, with no thematic or textual route into a gloss on Mt 6:28 — the gloss below it expounds sollicitudo and lilia agri, i.e. Matthew, and nothing else.
• (1) therefore stands. Migne's lemma prints the plural where the Clementine prints the singular. The divergence is small and I say so plainly; what makes it markable rather than orthographic is that it is a number change in a verse-lemma carrying its own VERS. address, confirmed on two witnesses, and Pattern 14 exists precisely so that a reader can see it without opening a cruces file.

§6 · THE PENDING TEI PATCH — @0105C eleborata

dum timet ne aut possessa dispereant aut eleborata non contingant.

Our twin prints eleborata, which is no Latin word. Both plates print elaborata (Gallica f53 line-broken as ela-borata; archive p.58 the same). This is the digitization's error, not Migne's, so what is owed is a patch and not a [sic:] — the marker would publicly accuse the plate of a defect neither copy carries.

Per translation-style.md §"What the English translates — THE PLATE WHERE READ", the English renders what Migne printed: "the things laboured for". Filed in data/briefs/9002-PENDING-TEI-PATCHES-0012.md.

⚠ A second patch is proposed there and I mark it LESS certain deliberately: @0097D our twin reads Ut sitis fiili where the sense and both plates want filii. The archive at 600 dpi puts the tall l in third position (f-i-l-i-i); Gallica's native resolution is coarse there and reads fili/filii but certainly not fiili. Neither witness shows our twin's form, and the same fii-for-fil shape occurs at 11062 @0427D (fiilialem for filialem, with Filialis printed correctly two sentences later) — a Corpus Corporum habit, not a PL 114 compositor's. The English renders filii, "sons". I would still like a third look before the patch lands, and the patch file says so.

§7 · THIS RANGE'S OWN CRUCES, BY COLUMN

• @0096C dum veritati non credit — the subject of credit is unexpressed. I supplied "he" = the alter whose infirmitas compels the oath (sed alterius infirmitas id cogit). The rival, "so long as it [the weakness] does not believe", makes infirmitas the believer, which the sentence's own contrast (thy oath is not from thy evil but from his) does not support. Named because the brief's dominant defect class is exactly this.
• @0097A Praebe illi. / Alteram. — see §0. Migne splits one clause of Mt 5:39 into two lemmata and glosses each; the English does not expand either into the other.
• @0097D fiili — §6.
• @0098A qui dicit eum hujusmodi nec cibum sumere — the sentence alludes to 1Cor 5:11 cum ejusmodi nec cibum sumere, and eum for cum is an obvious guess. Declined: eum is a real word, it construes as printed (dicit + accusative + infinitive, hujusmodi attributive to eum), and both plates print eum. Rendered as printed, "that such a one is not even to take food". No marker on a corpus argument, however good.
• @0098B Solem suum stands inside the italic run of the Amos quotation, followed by Migne's (Amos. IV) and etc. — but the words are Matthew's (qui solem suum oriri facit), not Amos's. Migne's italic run has run two quotations together. Reproduced exactly as printed, including where the note falls; no marker, because nothing is defective — the divergence marked at this column is the Amos wording (§3c), not the join.
• @0099C ut … oret Deum — a third-person subjunctive in a passage whose lemma is Tu autem (second person singular). I rendered the impersonal "that one should pray to God" rather than "that thou shouldst pray", because Migne's verb is third person and Pattern 17 makes person a thing we track, not smooth.
• @0100D ne a sanctificatione nominis primus in aliquo discrepemus — a nominative singular primus inside a first-person-plural clause. Rendered as printed ("that we may not, first, in anything depart…"), which is as close as the words allow. The natural conjecture is primum (adverbial), an ordinary medieval spelling; declined for a [cj:] because the English is awkward, not false, which is the wrong side of Pattern 18's bar. Both plates print primus.
• @0101D Vel sicut justis, ita in peccatoribus conversis — Migne prints no in before justis though the three parallel members all have one. An ellipsis carried from the preceding member, not a defect; rendered "Or as in the just".
• @0102C et peccantibus in se veniam promittit, et nobis et per nos aliis veniam promittit — veniam promittit twice. [d:] declined: Pattern 11 marks a compositor's doubling, and this is not one — each clause has its own dative (peccantibus in se / nobis et … aliis) and each says something the other does not. Both copies print both. Recorded because a later reader will see the repetition and wonder.
• @0102D Et ne nos, etc., deserens in tentatione sinas vinci — the ne that governs sinas vinci is printed inside the quoted lemma, and the clause it governs is Migne's own prose. A first draft rendered it "deserting us in temptation, suffer us to be overcome", which an English reader parses as a positive command: 7a′ exactly — quarantining a fragment does not quarantine what it governs. The English now reads "do not suffer us to be overcome". Supplying that negative is not an emendation; it renders the ne Migne prints, which was being dropped.
• @0103B Demoliri est quasi dissipare — a definition of demoliri inside a gloss whose lemma is Exterminant, with no demolitur anywhere near it (Mt 6:19–20's demolitur is eleven lines below, in a different verse). Rendered as printed; the abridgment has plainly spliced a note from another lemma. No marker: nothing is defective, and the usque ad formula removes words without suspending grammar.
• @0104A VERS 20.--, without the period after VERS that every other address in the range carries. Both plates print VERS 20. — the archive's apparent period is ink. Left exactly as printed, in the Latin and in the English. ⛔ This is the class 8950 proved cannot be corrected from one witness, and it is the reason I fetched Gallica before touching it.
• @0104B Si etenim illa in genere suo munda tamen sordescit res dignior — illa … munda has no verb; English cannot supply one without inventing it, so the rendering keeps the ellipsis ("For if those things are clean in their own kind, yet the worthier thing grows foul"). Both plates agree.
• @0104A charitatem amplectantes — not a form of amplector (amplectentes expected). Both plates print amplectantes, so it is Migne's, and corpus frequency is 1 against 3 for the correct form. Treated under Pattern 9 as an ordinary medieval a/e participle spelling and rendered normally ("embracing"): no italics, no [sic:], no crux status beyond this line. Recorded because it is the strongest non-word candidate in the range that I did not fire on.
• @0105A Quidquid facis, vel ex amore Dei, vel servitute diaboli is repeated four sentences later as Quidquid enim facis, aut ex amore Dei, vel ex servitute diaboli. Two different paragraphs, two different conjunction sets; not a doubling, and no marker. Both clauses are verbless in the Latin and are rendered verbless.
• @0105B Superbia eorum qui te oderunt is keyed (Psal. III) and is Ps 73:23. The words agree with the Clementine exactly, so no [var:]; Migne's number is passed through under convention 4 (his attributions and locators are not ours to correct). Same class: @0102A qui edunt me adhuc esurient keyed (Eccl. XLIV), which is Sir 24:29; and @0100C (Joan. VII) on a sentence found in Jo 14/16/20.
• @0106D qui novit nobis haec in via sua esse necessaria — "his way" or "our way"? sua is reflexive to qui novit, i.e. God's, and I rendered "on his way". The rival ("on our journey") reads better and has no morphology behind it. Both plates print in via sua as three words; the archive's apparent invia is letterspacing.
• @0107A Quid ergo amplius habet a gentili — the subject of habet is unexpressed and I supplied "he" = the solicitous man of the preceding lemma (facit similem gentili), not the Gentile.
• @0107B cura praesenti mentes obruitur — plural subject, singular verb, on both plates. I rendered "minds are overwhelmed by present care", which carries Migne's plural noun into the English and loses only the verb's number, for which English has no exponent (Pattern 9). ⚑ The rival I weighed and rejected: a [sic: mentes] on the 11632 @0304A precedent (ut signa prius ardeat, where a number disagreement took a marker because rendering it hid the defect). It loses because the two cases are not the same shape: at 11632 the English could show nothing at all, whereas here the reader meets the plural "minds" opposite Migne's plural mentes, and only the concord — which no English sentence can carry — is lost. Marking it would put a [sic:] on a sound word to record a fact the facing Latin column already shows.
• @0107B malitia Dei — §3b, the one [cj:].

⚑ The ship test, and what the gate will say

plate-gate.mjs will flag all fourteen markers in this range until the orchestrator merges data/briefs/9002-PLATE-READS-0012.json into data/plate-reads.json — every one of them stands on a column read on both witnesses, and none was written before that read was made. This stint did not touch the shared file (peers run concurrently).

§8 · DECLINED-ODDITY LIST — every suspect word I did NOT mark, in one place

Ordered by column. On John and Luke this list produced nearly every correction those works got.

Test 1 (strip every marker and read what is left) was run mechanically over all four chunks: the two markers that sit inside a sentence — [sic: nomo] @0097A and [cj: Dei] @0107B — leave "that a faithful man should stand in judgment" and "the malice of God ought not to be made to sprout". No stranded article, no stranded preposition, no doubled mark, and both clauses keep their subject as well as their verb.

