Actus Apostolorum

Anselm of Laon and his school, on the Acts of the Apostles. Translated 2026-08-25 by four translators working in parallel — 0000–0004 (0425C–0437C), 0005–0009 (0437D–0450A), 0010–0013 (0450B–0459D) and 0014–0017 (0460A–0470A) — and merged here. 18 chunks, 16,287 Latin words. verify-english.mjs clean across all 18.

Apparatus fired, work-wide: 36 [var:] · 1 [sic:] · 1 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 [cn:]. The single [sic:] is licensed by a plate read (§P). Zero em-dashes, zero guillemets, notes at exact parity throughout.

⭐ This is the most familiar narrative in the batch, and that was the stated danger. Pentecost, Ananias and Sapphira, Stephen, the Damascus road, the shipwreck — sentences every translator knows in English before reading them in Latin. Conformation is the defect this apparatus exists to catch, and it is least visible exactly where the text is best known.

§M · THE MERGE

Four translators. Each later stint read its predecessors off disk and inherited from them, and said where: the second changed four of its own renderings to the first's §V table (testamentum → testament, Divinitas → Divinity, lower-case spiritui → the spirit, sermonis → discourse); the fourth caught Isaiah → Isaias in its own draft from the third's file and repaired it rather than conforming silently. Measured across the finished work: six Isaias, zero Isaiah.

⚖ Non-words: RULED — render for the evident sense, do not carry the Latin

The first two stints split on this, and the second flagged the split for a ruling instead of harmonizing to its peer: one renders a non-word for its evident sense, the other carries it in Latin italics. Six sites, and neither fired a marker, so nothing asserted anything about the plate either way.

Ruled: render for the evident sense. It is the practice in every other book of this batch, and it is what test 1 already requires whenever a marker does fire — a [sic:] must not swallow the word it marks, so the rendering and the marker are both required. Consistency demands the same when no marker fires. Applied from the third stint onward and honoured by the fourth (Sceno factoriae, Tyrrhenc, Thelasson all rendered, all logged, none marked).

⚖ apostolus — NOT mirrored, and the plate is the reason

The second stint declined to mirror apostolus/Apostolus, counting 19 lower-case against 11 capitals, alternating for the same referent inside a single sentence at 0439A and again at 0446D–0447A. That is right, and it is the same evidence that settled In Regum/In regum in a sister book: where Migne's own casing is inconsistent for one referent, it carries no distinction, and mirroring it would transcribe an accident as a signal. The mirror rule applies where the case is semantic in English; Ecclesia is mirrored throughout this work accordingly.

⭐ And 0443C's sidus Dei vestri is mirrored, deliberately — Migne capitalizes a pagan star-god and the gloss then argues from the capital. The rule working in both directions.

The gentes seam — checked, and both readings are correct

The fourth stint flagged, without harmonizing, that the first renders omnes gentes "all nations" at 0432D while it has "All the Gentiles" elsewhere. No change: the series rule already decides this per occurrence — gentes is the gentiles where the contrast is with Israel and nations where the sense is simply peoples. At 0432D the words are Act. 3:25's promise to all peoples; at the omnia quadrupedia gloss the whole point is gentiles against Jews. Both right.

⭐ A defect class NEARLY MANUFACTURED, and the check that stopped it

The fourth stint filed 0469A's unde? patet beside the punctuation class — then scanned the whole work and found eight ? [a-z] sites, six of them Migne's own. Continuing in lower case after a mark is his ordinary habit here. Recorded as a checked non-defect so that nobody spends a plate read on it. That is the corpus-as-its-own-control move, used to kill a candidate rather than confirm one.

§P · THE PLATE READ AT 0435B — three defects in one line, and one of them reversed our assumption

The first stint flagged a full stop followed by a lower-case word and declined to capitalize, on the ground that capitalizing would silently repair the plate's typography — while saying plainly that it would read to a viewer as our build defect. Both instincts were right, and the answer was neither of the two we expected.

The plate prints a COLON. (On-disk PL 114 scan patrologiaecurs06saingoog, page map re-calibrated against the printed corner numbers before the read: p. 223 = cols 435/436.) A colon before a lower-case word is ordinary — so the sentence was never Migne's defect. The mark was ours.

⭐⭐ Reading that same line for that answer produced two more, which is the addendum's rule working exactly as written — the normalization class lives in the NEIGHBOURS of the word you came to read:

• a second colon corrupted, this time into a spaced hyphen, a mark Migne uses nowhere in the book;
• and a NORMALIZATION: the plate prints Chistus, missing its r, and our source had tidied it to Christus. Our Latin column was printing a word Migne did not print, in a line where he visibly slipped. The class now stands at six, and as always nobody could have flagged it from the file, because a normalized form reads perfectly.

All three patched at the TEI (data/tei-patches/8976.json) and re-chunked. The English now carries [sic: Chistus] — the work's only one — because a [sic:] asserts a plate defect and only a plate can license one.

⚑ The patch was written early and deliberately HELD until every stint had finished reading src/latin/8976, since restoring a colon changes token counts and could have shifted a chunk boundary under a working agent. It did shift one word count; chunk 0003's English frontmatter was re-synced and the work re-verified.

⚑ And the first attempt at the patch failed safe, which is worth recording: the find strings were written against the chunked text, matched zero times, and chunk-work.mjs refused to run rather than proceed. The TEI breaks both lines mid-sentence.

⭐ The fourth stint then predicted a second instance of the class at 0462A–B (fuit. ita, at a sicut … ita hinge, which is exactly where a colon belongs), wrote the patch text, and did not patch. ⬜ Unread and unpatched — see the resume note.

§Q · THE FLAGGED CRUCES, AND THE VALUE OF A CHECKED NEGATIVE

Three cruces were named in the briefs before anyone opened the book. All three came back as checked negatives, and that is a result worth as much as a finding:

• The apostolic decree (Act. 15:20, 29) — agrees with the Clementine exactly as Migne prints it. ⚑ But the negative golden rule is in the range, at 0458A, in the gloss prose (verum ea quae vobis fieri non vultis, alii ne faciatis) — roman, uncited, grammatically the gloss's own sentence. No marker: Pattern 14 covers citations. The rival is named and why it lost.
• Saul / Paul — checked at every occurrence; Migne prints Saulum at 11:25 and Saulus / Qui et Paulus at 13:9, all agreeing. No inconsistency in the book.
• Act. 20:28, the ecclesiam Dei / Domini crux, one of the most contested readings in the New Testament — is not glossed at all. Chapter XX's lemmata run 2·4·7·9·10·16·31·35; Migne skips v. 28. Recorded so the question is not reopened.
• The shipwreck numerals — a checked zero. Migne clips ducentae septuaginta sex out of Act. 27:37 and reasons from animae, not viri; every other number in the range checks.
• Act. 7:14's numbers — In animabus and In Hebraeo septuaginta tantum agree with Act. 7:14 (75) and Gn. 46:27 (70). The range's likeliest numeral defect, and it is not one.

§R · THE DIVERGENCES THAT CARRY THEIR OWN WITNESS

Several of this book's [var:] are self-witnessing — Migne's own text elsewhere supplies the control, so the divergence rests on the plate and not only on the Clementine:

• 0452A — his Act. 10:31 replaces in conspectu Dei with coram Deo, and his very next lemma is In conspectu Dei. The gloss expounds a phrase its own proof-text no longer contains.
• 0455C — dispergimini for disperdimini (scattered, not destroyed), glossed ab Assyriis et Chaldaeis, novissime a Romanis — and his own line above prints disperdimini.
• 0464B — Mt. 10:9 Nolite accipere aurum for possidere, in a paragraph proving beatius est dare quam accipere; this work prints possidere at 0434B, so both are his.
• 0458B — lemma Viduae filius where Act. 16:1 reads filius mulieris Judaeae, and the gloss's first words are Vel Judaeae, offering the Clementine's own word as its alternative.
• 0463D — Sosipater for Sopater, and the gloss etymologizes the name (salvans dispersos needs the Sosi-). Rom. 16:21 checked: it does read Sosipater.

Six misprinted citations across the four stints, each identified from the quoted words and never from the numeral, all left exactly as printed. ⬜ Not yet entered in data/citation-corrections.json — see the resume note.

First stint of four. Translated 2026-08-25, one translator, 5 chunks, 4,480 Latin words (frontmatter counts: 8 + 1,529 + 1,587 + 653 + 703). verify-english.mjs clean over this range — the only failures it reports are the peers' still-missing 0006–0017. English/Latin ratio 1.46–1.55 (0000 excepted: an eight-word file).

Apparatus fired: 6 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 38 [n:] (exact parity with the Latin twins: 0 + 5 + 14 + 9 + 10). Zero guillemets in either language. Zero em-dashes in the English. One ? in the Latin (0427D) and one in the English.

⚠ This file is cruces-0000.md, not cruces.md. Three peers hold 0005–0017; the merge is the launcher's.

§V · WORK-WIDE VOCABULARY FIXED HERE — later stints should inherit, not re-invent

Fixed at first occurrence and held for all five chunks. Where the pair is held apart deliberately, a later sweep that collapses it would destroy something the plate makes.

Proper names, form for form as Migne prints them — this is where Acts is most dangerous, and every one of these is a place the familiar English Bible would have overwritten him:

• 0429D Barnabae / Barnabas, not Barsabas. Act 1:23 reads Joseph, qui vocabatur Barsabas; Migne's gloss says In laude Barnabae immoratur and Barnabae laus cumulatur … quia cum Paulo gentium apostolus ordinatur, which is the Barnabas of Act 4:36. He has conflated the two men. Rendered "Barnabas" both times. Normalizing to Barsabas would have silently repaired his conflation and destroyed the evidence for it.
• 0429D Matthaeus, not Matthias. quia pro laude Matthaeus habet, quod sorte eligitur — the man chosen by lot is Matthias, named as Et Matthiam eleven words later in the same paragraph. Migne prints both forms. Rendered "Matthew" and "Matthias" respectively, so the plate's two forms stay two forms.
• 0429A–B Jesus for Josue. Moyse defuncto, Jesus successit, qui populum in terram promissionis induxit — this is Joshua. Rendered "Jesus", per the 8965 precedent. And at 0435A Migne sets the two names side by side himself (Josue de Ephraim, Jesum de Juda Deus suscitavit), which is the control: he knows the difference and uses both.
• 0427D the apostle list in the order Migne gives it as the Greek's — Petrus et Andraeas, Jacobus et Joannes, Simon Zelotes et Judas Jacobi, which is not the Vulgate order of Act 1:13 (Petrus, et Joannes, Jacobus, et Andreas…). Rendered in his order. Judas Jacobi → "Judas of James", not Douay's "Jude the brother of James": nothing in the Latin supplies "brother".
• 0428C Chananaeus · Chana kept; the gloss is the etymology (Chana zelus interpretatur). 0427D Zelotes kept untranslated as the name.
• 0437B Barjona kept.
• 0434A Joathan filius Oziae → "Joathan the son of Ozias" (Migne's Vulgate forms).
• ⚑ 0427D Andraeas is the one name I did NOT keep form for form: it is Andreas with the ordinary medieval ae-for-e, the same class as coelum for caelum and tanquam for tamquam, which stand all over this range. Pattern 9 governs (attested medieval orthography, translate normally) and the English is Andrew. The Latin spelling is recorded here so the decision is visible.
• ⚑ Hierosolyma / Hierosolymorum (0427D, 0428C, 0429C) and Jerusalem are both rendered Jerusalem. English has no exponent for the two Latin name-forms (Pattern 9 family). Recorded as a class rather than site by site.
• 0427D Galilaei → Galilaeans, holding Migne's ae.

Second person (Pattern 17). 21 archaic second-person tokens in this range (thou 4 · thee 2 · thy 6 · art 2 · hast 5 · wilt 1 · shalt 1), every one fixed by the Latin's own morphology and not by who is addressed: 0425C si tu sis amicus Dei, tibi scribitur → "if thou art a friend of God, it is written for thee"; 0432D Non derelinques · Notas mihi fecisti · Replebis me (2nd sing. futures of Ps 15); 0433C Sede a dextris; 0435B In semine tuo; 0436D Unxisti and Ps 44:8 Unxit te Deus Deus tuus … prae consortibus tuis. Every plural stays you — 0427A Vos autem · vos baptizabimini, 0427B–C Non est vestrum nosse, 0427C non potestis, 0430B Sedete hic in civitate, 0436B Qui estis judices, vestrum facite officium.

§1 · THE LEMMA COLLATION — every italic span in cols 0425C–0437C, walked

The launch brief's rule: render what MIGNE prints, never what the Vulgate, the Douay or the AV prints. Every span in data/briefs/8976-lemmata.txt for these bands was read word by word against sources/vulgate/clementine-flat.txt; the verse it actually glosses was fetched, not the pre-scan's "occurs somewhere" mark. The agreements are listed as well as the divergences, because a list of findings alone cannot be told from a report by an agent that never looked.

1.1 · FIVE [var:] FIRED — the divergences that change what the proof-text says

⭐ 0435B is the sharpest of the six, and it is self-witnessing. Migne's Beda abridgment ends in semine tuo benedicentur omnes gentes, and then, twelve lines later at 0435C, he prints as his own next lemma Omnes familiae, which is the Clementine's word. The English must carry both: "in thy seed shall all nations be blessed" in the tail, "All families" as the lemma. Conforming either to the other would have made the second lemma gloss a phrase that is not above it.

Six divergences, six markers — 0425D, 0430C, 0431C, 0432C, 0435B, 0436C. Distributed 0001 ×1 · 0002 ×3 · 0003 ×1 · 0004 ×1. The Gal 3:28 / Col 3:11 conflation is a single [var:] because it is a single quotation.

1.2 · DIVERGENCES RAISED AND NO MARKER FIRED — with the reason each declined

Every one of these is real and checked; none changes what the sentence asserts, and a [var:] is a public claim, so the crux is the right home (ship test B).