Test 2 (name the host of every printed negative) was run clause by clause. The sites where it bites in this range are all in gloss prose, not in the lemmata: Non tantum non repercutias (0097A, two negatives, two hosts), non probantur … opera (0098D, on the verb, not on simplici), ne non sit simplex ad Deum (0103B, a deliberate double negative kept double), Non vero precatur non tentari, sed nec ad modicum vinci (0102C, three), Haec exempla non prohibent providentiam ne laboret (0106A, two, and the second belongs to laboret, not to prohibent), and Non quod laborare et providere hoc non deceat (0106B, two). ⚑ The trigger condition in 2a — a negative on a verb with a quantifier or correlative nearby — fires at 0099B (non solum vera merces … sed vitium … augetur, where the non governs solum and the sed clause proves it) and at 0106A (non prohibent … ne laboret, sed sollicitudinem, where the sed shows non prohibent governs providentiam and the ne governs laboret). Both were resolved from the sentence's own structure, not from what reads well.

Test 4 (does the English ASSERT what he asserts?) was run last, paragraph by paragraph, with the counts already clean. It is what produced §3b, §7's @0107B entry and §7's @0099C entry; nothing else in the range changed under it.

Preserved from cruces-0016.md, unedited; headings demoted one level. Companion files: data/briefs/9002-PLATE-READS-0016.json, data/briefs/9002-PENDING-TEI-PATCHES-0016.md, data/briefs/9002-lemmata-0016.txt.

9002 — cruces, stint 0016–0021 (cols 0107C–0123D)

Glossa ordinaria, Evangelium secundum Matthaeum, PL 114. Chunks 0016–0021: CAPUT VII. through CAPUT XI., 6,577 Latin words, the Sermon-on-the-Mount close and the miracle and mission chapters.

Companion files from this stint: data/briefs/9002-PLATE-READS-0016.json (every page opened, both witnesses) and data/briefs/9002-PENDING-TEI-PATCHES-0016.md (seven sites where our Latin has departed from Migne).

(a) THE POLICY I APPLIED — stated before the findings, so it can be argued with

1. Markers. Zero [sic:], zero [cj:], zero [d:], zero [ed:], six [var:].

2. When a [var:] fires. I inherited the 0000–0005 stint's bar and did not loosen it: Migne's quoted scripture takes a [var:] where his words assert something different from the received text — a substitution. It does not fire where he merely shortens a verse, where he splices two verses, where the difference is orthographic (Nunquid/Numquid, quandiu/quamdiu, patrem familias/patremfamilias), or where the difference is the form of a transliterated proper name (Gerazenorum, Scariothis).

⚑ One extension I did make, and it should be argued with if it is wrong. The 0000 stint's list is six lexical substitutions. Three of my six are inflectional — a tense (Appropinquabit for appropinquavit), a mood (conteret … Dominus for the imperative contere), and a stem (dicite for discite). I fired them because in each case the English word on the page changes: "shall draw near" for "hath drawn near", "the Lord shall crush them" for "crush thou them", "say ye" for "learn". A reader who is told nothing would take Migne's Latin for the Vulgate's. The 0000 stint's own @0069A (invocavit for invocabit) is exactly this class and it fired, so I take the precedent to be established rather than new.

3. When a single-word [var:] fires. Launch brief §2 requires a witness of its own, because a one-word divergence has two possible owners. Two of my six are single words (@0116B dicite, @0118A Appropinquabit) and for both the second owner is eliminated: I read the word on both witnesses. Ownership is settled by the plate, never by the argument.

4. Every claim about the type rests on two witnesses. I read all nine pages of my range on the archive PDF and all nine leaves on Gallica, corner numbers first on each. This changed the answer nine times — see §(e). Nothing in this file rests on one copy, and in particular a verse number I would have altered on the archive alone is refuted at Gallica (§(d) 14).

5. Where the plate and our twin disagree, the English renders the plate and the site is filed as a pending TEI patch. Seven sites, all seven Corpus Corporum's, none of them Migne's.

6. The vocabulary axes bind. turba → crowd (5 occurrences; multitudo does not occur in this range at all). daemonium/daemones → demons, daemoniacus → demoniac; diabolus → devil. gentes/gentium → Gentiles in the Judaei/gentes opposition (0113B, 0115A, 0116A, 0117D, 0119A, 0121A, 0123B) and nations in the one plainly generic quotation, Docete omnes gentes at 0117D, which the axis names explicitly. gentilis → Gentile (0113B), gentilitas → the Gentile world (0116D ×2, 0122D). ethnicus → heathen (0116A). mysterium → mystery (0122B), sacramentum → sacrament (0108C, 0113B, 0120B). ⚑ natio does not occur in this range. The grep returns four hits and all four are the documented false positives — dignatione, incarnationem, indignatione, venatio. Recorded because the brief asks for the instrument check to be stated, not merely run.

7. The commentator's own voice is not archaised. dicit/ait → says; -eth and -est forms appear only inside quoted scripture (that saith, He raineth, He that findeth, leadeth, knoweth). thou is singular and you plural throughout — see §(d) 20.

(b) SPAN COUNT — 335 against 335, paragraph-bounded

data/briefs/9002-lemmata-0016.txt declares 335 spans for chunks 0016–0021, of which 0 are open-ended.

Counting my own Latin paragraph by paragraph (per the file's own warning that a bare total can agree falsely when two errors cancel):

335 = 335, exact. No paragraph in my range carries an odd number of asterisks, and there is not one guillemet in 6,577 words, so both failure modes that make a bare total lie are excluded. Matthew is an italic-lemma book, as the 0000 stint found.

⚑ A note on the brief's own classification, because it changes the arithmetic. The inventory's VERS. lines are those matching VERS. n.--; two verse lemmata in my range are printed by Migne WITHOUT the dash — VERS. 25. Suscitaverunt. (0114B) and VERS. 28. In regionem Gerazenorum. (0114C) — and they are therefore filed among the second lemmata. Counted correctly my range holds 134 verse lemmata, not the 132 a mechanical read of the brief gives. Both irregular addresses are Migne's and are reproduced verbatim under convention 1; both were read on both witnesses (§(e)).

The 335 split: 134 verse lemmata · 53 usque ad formula spans · 148 second lemmata, inline scripture, name-etymologies and technical tags.

(c) THE LEMMA DIVERGENCE LIST — all 335 spans, collated against sources/vulgate/clementine-flat.txt

Migne's citation sigla pass through verbatim under convention 4 and are not treated as divergences; a wrong siglum is his attribution practice. Three in my range are demonstrably wrong and are recorded, not corrected: (Psal. VIII) at 0113C on what is Ps 18:13; (Luc. IX) at 0119C on what is 1 Cor 9:9 (his own sentence says Apostolus); and (I Cor. IV) at 0123D on what is 2 Cor 4:17.

C1. THE SIX THAT FIRED

C2. DECLINED — the fifteen other verse lemmata that are not verbatim in their own Matthew verse

134 verse lemmata were collated against their own chapter and verse in the Clementine. 118 are verbatim. Sixteen are not, and three of those sixteen fired (above). Here are the other thirteen, with the fourteenth — @0122D — belonging to §(c) C4 because the divergence is ours, not Migne's.

• 0108B Mt 7:4 Aut quomodo dicis fratri tuo. — ⭐ the Clementine is the one in error here: clementine-flat.txt reads fratris tuo, which does not construe with dicis. Migne is right. Recorded because a mechanical collation reports this as a Migne divergence and it is the reverse.
• 0108C Mt 7:6 Nolite sanctum, etc. — Migne cuts dare; a fragment, and his own etc. marks the cut.
• 0110C Mt 7:16 Nunquid. — orthography only; the Clementine spells Numquid throughout and therefore every Nunquid in this work flags mechanically. Same at 0116C (Mt 9:15 Nunquid possunt), 0117C Nunquid usque in coelum exaltaberis? (Mt 11:23) and 0119C Nunquid de bobus cura est Deo (1 Cor 9:9).
• 0111C Mt 7:23 Nunquam novi vos. — same class (Numquam).
• 0112B Mt 7:27 Et cecidit fuitque ruina ejus magna. — Mt 7:27 has et cecidit, et fuit ruina illius magna. An enclitic for et, and ejus for illius: two ways of writing the same thing, and nothing in the gloss turns on either. ⚠ The nearest of my declines to one I could be argued out of, since it is a two-word difference inside a whole quoted clause and §2's prior favours firing on quoted clauses. It stays declined because neither word changes what the sentence asserts, and the gloss expounds ruina and conscientia, not the pronoun.
• 0113A Mt 8:7 Ego veniam et curabo eum. Et respondens centurio ait, etc. — the lemma splices across the 8:7/8:8 verse break, which is why the per-verse scan cannot match it. Not a divergence.
• 0114C Mt 8:28 In regionem Gerazenorum. — Clementine Gerasenorum. ⭐ Re-read on both witnesses at the coordinator's request, 2026-09-05, and adjudicated under §2 rather than swept under Policy 2 — see §(d) 6a for the ruling. Both copies print the z, twice each; our twin agrees; the divergence is Migne's beyond doubt and it is still declined, because it is the form of a transliterated place-name and nothing any sentence says turns on the letter. The English keeps his z (Gerazenes / Geraza) rather than normalising to the received English form.
• 0116C Mt 9:15 Lugere quandiu cum illis est sponsus. — quandiu for quamdiu; orthography.
• 0117D Mt 10:4 Et Judas Scariothis. — Clementine Judas Iscariotes. The ordinary medieval form of the name; nothing in the gloss touches the letters (it is about choosing a betrayer). Policy 2.
• 0119A Mt 10:25 Si patrem familias. — the Clementine sets patremfamilias as one word. A word division in the printing of the comparand, not a reading.
• 0120D Mt 11:9 Plusquam propheta. — Mt 11:9 has plus quam prophetam. Word division, plus the accusative brought to the nominative because the phrase is standing as a lemma head. Declined as a lemma-form adaptation; the English reads "More than a prophet."
• 0121A Mt 11:12 A diebus Joannis. — Mt 11:12 has A diebus autem Joannis Baptistae. Ellipsis at both ends of a two-word lemma.
• 0122B Mt 11:21 Vae tibi, etc. — verbatim as far as Migne's etc.; the ⚠ in the pre-scan is the truncation. ✓