• 0427A · Vos vero baptizabimini Spiritu sancto — Act 1:5 reads vos autem baptizabimini. A particle swap with no sense change, and Migne prints his own Vos autem as the lemma four words earlier, so both forms are already on the page.
• 0427A · Quia Joannes baptizavit aqua (inside the Beda abridgment) — Act 1:5 reads quia Joannes quidem baptizavit aqua. A clip, not a variant.
• 0427C · sed accipietis virtutem Spiritus sancti — Act 1:8 reads virtutem supervenientis Spiritus Sancti in vos. Migne prints Supervenientis Spiritus sancti as its own lemma in the very next paragraph; this is his own abbreviated re-quotation.
• 0427D · Cum haec dixisset, elevatus est — Act 1:9 reads Et cum haec dixisset, videntibus illis, elevatus est. A clip in a sentence whose whole point is the two evangelists' wording, and Migne's own lemma Et cum haec dixisset stands unclipped above.
• 0427D · Galilaei, quid statis… — Act 1:11 reads Viri Galilaei. A clipped lemma opening at the second word (convention 3), not a divergence.
• 0430B · Sedete hic in civitate — Lc 24:49 reads vos autem sedete in civitate, with no hic. An added deictic; the gloss (Sessio est indicium humilitatis) turns on sedete, not on hic, so nothing hangs on it. Raised, not marked.
• 0434B · Nolite possidere aurum et argentum — Mt 10:9 reads aurum, neque argentum. Copulative for disjunctive inside a negative; no sense change.
• 0435B · A Samuele et deinceps — Act 3:24 reads a Samuel et deinceps. Migne declines the indeclinable. A form, not a reading.
• 0432D · Et exsultavit lingua and 0433C · Sede a dextris and 0435C · Venite, benedicti Patris — each drops a possessive the Clementine prints (lingua mea, dextris meis, Patris mei), and each is followed in Migne by etc. or stands as a clipped lemma. Rendered without the possessive, exactly as printed. This is the class most likely to be quietly "restored" by a later reader; it should not be.

1.3 · THREE READINGS MIGNE HIMSELF ATTRIBUTES TO THE GREEK — no marker, by construction

Pattern 14 is for a quotation that diverges without saying so. These three announce the divergence in their own words, so a [var:] would tell the reader twice and blur whose observation it was (the same logic as Migne's own ( sic )):

• 0427D In Graeco ita ponitur: Cum introissent civitatem, in coenaculum ascenderunt, ubi erant manentes Petrus — against Act 1:13 Et cum introissent in coenaculum, ascenderunt ubi manebant Petrus. He is reporting the Greek's order, and the order is the point of the note.
• 0432B In Graeco sic est: Praescientia sic traditum accipientes — no Clementine counterpart; it is a rendering of Act 2:23's Greek participles.
• 0436C In Graeco: Per nomen pueri tui Jesu — Act 4:30 reads per nomen sancti filii tui Jesu. Rendered "thy child Jesus" (παῖς), which is the whole reason Migne cites it: quod magis apostolorum votis congruit.

Also in this class, and also unmarked: 0436B In Graeco: Omnes clarificabant in eo, quod factum est: et non habetur in eo quod acciderat — Migne noting that the Greek lacks the Vulgate's closing phrase. The English sets that phrase in our own " " (Pattern 6) because it is being mentioned, not used: and "in that which had befallen" is not had.

1.4 · THE LEMMATA CHECKED AND FOUND TO AGREE WITH THE CLEMENTINE EXACTLY

Named, because a divergence list without its negatives is not a report. Each was fetched at its own verse, not matched "somewhere in the Clementine".

Act 1 (chunk 0001): Sermonem feci · De omnibus · Theophile · Quae coepit Jesus facere · usque in diem · Praecipiens apostolis · per Spiritum sanctum · Elegit · Assumptus · Vivum · Per dies · Convescens · Promissionem · Patris · Quia Joannes · Vos autem · Non est vestrum nosse · non est · Quae Pater · In sua potestate · Sed accipietis · Supervenientis Spiritus sancti · In Jerusalem et in omni Judaea · In Jerusalem · Et cum haec dixisset · Tunc reversi sunt · Et cum introissent · Petrus et Joannes · Zelotes · Maria · In medio fratrum (Act 1:15, glossed under Migne's VERS. 14 — see §2.1) · Possedit agrum · possedit · Viscera · Eorum · Deserta, et non sit · Duos · Qui cognominatus · Justus. Also the inline Jo 3:5 Nisi quis renatus fuerit ex aqua et Spiritu sancto (verbatim to the clip) and Mc 16:19 Et Dominus quidem Jesus postquam locutus est eis, assumptus est in coelum (verbatim; the [n:] misattributes it — §2.2).

Act 2 (chunks 0002–0003): In eodem loco · De coelo · Sonus · Spiritus · Vehementis · Totam domum · Sedentes · Dispertitae · Tanquam ignis · Et coeperunt · Variis linguis · Ex omni natione · Confusa · Unusquisque · Audivimus · Cappadociam · Advenae Romani · Viri Judaei · Effundam · De Spiritu · Super omnem carnem · Et prophetabunt filii · Et prophetabunt · Et dabo · Prodigia in coelo · Sanguinem · Vaporem fumi · Sol convertetur · Definito consilio · Et praescientia Dei · Suscitavit · Solutis doloribus · Providebam Dominum · Ne commovear · Non derelinques · Notas mihi fecisti · Replebis me · Quoniam a dextris · Et exsultavit · In inferno · Promissione Spiritus sancti accepta · Effundam · Effudit hunc · Sede a dextris · Dominum eum · Et Christum · Jesum quem · Compuncti · In nomine · Et omnibus · Testificatus est · Salvamini · A generatione. Also inline Jo 20:22–23 Accipite Spiritum sanctum; quorum remiseritis peccata, remittuntur eis · Ps 108:6 Et diabolus stet a dextris ejus · Mt 26:38 Tristis est anima mea usque ad mortem · Act 2:21 Omnis quicunque invocaverit nomen Domini salvus erit — all four verbatim.

Act 3 (chunk 0003): Portam templi · Argentum · Et exsiliens stetit · Stupore et exstasi · Cucurrit · Deus Abraham · Interfecistis · Integram sanitatem · Per ignorantiam · Per os omnium · Ut deleantur peccata · Ut cum venerint · ut cum venerint tempora refrigerii a conspectu Domini (verbatim, in full) · Quem oportet · Moyses quidem · Quoniam prophetam · De fratribus vestris · Exterminabitur de plebe · Testamenti · Patres · patres · patrem · In semine tuo · Omnes familiae · Benedicentur. Also inline Jo 1:9 Illuminat omnem hominem venientem in hunc mundum — verbatim.

Act 4 (chunk 0004): Sacerdotes et magistratus · Sadducaei · Dolentes · Resurrectionem · Et factus est · Hic est lapis · Reprobatus · Sine litteris · Idiotae · Petrus vero · Non enim possumus · Clarificabant · Annorum enim · Unanimiter · Fremuerunt gentes · Meditati sunt inania · Convenerunt · Unxisti · Cum omni fiducia · Per nomen · Motus est locus · Erat cor unum · Nec quisquam eorum · Communia · Reddebant apostoli · Joseph · filius consolationis · Quod est interpretatum · Filius consolationis. Also inline 1Cor 1:17 Non in sapientia verbi, ut non evacuetur crux Christi · Ps 44:8 Unxit te Deus Deus tuus oleo laetitiae prae consortibus tuis · Mt 5:16 Videant opera vestra bona, et glorificent Patrem vestrum, qui in coelis est — all three verbatim (the third clipped of its opening ut).

1.5 · UNMARKED SCRIPTURE THE GLOSS QUOTES WITHOUT ITALICS — checked, all loose, none marked

Migne sets these in roman, so they are not lemmata and Pattern 14 does not reach them; they are recorded because a later reader will meet them and wonder.

• 0431A In aliis linguis, et labiis aliis, loquar populo huic, attributed secundum illud Prophetae. It is 1Cor 14:21's form of Is 28:11, not Isaiah's own words (In loquela enim labii, et lingua altera loquetur ad populum istum). Rendered as printed.
• 0434C Qui occiderit hominem volens, et ipse occidetur, attributed to the law. A loose memory of Lv 24:17 (Qui percusserit, et occiderit hominem, morte moriatur).
• 0436C Si duo vel tres convenerint in unum in nomine meo, de omni re quam petierint, fiet illis — a conflation of Mt 18:19 and 18:20. Rendered as printed.
• 0428D duodecim fontes deserti, quos Israel reperit in Elim agrees with Ex 15:27.

§2 · MIGNE'S OWN PRINTING, REPRODUCED AND NOT TIDIED

2.1 · Verse numbers that do not match the verse glossed — HIS, reproduced as printed

• 0427D · VERS. 14.-- Maria, etc. The chapter's Maria is Act 1:14 and agrees. But the same paragraph's second lemma In medio fratrum is Act 1:15, and so is the centum et viginti the gloss goes on to expound at 0429A. Mid-gloss second lemmata are not addressed by their own VERS., so this is not strictly a mis-numbering; recorded so the merge does not read it as one.
• 0429D · VERS. 23. covers Duos, Qui cognominatus, Justus and Et Matthiam, all of Act 1:23 — correct.
• 0433C · VERS. 34.-- Sede a dextris. Act 2:34 is Dixit Dominus Domino meo: Sede a dextris meis; correct, clipped.
• No out-of-sequence verse address anywhere in cols 0425C–0437C. VERS. 1, 2, 3, 4, 5, 7, 8, 9, 11, 12, 13, 14, 18, 19, 20, 23 (ch. I) · 1, 2, 3, 4, 5, 6, 8, 9, 10, 14, 17, 19, 23, 24, 25, 26, 27, 28, 33, 34, 36, 37, 38, 39, 40 (ch. II) · 2, 6, 8, 10, 13, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25 (ch. III) · 1, 2, 4, 11, 13, 19, 20, 21, 22, 24, 25, 26, 27, 29, 30, 31, 32, 33, 36 (ch. IV). Every one ascending. A checked negative.

2.2 · 0427D · (Matth. XXVIII) on a quotation of Mark 16:19 — HIS, and it is inside [n:]

The words Et Dominus quidem Jesus postquam locutus est eis, assumptus est in coelum are Mc 16:19 verbatim, and Migne's own sentence introduces them Aliter Marcus commemorat dicens. The note nevertheless reads [n: (Matth. XXVIII)]. Notes pass through verbatim, so it stands as printed. The right home for the repair is data/citation-corrections.json (refDisplay verbatim, refKey → Mark.16.19); no Ibid. depends on it. Not raised as a plate candidate — 8965's (Rabac. III) precedent shows an abbreviated book-name in a note is exactly the class where the plate has gone Migne's way.

2.3 · 0432C · (Zach. V) for Zach 9:11 — same class, same treatment

The words are Zach 9:11 (with the divergence marked at §1.1). The chapter numeral is Migne's. Left as printed; referred to citation-corrections.json.

2.4 · 0435B · A LOWERCASE SENTENCE-OPENING, MIRRORED — the sharpest plate candidate here

…de spirituali exterminantur. utramque inobedientiam mors sequitur sive carnalis sive spiritualis. A full stop followed by a lowercase word. The English mirrors it: "…are exterminated from the spiritual. death follows either disobedience…".

⚠ Mirroring is deliberate and I want it re-examined at the merge. Capitalizing it would be a silent repair of the plate's typography (Pattern 8's channel), and there is no marker that reaches it: nothing is carried type, so [sic:] cannot wrap it, and it is not our source's loss, so [ed:] cannot either. But it will read to a reader as our build defect, which is the one thing the marker conventions exist to prevent. This is the cheapest plate read in the stint: one crop of PL 114 col. 0435 band B settles whether Migne prints Utramque (then it is our digitization, and a data/tei-patches/8976.json fix) or utramque (then the mirror is right and the cruces line is the whole apparatus). ⛔ It is the only . [a-z] in all eighteen chunks of this work except one at 0015, so it is not a corpus habit.

2.5 · 0435B · a bare hyphen in the Latin — reproduced

Semen Abrahae Christus est - in fide cujus omnibus familiis… Migne prints no em-dashes anywhere in the Latin corpus, and this is not one; it is a spaced hyphen. Reproduced in the English exactly (The seed of Abraham is Christ - in whose faith…), on Pattern 8's never-delete rule. Also worth one look at the same crop as §2.4, since the two sit two lines apart in the same column band.

2.6 · 0437B · a doubled period after a marker — reproduced

Filius consolationis. [n: (RAB.)]. — the period inside the italic run and a second one after the note. Carried verbatim.

§3 · PLATE CANDIDATES — RAISED, NEVER PATCHED, NO MARKER FIRED

Per the 2026-08-23 addendum: a non-word is worth raising and has been our digitization's error ten times out of ten; a real, well-formed word that reads oddly is MIGNE'S until a plate says otherwise, refuted five times out of five. Nothing in this section carries a marker, so the five chunks assert nothing about the plate.

3.1 · Two NON-WORDS, both rendered for their evident sense

⚠ Both are single-letter-class faults in well-formed neighbourhoods, and neither is patchable from the file: a tei-patch needs the plate. The English is readable without one.

3.2 · Real words that read oddly — DECLINED as Migne's, with the argument that lost

Recorded because the addendum's finding is that the argument loses, not that raising it was wrong.

• 0428B · conversationem activalem. activalis is not classical, but it is a well-formed medieval adjective and the sentence needs exactly it (de activa in contemplativam ascendens, iterum descendit in conversationem activalem). Rendered "the active manner of life". No marker, no conjecture.
• 0429D · Merito parvi transcenditur ille laude hominum qui justus erat, inquit Arator. Every word is real; the sentence construes (parvi genitive of value) and yet it asserts something odd — that the man who was just is deservedly passed over. Rendered literally: "Deservedly is he passed over at small worth in the praise of men, who was just, says Arator." ⛔ I considered a [cj:] on Merito and declined it: Pattern 18's bar is a false assertion produced by a word, and the word here is not the trouble — Arator is being quoted, in verse, out of its own context, and the abridgment has taken whatever qualified it. Naming the rival: the reading that loses is immerito, which nothing on the page supports and which would be our invention rather than a conjecture on his type.
• 0437B · Auctus Barnabas filius consolationis interpretatur, actus incremento et profectu spiritualis boni. actus (from ago, "driven") is a real participle and takes a plain ablative of means; rendered "being driven by the increase and the advance of spiritual good". The rival, auctus repeated from the sentence's own first word, loses because it would make the clause say twice what it already said and would leave Auctus at the head unexplained. Declined.
• 0437A · Nam multitudo credentium … copula charitatis invicem jungebantur.** Plural verb on a singular collective subject. English carries a collective plural idiomatically ("the multitude … were joined"), so there is no exponent for the disagreement and Pattern 9 governs. No marker, no crux beyond this line.

3.3 · ⭐ THE SUSPICIOUSLY TIDY FORMS — a checked list, and it came back empty

Brief §4 and addendum §C: the digitization silently corrects as well as corrupts, and the only handle is a form that reads too tidy in a rough neighbourhood. I read the whole lines around both §3.1 sites and around every [n:] in the range, and found nothing of that shape. Recorded as a negative, because the class is only credible when the zeros are reported too. The one place I would still ask a reader with the plate to glance is 0430D's two four-verb catalogues (Spiritus purgat, illuminat, accendit, urit / urit, purgat, calefacit, illuminat), where the second reorders the first — but that reordering is the gloss's own rhetorical point (fire's four natures matched to the Spirit's four works) and reads as deliberate, not as damage.

§4 · usque ad CENSUSED PER OCCURRENCE — 11 tokens, 8 formula, 3 not

Decided from what stands on either side, never by rule (addendum §D).

The abridgment formula (8), rendered as far as: 0427A (Beda, usque ad quod in baptismo Joannis non fiebat) · 0430C (Greg.) · 0431B (Beda) · 0433C (Rab.) · 0434A (Beda, Porta Speciosa) · 0434C (Beda, Cucurrit) · 0434C (Rab., Deus Abraham) · 0435B (Beda, Testamenti — printed with a comma, usque ad, in semine tuo…).

NOT the formula (3), and each is a trap the formula reading would have swallowed:

1. 0425D · usque in diem — Act 1:2's own words, italicized as a lemma. Rendered "until the day".
2. 0431C · Lex et prophetae usque ad Joannem — Lc 16:16's own words, in roman, an ordinary prepositional phrase. Rendered "The law and the prophets until John".
3. 0432D · Tristis est anima mea usque ad mortem — Mt 26:38 verbatim, inside an italic scripture run with its own [n: (Matth. XXVI)]. Rendered "even unto death". ⚑ This is the exact shape of the trap 8969 hit at Sap 8:1 and 8954 hit at A labio usque ad labium: an abridgment formula sitting inside a verse whose own words are usque ad.

⛔ No verb was ever supplied to complete a formula's host sentence. 0433C is the case: Migne prints Manifeste Petrus ex hoc psalmo regnum Christi non terrenum, sed coeleste, etc., usque ad… with no finite verb at all, because the abridgment cut it. The English stops where he stops: "Manifestly Peter, from this psalm, the kingdom of Christ not earthly, but heavenly, etc." It reads abrupt because the plate is abrupt (the 8970/L4 rule).

§5 · THE UNEXPRESSED SUBJECT — every supplied pronoun, named

The measured dominant defect class. Each site below is a "he/she/it/they" the Latin does not print. Where two readings both construe I say so rather than choosing silently.

• 0425D · Bonum doctorem instituit vel instruit, qui faciat quod docet. → "He appoints or instructs a good teacher, one who does what he teaches." ⚠ The subject is genuinely open. The paragraph is glossing facere in quae coepit Jesus facere et docere, so the two readings are (a) Christ, who by doing before teaching forms the good teacher, and (b) Luke/the evangelist, who by writing facere first instructs him. The gloss's own preceding clause is about Luke (Lucas sermones texuit), the lemma is about Jesus. Both construe; I supplied a bare "he", which carries both, and did not resolve it.
• 0426C · propter Spiritum praecepit → "he gave the commandment on account of the Spirit" — Christ, from Praecipiens apostolis in the lemma above.
• 0426C · Elegit. Hoc ad laudem apostolorum, quorum actus scribere proponit → "whose acts he purposes to write" — Luke, not Christ: the acts written are the book's, and Act 1:1 has just named Luke the writer.
• 0426C · eo quod post mortem vivebat → "in that after death he was living" — Christ, from Vivum (Act 1:3 praebuit seipsum vivum).
• 0427D · completis illis sermonibus quos commemoraverat Dominum ascendisse coelos ostendit → "he shows that … the Lord ascended" — Luke, named as the subject of adjecit in the same sentence.
• 0428C · Hic post Jacobum fratrem suum, rexit Ecclesiam Hierosolymorum → "He, after James his brother, ruled the Church of Jerusalem" — Simon Zelotes, the lemma.
• 0429A · Filio pereunte perditionis → "The son of perdition perishing" — no pronoun supplied; the genitive is displaced in Migne's word order and rendered where it stands.
• 0429C · qui traditionis crimen in se protinus mulctavit → "who forthwith punished in himself the crime of the betrayal" — Judas, named two sentences above.
• 0429D · In laude Barnabae immoratur → "He lingers in the praise of Barnabas" — Luke again (the narrator of Act 1:23), not Peter: Peter's speech ends at v. 22.
• 0432D · Non iterum movear → "Let me not again be moved" — the psalmist's voice, which Migne has just named (Vox Christi post resurrectionem exsultantis). First person, not third; the whole 0432C–D sequence is Christ speaking, and the English keeps it so.
• 0433B · Qui prius Jesum suscitatum a Deo dixit → "He who before said" — Peter, from the sermon the chapter is glossing.
• 0435A · ut sicut ille … introduxit, ita et Jesus … duxit → "as he brought the carnal Israel … so also Jesus" — ille is Moses, fixed by in similitudine Moysi six words earlier.
• 0436C · quod magis apostolorum votis congruit → "which agrees better" — the Greek reading, not a person.
• 0437C · quia dignus habitus est qui pro Juda consolaretur apostolos → "because he was held worthy to console the apostles in Judas' stead" — Barnabas, the lemma.

§6 · SENTENCES THAT DO NOT CONSTRUE CLEANLY — rendered as printed, resolved nowhere

• 0427A · Quando Petrus negavit, non erant apostoli baptizati; non aqua, sed spiritu. Rendered "When Peter denied, the apostles were not baptized; not with water, but with the spirit." ⚠ As printed it is self-contradictory: the first clause denies they were baptized, the second says with what they were. Both negatives are carried on their printed hosts (erant baptizati and aqua), per 7a, and the contradiction stands on the page. The neighbouring sentences (post resurrectionem enim dictum est: Vos vero baptizabimini) suggest the sense is "not yet with water, but [afterwards] with the Spirit", but that is a reading, not a repair, and it is not in the text.
• 0427D · et in his qui se in alienos fecerint a terra praedicaturos eos praedixit, per quos quasi rotae radios de medio terrae ubi salutem operatus est… Rendered "and he foretold that they would preach among those who shall have made themselves strangers from the earth, through whom, as through the spokes of a wheel, from the middle of the earth where he wrought salvation, spiritual grace and saving doctrine flowed out through the whole world." The quasi rotae radios is an accusative with no verb governing it in the printed sentence; rendered as an appositive comparison, which is as close as the printed words allow. Logged, not repaired.
• 0428C · Primus (si fas est dici) plus caeteris dilexit: alium, plurimum Jesus amavit. Rendered "The first (if it is right to say it) loved more than the rest: the other, Jesus loved most." alium is accusative and takes amavit; the asyndeton is Migne's. The parenthesis count is 1:1 with the Latin.
• 0434D · cum virtute resurrectionis mors destruetur novissima → "when by the power of the resurrection death shall be destroyed last." novissima agrees with mors and echoes 1Cor 15:26 (novissima … inimica destruetur mors). The rival reading, "the last death shall be destroyed", loses because the paragraph is about when sins are fully blotted out, i.e. about the order of destruction, not about a second death.

§7 · THE TESTS, RUN ON THESE FIVE FILES

Test 1 — a [sic:] must not swallow the word it marks. Vacuously clean: zero [sic:] in the range. The English was nevertheless read once with all markers stripped; no stranded article or preposition, no doubled mark, every sentence readable aloud.

Test 2 + 2a — each printed negative named to its single Latin host. 91 negative particles in the four prose chunks (0001 ×34 · 0002 ×31 · 0003 ×10 · 0004 ×16, counting non · nec · neque · nisi · nullus · ne · nondum), extracted mechanically with context and then each one located in the English and checked to be ON its host — not merely present. The 2a trigger population (a negative resting on a verb with a quantifier, correlative or predicate nearby) fired eight times, and each host was proved from the sentence's own structure, never from what reads well:

⚑ 0436A is the one that could have gone the 8990 way: omnino sits adjacent to the verb nescirent and would have produced the plausible-but-wrong "because they were not wholly ignorant of letters". The sed quia correlative is what settles it, and it is structure, not sense.

Test 2b — every anchor against the WORD it stands beside. All 46 anchors in the range (15 + 16 + 7 + 8) were printed with 45 characters of Latin context and the matching English context and compared one by one. Zero displaced. Six are deliberate mid-construction splits, matching Migne's: 0426C (Spiritus | sanctus → "the Holy | Spirit"), 0429C (non | valuerunt → "were not | able"), 0429A (numerus | apostolorum → "the number | of the apostles"), 0427D (non regnum | Israel → "the kingdom | of Israel"), 0430B (soli Petro | prius → "to Peter alone | beforehand"), 0433D (Spiritu | sancto → "the Holy | Spirit").

Test 3 — punctuation compared BY COLUMN BAND, never by total. A band-partitioning script compared colons, semicolons, question marks, guillemets and parentheses across all 46 anchor-delimited bands, plus asterisk counts, [n:] counts and paragraph counts. First pass returned 2 mismatched bands, both mine, both supplied semicolons — exactly the class the brief names:

• 0427D — I had written "…the power of bearing witness of me; and the time…" where Migne prints a comma (praestabit, tantumque). Repaired to a comma.
• 0430C — I had written "in pride, dispersion; in humility, concord is made" where Migne prints in superbia dispersio, in humilitate fit concordia with one comma. Repaired.

Two further supplied semicolons were caught during drafting, before the script ran, and both were the same reflex: 0432B (eligerent, sed latroni concessa vita) and 0433A (rediit, nec caro est corrupta). Final state: 46 bands, zero mismatches, on every mark counted. ⛔ No comma→semicolon sweep was performed: Migne prints 26 semicolons of his own in this range (0001 ×15 · 0002 ×9 · 0003 ×2 · 0004 ×0), so Tobit's narrow exception — which rests on a book printing zero — does not apply here and was not invoked.

Test 4 (the fourth test) — does the paragraph ASSERT what he asserts? Run last, with the counts already clean, on every paragraph. It is what produced §1.1's six divergences (each found by reading the lemma against the Latin, not against the pre-scan), §3.2's four declines, and the three name decisions in §V that would otherwise have gone the familiar English Bible's way — Barnabas not Barsabas, Matthew not Matthias, Jesus not Josue. Those three are the exact shape the launch brief warned of (Absalon, Scarioth, Hebrun), and all three arrived over the plate in my own first reading of the paragraph before the collation caught them.

⚑ Two smooth-where-the-Latin-is-rough suspects were re-examined under this test and both were left rough: 0427D's quasi rotae radios (§6) and 0433C's verbless abridgment head (§4).

§8 · WHAT I DECLINED, AND WHY

1. A [cj:] at 0429D on Merito (Arator). Declined — the bar is a false assertion produced by a word, and the trouble here is the abridgment, not the type. Rival named and rejected in §3.2.
2. A [var:] on the eleven §1.2 divergences. Declined — each is a particle, a clip, or a dropped possessive with no sense change, and a [var:] is a public claim. The crux is always available and always safe.
3. A [var:] on the four §1.3 Greek-attributed readings. Declined — the page already glosses them, in Migne's own words.
4. A [sic:] on subtsantiae and crediderent. Declined — the non-word class has been our digitization's error ten times out of ten, and a [sic:] would publicly blame Migne's plate for it. Both rendered for their evident sense, both raised in §3.1.
5. Capitalizing utramque at 0435B. Declined, and flagged for the merge (§2.4). This is the one decision in the stint I would most like a second reader on: mirroring preserves evidence, and also makes the page look as though we broke it.
6. Renaming Hierosolyma apart from Jerusalem (§V). Declined — English has no exponent, and manufacturing one ("Hierosolyma") would be our editorial hand, not Migne's.
7. A plate read. Not attempted: the launch brief records PL 114's spot check for this work as a checked zero, with both mid-page divisions (John→Acts at 425/426, Acts→Romans at 469/470) examined. §2.4 and §3.1 are the three candidates I would spend one crop on if the merge wants them; none blocks the range.

§9 · READ CLEAN — the sections where I found nothing

Named because a findings list without them is not a report.

• Chunk 0000 (eight words, one italic span). Vide inter opera ejus. is not scripture at all — it is Migne's own cross-reference standing in for Jerome's prologue, which this recension does not reprint. Rendered "See among his works.", with no object supplied. The pre-scan's ⚠ against it is correct and innocent.
• Chunk 0003, cols 0434A–0434D — 15 verse addresses, 25 lemmata, nine [n:] notes; the only entries are the usque ad census (§4) and the §1.2 Mt 10:9 clip. Otherwise clean.
• The [n:] sigla. All 38 pass through verbatim: (RAB.) ×12, (BEDA.) ×9, (GREG.), (AUG.), (ID.), and 14 scripture locators — (Joan.) ×4, (Matth.) ×4, (Psal.) ×2, (Gal. III), (Jer. XII), (Zach. V), (I Cor. I). None is editorial prose, so no Pattern 15 [nt:] fires anywhere in the range — each was judged by content, not by length, and the longest ((I Cor. I)) is three tokens.
• Guillemets: zero in both languages, all five chunks. Pattern 5 and 6 have nothing to govern here.
• Dittography: none found. The two four-verb catalogues at 0430D repeat urit and illuminat across two lists, but they are two different lists (fire's natures, then the Spirit's works), not a compositor's doubling — see §3.3.
• Migne's own ( sic ): none in this range, so nothing to stack ours on.

Stint: chunks 0005–0009, cols 0437D–0450A (4,863 Latin words), CAPUT V – CAPUT IX. Ananias and Saphira · the choosing of the seven · Stephen's speech and stoning · Simon Magus · Philip and the eunuch · Saul on the Damascus road.

Three peers hold 0000–0004, 0010–0013 and 0014–0017. This file is a stint file. It is not cruces.md; the merge is the launcher's.

Marker tally for this stint: 5 [var:] · 1 [cj:] · 0 [sic:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 [cn:] · 115 [n:] (exact parity with the Latin twins). Zero guillemets in either language, in all five chunks.

Mechanical state, all five chunks: column anchors verbatim and in order and checked against the word each stands beside (test 2b); [n:] contents verbatim and in order; paragraph counts 1:1; verify-english.mjs reports no error on 0005–0009 (it fails only on peers' unwritten chunks). Punctuation compared band by band, never by totals — see §T3.

§0 · WHAT I INHERITED, AND FROM WHERE

⚑ Timing, stated because it changes what this file can claim. When I began, src/english/8976/cruces-0000.md was not on disk — the peer holding 0000–0004 had written 0000.md only, eight words (Vide inter opera ejus.) with no vocabulary in it. His file appeared while I was drafting, and I read it before filing and reconciled against it. §S below records what changed in my five chunks as a result, and the one place where I did not conform. Everything I settled independently and that turned out to agree with him is marked ✓.