C3. DECLINED — second lemmata and inline scripture that diverge

Thirty-three of the 148 non-verse spans did not match the Clementine verbatim. Six of them are the [var:] sites or the patched site; the rest are these, every one collated word by word:

• 0109A Ecce ego sto ad ostium et pulso [n: (Apoc. III)] — Apc 3:20 has Ecce sto. One inserted pronoun in a quotation adduced for the act of knocking. Declined; the gloss argues from pulso.
• 0109C Domini enim est terra [n: (Psal. XXIII)], etc. — Ps 23:1 has Domini est terra. enim is Migne's splice into his own sentence, exactly like the quia the 0000 stint declined at @0067B.
• 0109D Omnia ergo quaecunque, etc. — Mt 7:12 quaecumque; orthography.
• 0109D Haec est lex et prophetae, etc. — Mt 7:12 has haec est enim lex, et prophetae. Ellipsis.
• 0111B Qui dicit: Domine, Domine. — Mt 7:21 has qui dicit mihi, Domine, Domine. Ellipsis in a lemma the gloss expounds for the dicere, not the mihi.
• 0112A Inimicus homo hoc fecit, cujus — Mt 13:28 reads Inimicus homo hoc fecit exactly; the cujus is the gloss's own relative, caught inside Migne's italic run. Not a divergence, and the English keeps the run: "An enemy hath done this, whose works are all unstable".
• 0112D Non veni solvere legem, sed adimplere, etc. — Mt 5:17's second clause is non veni solvere, sed adimplere; legem stands in the first half of the same verse. A compression of one verse into one clause, not a substitution. ⚑ Recorded with care because the gloss around it argues about lex — but the word it argues from is on the page in the verse either way.
• 0114C In regionem Gerazenorum. — §C2.
• 0115A vestitum et sana mente sedere ad pedes Jesu [n: (Luc. VIII)] — Lc 8:35 has vestitum ac sana mente, ad pedes ejus, and sedentem where Migne has the infinitive. Migne has recast the verse into the indirect construction his own sentence needs (invenerunt hominem … sedere). Declined as an adaptation; ad pedes Jesu for ad pedes ejus asserts the same thing and the antecedent is Jesus in both.
• 0116A In viam gentium ne abieritis. Peccatores venientes discumbebant. — two verses in one italic run: Mt 10:5 verbatim, then Mt 9:10's peccatores venientes, discumbebant. Not a divergence; the run's composition is why the scan cannot match it.
• 0116A quia multi sequebantur [0116B] eum [n: (Marc. II)] — Mc 2:15 reads erant enim multi, qui et sequebantur eum. Migne compresses to a quia clause for his own syntax. Declined; and note the column anchor stands inside the span, which is a second reason the scan misses it.
• 0116C Lugere quandiu cum illis est sponsus. — §C2.
• 0118B non nisi virgam: — not a quotation at all. It is Migne's own summary of Mark, set in italic beside neque virgam precisely so the two can be contrasted (Matthaeus et Lucas dicunt neque virgam; Marcus non nisi virgam). Mc 6:8 reads nisi virgam tantum. No [var:] can be written on a paraphrase the author offers as his own.
• 0119A In via gentium ne abieritis, — Mt 10:5 has In viam gentium, and Migne himself prints viam at 0116A. Read on both witnesses at both columns: he prints viam in the one place and via in the other. This is his own orthographic variation inside one work — the class the 0000 stint declined for Nazareus/Nazaraeus — and the English renders "the way of the Gentiles" at both, since English has no exponent for the case.
• 0119B Sufficit diei malitia sua. Beelzebub. — Mt 6:34 sufficit diei malitia sua verbatim; Beelzebub is the next lemma, swallowed into the same italic run. Not a divergence.
• 0119C Nunquid de bobus cura est Deo [n: (Luc. IX)] — 1 Cor 9:9 Numquid de bobus cura est Deo? Orthography, plus a wrong siglum which is Migne's and passes through.
• 0119D Erunt quinque in una domo divisi, tres in duo et duo in tres dividentur. — Lc 12:52 reads erunt enim ex hoc quinque in domo una divisi, tres in duos, et duo in tres. Word order, duo for duos, and a supplied verb to close a sentence the abridgment has detached. Declined as quoting practice; nothing in the gloss turns on any of the three.
• 0120C Qui ponunt pulvinaria sub cubito, — closest to Ez 13:18 quae consuunt pulvillos sub omni cubito manus, but the vocabulary is wholly different and Migne gives it no siglum and does not introduce it as a quotation: it stands as a second italic lemma beside Ecce qui mollibus. Declined. ⛔ I record this as the decline I am least sure of — if it is a scripture citation it is a heavy one, and a [var:] naming Ez 13:18 would be defensible. It stays declined on rule B: when unsure and unable to settle it from the file, the crux is always available and always safe.
• 0121D Convertimini ad me in toto corde vestro in jejunio et fletu et planctu [n: (Joel. II)] — Jl 2:12 reads in jejunio, et in fletu, et in planctu. Two dropped prepositions; a contraction that asserts nothing new.
• 0121D et hujusmodi — not scripture; Migne's own "and the like". ⚑ Note that he sets it roman after the Psalm quotation twelve words earlier and italic here; our twin has both right, and both witnesses confirm it. The English mirrors the change.
• 0122B mysterium meum; · domus fructuum · venatorum. Si in Tyro, — the first two are Jerome's onomastic glosses of Corozain and Bethsaida, not quotations; the third is the tail of an etymology with the next lemma (Mt 11:21 Si in Tyro) run into the same span.
• 0122C villa pulcherrima, — the etymology of Capharnaum. Not a quotation.
• 0122C Justificata est Sodoma soror tua ex te, — not in the Clementine anywhere. The nearest is Ez 16:51–52 (justificasti sorores tuas … justificatae sunt enim a te), which Migne has recast into a sentence addressed to Jerusalem. Declined because a [var:] names one received reading and there is no single verse here to name — the 0000 stint's conflation-decline at @0068A, applied to the same shape. Recorded so that a later reader does not read the ⚠ as an unexamined site.
• 0122D Domine coeli et terrae. — this scores ✓ Mt 11:25 in the pre-scan against the ablative form, and the reason it fails my stricter per-verse test is that Mt 11:25 spells caeli. Orthography (coelum/caelum) throughout this work; not reported again.
• 0123B Nisi Filius, et cui Filius voluerit, etc. — Mt 11:27 has nisi Filius, et cui voluerit Filius revelare. Word order in a two-word span. Declined.
• 0123B Tollite super vos. — the gloss's own re-quotation of its lemma (Mt 11:29 Tollite jugum meum super vos), a fragment with the governed noun cut. Declined.

C4. THE ONE DIVERGENCE THAT WAS OURS

• 0116B Non enim veni vocure justos. — a non-word. Both witnesses print vocare. Corpus Corporum's corruption, filed as TEI patch 2, and the English renders "to call". ⭐ Had the corruption stood, the 7a″ collation would have reported a Vulgate divergence at Mt 9:13 that does not exist — the class launch-brief §8a names (the digitization manufacturing a divergence a collation then dutifully finds). This is that class caught in the act, and it is what reading the plate inline buys.
• 0122D Respodens Jesus. — same class; both witnesses print Respondens; TEI patch 6. Mt 11:25 agrees exactly once the patch lands.
• 0117A canantes (in the gloss, not a lemma) — both witnesses print canentes; TEI patch 7.

C5. CHECKED AND AGREEING WITH THE CLEMENTINE EXACTLY

118 of the 134 verse lemmata are verbatim in Matthew's own chapter and verse. Rather than print 118 fragments, I record the collation as it was run: each lemma was normalised (markers and column anchors stripped, punctuation and case ignored) and searched inside the text of its own verse, the chapter taken from the CAPUT head and the verse from the VERS. address. That is the test launch brief §3 asks for, and it is the one a bare ✓ cannot pass: it cannot be satisfied by a harmony parallel in another gospel, because the other gospel's verse is never consulted.

Verse lemmata verbatim in their own verse, in order: Mt 7:1 Nolite judicare · 7:2 In quo enim · 7:3 Quid autem vides · 7:5 Hypocrita, ejice · 7:7 Petite · 7:8 Omnis enim · 7:9 Aut quis · 7:10 Si piscem · 7:11 Si ergo vos · 7:12 Omnia ergo · 7:13 Intrate · 7:14 Quam angusta · 7:15 Attendite · 7:17 Sic omnis arbor · 7:18 Non potest · 7:19 Omnis arbor · 7:20 Igitur ex fructibus · 7:21 Non omnis qui dicit · 7:22 Multi dicent · 7:24 Omnis ergo · 7:25 Et descendit · 7:26 Qui audit verba mea · 7:28 Turbae · 8:1 Cum autem descendisset · 8:2 Adorabat eum dicens: Domine · 8:3 Extendens · 8:4 Vide, nemini · 8:8 Non sum dignus · 8:9 Nam et ego · 8:11 Ab oriente et occidente venient et recumbent · 8:12 Filii autem regni ejicientur · 8:14 Et cum venisset · 8:15 Et tetigit · 8:16 Vespere autem facto · 8:18 Videns autem Jesus · 8:19 Et accedens · 8:20 Vulpes foveas habent · 8:21 Alius autem · 8:22 Mortuos suos · 8:23 Et ascendente · 8:24 Ecce motus · 8:25 Suscitaverunt · 8:26 Modicae · 8:27 Qualis est hic · 8:29 Jesu fili Dei · 8:31 Mitte nos · 8:33 Pastores autem fugerunt · 8:34 Exiit obviam · 9:1 In civitatem · 9:2 Offerebant · 9:5 Quid est · 9:6 Tunc ait paralytico · 9:7 Et surrexit · 9:9 Vidit hominem · 9:10 In domo · 9:11 Quare cum publicanis · 9:12 At Jesus audiens, ait · 9:14 Discipuli autem tui non jejunant · 9:16 Nemo autem · 9:17 Neque mittunt · 9:18 Ecce princeps unus accessit · 9:20 Mulier quae · 9:21 Si tetigero · 9:22 Confide, filia · 9:23 Vidisset · 9:24 Non est · 9:25 Et cum ejecta esset · 9:27 Transeunte · 9:32 Hominem · 9:34 Pharisaei autem · 10:1 Et convocatis · 10:2 Petrus et Andreas · 10:3 Thomas · 10:5 In viam · 10:8 Infirmos curate · 10:9 Aurum neque argentum · 10:10 Neque calceamenta · 10:11 Interrogate quis · 10:14 Excutite pulverem de pedibus · 10:16 Ecce ego mitto · 10:18 In testimonium · 10:19 Dabitur enim vobis in illa hora quid loquamini · 10:21 Tradet autem frater · 10:22 Qui autem · 10:23 Cum autem · 10:24 Non est discipulus · 10:26 Ne ergo timueritis eos · 10:27 Quod dico · 10:29 Super · 10:30 Numerati · 10:32 Qui confitebitur · 10:33 Qui autem · 10:34 Pacem · 10:35 Veni enim · 10:37 Qui amat patrem · 10:38 Et qui non accipit · 10:39 Qui invenit · 10:40 Qui recipit · 10:42 Et quicumque · 11:2 Joannes autem · 11:3 Tu es · 11:6 Qui non fuerit · 11:7 De Joanne · 11:8 Ecce qui mollibus · 11:11 Major Joanne · 11:13 Omnes enim · 11:14 Et si vultis recipere · 11:16 Cui autem · 11:17 Cecinimus vobis · 11:18 Venit enim · 11:19 Et justificata est sapientia · 11:20 Tunc · 11:23 Et tu, Capharnaum · 11:26 Ita pater · 11:27 Omnia mihi · 11:29 Tollite jugum · 11:30 Jugum enim meum.

Second lemmata and inline scripture verbatim in the Clementine (79 spans; the reference given is the one the gloss is using, not merely the first hit): In qua mensura ×2 (Mt 7:2) · Et conversi (Mt 7:6) · Nostis bona · Dabit bona · Petentibus se (Mt 7:11) · Ut faciant vobis homines (Mt 7:12) · Per angustam · Et spatiosa (Mt 7:13) · Et pauci (Mt 7:14) · Ad vos in vestimentis ovium (Mt 7:15) · A fructibus eorum (Mt 7:16) · Nemo potest dicere: Dominus Jesus, nisi in Spiritu sancto (1 Cor 12:3, under Migne's own correct siglum) · Multi dicent (Mt 7:22) · Qui operamini iniquitatem (Mt 7:23) · Qui autem (Mt 7:26) · Assimilabitur · Viro · Qui aedificavit · Supra petram (Mt 7:24) · Pluvia (Mt 7:25) · Viro stulto · Super arenam (Mt 7:26) · Cum autem descendisset · Secutae sunt (Mt 8:1) · Si vis, potes me mundare (Mt 8:2) · Tetigit ×2 · Volo · Mundare (Mt 8:3) · Nemini dixeris (Mc 1:44, Migne's siglum correct) · Vade · Ostende te sacerdoti · Offer · In testimonium ×2 (Mt 8:4) · vir justus et timens Deum (Act 10:22) · Pluit super justos et injustos (Mt 5:45) · In tenebras exteriores · Fletus (Mt 8:12) · Et febricitantem (Mt 8:14) · Delicta quis intelligit: ab occultis meis munda me (Ps 18:13) · Et surrexit (Mt 8:15) · ambulantibus illis in via (Lc 9:57) · Magister ×2 (Mt 8:19) · Filius autem hominis non habet (Mt 8:20) · Domine (Mt 8:21) · Dormiebat (Mt 8:24) · Imperavit (Mt 8:26) · Duo habentes · Saevi (Mt 8:28) · Jesu fili Dei · Ante tempus · Torquere nos (Mt 8:29) · Et invenerunt hominem (Lc 8:35) · Rogabant eum (Mt 8:34) · Fidem · Confide, fili, remittuntur tibi (Mt 9:2) · Surge · Tolle lectum tuum et vade in domum tuam (Mt 9:6) · In telonio · Matthaeum · Surgens (Mt 9:9) · Ecce multi publicani (Mt 9:10) · Publicanis et peccatoribus (Mt 9:11) · Non est opus valentibus medicus, sed male habentibus (Mt 9:12) · Et tunc jejunabunt · Jejunabunt (Mt 9:15) · Tollit enim (Mt 9:16) · Duodecim annis · Accessit (Mt 9:20) · Et turbam (Mt 9:23) · Et tenuit manum (Mt 9:25) · Duo caeci (Mt 9:27) · Duodecim (Mt 10:1) · Docete omnes gentes (Mt 28:19) · Regnum Dei intra vos est (Lc 17:21) · Gratis accepistis ×2 (Mt 10:8) · Neque pecuniam in zonis vestris (Mt 10:9) · Neque virgam ×2 (Mt 10:10) · Dignus est · Dignus sit · Ibi manete (Mt 10:11) · Excutite (Mt 10:14) · Sicut serpentes (Mt 10:16) · praedicate (Mt 10:27) · Nonne duo passeres? · Et unus ex illis (Mt 10:29) · Nonne quinque passeres veneunt dipondio (Lc 12:6) · Separare (Mt 10:35) · Et sequitur me (Mt 10:38) · Uni ex minimis (Mt 10:42) · qui venisti (Migne's own counterfactual, not a quotation — see below) · Lamentavimus (Mt 11:17) · Exsultate Deo adjutori nostro, jubilate Deo Jacob (Ps 80:2) · angustia · venatio (etymologies) · exstitisti · Parvulis (Mt 11:25) · Quoniam sic fuit placitum ante te (Mt 11:26) · Et nemo novit Filium (Mt 11:27) · Discite a me (Mt 11:29) · Arcta via est quae ducit ad vitam (Mt 7:14, quoted back in CAPUT XI with Migne's own supra dicitur).

⚑ Three ✓ in the pre-scan that a careless reader would have taken as clearance, checked and recorded:

• 0120B qui venisti, scores ✓ 3 Rg 13:14. It is not a quotation of 3 Kings at all: the sentence is Non ait qui venisti, Migne saying that the text does not read those words (Mt 11:3 has qui venturus es). A mention, not a use.
• 0113B Pluit super justos et injustos scores ✓ Mt 5:45 inside a gloss on Mt 8:12. That is correct — Migne is adducing the Sermon on the Mount as a proof-text, three chapters back in his own gospel. Not a harmony parallel answering for the wrong evangelist.
• 0123D Arcta via est quae ducit ad vitam scores ✓ Mt 7:14 inside CAPUT XI. Also correct, and Migne signposts it himself (cum supra dicitur).