Corpus controls run before inventing anything (src/english/): Moses 148 : Moyses 11 → Moses · Saul 187 : Saulus 1 → Saul · Philip 943 : Philippus 168 → Philip · Stephen 5 : Stephanus 0 → Stephen · Sadoch 2 : Zadok 0 → Sadoch · "by antiphrasis" and "anticipation" both attested · "catechumens" attested.

§S · THE SEAM WITH 0000–0004 — three terms changed, one divergence NOT harmonized

S1 · Changed in my chunks to match the opening stint

Three of my independent choices were wrong for this work once his table was on disk, and all three are repaired. A work that renders one term two ways is incoherent on its own page.

⚑ sermo has two sense-distinct occurrences in my range that I did NOT flatten, on the per-occurrence model his own virtus entry uses: 0439B Prius sunt sermone prohibiti, modo flagellis caesi → "forbidden by word", where the whole contrast is words against whips; and 0449A Implet sermonem Domini dicentis → "the word of the Lord who says", a formula introducing a dominical saying. "Discourse" in either would say something Migne does not. Declared, not smuggled.

S2 · ⛔ ONE DIVERGENCE DELIBERATELY NOT HARMONIZED — the non-words

His §3.1 RENDERS its two non-words for their evident sense (subtsantiae → "of the substance", crediderent → "they believed"), with no marker and a crux. My §W CARRIES its four into the English untranslated, in italics, with no marker and a crux.

I did not conform to him, and I did not conform him to me. The merge should rule. Both arguments, so the ruling can be made on them rather than on which file was read second:

• For carrying (mine). Pattern 7's non-word class is explicit and closes with "Reading the typo was considered and rejected: it is the same silent correction this pattern exists to forbid… Pattern 7 has no exceptions." A reconstruction leaves the reader no way to know the file is defective, and it is the one error class nobody downstream can find.
• For rendering (his). Addendum A's finding is that a non-word is our digitization's error ten times out of ten. If the fault is ours, there is no type of Migne's to carry, and carrying it prints our own corruption on the page dressed as the plate's — which Pattern 13's principle ("a defect in our files is never attributed to Migne's plate") argues against as strongly as Pattern 7 argues the other way.

⚑ Neither of us fired a marker, so neither set of chunks asserts anything about the plate, and the disagreement is confined to four words in mine and two in his. Whichever way it goes, the change is a search-and-replace on six sites; it is not a re-translation. ⭐ A plate read of PL 114 pp. 219, 224, 226, 227, 228 and 230 would end it, and would also decide whether the fix is a data/tei-patches/8976.json entry rather than either treatment.

S3 · Checked across the seam and sound, needing nothing

• The [var:] marker form — his §1.1 fires five with book, chapter and verse explicit; mine fires five the same way. Ten markers, one form, and no repeat of 8964's two-form drift.
• Misprinted citations left exactly as printed and referred to citation-corrections.json — his §2.2 and §2.3 ((Matth. XXVIII) for Mark 16:19; (Zach. V) for Zach 9:11), my §P ((Matth. VI) for Matt 5:10; (II Tim. II, 9) for 2 Tim 2:19). Same class, same treatment, and all four identified from the quoted words, never from the numeral.
• The [cj:] bar. He considered one at 0429D and declined it as outside Pattern 18a's scope; I fired exactly one, at 0438A, on a coordinating conjunction producing a false assertion, which is the class the 2026-08-15 extension admits. The two decisions are consistent with each other and with 18a; see §R1.
• His §2.1 records verse numbers that do not match the verse glossed. My range has none — every VERS. address in 0437D–0450A is in sequence, and the two that looked displaced (Act 7:58/59, Act 9:7) check out against the Clementine's own versification. Recorded as a checked negative, since his half of the book has the defect and mine does not.

§V · THE DIVERGENCE LIST — every lemma in 0437D–0450A walked

⛔ The point of this section is that the agreements are listed too. A findings-only list cannot be told apart from no look at all. Every reading below was checked against sources/vulgate/clementine-flat.txt, cited by book chapter:verse, never from recollection.

V1 · The five that took a [var:]

All five [var:] contents were checked not to occur verbatim in their Latin twin (the Pattern 14 mirror test): propter justitiam, efferte, fundamentum Dei, tabernaculum domus meae, venit hora — none is in the chunk it annotates.

V2 · Divergences found, declined a marker, with the reason

Pattern 14 fires on a divergence that changes the sense. These do not, and a marker that fired on a trimmed enim would be spam on every page. All are recorded because a reader is entitled to know they were seen.

• 0438A Utique convenit inter vos tentare Spiritum Domini? — Act 5:9 has convenit vobis. This is the gloss's own re-punctuation of the verse (it is arguing about where the question begins), and inter vos says what vobis says.
• 0439D / 0440B Omni die non cessabant docentes — Act 5:42 Omni autem die non cessabant in templo et circa domos, docentes. A re-quotation closed with etc.; the full lemma at VERS. 42 (Omni autem die, etc.) agrees exactly.
• 0440D Exi de terra et de cognatione tua — Act 7:3 Exi de terra tua, et de cognatione tua. One tua trimmed in a re-quotation; the lemma above it agrees.
• 0441D In sepulcro — Act 7:16 sepulchro. Orthographic.
• 0442A Rex alius — Act 7:18 alius rex. A transposition; no sense turns on it.
• 0442C O homo, quis me constituit… — Lc 12:14 has bare homo. One vocative particle.
• 0443B Qui fecerit ea, vivet in eis — Gal 3:12 has in illis, Lv 18:5 has quae faciens homo, vivet in eis. A conflation of the two, and it is unattributed (no [n:]), so there is no single host verse for a marker to name. Recorded here instead.
• 0443C Nunquid victimas / militiae coeli / Coelum mihi — Act 7:42, 7:49 print numquid, caeli, caelum. Migne's coe- orthography throughout; not divergences.
• 0445A Educ eum foras extra castra, et lapidet eum populus — Lv 24:14 Educ blasphemum extra castra, et ponant omnes … et lapidet eum populus universus. An abridged quotation; and Migne's own prose two words earlier supplies de blasphemante, so nothing the substitution removed is lost to the reader.
• 0445C Cum persequentur vos — Mt 10:23 Cum autem persequentur vos.
• 0445C Benjamin lupus rapax mane comedet praedam, sero dividet spolia — Gn 49:27 mane comedat praedam, et vespere dividet spolia. sero and vespere are the same evening; the gloss turns on lupus rapax, which agrees.
• 0446B Spiritus sanctus disciplinae effugiet fictum — Sap 1:5 opens Spiritus enim sanctus. A trimmed connective.
• 0447B humiliavit se factus obediens usque ad mortem — Phlp 2:8 Humiliavit semetipsum.
• 0447C De quo propheta dixit? — Act 8:34 de quo propheta dicit hoc? The gloss is recalling the eunuch's question, not quoting it as a proof-text.
• 0448A Saulus adhuc devastabat Ecclesiam — Act 8:3 Saulus autem devastabat ecclesiam. Migne's back-reference has taken adhuc from Act 9:1 (Saulus autem adhuc spirans), which is the verse he is glossing. Recorded because it is an intelligible contamination, not a defect.
• 0448A Quandiu uni ex minimis istis fecistis, mihi fecistis — Mt 25:40 quamdiu fecistis uni ex his fratribus meis minimis, mihi fecistis. Reordered, and fratribus meis dropped; the argument (persecuting the least is persecuting Christ) is untouched.
• 0448B Apertis oculis — Act 9:8 apertisque oculis. A fragment cut at the enclitic.
• 0449A Qui vult venire post me, abneget semetipsum — Lc 9:23 Si quis vult post me venire, abneget semetipsum.
• 0444B Fundamentum aliud nemo potest ponere praeter illud quod positum est — 1 Cor 3:11 praeter id quod positum est, and enim trimmed. ⚑ Worth naming because I had qui est Christus Iesus in my head from the Greek and the Clementine reads quod est, exactly as Migne prints it. The check overturned my recollection, which is the whole reason the file is on disk.

V3 · Lemmata checked and found to AGREE with the Clementine exactly

Named, because the brief requires it. Act 5:3 Tentavit · 5:3 Et fraudare · 5:4 Sed Deo · 5:5 Cecidit, Et factus est timor · 5:8 Dic mihi · 5:9 Quid utique · 5:11 Et factus est · 5:13 Caeterorum (Clem. Ceterorum, orthographic) · 5:14 Augebatur · 5:17 Haeresis, Sadducaeorum repleti · 5:20 Vitae hujus · 5:23 Carcerem quidem · 5:24 Magistratus, Ambigebant · 5:26 Tunc abiit magistratus cum ministris · 5:28 Praecipiendo, Replestis Jerusalem, Super nos sanguinem · Mt 27:25 Sanguis ejus super nos · 5:30 Deus patrum · 5:32 Spiritus sanctus, Omnibus · 5:37 Judas Galilaeus · 5:40 Caesis · 5:41 Quoniam digni · 5:42 Omni autem die · 6:1 Graecorum, Despicerentur · 6:2 Non est aequum nos · 6:3 Considerate · 6:4 Nos vero · 6:5 Et elegerunt, Stephanum · 6:6 Et orantes, Orantes imposuerunt eis manus · 6:7 Et verbum, Et verbum Domini crescebat · 6:8 Stephanus · 6:13 Homo iste · 7:2 Viri fratres et patres · Lc 6:27 Diligite inimicos vestros · 7:2 In Mesopotamia · 7:3 Et dixit, Exi de terra tua · Gn 12:1 et de domo patris tui · 7:4 Tunc exiit, Pater ejus · 7:4 Vos habitatis · 7:6 Quia erit semen, Annis quadringentis, Accola · 7:7 Judicabo · Ps 42:1 Judica me, Deus, et discerne causam meam · Ps 5:11 Judica illos, Deus · 7:8 Testamentum circumcisionis, Die octavo · 7:9 Vendiderunt in Aegyptum · 7:14 In animabus · 7:15 Et descendit · 7:16 Translati · 7:19 Genus nostrum · 7:20 Gratus Deo · 7:22 Moyses · 7:24 Vindicavit illum · 7:26 Apparuit illis · 7:27 Quis te constituit · 7:28 Nunquid · 7:29 Fugit · 7:30 In igne flammae · Ex 3:14 Ego sum qui sum, Qui est, misit me ad vos · 7:33 Solve calceamentum (Clem. also calceamentum — checked, they agree) · 7:34 Videns, Vidi afflictionem · 7:35 Quem negaverunt, Principem · 7:37 Hic est Moyses · 7:38 Verba vitae · 7:40 Ad Aaron, Fac nobis deos · 7:43 Figuras, Et transferam · 7:44 Tabernaculum, Secundum formam · 7:46 Ut inveniret tabernaculum · 7:48 Sed non Excelsus · Mt 18:16 in ore duorum vel trium testium stet omne verbum · 7:49 Terra autem · 7:51 Dura cervice, Et incircumcisis · 7:52 Patres vestri, patres nostri, Et occiderunt · 7:55 Stantem, Video coelos, Filium hominis · 7:57 Et ejicientes · 7:58 Domine Jesu, suscipe spiritum meum · Lc 23:46 Pater, in manus tuas commendo spiritum meum · 7:59 Obdormivit, Saulus autem · 8:1 In illa die, Ecclesia, Dispersi sunt · 8:2 Curaverunt autem Stephanum viri timorati · 8:3 Saulus, Devastabat, Trahens, Dispersi · 8:5 Philippus · 8:6 Intendebant · 8:11 Seducens, Dementasset · 8:12 Philippo, Jesu Christi · 8:13 Tunc Simon · 8:14 Miserunt · 8:16 In nomine, Jesu · 8:17 Imponebant manus · 8:19 Potestatem · 8:20 Pecunia tua, sit tecum, Pecunia · 8:22 Poenitentiam age, Roga Deum · Ps 72:1 Quam bonus Israel Deus · 8:24 Precamini, Quae dixistis · 8:26 Angelus autem, Ad viam · 8:27 Aethiops, Potens, Candacis reginae · Ps 67:32 Aethiopia praeveniet manus ejus Deo · Ps 44:11 Audi, filia, et vide et inclina aurem tuam · 8:28 Revertebatur · 8:30 Putasne intelligis · 8:31 Quomodo · 8:33 In humilitate, Humilitate, Tolletur (⚑ checked with care — the Clementine also reads tolletur, not tollitur) · 8:35 Aperuit, Evangelizavit · Ps 80:11 Dilata os tuum, et implebo illud · 8:36 Ad quamdam aquam · 8:37 Si credis · 8:39 Spiritus Domini · 8:40 Inventus est, Caesaream · 9:1 Saulus autem, Minarum, Spirans · 9:4 Et cadens, Quid me · 9:5 Ego sum Jesus · 9:7 Surge, Ingredere, Et dicetur · 1 Cor 9:9 Non alligabis os bovi trituranti · 9:8 Apertis oculis · 9:9 Et erat, Tribus diebus, Non videns, Non manducavit · 9:10 Anania · 9:11 In vicum · 9:15 Vade · 9:17 Introivit · Is 11:6 Habitabit lupus cum agno · 9:18 Ceciderunt · 9:20 Et continuo · 9:21 Nonne · 9:24 Custodiebant · 9:26 Cum autem venisset · 9:27 Barnabas · 9:29 Cum Graecis · 9:30 Caesaream · 9:31 Ecclesia, Et consolatione · 9:33 Nomine Aeneam · 9:34 Surge · 9:35 Lyddae et Saronae · 9:36 Nomine Tabitha · 9:39 In coenaculo · 9:40 Ejectis, Aperuit, Aperuit oculos · 9:41 Dans autem.

V4 · ⚑ THE NUMBERS, because the brief said a number the gloss reasons from is the prize

• 0441D, Act 7:14 — the seventy-five. Migne's lemma is the bare In animabus; his gloss is Hoc secundum editionem Septuaginta interpretum dicitur. In Hebraeo septuaginta tantum reperiuntur. Both numbers check. Act 7:14 Clem. reads in animabus septuaginta quinque; Gn 46:27 Clem. reads fuere septuaginta. So the famous divergence is real, the gloss states it correctly, and there is nothing to mark. A checked zero on the one number this range was most likely to carry a defect in.
• 0441C, Annis quadringentis — Act 7:6 Clem. annis quadringentis. Agrees.
• 0440A, the seven (non sine septenarii numeri mysterio) and 0448B–C, the three days (triduum, tres dies, tertia die, tres gradus) — all internally consistent and all agreeing with Act 6:3, 9:9 and Act 7:52's three degrees. No numeral in this stint diverges.

§P · MIGNE'S CITATIONS — two misprinted, both left EXACTLY as printed

Identified from the quoted words, never from the numeral, per the standing rule. Neither is touched in the text; both are referred to citation-corrections.json here.

1. 0439A — [n: (Matth. VI)] on Beati qui persecutionem patiuntur propter Deum. The words are the eighth beatitude, Matt. 5:10. Matthew VI is the Lord's Prayer and the lilies; it contains no beatitude at all.
2. 0444A — [n: (II Tim. II, 9)] on Firmum fundamentum Domini stat. The words are 2 Tim. 2:19. 2 Tim. 2:9 is usque ad vincula quasi male operans.

⛔ Neither is a candidate for a plate read that changes anything, because both are left as printed either way; they are recorded so the index resolves refKey correctly while refDisplay keeps what Migne set.

Two references in Migne's PROSE, not in an [n:], that name the wrong book. These take no apparatus at all — they are the author's sentences, rendered as written:

• 0449A, Act 9:24 — Non Judaei, sed cives cum rege suo Aretha, ut ipse scribit ad Galatas. Paul writes the Aretas episode in 2 Cor. 11:32; Galatians 1:17 mentions Damascus and Arabia but not Aretas. Rendered "as he himself writes to the Galatians."
• 0448B, Act 9:7 — In lege scriptum est: Non alligabis os bovi trituranti [n: (I Cor. IX)]. The citation is right and the description is loose: 1 Cor 9:9 does print those exact words, and it is itself quoting the law (Deut 25:4, which reads Non ligabis os bovis terentis). Recorded as a checked non-defect so nobody re-raises it.

§W · THE FOUR NON-WORDS — raised as plate candidates, NO MARKER FIRED

⛔ This is the decision in this stint most likely to be revisited, so here is the whole of the reasoning rather than the conclusion.

Four runs in my range are not Latin words at all:

All four are carried into the English untranslated, in italics, with no [sic: …]. Two rules pull opposite ways here and both were read:

• Pattern 12 says type carried under Pattern 7 or 10 is wrapped in [sic: …].
• The ship test (brief addendum B) says a [sic:] is a public claim that Migne's plate is defective, and that an unread candidate is safe iff it fires no marker — seven false [sic:] were withdrawn in one batch, each publicly blaming Migne for the digitization's error.

Addendum A decides which way the evidence points, and it points away from the marker: every suspicion in that batch aimed at a non-word was confirmed at the plate as our corruption, ten for ten — and quisbusdam is on that very list, found and confirmed in a sibling book. A [sic:] here would therefore very probably be false, and false in the direction that defames the source. So: carry the run, log the crux, fire nothing.

⚑ What a plate read would change, stated so the reader of this file can act on it. If the plate prints the sound word in any of the four, the fix is a data/tei-patches/8976.json entry — not a [sic:], and the English then renders the word normally. If the plate prints the broken run, the [sic:] becomes correct and should be added. Either way the change is cheap, and either way nothing in the running text asserts anything false today. PL 114 = archive.org patrologiaecurs06saingoog, PDF page = (column + 11) / 2 — cols 0441, 0444, 0446 and 0449 are pp. 226, 227, 228 and 230.

⚑ 7a⁗ was run on each of the four, because a carried run takes whatever it governs out of the English with it:

• posili — the run is the participle of a lemma address, not of an assertion the gloss rests on; the clause keeps Et … sunt, and the very next lemma (In sepulcro) supplies the place. Nothing is negated, nothing is governed.
• ite rum — the clause keeps its negative (non) and its predicate (casurus): "because he is not ite rum to fall." The quarantine takes only the adverb.
• credeban — ⚠ this one takes the verb, and it is the only one that does. The clause reads "they credeban in the Father and in the Holy Spirit." It was left this way rather than glossed because the run is 90% legible as credebant to any reader, because inserting an English verb beside it would be the silent repair Pattern 7 exists to forbid, and because brackets are already overloaded (Pattern 11's own note). This is the site to revisit first if the treatment is reconsidered.
• quisbusdam — an adjective in an adjective's slot; nothing governed.

⭐ Recorded as a checked negative as well: I looked for the other direction of brief §4 — a form that reads suspiciously TIDY in a rough passage — and found none in 4,863 words. The nearest candidate was credulitate at 0446A (credulitate aliarum duarum carere non potest), and it is the refuted class, by name: 8958's credulitatem was raised on an excellent argument and the plate said Migne printed it. Rendered "belief", no crux beyond this line.

§R · REAL WORDS THAT READ ODDLY — rendered as printed, one glossed

Per the standing rule, a real, well-formed word that reads oddly is Migne's until a plate says otherwise (refuted five of five). All of these are rendered as printed.

R1 · The one [cj:] — 0438A, Ananias a Saphira … mortui sunt

Prius Ananias a Saphira in conspectu Ecclesiae cito mortui sunt

a is a perfectly good Latin preposition, so Pattern 12 has nothing to wrap. But the verb is plural and there is only one subject: the sentence as printed says Ananias died by Saphira, which the whole book denies — she outlived him by three hours and the gloss two paragraphs above is explaining exactly that gap. That is Pattern 18's bar: the printed word makes the sentence assert what the author is arguing against, and the failed word is a coordinating conjunction, which is the correlative family the 2026-08-15 extension admits.

Marked First Ananias by [cj: a; read ac, "and"] Saphira died quickly… — additive, with Migne's word rendered and standing first, so the verbatim-in-the-Latin-twin guard holds.

⚑ The rival, and why it lost. The other reading is that this is not Migne's at all but our digitization dropping the c of ac — a tei-patch, not a conjecture. It lost only because I cannot read the plate, and [cj:] is the marker that survives either outcome: it claims nothing about the plate, only that ac is the sense. If the plate is read and shows ac, the right fix is the patch and the [cj:] comes out. Named here because the apparatus records only what the marker says.

R2 · Rendered literally, marker declined, reasons given

• 0448B, dehinc Dominus jubetur ingredi. A nominative Dominus against a passive jubetur whose subject can only be Saul; the sense wants a Domino. Rendered as printed — "thereafter the Lord is bidden to enter" — and the [cj:] is declined because 18a scopes that ruling narrowly and explicitly leaves case, person, voice and mood anomalies that merely fail to construe outside it (11064 @1154C is the named parallel). The conjecture is a Domino, "by the Lord," and it lives here rather than on the page by design.
• 0438C, sed nec angelum, neque spirituum ullum esse credebant. A genitive plural with a masculine accusative. Rendered "neither an angel nor any of spirits." Migne's.
• 0442C, Moyses fuit in Madian. The narrative wants fugit, and the lemma four words earlier is Fugit. Rendered "Moses was in Madian." No marker: fuit construes, and the clause asserts nothing false — Moses was in Madian.
• 0442B, a filiis Emor filiis Sichem emit Abraham. filiis where the same sentence prints filii Sichem twice more. Rendered as printed. The whole paragraph is about that genitive chain, so the variation is visible to a reader without help from us.
• 0440A, the anacoluthon at Act 6:6 — ut ministri eligerentur, qui ut digni inventi sunt … placuit eosdem ipsos … ordinari. The relative has no verb of its own and placuit takes over. The English is as broken as the Latin, deliberately.
• 0443B, Qui licet aliquem tentari permittam… — a connecting relative in the first person with no main clause to attach to. Rendered "I who, although I permit someone to be tempted, yet it is not to be thought that I am ignorant of it." Migne's.

§T · THE FOUR TESTS, AND WHAT EACH RETURNED

T2a · The negation sites — host proved from STRUCTURE, per the known hole

Every negative in the stint was attached to a named host. These are the sites that fired the 2a trigger condition (a negative on a verb with a quantifier or correlative adjacent), and each was decided from the sentence's own sed / sicut … ita pivot rather than from what reads well:

⚑ The one that took real work: 0439A, qui nec territi minis cessant praedicare. The nec stands immediately before territi, and the reading it invites — "who, not terrified by threats, cease to preach" — makes the apostles stop, in a sentence whose own main clause says the council is praising them. The host is cessant, and the proof is not the sense but Migne's own idiom two columns later: at 0440C he prints Homo iste non cessat loqui verba, the same verb, the same negative, the same infinitive complement. Rendered "who, terrified by threats, cease not to preach." Test 2 alone would have passed the wrong reading, which is exactly 2a's point.

T2b · Anchors checked against the word each stands beside

All 49 anchors in the five chunks were placed against the Latin word they precede, not merely reproduced in order. The ones where English word order forced a decision, and what it was:

• 0437D — Latin semper [0437D] vindictis commendantur; English "always commended by [0437D] punishments," so the anchor sits on vindictis.
• 0441B — split inside an italic lemma, exactly as 8995 @0605B did: et [0441B] de domo patris tui → "and [0441B] from the house of thy father:".
• 0444A — Latin quam verbum [0444A] posuit; the verb fronts in English, so the anchor rides immediately after "word" (verbum), the word it stands beside, rather than travelling with the verb.
• 0444D, 0446D, 0448C — same class, each placed at the Latin word rather than the English verb: malitiae eorum [0444D] ponit, commercium [0446D] fieri posse, quae turbatur, [0448C] excludens.

T3 · Punctuation, band by band — TWO of my own supplied marks caught

⛔ Both were found by the band comparison and neither by reading. Both were mine.

• 0005, the HEAD band — I had written "carried on in three ways: by persuasion," against Migne's Tribus modis tentatio agitur, suasione. A supplied colon. Removed.
• 0005, band [0439A] — "They question; for there is wont to be doubt" against Migne's Interrogant, solet enim. A supplied semicolon. Removed.

After repair: all bands in 0005, 0007, 0008 and 0009 match exactly (colons, semicolons, question marks, guillemet-opens, exclamations), and 0006 matches in every band but one, by design — see the next item.

⚑ 0006, band [0443C]: one colon in the English that Migne does not print, and it is Pattern 8b. Quaeritur quomodo non obtulerint Deo … non voluntate, sed metu poenarum? is an indirect question closed with ?, and English cannot end that construction with a mark. Recast direct behind a colon: "It is asked: how did they not offer to God, when it is read that they offered certain things to God, not of will, but from the fear of punishments?" The mark is kept, nothing is deleted, and the extra colon is the convention's own cost. This is the only band mismatch in the stint and it is intentional; the merge should not "fix" it.

Question-mark parity across the stint: 7 in the Latin, 7 in the English, and each is on the sentence Migne puts it on. Three of them are worth naming because they look like errors and are not:

• 0438A Increpative dictum? — a ? on what reads as a statement. Kept (Pattern 8).
• 0438A quid? — a one-word question inside a discussion of how to punctuate the verse. Kept.
• 0447D … usque ad in quos Spiritus descenderat? — a ? closing an abridgment tail. Kept exactly there.

T4 · The fourth test — does the paragraph ASSERT what he asserts

Run last, with the counts already clean, on all 176 paragraphs. It returned one repair, and the repair was invisible to every other check in this edition:

⛔ 0446C, Act 8:19. Migne prints Non baptizandi tantum in aqua et praedicandi, sive virtutes faciendi quam habebat Philippus, sed cuicunque imponendi manus. My draft read "or of working the miracles which Philip had" — smooth, and false: quam is feminine singular and cannot agree with virtutes; it reaches back to potestatem, the lemma. The draft made Philip's possession the miracles; Migne's sentence makes it the power, which is the entire point of the paragraph, because what Simon is refused is a greater power than the one Philip had. Nothing was missing, nothing was added, the band matched, the negation was on its proper host. Repaired to "or of working miracles, which Philip had," which leaves the relative as loose in English as it is in Latin rather than binding it wrongly. This is 7a's fifth mechanism — reference mis-bound — and it is the one no negation-aware check will ever see.

Three further paragraphs were read twice under this test and stand as written, with the reasoning recorded because a blind reader will stop at each:

• 0441A, Act 7:3 — post mortem patris sui locutus est Dominus Abrahae qui in Charra mortuus est. The relative sits next to Abrahae and belongs to the father: the next clause is ubi cum eo filius habitavit, "where the son dwelt with him," which only parses if the man who died there is the father, and Gn 11:32 has Thare dying in Haran. The English moves the clause to bind it correctly: "after the death of his father, who died in Charra where the son dwelt with him, the Lord spoke to Abraham."
• 0448C, Act 9:9 — suo exemplo instruitur, qui tenebras tridui luce reversa mutavit. Both readings construe and I did not choose: it is Christ's own example (three days in the tomb) or Saul's own (three days blind, then light). The English — "he is instructed by his own example, who changed the darkness of the three days by the returning light" — is ambiguous in precisely the same way and in the same place. Said here rather than decided silently.
• 0438C — qui virtute et signis reddebant testimonium resurrectionis Christi attaches to apostolorum, not to principibus: it is the apostles who witness to the resurrection, and the Sadducees are being described as denying it. English word order happens to put the relative beside "apostles," so nothing had to be moved.

§M · THE HOUSE CONVENTIONS AS APPLIED, AND ONE DECLARED EXTENSION

M1 · The mirror rule — Ecclesia mirrored, apostolus NOT, and the reason

Ecclesia is capitalized at every one of its 13 occurrences in my five chunks (checked mechanically, plus one lower-case ecclesiasticae at 0449B). All render the Church; the adjective renders "the ecclesiastical faith." Nothing to adjudicate.

⚠ Apostolus is where his case demonstrably carries no distinction, and I did not mirror it. Mechanical count over the five chunks: 19 lower-case against 11 capitals, and they alternate for the same referent inside a single sentence:

• 0439A Hic erat discipulus apostolorum … remanens apud Judaeos consilio Apostolorum — Gamaliel, one sentence, both casings.
• 0446D–0447A sed apostolos rogat … Apostolis etiam ficte loquitur — Simon, two sentences apart, both casings.

That is 8964 §M's In regum / In Regum case exactly: his case is not carrying a signal there, so mirroring it would transcribe an accident as though it were one. Rendered lower-case throughout, except Apostolus used as Paul's title (0444A, de qua Apostolus) → "the Apostle," which is a title and not a plural.

⛔ This is an EXTENSION beyond the two cases 8964 §M names, and I am declaring it rather than smuggling it. If the merge prefers absolute mirroring, the change is mechanical and the two sentences above are the evidence either way.