C6. NOT SCRIPTURE — name-etymologies and technical tags

Of the 38 single-word non-verse spans, most are ordinary Vulgate words re-quoted from the lemma above and are covered in C5. The genuinely non-collatable ones are Jerome's onomastics — angustia (Tyre) · venatio (Sidon) · mysterium meum (Corozain) · domus fructuum / venatorum (Bethsaida) · villa pulcherrima (Capharnaum) — together with the Greek words Migne prints in Greek, Τέλος at 0115D and Κωφὸς at 0117C, both reproduced verbatim and not transliterated.

C7. usque ad — the etc. splice test, run per occurrence

55 occurrences of the words in my range. 53 are the formula; 2 are ordinary Latin.

All 53 formula occurrences stand inside an italic span AND follow an etc., immediately — the etc. splice test is satisfied 53 times out of 53, with no borderline case anywhere in the range. Two of the 53 have a column anchor between the etc., and the usque ad (@0109A, @0117B); a raw adjacency grep misses both, and they are the formula. All 53 render "as far as".

The two ordinary uses carry no etc., stand outside every italic span, and are rendered by sense. The brief requires a departure from the axis to name its Latin:

• @0113B occidens, qui passiones usque ad mortem tolerant — temporal, rendered "who bear sufferings even unto death".
• @0122B ut Paulum Tyro abeuntem omnes cum uxoribus et liberis usque ad naves prosequerentur — spatial, rendered "followed Paul as far as the ships". ⚑ The English words coincide with the formula's, by accident of sense and not by application of the axis; the same coincidence the 0000 stint recorded at @0065D.

(d) CRUCES OF THIS RANGE, BY COLUMN

1. @0107C — a mark Migne prints and our twin has lost (the one band-test mismatch)

temere judicant haec judicia prohibet in our Latin; both witnesses print a mark, the archive an unambiguous colon and Gallica the upper point of one. The English carries a colon: "they judge rashly: these judgments he forbids." Filed as TEI patch 1, and this is the single band in my range where the punctuation test reports a mismatch against the twin ([0107C], colons 2 vs 3). It is expected and it should clear when the patch lands. No [sic:] — nothing on the plate is defective, and none could be written for an absence in any case.

2. @0108B — Salutem in eo quod temere judicant in oculo habent trabem

Read on both witnesses, both clean: an accusative Salutem governed by nothing at all, standing as its own one-line paragraph between two well-formed ones. Migne prints it so. Pattern 7 renders it as literally as English can hold: "Salvation, in that they judge rashly they have a beam in the eye." No marker — a real word, both witnesses, and the sentence is opaque rather than false, which is 18a's explicit ❌. I record it because it is the roughest sentence in the range and a reader will stop at it.

3. @0112D — Non quia sine tactu sanare posset, and no second non on either copy

Tetigit. Non quia sine tactu sanare posset, sed ut indicet se non subjectum esse legi…

The argument plainly wants non quia sine tactu sanare non posset — "not because he could not heal without touching". Both witnesses print exactly what our twin prints, at normal letter spacing, with no room for a lost word: archive p.61, Gallica f56. So the English carries what is printed: "Not because he could heal without touch, but that he might show himself not subject to the law."

⛔ A [cj:] was weighed and declined, and here is why the rival lost. 18a's ✅ extension does reach a dropped negation (7561 @0648, where Migne's setting loses the non of non solum … sed etiam). But that extension opens the gloss on a word Migne did print, and here there is no such word: the clause reads non quia … posset, and the conjecture would have to be attached to posset while asserting that a different word is missing beside it. More decisively, the printed clause is not false — "he touched, not for the reason that he could heal without touch, but in order to show…" construes, awkwardly, as a denial of one motive. Opaque, not false; 18a's ❌ covers it. The conjecture lives here and the printed word stands in the text.

4. @0112A — quae sensum rigorem fidei emollit, a double accusative

Read on both witnesses. Two accusatives with one verb and no conjunction. Rendered as apposition — "which softens the sense, the rigour of faith" — which is the closest English the printed words allow. No marker; both words are real and the defect (if it is one) has no English exponent beyond the apposition. The reader has the Latin opposite.

5. @0112C — the orphaned pluperfect subjunctive fecisset

Secundum Lucam, et in campestri loco sermonem quem jam forsitan supra habuerat ei fecisset. The governing clause has been cut away by the abridgment; fecisset has nothing to depend on. Rendered as an indicative main verb, "he had delivered to him the discourse which perhaps he had already held above", which is as close as English gets without supplying a conjunction. Both witnesses print it. Logged rather than smoothed.

6. @0113A — discendat, and which reading I took

In testimonium. Id est, Domino credant, si impietatis lepra discendat. Read on both witnesses; both print discendat. Under Pattern 9 this is an attested medieval spelling, not broken type — the de-/di- interchange is a convention of the period — so no [sic:] and no italics.

⚑ Two readings, and I say which I took and why the other lost. discendat stands either for discedat ("depart") or for descendat ("come down"). I rendered "if the leprosy of impiety depart", i.e. discedat. The rival loses on two grounds: the clause is the condition on which the priests are to believe, and it is the removal of the leprosy that is the sign, not its descending; and Migne uses apparet elsewhere in this very chunk (@0112D) rather than paret, so he is not a writer who drops prefixes casually. The reading is recorded rather than concealed.

6a. @0114C — Gerazenorum, the span the ownership test never saw, adjudicated

⚑ Added 2026-09-05 at the coordinator's request, after lemma-inventory.mjs was found to require the address form VERS. n.--. My range holds two lemmata Migne prints without the dash, and the inventory therefore filed both as second lemmata — which I reported in §(b) as an arithmetic correction. What I could not see from my own file is the consequence: the ⚑ MATCHES ONLY ELSEWHERE ownership test fires only on a span carrying a VERS. address, so it silently skipped both. The count was right and the instrument was wrong.

⭐ The per-verse collation in §(c) C5 was not blind to it. That collation takes its chapter from the CAPUT head and its verse from the VERS. address in my own Latin, not from the brief, so both dashless lemmata were in its 134 and this one is already among its sixteen misses. That is the argument for rebuilding the check rather than reading the pre-scan: a tool bug in the brief did not propagate into the collation, because the collation did not consume the brief.

The read. Archive p.62 and Gallica f57 — col. 114, not p.63/f58; corner numbers 113/114 confirmed on both, as in my original pass.

All three agree, and the z stands twice on each copy — once in the italic lemma, once in the roman gloss four words later. Both copies also print the address as the bare VERS. 28. with no dash, so the dashless form is Migne's and not an artefact either.

What the read settles, and what it does not. It settles ownership completely: two witnesses eliminate Corpus Corporum, and s→z is a different sort, not a filled counter or a thickened stroke, so it is not the over-inking class that produced my @0122B verse number. The divergence is Migne's. What it does not settle is whether the divergence is worth a public claim, and that is §2's question.

⛔ DECLINED, and here is why the rival lost. §2 lets a single-word [var:] stand on either of two limbs, and it fails both:

1. The gloss does not argue from the word. It says only Geraza urbs est Arabiae — it names the city and spells it the same way. There is no etymology, no play on the letter, nothing the z carries. Contrast the 0000 stint's @0069B, where the lemma Adjutus is the etymology that answers the verb's aspect and the marker fired on one letter; here the second occurrence is corroboration of Migne's spelling, not an argument from it.
2. It is the form of a transliterated proper name, which the 0000 stint's Policy 4 declined across five instances (Esrom · Josaphat · Eliachim · Sadoch · Nazareus) on the ground that such spellings "are not different readings, and none changes what any sentence says." Γερασηνῶν has no settled Latin transliteration; the Clementine's uniform s across its four occurrences is evidence about that edition's normalising hand, not about Migne being wrong. The English is the same word either way.

⚑ The one consideration that cuts the other way, stated so it can be argued with. Nazareus was declined partly because Migne varied within his own column (Nazaraeus twice on the next two columns), which made it visibly his own loose practice. Here he does not vary: he is uniform in z. Uniformity is the harder case, because a consistent spelling looks more like a deliberate reading than a slip. I take it the other way — a consistent orthography is house style, not a reading — and I would rather be argued out of that than have applied it silently. Firing here would also breach the standing reason for the class: five name-spellings marked in this work would bury the six substitutions that are the range's real payload.

No [var:] fired; 0018.md is unchanged. The English continues to render Migne's own transliteration, "Into the country of the Gerazenes … Geraza is a city of Arabia".

⚑ The other span the bug hid in my range clears. VERS. 25. Suscitaverunt. @0114B is verbatim in Mt 8:25 and is already listed among §(c) C5's 118; it was read on both witnesses in my original pass (the archive prints a dash before it, Gallica and our twin do not — §(d) 15).

7. @0114D — Et ecce magno, and how far convention 3 expanded it

The lemma is a three-word fragment of Mt 8:32 as Migne's text reads it, et ecce magno impetu abiit totus grex. Convention 3 renders a fragment as the phrase it stands for, so the English reads "And behold with great violence", supplying impetu's English from the verse the fragment opens — exactly the move Ruth's Non te. → "Let no man see thee" licenses. The [var:] then names what the Clementine has. ⚑ Stated explicitly because the expansion and the marker interact: without the expansion the divergent word magno would be the only word on the page and the marker would have nothing to sit beside.

7a. Convention 3 — every fragmentary lemma I expanded, and how far

Convention 3 renders a fragment as the phrase it stands for, and the launch brief's own control warns that contrasting an expanded fragment with a complete one-word lemma manufactures an inconsistency that is not there. So: the complete one-word lemmata in my range are rendered as they stand (Ask. · Beware. · Enter ye. · Thomas. · Peace. · Then. · Numbered. · Weeping. · Master. · Lord. · Faith. · Arise. · Twelve. · Matthew. · Rising up.), and the clipped ones are expanded from the verse each opens. Named in full, since only a list can be audited:

• @0110C Nunquid. → "Do they gather." (Mt 7:16 Numquid colligunt de spinis uvas), on Luke's precedent at 9000 @0265A where Nunquid potest. became "Can he."
• @0111A Non potest. → "A good tree cannot." (Mt 7:18), the subject supplied because English cannot leave a finite verb bare.
• @0109B Si piscem. → "If he ask a fish." (Mt 7:10 si piscem petierit).
• @0108C Nolite sanctum, etc. → "Give not that which is holy, etc." (Mt 7:6 Nolite dare sanctum canibus).
• @0108D Et conversi. → "And turning upon you." (Mt 7:6 et conversi dirumpant vos).
• @0114B Modicae, etc. → "O ye of little faith, etc." (Mt 8:26 modicae fidei).
• @0114C Duo habentes. → "Two having demons." (Mt 8:28 duo habentes daemonia), and Saevi. → "Exceeding fierce." (saevi nimis).
• @0114D Et ecce magno. → "And behold with great violence." — §(d) 7, and the one where the expansion and a [var:] interact.
• @0117B Et cum ejecta esset. → "And when the crowd was put forth." (Mt 9:25 cum ejecta esset turba).
• @0120B Tu es. → "Art thou he." (Mt 11:3 Tu es qui venturus es).
• @0120C Qui non fuerit. → "He that shall not be scandalized." (Mt 11:6 qui non fuerit scandalizatus in me).
• @0113D Et accedens. → "And there came." — deliberately not expanded to "one scribe", because the gloss immediately names him and the addition would be redundant rather than clarifying.

⚑ Where a fragment's continuation is a second lemma Migne is about to gloss separately, I did not swallow it: Et descendit. stays "And there came down" and does not take pluvia, because Pluvia. is the next lemma four words later.

8. @0115C — the [var:] sits after the span, not inside it

Migne's italic run is Duplici contritione conteret eos Dominus. [n: (BED.)] Quid est facilius, — one span carrying a quotation, a siglum, and the next lemma. Nested markup is not expressible, so the [var:] stands immediately after the run closes and names its verse (Jer 17:18) so that a reader cannot mistake which of the two quotations it is about. Recorded because a marker that floated to the wrong lemma is the 8995 @0601D failure mode with a different marker.

9. @0117B — in tuam domum against sua domus four words later

Id est, de domo principis in tuam domum. Domus principis Synagoga est, sua domus coelestis Jerusalem. A second-person possessive in a sentence whose next clause uses the third. Both witnesses print tuam (archive p.64, Gallica f59). A real word occupying its slot; rendered as printed, "from the ruler's house into thy house". No marker: the sentence is not false, only inconsistent, and Migne addresses his reader in the second person freely elsewhere in this range (Ne putes, Etsi difficile sit ut aliis facias, si juxta fragilitatem nostram accipis).

10. @0119C — Dipondio enim est ex duobus assibus, an ablative as subject

Both witnesses. Rendered literally, "For a dipondius is out of two asses", keeping the coin-names as Migne prints them (as, dipondius) rather than converting to farthings, since the whole gloss is an arithmetical argument about the two of them and a conversion would delete it. The Douay's "farthings" appears only inside the quoted Lc 12:6, where Migne quotes dipondio himself and the English carries "two farthings" for the received phrase — the two words stand three lines apart on the page and the Latin column shows why.

11. @0121D — cum et jejunium et saturitatem petitur spernunt

Read on both witnesses; both print petitur, a third-person passive with two accusatives beside it and no grammatical slot of its own. Rendered "when they spurn both fasting and the fullness that is sought", which renders every printed word and attaches petitur to the nearer accusative. A [sic:] is impossible (the form is a real word) and a [cj:] fails 18a's bar (the English asserts nothing the author is arguing against). Logged; the conjecture petitam is the obvious one and it is here, not in the text.

12. @0123A — coomnipotentem, a coinage, and the punctuation I supplied

…incipiet non habere qui dedit per omnia coomnipotentem nota, per tradita, Filium… Both witnesses print coomnipotentem, closed up at normal letter spacing. It is a coinage, not a non-word — co-omnipotens, "co-omnipotent" — and it is renderable, so Pattern 7's main rule governs and no marker fires.

⚠ Migne prints no mark at all between qui dedit and per omnia, and English cannot hold that. I supplied a comma, not a colon or a semicolon, and used our own double quotes (Pattern 6: " " is ours) round the two words Migne is glossing as words: "by "all things" note the co-omnipotent, by "delivered", the Son". The comma keeps the punctuation band test clean; the supply is recorded here because it is a supply.

13. @0121A, @0122D — the two anacolutha of CAPUT XI, and what I supplied

• @0122D Respondens tacitae objectionis quasi quaereretur… — a genitive where the sense wants an accusative or a dative. Rendered "Answering of a tacit objection, as though it were asked what the reason was…", keeping the genitive visible.
• @0121C Coaequales contribules quos solebant exprobrare… — two nominatives in apposition with no verb, the sentence carried over from Similis est pueris. Rendered as apposition.

14. ⛔ @0122B — THE VERSE NUMBER, and why it stands untouched

The archive copy prints VERS. 24. — Vae tibi. Our twin prints VERS. 21., which is the correct address for Mt 11:21, the verse the gloss expounds. Gallica f61 prints VERS. 21. — the archive's 4 is a 1 whose counter has been filled by over-inking.

⚑ This is the 8950 episode reproduced in my own range: a digit, on one witness, in the direction that reads like a printer's habit. Under the standing rule a verse address may not be altered on one witness, and here the second witness settles it in our twin's favour. No change, no marker, and the English carries VERS. 21.-- verbatim under convention 1.

15. Verse addresses Migne prints irregularly — three, all carried

• @0114B VERS. 25. with no .--, where every neighbour has the dash. Our twin has it so; Gallica agrees; the archive prints a dash. Witnesses disagree on a dash, which is exactly the case the brief says may not be adjudicated on one copy. Left as our twin has it.
• @0114C VERS. 28. with no .--. Both witnesses agree — no dash on either.
• @0112B our twin's VERS. 27.-- against the archive's VERS. --27. (dash before the number). Gallica prints the normal form and our twin follows it.

Convention 1 makes the address pass through verbatim in any case, and nothing machine-readable consumes it.

16. @0121D, @0122A — two press variants, both terminal stops

The archive prints a full stop after et hujusmodi (@0121D) and after sed praedicit ventura (@0122A); Gallica prints neither, and our twin has neither. Both copies are clean at both sites. That is a press variant in the 8950 Mititte/Mittite class and the 0000 stint's Bethehem/Bethlehem class: two printings, no [sic:], no patch, crux only. The English paragraphs end without a mark, per Pattern 8's "never supply".

17. Sentence-type marks Migne prints — all sixteen carried, none supplied, none deleted

Sixteen question marks in my Latin and sixteen in my English, band for band: @0114C (cur non creditis etiam in mari?), @0118B (nunquid defuit vobis aliquid?), @0118C (Unde ergo necessaria?), @0119B (Nonne duo passeres?), @0119C (Nonne quinque passeres veneunt dipondio?, Nunquid de bobus cura est Deo?), @0120C (an alium ad haec sacramenta missurus es?), @0121C (qui noluistis credere ei?), @0121D (cur Joannes displicet?, cur Filius hominis?), @0122C (Nunquid usque in coelum exaltaberis?, superbissime restitisti?), @0122D (quem gentilitas tam prona recepit?), @0123C (Et quid jugo ejus suavius?, quid onere levius?), @0123D (hic et concupiscentia damnatur?).

⚑ Pattern 8b was not needed once. The two indirect questions in the range — quis ordinem rei gestae teneat, non apparet (@0120C) and Quando hoc sit factum non expressit Matthaeus (@0113C) — are both closed by Migne with a full stop, so nothing had to be recast and nothing was dropped.