Mirrored where the case is semantic and the plate is consistent: Scriptura / Scripturas → the Scripture, the Scriptures · Gentibus → the Gentiles · Propheta → the Prophet · Divinitatis → the Divinity. Not mirrored where English has no lower-case option: Spiritui sancto → the Holy Spirit, samaritanae (0445D) → the Samaritan woman. ⚑ Except spiritus itself, which the opening stint mirrors like Ecclesia and I have followed: 0444D’s lower-case spiritui renders the spirit — see §S1.

⭐ One capital that is NOT an accident and does real work — 0443C, sidus Dei vestri. The Clementine prints sidus dei vestri at Act 7:43, of a pagan star-god. Migne capitalizes it, twice, and the gloss then argues from the capital: Quidam sic dicunt sidus Dei vestri, ut ad Deum pertineat verbum, quia non ipsum Dominum, sed ejus sidus susceperunt — "so that the word refers to God, because they received not the Lord himself but his star." Mirrored, in all three places. No [var:]: capitalization is not a textual variant, and the reading itself agrees with the Clementine word for word.

M2 · usque ad — censused per occurrence, and the trap was live

23 occurrences in my range. 22 are the abridgment formula → as far as. One is not.

⛔ 0447B, humiliavit se factus obediens usque ad mortem [n: (Philip. II)] — Phil 2:8's own words, inside a scripture quotation, rendered "unto death." This is warning D's exact shape, and it is the more dangerous because usque ad is set in the same italics as the formula five lines above it. Confirmed mechanically: grep "as far as" returns 22 across the five English chunks, and grep "usque ad" returns 23 across the five Latin ones.

⚑ No verb is ever supplied to complete a formula tail (Tobit's rule), and each tail was parsed as strictly as ordinary prose. Two tails needed that discipline:

• 0441D …etc., usque ad quam pro Hebraica veritate sermonem suum suspectum facere — a comparative quam with nothing to complete it. It is not a stranded correlative and takes no [cj:]: the abridgment removed the maluit it answers, which is the formula working correctly, not a plate defect. Rendered as it stands.
• 0447D …usque ad in quos Spiritus descenderat? — the tail keeps its question mark. See T3.

M3 · The verse addresses, reproduced as printed

• ⚠ 0443A prints VERS 34.--, with no period after VERS — the only one of the 123 verse marks in my range set that way. Reproduced exactly, the same class as 8964's malformed VERS 9.--. Not patched, not tidied.
• The verse numbers are in sequence throughout my five chunks and match the Clementine's versification where I checked it (Act 7:58/7:59, Act 9:7). No displaced number in this range — recorded as a checked negative, since sibling stints found several.
• Act 9's VERS. 7.-- Surge. looked out of place against Act 9:6 and is correct: the Clementine puts Surge, et ingredere civitatem, et ibi dicetur tibi at 9:7.

M4 · Two sentences that simply stop — no terminal mark supplied

Pattern 8, and both are Migne's:

• 0441B — et Stephanus, qui jam quasi in morte positus erat (no full stop).
• 0444A — ostendit quod in sanctum locum non est locutus blasphemiam (no full stop).

The English stops where he stops, in both.

M5 · Lemma expansion — the class, stated so it is not misread as inconsistency

Per the 2026-08-19 addendum, a control has to be the same class as the thing it controls, so here are both classes:

• Clipped lemmata left as legible stumps (nothing is elided that English cannot carry, and the Latin stands opposite): Horarum, → "Of hours," · Coelum mihi. → "Heaven to me." · Terra autem. → "And the earth." · Deus patrum. → "The God of the fathers." (not "our fathers" — nostrorum is not printed) · Solve calceamentum. → "Loose the shoe." (not "thy shoe" — tuorum is not printed).
• Clipped lemmata expanded, because the stump is not English at all: Nunquid, → "Wilt thou, etc." (a bare interrogative particle has no English exponent) · Nonne. → "Is not this he." · Caesis, → "Having beaten them, etc." · Translati. → "They were carried over."

⛔ Contrasting the two lists would manufacture an inconsistency that is not there — the first group has nothing elided that English needs, the second does. Ruth's convention 3 governs both.

M6 · Two glosses where the Latin word had to survive into the English

The Song-of-Songs case 2 (where the gloss IS the etymology, an English word sharing no root makes the gloss explain nothing):

• 0438D, magistratus. The gloss's whole content is that magistratus is to be read as a singular — a fact with no English exponent, since "magistrate/magistrates" is not what is at issue. The Latin word is kept inside the gloss; the lemma above it is rendered plural ("The magistrates.") because the plural is the reading the gloss is correcting, and the gloss's own example sentence is rendered singular ("Then the magistrate went away with the ministers"), as Migne's argument requires.
• 0449C, dorcades. unde ab acumine visus dorcades Graece vocantur — the Greek word is kept, roman as Migne sets it, untranslated. Englishing it would delete the etymology.

⚑ And one where the Latin word did not need to survive: 0442C, Moyses idem valet quod aquaticus … vel liniens. Rendered "of the water" and "daubing", because the gloss's own reasons (quia de aqua est assumptus, quia inventus est in fiscella, bitumine lita) carry into English intact. Nothing is lost.

M7 · One deliberate reading, declared: vicus at 0448D

Act 9:11's vicus is the street called Strait, and every English Bible says "street." Migne's gloss defines it as something else: Vici dicuntur agri, aut civitatibus, aut sibi intrinsecus in civitate proximi — lands lying next to cities or to each other, which is a village, not a street. Rendered "Into the village." and "Villages are called lands lying next…", so that the gloss explains something. Rendering "street" would have been conformation to the familiar English against Migne's own argument — the 11551 stylus failure, in miniature.

§X · VOCABULARY FIXED IN THIS STINT — for the merge to reconcile

Inherited terms are marked ⟵; the rest are this stint's choices, declared rather than assumed.

Proper names, form for form, as Migne sets them — and this range is a genealogy, which is where conformation is least visible: Theodas (not Theudas) · Aretha (not Aretas) · Saphira (not Sapphira) · Sadoch (not Zadok) · Tharsus (not Tarsus) · Charra (Migne's Charram / Charra; the Clementine has Charan) · Seor the Ethite (Migne's Ethaei; Gn 23 Hetheus) · Hebron, Arbe, Sichem, Emor, Putiphar, Ismaelites, Madianites, Madian, Chanaan, Sina, Lydda and Sarona, Candace / Candaces, Tabitha, Dorcas, Damula, Aeneas, Oza, Lia, Isaias, Fadus the procurator, Jesus at 0444B (Act 7:45's Josue, and Migne prints Jesu). Anglicized only where the corpus has already settled it: Moses, Saul, Stephen, Philip, Ananias, Bede. Jerusalem renders both Jerusalem and Hierosolymam (0439B), English having no second form.

§Z · WHAT I DID NOT DO, AND WHY

• No plate was read. PL 114's spot check for this work is a recorded checked zero — no numbered Forte apparatus, both mid-page work divisions examined — so no [cn:] is owed and none is missing. That zero licenses shipping; it does not license the claim that this work has no notes, and I make no such claim.
• No [sic:] — §W and §S2, and the ship test is the reason.
• No [ed:] — nothing in these five chunks is a hole in the digitization. The four non-words are corruptions of text, not losses of text, and Pattern 13 is for the second.
• No [nt:] — all 115 notes in my range are attribution sigla ((RAB.), (ID.), (BEDA.)) or scripture locators. Not one is editorial prose, so Pattern 15 never fires. Judged by content; the longest is nine characters.
• No [f:] — runbook false positive 4. This is lemma-and-gloss commentary, not a florilegium.
• No [d:] — no run is set twice anywhere in the five chunks.
• (cont.) on 0007's CAPUT VII. head is left untranslated — runbook false positive 7; the builder merges it at assembly.
• Column-band arithmetic not reported — runbook false positive 1. The bands in my range skip letters freely (0437D → 0438A → … → 0439D → 0440A) and that is normal.
• I did not write cruces.md, and I touched no file outside 0005–0009.

Third stint of four. Translated 2026-08-25, one translator, 4 chunks, 3,723 Latin words (frontmatter counts: 1,216 + 558 + 825 + 1,124) → 5,430 English words (markers stripped), 1.46× — 0010 1.47 · 0011 1.45 · 0012 1.42 · 0013 1.48. In range with the first stint's 1.46–1.55.

Chapters X–XVI: Cornelius and the sheet, Peter's deliverance, the first missionary journey, the Jerusalem council, the call to Macedonia.

Apparatus fired: 21 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 [cn:] · 107 [n:] (exact parity with the Latin twins: 31 + 12 + 32 + 32). Distribution of the [var:]: 0010 ×8 · 0011 ×2 · 0012 ×5 · 0013 ×6.

Zero guillemets in either language. Zero em-dashes in the English. Question marks: Latin 3 / English 3, in the same bands (0450C, 0455B, 0456A). verify-english.mjs clean over this range — its only complaints are the fourth stint's still-missing 0014–0017.

⚠ This file is cruces-0010.md, not cruces.md. Three peers hold 0000–0009 and 0014–0017; the merge is the launcher's.

§V · VOCABULARY — inherited from cruces-0000.md and cruces-0005.md, and where I extended it

⭐ Both peer cruces were read off disk before a word of this stint was written, and the table in cruces-0000.md §V was taken as binding rather than re-derived. Inherited and held:

Extensions declared here rather than smuggled (three, all forced by words the first two stints did not meet):

• designat → designates, held apart from significat → signifies and signatur → is betokened. All three stand in this range and two of them within eight lines: 0451B Designat autem in hoc coelos gentibus esse apertos and Vas illud designat Ecclesiam against 0451C quam Petrus significat and 0459A Purpura recte sanguinem colore et natura designat against 0455D Significat autem eos qui natura gentiles erant, with 0459C Velocitas obedientiae signatur → "is betokened". The signat/ significat split is 8964 §M's; designat is the third verb and it needed its own word.
• gentes / gentibus / gentium → the Gentiles, and gentilis → gentile (adj.), throughout. Fixed at 0450C and held for all four chunks. The reason is that this range is about the Gentile mission and the same noun is re-quoted across chapters; the first stint's gentium apostolus → "apostle of the Gentiles" (0429D) is the anchor and 0453C reproduces that exact phrase, so the seam is checkable. ⚠ Note for the merge: the first stint renders omnes gentes "all nations" at 0432D in a Pentecost-tongues context. I have not touched it and do not think it should be swept — the sense there is peoples-with-tongues, not the Gentile church — but the divergence is real and is flagged rather than hidden.
• quadragenarius → quadragenary (0452C, in illo quadragenario dierum numero → "in that quadragenary number of days"), after 8965's substantival number-names (the senary, the quaternary, the octonary) as carried into 8964 §M. The neighbouring quaternio is kept as "quaternion" at 0453D–0454A for the same reason the coenaculum pair is kept apart: Quaternio dicitur princeps quatuor militum. Sicut enim centurio centum, ita quaternio quatuor sub se milites habet is an etymology, and "a file of four soldiers" would leave the centurion parallel explaining nothing. The lemma therefore reads To four quaternions, not Douay's "to four files of soldiers".

Biblical name forms — Douay-shaped, and consistent across all four chunks, matching the peers: Isaias · Ezechiel · Chanaan · Jonas · Judaea · Moses · Jerusalem. (Judea was written twice in draft and corrected to the peers' Judaea before filing; recorded because a silent conformation to a peer's form leaves no trace that the two ever differed — 8964 §M.)

Proper names form for form as Migne prints them, and these are the places the familiar English Bible would have overwritten him — every one was checked against the Latin, not against the Acts I know in English:

• 0454D Salamina, not Salamis. Abierunt Seleuciam, quae est Salamina. Rendered "Salamina".
• 0454D Niger kept untranslated as the name, because the gloss is the etymology (A colore, si Latinum est): "Niger" is the only English in which "From the colour" explains anything.
• 0454D–0455A Sergius and Paulus for the proconsul, Paul for the apostle. Migne prints Paulus for both and the gloss turns on their being the same word (A Sergio Paulo … sortitus est nomen Pauli), so the English keeps Sergius Paulus as the man's name and Paul as the apostle's, which is the received English of each. The identity is not lost: "From Sergius Paulus the proconsul … he obtained by lot the name of Paul."
• 0455B Chanaan (not Canaan) · 0455B–C Isaias (not Isaiah) · 0457C Ezechiel (not Ezekiel) · 0450C Jonas (not Jonah) · 0453C Caesarea for Migne's Caesaraeam (the ordinary medieval ae, Pattern 9, the class the first stint ruled on at Andraeas).
• ⭐ 0459A Lydda, not Lydia — see §2.4. This is the sharpest conformation trap in the range and it is the reason the fourth test is run.
• 0459A Thracia kept as printed, although the lemma two words earlier is Samothraciam. Migne's own shortening; not normalized.
• 0452A Ananias and Sapphira · 0454A Herod · 0458B Silas, Mark, Judas · 0458C Timothy, Titus — the received English of Migne's forms, no divergence in any.

Second person (Pattern 17) — 5 archaic tokens in this range, every one fixed by the Latin's own morphology: 0450C misereatur tui → "have mercy on thee" and qui misericors est → "thou art merciful"; 0451C ut hostias Deo facias → "that thou mayest make"; 0455C Eris caecus → "Thou shalt be blind"; 0458C quod non feceris → "what thou hast not done". Every plural stays you: 0452B Vos scitis → "You know", 0452D Vos venistis … ut testes sitis → "You have come … that you may be witnesses", 0453A Vos autem baptizabimini → "you shall be baptized", 0458A quae vobis fieri non vultis → "which you would not have done to you", 0455C Non credetis → "You will not believe".

§1 · THE LEMMA COLLATION — every italic span in cols 0450B–0459D, walked

data/briefs/8976-lemmata.txt was walked line by line for these bands. Every reading below was fetched at its own verse in sources/vulgate/clementine-flat.txt, cited by book chapter:verse — never from the pre-scan's "occurs somewhere" mark and never from recollection. The agreements are listed as well as the divergences, because a findings-only list cannot be told apart from a report by an agent that never looked.

1.1 · THE TWENTY-ONE [var:] FIRED

All in the house form — [var: Vulg. <Book> <ch>:<v> word …], book, chapter and verse explicit, per 8964 §M's ruling against the bare form.

Every [var:] content was checked against the Pattern 14 mirror test — none of the Clementine readings quoted inside a marker occurs verbatim in the Latin twin of the chunk it annotates. Checked individually for the eight in 0010, the two in 0011, the five in 0012 and the six in 0013.