18. Punctuation, band by band — the mechanical test, and what it leaves

Column anchors are verified 1:1, so they partition both twins identically. Comparing marks band by band across all 72 bands of my six chunks (66 column anchors plus each chunk's head band), with [var:] contents stripped:

• question marks: zero mismatched bands (16 in the Latin, 16 in the English);
• guillemets: zero, in both — this book prints none;
• colons: one mismatched band, and it is the TEI patch (@0107C, §(d) 1). Every other band is exact, after I removed thirteen colons I had supplied and restored seven that Migne prints and I had softened — the destructive direction, and the one this test exists to catch. The seven: @0110B (adjecit: Attendite:), @0113D (trans fretum: Lucas vero … in via: non est contrarium:, two of the three), @0114B (et rogent: et rogantibus), @0118B (illud unde emuntur:, which I had softened to a semicolon), @0120C (descensurus sum: debeo), @0121C (Joannis moti sunt: vel).
• semicolons: zero mismatched bands, after removing seventeen I had supplied and restoring one Migne prints (@0110B, possit accipi; tamen). ⚑ I have not applied the standing comma→semicolon licence at all in this range: Migne prints 35 semicolons of his own here, so the Tobit exception (which rests on a book printing zero) does not apply, and every semicolon in my English stands where he prints one.

19. pietas — one Latin word, two English words, disclosed

pietas is not on any settled axis and the 0000 stint recorded that it splits. In my range it occurs three times and I rendered it twice as "mercy" and once as "piety":

• @0112C non quasi pietatis incredulus → "not as one incredulous of his mercy" — the leper doubts the will, not the power, so the word is Christ's compassion;
• @0116B ad humilitatis et pietatis gratiam provocat → "the grace of humility and of mercy" — the Pharisees are being provoked toward the sinners Christ has just received;
• @0120A odium in suos, pietas est in Deum → "hatred toward one's own is piety toward God" — here the word is dutifulness toward God and "mercy" would be simply wrong.

The related pium affectum at @0120B is rendered "the devout affection", matching the 0000 stint's pio affectu → "devout affection" at @0075C. I record the split rather than let it pass, since that is how the axes got made.

20. Pattern 17 — where the singular and the plural change hands

The range moves between them constantly and a reader can see it with no apparatus. Singular (thou): the whole of CAPUT VII's second person — benefeceris, liberaberis, punieris, tuleris, incurras, quid autem vides, fratri tuo, corde tuo, ejice, si difficile sit ut aliis facias quod tibi vis fieri; and in CAPUT XI Tu es, debeo te nuntiare, exaltaberis, restitisti, soror tua ex te, exstitisti ingrata, si juxta fragilitatem nostram accipis, quod pati non vis. Plural (you): the apostolic commission throughout CAPUT X — Gratis accepistis, in zonis vestris, Excutite pulverem de pedibus, Ecce ego mitto, Dabitur enim vobis, Quod dico, Ne ergo timueritis eos — and in CAPUT VII si petieritis patrem vestrum, Intrate, Attendite, Nolite, Ut faciant vobis homines, and in CAPUT XI Cecinimus vobis, Convertimini ad me in toto corde vestro, ut minus excusabiles sitis.

⚑ The visible payoff is at 0109B–0110A, where Migne moves from si petieritis patrem vestrum (plural, the crowd) to Etsi difficile sit ut aliis facias quod tibi vis fieri (singular, the individual reader) and back to Intrate within four lines. Under a flat "you" that is invisible.

21. Supplied subjects and objects — every one named

The brief requires that each supplied pronoun be named. Mine are:

• @0108A praeventus majoribus, alium non judicet — the subject is the participle itself; rendered "one who is himself forestalled". Nothing supplied.
• @0113B Per istas duas partes mundi ex quibus gentes vocandae erant designantur — the subject of designantur is unexpressed and cannot be gentes (which is the antecedent of quibus). Rendered "they are designated", i.e. the two quarters. Both readings construe; I say so rather than choose silently.
• @0114C Marcus dicit quod nullus jam daemoniacum, etc. — the verb has been cut away by Migne's etc.; I supplied [could bind] in brackets, from Mc 5:3 (nemo poterat eum ligare), which is the sentence Mark's own text supplies.
• @0112B Vel aliqui in turba, quia infideles, vel potius omnes de excellentia sapientiae — no verb; supplied [wondered] in brackets, from the lemma's own verse (Mt 7:28 admirabantur).
• @0107D Ita facit Deus, et non inique ut illi in aeternum — no verb in the ut clause; supplied [it be] in brackets. The paragraph below (aeternam poenam et inveniunt) fixes the sense.
• @0117B Quia Synagoga laetitiam Dominicae inhabitationis, etc. — the verb is inside Migne's cut; supplied [lost] in brackets, from laetitiam being accusative and the Synagogue being the losing party throughout the paragraph.
• @0120C Beatus Joannes, despectis quoque ovium pilis … pilos camelorum, etc. — same shape; supplied [chose] in brackets.
• @0120C hoc non contra Joannem dixisse putandus est — the subject is Christ, not the qui clause that precedes it; rendered with a resumptive "he". Both parses construe and I record it.
• @0117B paruit in susceptione — rendered "it was obedient in the receiving". ⚑ The rival is pareo = "appear", and it lost on Migne's own usage: he writes apparet for "appear" eleven columns earlier (@0112D, Apparet supra legem … esse gratiam), so the bare paruit is the obedience verb.

22. Test 2a — negatives resting on a verb with a quantifier or correlative beside them

The trigger population in my range, each with the host proved from the sentence's own structure:

• @0122C Non solum Corozaim et Bethsaidae vae erit, sed et tibi Capharnaum — host is Corozaim et Bethsaidae, proved by the sed et tibi. Rendered "Not to Corozain and Bethsaida only shall there be woe, but to thee also".
• @0110D non hoc faciunt mali, sed fit de illis consilio Dei — host is the agency, not the verb; and the next clause quia non eorum fructus est, sed vitis puts the second negative on eorum, which the English keeps ("it is not their fruit, but the vine's").
• @0111A Arbor bona vel mala non natura … sed voluntas; host natura.
• @0110C Non a veste, sed ab operibus; host a veste.
• @0118B viderentur non causa salutis praedicare, sed lucri; host causa salutis.
• @0118D mala inferre non debet, sed tolerare; host inferre.
• @0119A quod persecutiones non debeant timere, sed declinare; host timere.
• @0122A quibus non optat mala, sed praedicit ventura; host optat mala.
• @0122D non opponit insipientes et hebetes, sed parvulos; host insipientes et hebetes.
• @0123A Tradita autem omnia non coelum et terra … sed hi qui; host coelum et terra.
• @0116C A mundi amore, non solum a cibis; host a cibis.
• @0119D non quod non sint honorandi pater et mater, sed post Deum — both negatives carried, and the sed shows the second is a real denial and not a scope error.

And the one that is not a negation at all, recorded because a negation-counting check will see it: @0112C De voluntate non quasi pietatis incredulus, dubito; sed quasi meae colluvionis conscius non praesumo. The first non is on the quasi phrase, not on dubito — the leper does doubt, and says so. Rendering it "I do not doubt" would reverse the sentence and read better, which is the tell.

(e) DECLINED-ODDITY LIST — every suspect word I did NOT mark, with its column

Eight of the eighteen entries below are archive-only readings that Gallica refutes; one is a Gallica-only reading the archive refutes. They are listed in full because the count is the finding: read on one witness this file would have carried eight or nine confident claims about Migne's type, and every one of them would have been false — including a changed verse number.

And the suspicions that were NOT about either copy's ink

Suspiciously TIDY forms — brief §4's other direction, and the checked negatives

The class that leaves nothing to notice is the one where the digitization has corrected Migne. I looked for it deliberately at the places where our Latin reads smoothly in rough surroundings, and I read whole lines, not target words. In this range the answer is: none found. All seven defects run the other way — the digitization corrupting, not tidying.

⭐ The negatives worth recording, because the brief asks for them explicitly and because a checked negative is only worth having when the positives have been counted. The five passages in my range where our Latin is rough enough that a transcriber might have tidied it are @0108B (Salutem with no verb), @0112C (fecisset with no governing clause), @0112A (sensum rigorem), @0117B (in tuam domum) and @0121D (petitur). All five are printed exactly as rough on both witnesses as they are in our twin. Our transcription is faithful at every one of them. This volume's digitization drops and swaps letters; in my range it added nothing and repaired nothing.

The ⚑ MATCHES ONLY ELSEWHERE candidates — none in this range

The launch brief lists eight; the nearest are @0106B (the previous stint) and @0157D (a later one). My range is clean of them, and I confirm it rather than assume it: no span in cols 0107C–0123D carries that flag in 9002-lemmata-0016.txt, and the per-verse collation in §(c) C5 is the stronger check that would have caught one had the flag missed it.

Preserved from cruces-0022.md, unedited; headings demoted one level. Companion files: data/briefs/9002-PLATE-READS-0022.json, data/briefs/9002-PENDING-TEI-PATCHES-0022.md, data/briefs/9002-lemmata-0022.txt.

9002 — cruces, stint 0022–0026 (cols 0124A–0137C)

Glossa ordinaria, Evangelium secundum Matthaeum, PL 114. Chunks 0022–0026: CAPUT XII (the ears of corn, the withered hand, the Beelzebub controversy, the sin against the Spirit, Jonas, the true kindred), CAPUT XIII (the parables) and CAPUT XIV (the Baptist's death, the five loaves, the walking on the sea) — 5,551 Latin words, 16 columns, both plates read on every one of them.

Companion files from this stint: data/briefs/9002-PLATE-READS-0022.json (every page opened, both witnesses, corner numbers first) and data/briefs/9002-PENDING-TEI-PATCHES-0022.md (three sites where our Latin has departed from Migne).

(a) THE POLICY I APPLIED — stated before the findings, so it can be argued with

1. Plates first, both of them, and on every column. All eight archive pages (pp. 67–74, 300 dpi) and all eight Gallica leaves (f62–f69) were read column by column, corner numbers checked before anything else on each. This was not a candidate-driven second-witness check: I read every Gallica column whether or not the archive raised anything, because a candidate-driven check can only see divergences the archive manufactures and is blind to the ones it hides. That is how @0125A surfaced.

Sixteen candidates raised on the archive; fourteen refuted at Gallica. The 12:1 measured by the 0000–0005 stint and the 20:1 by the 0006–0011 stint hold on pp. 67–74 as well.

2. Markers. 13 [var:], 2 [sic:], zero [cj:], zero [d:], zero [ed:]. Both [sic:] stand on two witnesses and quote a word still verbatim in our Latin (@0131A Excaevavit, @0135B Constristatus); each was re-fetched from Gallica at native resolution before it was written. This is the first stint of Matthew to fire one, and the reason is not boldness: it is that these two are non-words that our transcription faithfully carries, which is exactly the narrow class the corrected one-witness rule leaves open.

3. When a [var:] fires, and it is the previous stint's threshold, not a new one. Launch brief §2 requires a single-word [var:] to have a witness of its own — the gloss argues from the word, or the form is one letter-level corruption cannot manufacture. I applied that mechanically, and I adopted the 0006–0011 stint's own worked application of it so that the two ranges can be read together:

• fires — a different word (Macta/occide, plurimis/Psalmis, sed/sic, sed/et, enim/autem, autem/enim), a changed construction (Si vos persecuti sunt for Cum autem persequentur vos), an added or substituted phrase (ad vos, de semine vestro, inter vos for vestrum), a finite verb for a participle inside a quotation that stands on its own.
• declines — a one-letter difference in a form that means the same or better (appropinquabit/appropinquavit, Simile/Similis, fructum/fructus, quicunque), word order alone (Dominus est enim, Fulgebunt justi), an ellipsis, a word division (Ne quando), and a divergence that is exactly what Migne's own host syntax requires (@0136C venturi erant, @0136D Volens praeterire).

⚑ The last of those is a test I had to state, because the range forced it twice in one column. A divergence fires where Migne introduces the words as a quotation standing on its own — @0136D Unde Marcus ait: Vidit eos laborantes remigando., closed by a full stop before his comment — and declines where the words are woven into a clause of his own, so that the indicative or the participle is his construction and not a reading. That is the 0000–0005 stint's @0069B ruling (habebit for habeat once ut is gone) applied to a harder pair of cases.

4. Where the plate was read and our twin diverges, the English renders the plate. Three sites, all three Corpus Corporum's, all three filed as pending patches: @0130C Scaagesimum → plate Sexagesimum; @0134A a comma our file inserts in ad, coelum; @0134C a semicolon where both plates print a colon. The third is this stint's single deliberate band-level punctuation mismatch and is declared as such in §(d) 12.

5. Vocabulary axes, applied from the first chunk. turba → crowd (0123D pro turbis, 0130A Turba stabat, 0133C/0135C/0136B turbas); no multitudo except the AUG. note at 0132C (Multitudo non est excommunicanda) → multitude, so the two words do stand in one range and are kept apart. daemonium → demon (0125D, 0126A–D, 0129B, 0130B), diabolus → devil (0126C, 0127B, 0128A, 0129A, 0132A), satanas → Satan (0129A) — three words, never crossed. mysterium → mystery / sacramentum → sacrament, 1:1 (0130A mysteriis, 0130D mysticarum, 0134A mysteria, 0136B mysteria legis; 0130A sacramenta, 0133C sacramenta, 0134B sacramenti) — with one declared exception at 0135B, argued in §(d) 1. The gentes family: gentes → Gentiles in the Judaei/gentes typology (0125D, 0129B, 0134C ×2, 0135D Fastidiunt Judaei gentes recipi, 0135D In viam gentium); nations in the two plainly generic quoted Isaiah verses at 0135D; singular gens → nation (0126C and 0132D genti facienti fructum ejus, 0128D genti incredulae, 0129B in hac tota gente); gentilitas → the Gentile world (0128D). No natio, no ethnicus, no gentilis. The gloss's own voice is not archaised: dicit/ait → says throughout.

6. usque ad — the etc. splice test, run per occurrence. 11 in the range: 10 formula, 1 ordinary. Every one of the ten is preceded immediately by etc., and every one is rendered "as far as": @0127A, @0128A, @0129C, @0130A, @0131C ×2, @0131D, @0135A, @0135B, @0136C. The one ordinary use carries no etc. and is named here as the axis requires: @0124D manus … arida erat usque ad tempus Christi, temporal, rendered "until the time of Christ". ⚑ A twelfth occurrence is bare usque, not usque ad: @0134A quae usque hodie trahitur, rendered "which is drawn even to this day". ⚑ And @0130A is worth a line of its own: the formula there is followed by a resumption tail that itself begins ad quae — etc., usque ad ad quae Judaei non intraverunt. Both witnesses print the doubled ad, and it is not a dittography: the first belongs to the formula, the second to the resumed sentence. No [d:].

7. Second person by number, not by reverence (Pattern 17). The range moves between the two and the Latin decides every time: plural you for the Pharisees (0124B probatis, 0128A vos qui … dicitis), for the disciples (0130D Vobis datum est, 0131B Vos ergo audite, 0134A–B intelligitis … faciatis), and for the Jews addressed at 0128D–0129A (Vos non estis similes … non poenituistis … non didicistis); singular thou for the individual reader (0128C habes tres dies) and for Peter to Christ (0137A Domine, si tu es … ad tuam imitationem). ⚑ "ye" is used only inside a quoted scripture lemma, following this work's shipped practice (0009, 0019): Have ye not read, Ye shall not do any servile work therein, Go ye not into the way of the Gentiles, It is I, fear ye not, And bind ye. The scripture-echo licence covers the quoted words and stops where they stop.

8. Nothing supplied but one bracketed word, and it is named. No bracketed conjecture stands in the running text anywhere in this stint, and where Migne's sentence lacks a word English cannot do without, the English is left as broken as the Latin and the site is logged in §(d). The single exception is @0132B [good], argued in §(d) 24 — a supplied English complement in the corpus's bracket convention, not a reading.

(b) SPAN COUNT — 211 against 211, paragraph-bounded, and 0 guillemets

data/briefs/9002-lemmata-0022.txt heads its inventory 211 spans, of which 0 open-ended.

Counted in my own Latin both whole-file and paragraph-bounded (the brief requires both, because an unmatched delimiter defeats a counter in either direction):

211 = 211 on both counts, with 422 asterisks and not one guillemet character in 5,551 words, so the two failure modes that let a bare total agree falsely are both excluded. Matthew is an italic-lemma book. The cut is clean at both ends: the file opens at [0123D] VERS. 1.-- In illo tempore., which is chunk 0022's first span, and closes at [0137C] VERS. 36.-- Fimbriam vestimenti,, which is chunk 0026's last.

(c) THE LEMMA DIVERGENCE LIST — all 211 spans, in order, declines and reasons included

The ✓ / ⚠ column is the brief's mechanical pre-scan; the last column is my verdict after reading the line and the verses the pre-scan names. ⛔ A ✓ was never taken as clearance. Two entries below exist only because I read the printed references rather than the ticks: @0124C Dico autem vobis (whose reference list leads with three Lucan verses and whose true match, Mt 12:6, is fourth on the line — and which is the very span that fired a [var:] at @0093A under a Mt 5 address) and @0132D Simile est regnum (whose ✓ points at Mt 13:44/45, not at the Mt 13:31 the address names).

Migne's own citation sigla pass through verbatim under convention 4 and are not treated as divergences. In this range that covers (Isa. VI) for Jo 12:40 @0127A, (Joan. X) for Jo 12:39 @0127A, (Matth. III) for Mc 3:30 @0127B, (Joan. XI) for Jo 12:39–40 @0131A, (Matth. XXI) for Mt 3:2/4:17 @0126C, (Joan. IV) for Mt 14:2 @0134C, (I Cor. I) for 1 Cor 2:14 @0136A, and (Eccl. IV) for Sir 4:36 @0125A. A wrong siglum is Migne's citing practice; refDisplay keeps it and refKey resolves it.