1.2 · DIVERGENCES FOUND, MARKER DECLINED, with the reason each lost

Pattern 14 fires on a divergence that changes the sense, and Addendum B's ship test says the crux is always available and always safe. These are real, checked, and recorded here because a reader is entitled to know they were seen.

• 0450C · Nunquid aquam — Act 10:47 prints Numquid. Migne's nunqu- orthography stands all over the corpus (with coelum, tanquam, sylva). Pattern 9: not a divergence.
• 0450C · Joppen — Act 10:5 has in Joppen; the lemma is clipped at the noun (convention 3). Englished "Joppa".
• 0450D · Hospitatur — Act 10:6 hic hospitatur apud Simonem quemdam coriarium. A one-word lemma clipped at the verb; nothing elided within it.
• 0451A · Apparentibus illis — this is Migne's own alia littera, introduced as such. He is reporting a variant, not misquoting the Vulgate; a [var:] would tell the reader twice and blur whose observation it is (first stint §1.3, same ruling).
• 0451A · in coenaculum — likewise an alia littera Migne names as one.
• 0452D · In nomine Christi (Act 10:48 in nomine Domini Jesu Christi) — a compressed lemma, and the gloss's own next sentence prints the fuller form (Lucas tamen testatur in nomine Jesu Christi illum dari), so both are on the page already.
• 0453B · dispersi erant, pertransibant evangelizantes verbum — not Act 11:19 at all but Act 8:4, which Migne says himself (Revertitur ad id quod dixerat). Clementine Act 8:4 reads Igitur qui dispersi erant pertransibant, evangelizantes verbum Dei; the Dei falls after Migne's stop. A clip, not a variant.
• 0453D · Et prophetabunt filii vestri [n: (Joel. II)] — Joel 2:28 continues et filiae vestrae. Migne stops where the argument stops.
• 0455C · In Graeco: Audite contemptores, et respicite et disperdimini — Migne's own report of the Greek, announced as such. Unmarked by construction; and it is the control that makes the dispergimini marker four lines later worth firing, because it shows he can print disperdimini when he means it.
• 0455D · colentium Deum, sicut in Graeco habetur — again his own textual note against Act 13:43's colentium advenarum. He is doing the criticism; we are not.
• 0456A · Posui vos — the gloss's own counterfactual (Non ait, Posui vos), not a quotation.
• 0456A · Religiosas mulieres — Act 13:50 prints mulieres religiosas. A transposition in a two-word lemma with no sense change; logged, not marked, and this is the one place I am aware my line sits close to the Gal 2:4 marker two chunks later. The difference I acted on: Gal 2:4 is a quotation of Paul introduced by quod dicit, where the wording is the point; this is a bare lemma address, where the order of an adjective is not.
• 0456B · Deus autem pacem fecit — Migne's own note that the Greek has a clause quidam Latini codices minus habent. His observation, unmarked.
• 0458A · Solus autem Judas abiit Jerusalem — Act 15:34 Judas autem solus abiit Jerusalem. Word order only.
• 0458C · Custodire dogmata, quae erant — Act 16:4 reads custodiri. ⛔ Declined deliberately, and the reason is the file, not the reading: sources/vulgate/README.md records that later upstream corrections are not applied, so a call turning on a single letter has to say so rather than trust this copy to that precision. Nothing in the gloss turns on active against passive. Raised, not marked.
• 0459D · Cumque perduxisset, apposuit eis mensam — Act 16:34 Cumque perduxisset eos in domum suam. An ellipsis in the lemma; and the gloss under it names what was elided (ingressus domus, et mensae appositio), so nothing is concealed.

1.3 · THE LEMMATA CHECKED AND FOUND TO AGREE WITH THE CLEMENTINE EXACTLY

Named, because the brief requires the negatives. Each was fetched at its own verse.

Act 10 (chunk 0010): Vir autem (10:1) · Cornelius · Religiosus (10:2) · Is vidit (10:3) · Angelum Dei · Introeuntem · Corneli · Et nunc mitte viros (10:5) · Hospitatur (10:6) · Et cum discessisset (10:7) · Militem · Appropinquantibus (10:9) · in superiora · Cum esuriret (10:10) · Coelum apertum (10:11, Clem. caelum) · Vas quoddam · Linteum · De coelo in terram (Clem. de caelo) · Omnia quadrupedia (10:12) · Occide (10:13) · Descendens (10:21) · Ecce ego · Quae causa est · Responsum accepit (10:22) · In domum · Introducens (10:23) · Convocatis cognatis (10:24) · Et loquens (10:27) · Et ecce vir (10:30) · In conspectu Dei (10:31) · Non est personarum acceptor (10:34) · Vos scitis (10:37) · Qui pertransiit (10:38) · Et bibimus (10:41) · Spiritus (10:45). Also the inline Job 1:1 Vir erat in terra Hus, nomine Job — verbatim.

Act 11–12 (chunk 0011): Quadrupedia terrae (11:6) · Ergo et gentibus (11:18) · Et illi quidem (11:19) · Erant autem quidam (11:20) · Multusque numerus (11:21) · Et miserunt Barnabam (11:22) · Et hortabatur (11:23) · Tarsum (11:25) · Ut quaereret Saulum (11:25) · Discipuli, Christiani (11:26) · In his (11:27) · Prophetae (11:27) · Quatuor quaternionibus (12:4) · Volens · Milites, vinctus (12:6) · Lumen (12:7) · Percussoque (12:7) · Ad se reversus (12:11) · Processit puella (12:13) · Angelus ejus est (12:15) · Tacerent (12:17) · Confestim (12:23).

Act 13–14 (chunk 0012): Erant autem in Ecclesia (13:1, and see §5 on the capital) · Niger · Missi a Spiritu sancto (13:4) · Abierunt Seleuciam · Sergio Paulo (13:7) · Elymas (13:8) · Avertere · Saulus (13:9) · Qui et Paulus · Et nunc (13:11) · Eris caecus · Antiochiam Pisidiae (13:14) · Sederunt · Terra Chanaan (13:19) · Dedit (13:20) · Vobis (13:26) · Judicantes (13:27) · Sancta David fidelia (13:34) · Non dabis (13:35) · Videte contemptores (13:41) · Admiramini · Non credetis · Et colentium (13:43) · Primum (13:46) · Aeternae vitae · Sic enim (13:47) · Posui te · Praeordinati (13:48) · Excusso pulvere (13:51) · Confugerunt (14:6) · Et universam · Levaverunt (14:10) · Coronas (14:12) · Implens cibo (14:16) · Traxerunt (14:18) · Ut permanerent in fide (14:21) · Quoniam per multas tribulationes (14:21). Also the inline Ps 1:4 pulvis quem projicit ventus a facie terrae · 2Cor 11:25 Semel lapidatus sum · Act 9:4 Saule, Saule, quid me persequeris? · Is 55:3 Et feriam vobiscum pactum sempiternum misericordias David fideles (Migne's In Hebraeo quotation — verbatim, which is what convicts the quotation above it).

Act 15–16 (chunk 0013): Seditione (15:2) · Statuerunt · Conveneruntque apostoli (15:6) · Per os meum (15:7) · Fide purificans (15:9) · Tacuit autem omnis multitudo (15:12) · Et postquam (15:13) · Revertar et reaedificabo (15:16) · Super quas invocatum (15:17) · Invocatum est · Ego judico (15:19) · ⭐ Et suffocatis (15:20) · ⭐ Et sanguine (15:20) · Tunc placuit (15:22) · Apostoli et seniores (15:23) · Fratribus ex gentibus · Salutem · Evertentes animas (15:24) · Visum est (15:28) · Spiritui sancto · Et nobis · Quam haec necessaria · ⭐ Ut abstineatis vos (15:29) · Visum est autem (15:34) · Visum est autem Silae ibi remanere (15:34) · Dissensio (15:39) · Assumpto Marco · Electo Sila (15:40) · patre gentili (16:1) · Et circumcidit (16:3) · Et ecclesiae quidem (16:5) · Tentabant ire (16:7) · Vir Macedo (16:9) · Adjuva nos ×2 · Samothraciam (16:11) · Aliquot, conferentes (16:12) · Purpuraria (16:14) · Coegit nos (16:15) · Subsecuta Paulum (16:17) · Miserunt eos in carcerem (16:23) · Media autem nocte (16:25) · Terrae motus factus est magnus (16:26) · Procidit (16:29) · Lavit plagas eorum (16:33) · Continuo (16:33).

1.4 · ⭐⭐ THE APOSTOLIC DECREE — the crux the brief named, and what the check actually found

The brief flagged Act 15:20 and 15:29 as a place where Migne and the received text may part. They do not part in the lemmata, and that is the finding.

• Act 15:20. Clementine: sed scribere ad eos ut abstineant se a contaminationibus simulacrorum, et fornicatione, et suffocatis, et sanguine. Migne glosses two clauses of it, Et suffocatis and Et sanguine, both verbatim. He prints no et a fornicatione lemma and no golden-rule clause here. The four-clause Clementine decree is what stands.
• Act 15:29. Clementine: ut abstineatis vos ab immolatis simulacrorum, et sanguine, et suffocato, et fornicatione : a quibus custodientes vos, bene agetis. Migne's lemma is Ut abstineatis vos, etc. — agreeing exactly as far as it is printed, and the gloss under it is a Tertullian abridgment about food.

⭐ But the Western decree is in the range — in the GLOSS PROSE, not in a lemma. At 0458A, glossing Quam haec necessaria, Migne writes: Haec necessaria sunt sine quibus non est salus, sed circumcisio necessaria non est: verum ea quae vobis fieri non vultis, alii ne faciatis. That is the negative golden rule, the clause the Western text adds to the decree at 15:20 and 15:29, and the gloss plainly knows it as part of haec necessaria — it supplies the fourth item where the Clementine's fourth item is fornicatione.

⛔ No [var:] fired, and the reason is the pattern's own scope. Pattern 14 marks Migne's scripture citation diverging from the received text. This is not printed as a citation: it is roman, not italic, carries no [n:], and is grammatically part of the gloss's own sentence (verum … ne faciatis, continuing sed circumcisio necessaria non est). Firing a [var:] on it would be asserting that Migne quotes Acts here, which he does not — he reasons from a form of the decree he has. The rival I weighed and rejected was marking it against Act 15:29 with a note naming the Western reading; it lost because the marker would misdescribe the typography, and because the crux can say the whole thing and a marker cannot.

Both negatives are rendered on their printed hosts: non on vultis ("which you would not have done to you"), ne on faciatis ("do not to another"), with alii dative singular kept singular.

§2 · PLATE CANDIDATES — five raised, ZERO markers, following 8964 §W

Addendum A's split governs: a non-word is worth raising and has been right every time; a real, well-formed word that reads oddly is MIGNE'S until a plate says otherwise — render as printed, log, fire nothing. None of these blocks the work, which is the point of the ship test.

2.1 · 0451C · Mundacavi — a NON-WORD, and the likeliest of the five

VERS. 14.-- Mundacavi, etc. Act 10:14 reads numquam manducavi omne commune et immundum, and Migne's own gloss underneath is built on manducare: Qui manducat cibum foris positum in suum trajicit corpus. Mundacavi is no Latin word at all — a metathesis of manducavi. Rendered for its evident sense, "I have eaten, etc.", logged, no marker. A [sic:] would blame Migne's plate for what Addendum A says has been the digitization six times out of six; the corpus-frequency argument is exactly the argument that loses on real words and wins on non-words, and this is a non-word.

2.2 · 0453A · paritet — the second non-word

qui etiam Patrem dicit, et Filium et Spiritum paritet ostendit. For pariter. Rendered "alike", logged, no marker, same reasoning.

2.3 · 0455A · e c. — split type in the abridgment formula

A Sergio Paulo dictus est apostolus, Paulus, e c., usque ad, Ego sum minimus apostolorum. Every other resumption in this range is introduced by etc.; this one is broken across a space. Pattern 10 would carry the whole run in a [sic:], and I declined that: the run is etc. and nothing else, the class is the digitization's per Addendum A, and 8964 §W's thirteen candidates took zero markers for the same reason. Rendered "etc.", logged.

2.4 · ⭐ 0459A · Lydda — a REAL word, and it stays

Lydda apostolum, cum sociis suis domum intrare coegit. The woman of Act 16:14 is Lydia; Lydda is the town of Act 9:32–38, a different place and a real, well-formed name. Addendum A is decisive: a real word that reads oddly is Migne's. Rendered "Lydda", logged, no marker. ⛔ This is the conformation trap of the range — "Lydia" is what the English Acts supplies before the Latin is read, it would have parsed perfectly, and nothing mechanical would have seen it. It is the same class as the peer batch's Absalom→Absalon.

2.5 · 0458D · Hoc es — where the evidence points at OUR file, and I say so

Hoc es [0458D] superioribus capitulis saepius expositum. es is a real word ("thou art") and cannot stand here; est is what the sentence needs. ⚑ The reverse rule the brief names applies, and I am naming it: the defect sits immediately before the column-break anchor, which is where our archive copy's transcription is weakest, and Migne prints Hoc est correctly in the ordinary way everywhere else in the work. On that evidence this is more likely ours than his, so I decline any marker and do not attribute it to the plate. Rendered for its evident sense ("This has been more often expounded"), logged, and offered as a data/tei-patches/8976.json candidate for whoever reads PL 114 p. 231 — not as a [sic:].

2.6 · Suspiciously TIDY forms — a checked negative

Brief §4 and Addendum C ask for forms that read too well in a rough passage. I looked at the neighbours of every line I examined closely and found none to raise. Recorded as a checked negative, because the class is invisible from the Latin alone and a silent pass is indistinguishable from not looking. The forms I considered and cleared: praeposteravit (0452D, rare but well formed and exactly right), effoetandi-class gerunds (none here), conchyliorum (0459A), feminalibus (0458C — the priests' linen breeches, Ex 28:42's word).

2.7 · The . [a-z] class the first stint found at 0435B — ZERO in my range

Searched all four chunks for a full stop followed by a lower-case word (the shape that turned out to be our punctuation rather than Migne's colon). No occurrences in 0450B–0459D. Recorded as a checked negative so the merge does not have to re-run it.

§3 · MIGNE'S OWN TROUBLE — rendered as printed, logged, nothing repaired

3.1 · ⚠ 0452D · A SENTENCE WITH NO TERMINAL MARK, and the English has none either

nunquid quod minus est, potest prohibere Vos venistis ad hoc ut testes sitis, ergo judicate. The capital V shows a sentence boundary; no mark stands before it in our Latin. Pattern 8 forbids supplying one, so the English reads "can he forbid what is less You have come for this, that you may be witnesses, therefore judge." ⛔ It reads as a defect because it is one, and it is visible in the facing column. A plate read of PL 114 p. 228 would settle whether Migne prints a ? there (the sense wants one) or whether the digitization dropped it; on the first stint's 0435B precedent — where the plate turned out to print a colour our file had flattened — our file is the likelier culprit, and that is why nothing was supplied and no [ed:] was fired (Pattern 13 wants demonstrated loss, and this is inferred loss).

Same paragraph: the unexpressed subject of potest prohibere. Two readings construe. (a) Deus, the nearest subject, carried on from Quasi Deus … prius dedit — "can he forbid what is less". (b) An indefinite quis, as Act 10:47 itself prints (Numquid aquam quis prohibere potest) — "can anyone forbid". (a) was taken, because it is the subject the sentence prints and (b) requires importing a word from the verse the gloss is abbreviating. The rival is named here because a reader will weigh it.

3.2 · 0452A · zelus ultionis ejus aperuit potestatis — a bare genitive with nothing to govern it

Communionem aequalitatis meruit sanctitas actionis, sicut in Ananiae et Sapphirae culpa zelus ultionis ejus aperuit potestatis. potestatis has no head noun. Real words throughout, so Addendum A holds and nothing is marked. Rendered by taking potestatis with an elided zelum carried over from the same sentence — "the zeal of his vengeance laid open that of his power" — which is the one construction the printed words allow. ⛔ The rival, and why it lost: [magnitudinem] potestatis aperuit, "disclosed the greatness of his power", is the smoother conjecture and is probably what a modern editor would print; it lost because it requires supplying a noun that is nowhere in the sentence, whereas zelum is four words back. No conjecture is put in the text.

3.3 · 0450C · vocantem cum — cum standing where the object must

Introeuntem, ad se forsitan, vocantem cum. A preposition with no object, in the accusative slot of vocantem; eum is the evident reading. Rendered "calling him" for the evident sense, on 8964 §W's precedent (a stranded est handled the same way), logged, no marker and no patch. ⚠ It is a real word, so Addendum A's presumption is Migne's; the reason I rendered the sense rather than "calling with" is that cum here is not odd, it is syntactically impossible, and Pattern 7's "render as closely as the printed words allow" allows nothing else. Flagged as a plate candidate.

3.4 · 0453D–0454A · catenae carcer — a missing comma inside a list of four

Quatuor difficultates cumulant custodiam: milites, catenae carcer, custodes. The gloss counts four and prints three commas' worth of separation for four items. Rendered exactly: "soldiers, chains prison, guards". Pattern 8: never supply. Logged; a plate read would settle it and nothing depends on the answer.

3.5 · 0452D · Spiritus, domum Dei est — the accusative where a nominative belongs

Spiritus, domum Dei est, inquantum eis datur qui per eum diligunt Deum. Two sentences later Migne prints the word the argument needs: Ita enim datur sicut donum Dei. Rendered as printed — "The Spirit is the house of God" — because domum is a real, well-formed word and Pattern 7 renders what is printed; Latin case has no English exponent in a predicate nominal, so nothing else was normalized. ⛔ The rival, and why it lost: donum, "the gift of God", which the same paragraph prints and which is plainly the sense. It lost the text under Pattern 7 and it is recorded here, which is where a conjecture belongs. No marker: Addendum A refutes the corpus argument on real words, and the reverse rule cuts the other way too — Migne's own next clause printing donum correctly makes an n→m misreading by the digitization at least as likely as a defect on his plate, which is a reason to decline, not to fire.

3.6 · 0459B · The objectless Misit in interiorem carcerem

Already marked at §1.1. Recorded again here because it is the one place in the range where the English is deliberately ungrammatical: "He put into the inner prison." Supplying "them" would have been the silent repair — the sentence would have read perfectly and the divergence would have vanished — and the [var:] is what tells the reader the object is missing from Migne, not from us.

3.7 · 0455B · Sancta David fidelia glossed under VERS. 27.

The lemma is Act 13:34; the address above it is VERS. 27, and VERS. 35. follows immediately. Migne's own placement. Reproduced exactly as printed, not moved (the Jude precedent in the brief). ⚑ Recorded as a checked datum for the merge: the first stint's §2.1 found displaced verse numbers in cols 0425C–0437C, the second stint found none in 0437D–0450A, and my range has exactly this one.

3.8 · 0459A · aliquos doctores coepisse doceri

Ostendit Macedones vel angelos a Deo inspirari, vel aliquos doctores coepisse doceri. A passive infinitive where an active (docere, "to teach") would be expected of teachers. Rendered as printed — "or that certain teachers had begun to be taught" — real words, logged, no marker. It is not nonsense: the gloss has just made the Macedonians and their angels the ones being inspired, so teachers being taught is intelligible, if odd.

§4 · THE usque ad CENSUS — decided per occurrence, never by rule

18 occurrences in the four chunks. 17 are the abridgment formula; 1 is not.

• 0010 — 8 occurrences, 7 the formula (0450B ×2, 0450B, 0451B, 0451C, 0451C, 0452B) and one that is not: 0452C docens per hoc quod nobiscum est usque ad consummationem saeculi. Those are Matt 28:20's own words, inside the gloss's own sentence, with no italics and no resumption after them. Rendered "until the consummation of the world". ⛔ This is Addendum D's trap exactly, and it is the only one in the range.
• 0011 — 3 occurrences, all 3 the formula (0453A, 0453B, 0454B).
• 0012 — 2 occurrences, both the formula (0455A, 0455B). ⚑ Note that 0455A's is the comma form, usque ad, followed by the resumption inside the same italic span — usque ad, Ego sum minimus apostolorum [n: (I Cor. XV)] — as is 0010's at 0451C. The italic structure is preserved 1:1 in both.
• 0013 — 5 occurrences, all 5 the formula (0457A, 0457A, 0457C, 0458A, 0458C).

No verb is ever supplied to complete a resumption tail (Tobit §1(b), inherited). Four tails are left with an unexpressed subject rather than an invented one: 0451C sed in sola Dei dilectione constringat → "but may bind fast in the love of God alone"; 0452B sed omnes credentes acciperent → "but that they should receive all that believe"; 0455B et ideo maledici meruit → "and therefore he deserved to be accursed"; 0457A ad fidem venisse probatur → "he is proved to have come to the faith". Each is named in §6.

§5 · THE MIRROR RULE, within its ruled scope

8964 §M's boundary was applied and not re-opened: mirror the plate where the case is semantic in English; do not mirror where English has no lower-case option for the referent.

• Ecclesia is capitalized at every occurrence in this range and is "the Church" at every one: 0450D, 0451A, 0451B ×2, 0451C, 0454D (Erant autem in Ecclesia — where the Clementine prints lower-case ecclesia and Migne's capital is his own), 0457A, 0459A. ⚑ 0458D Et ecclesiae quidem, lower case, is the one exception, and it is rendered "the churches" — mirrored, verified at the source, not taken on trust. That is the pair the founding case is about.
• Spiritus mirrors the plate the same way, per the first stint. Lower-case spiritu at 0452B (sed spiritu et virtute signorum) and lower-case spiritus at 0451D (prophetiae spiritus) are rendered lower-case; every capitalized Spiritus is "the Spirit"/"the Holy Spirit".
• Synagoga capitalized at 0451A → "the Synagogue", set against gentilis populus.
• Aquilone capitalized at 0453B → "the North". English has a lower-case option for a cardinal direction, so this one is inside the mirror's scope.
• Gentiles capitalized at 0450B (quia Gentiles grosso sensu) and lower-case gentilis/gentiles elsewhere. The English holds "the Gentiles" as the fixed rendering of the noun (§V) and "gentile" for the adjective, which is a vocabulary decision, not a mirror one; recorded so the two are not confused.
• Not mirrored, by the ruled exception: the divine name and Deus/Dominus forms. There is no lower-case dominus or deus in this range, so the exception was not needed.

Chapter heads mirror what Migne SETS, per the house convention: CAPUT X. → CHAPTER X. · CAPUT XI. → CHAPTER XI. · CAPUT XII. → CHAPTER XII. · CAPUT XIII. → CHAPTER XIII. · CAPUT XIV. → CHAPTER XIV. · CAPUT XV. → CHAPTER XV. · CAPUT XVI. → CHAPTER XVI. Seven heads, all numerals, none spelled out.

§6 · EVERY SUPPLIED PRONOUN OR SUBJECT, NAMED — the dominant defect class

The brief's rule: every "he", "she", "it" or "they" the Latin does not print is named here with what fixes it. Where two readings construe, both are stated.

Fixed by a printed antecedent or by the verb's own morphology — no judgment:

• 0450D contemnunt → "they despise" = the obedientes of the preceding clause (3 pl.).
• 0451C facias → "thou mayest make" = Peter, from the lemma Occide addressed to him.
• 0451D jubetur → "he is bidden" = Peter (lemma Descendens, Act 10:21).
• 0452A descenderat → "he had gone down" = Peter (lemma Introducens, Act 10:23).
• 0452A alienus erat → "so far was he from disdain" = Cornelius, the speaker of Act 10:30 whom Non dicit angelum is about.
• 0452C nobiscum est → "he is with us" = Christ, the subject of the whole paragraph.
• 0455A subegit → "whom he subdued to the faith" = Paul.
• 0455D praeripuit sanitatem → "snatched the healing before her" = the ruler's daughter, from filiam principis in the same sentence.
• 0456B solverent → "they would loose their feet" = the subject of the printed protasis Si vero apostolos reciperent.
• 0458B conjunctus → "He was joined to Barnabas" = Mark (lemma Assumpto Marco).
• 0458B remanere eum fecit → "made him remain" = Silas (lemma Electo Sila).
• 0458C videretur → "lest he should seem" = Paul, subject of circumcidit.
• 0459B obviasset → "although she had met them" = femina, printed in the same sentence.
• 0459C illis … canentibus → "while they sing hymns" = Paul and Silas (Act 16:25).
• 0459C cur se vellet interficere / procidit → "he wished to kill himself" and "he fell down" = the jailer (Act 16:27, 29).
• 0459D amisit → "he lost the stripes of his crime" = the jailer.

Supplied with a stated reason, where English could not leave it unexpressed:

• 0450C "occasion is given to Peter of excusing himself" — excusandi is a bare gerund; the reflexive comes from excusantur apostoli two clauses above.
• 0450C "calling him" — the object cum stands where eum must; §3.3.
• 0451A significavit → "he signified" = the writer of in superiora, i.e. Luke, not Peter. Fixed by the sentence's own frame: Dicens in superiora is about the word chosen, and the alia littera sentence before it is about copies of the text.
• 0451C Praecepit → "He commanded therefore" = the voice of the vision, Act 10:13's facta est vox ad eum, not Peter. ⚠ Peter is the nearer noun (quam Petrus significat) and would be the wrong choice: Peter is the one commanded.
• 0451C "that thou mayest make them victims" — object of facias, from the omnia quadrupedia of the lemma two paragraphs above.
• 0451D "By this they persuade him" — suadent 3 pl. = Cornelius's messengers; the object is Peter, from the narrative and from quia angelus jussit introire in domum ejus (Cornelius's house).
• 0456B "some Latin copies have it not" — "it" = the Greek clause Deus autem pacem fecit, named in the same sentence.
• 0456C imponeret → "or might set them upon them" — the crowns upon the victims, or upon the apostles. ⚠ Both construe and the Latin's eis does not decide: quasi diis immolaturus points at the apostles-as-gods of Act 14:12, while victimis simul coronatis in the next sentence points at the victims. The English keeps both pronouns bare, which is as undecided as the Latin. Named rather than resolved.
• 0457B Domino revelante → "when the Lord revealed it" — object supplied: the shadow of the law, subject of the sentence.
• 0457D ut … intelligantur → "so that they may be understood to wish them all good things" — the apostles (printed two clauses back) wishing the brethren.
• 0458C "But that she was by race a Jewess … that he says he was sprung of a gentile father" — three: the mother (from Vel Judaeae), the speaker of ait (the text of Acts), and Timothy. All three are forced by the sentence.
• 0459A Ostendit → "He shows that the Macedonians …" — ⚠ two readings. (a) the passage/the evangelist, continuing the impersonal exposition; (b) the angel of the previous paragraph. (a) was taken, because Ostendit opens a fresh gloss on a re-quoted lemma (Adjuva nos a second time) and every other gloss-opening third singular in this range is the expositor's. (b) is named because Vel hoc dicit angelus stands eight words earlier.
• 0459D "when he had brought them" — from eis in the same clause; Migne's lemma has dropped Act 16:34's eos in domum suam (§1.2).

Left unexpressed on purpose — the four abridgment tails in §4, where the subject is off the page and inventing one would be the silent repair.

§7 · THE TESTS, AND WHAT EACH RETURNED

Test 1 — a [sic:] must not swallow the word it marks. ✅ Vacuous here: zero [sic:] fired. Ran the mechanical form anyway (strip every marker, read what is left) across all four chunks: no stranded article or preposition sitting on punctuation, no doubled mark, and every sentence reads aloud without its brackets. The 21 [var:] are all additive notes standing after a complete sentence, never inside one, so nothing of Migne's is quarantined anywhere in the range.

Test 2 + 2a — name the host of every printed negative, then find it in the English. Ran on every non · nec · neque · nisi · ne · nihil · nullus · nunquam · minus. Six sites fired 2a's trigger condition (a negative resting near a quantifier, correlative or predicate) and each was proved from the sentence's own structure, never from what read well:

1. 0450B Non virtutibus ad fidem, sed fide pertingebat ad virtutes. Host is virtutibus, not pertingebat — the sed fide correlative pivots on the ablative. English: "Not by virtues did he attain to faith, but by faith he attained to virtues."
2. ⭐ 0453C qui non solus praesumit docere plebem, sed apostolatus quaerit consortem. This is the 8990 @0614B class exactly — a negative on a verb with solus adjacent. Host is solus: the sed clause says he seeks a partner, so the point is not that Barnabas refuses to teach but that he will not teach alone. English: "who does not alone presume to teach the people, but seeks a partner of the apostleship." Rendering it "does not presume to teach the people" would have passed every count and asserted the opposite of the sed.
3. 0453C non solum Judaeis, sed etiam gentibus. Host Judaeis.
4. 0457A haec non dicunt solum pro gentibus … sed etiam illos a lege abducere cupientes. Host is solum pro gentibus, not dicunt.
5. 0458B quia non omnis commotio est peccatum, sed irrationabilis. Host is omnis, the 8990 @0611D and 8960 @0734B class. English: "not every commotion is a sin, but the unreasonable one."
6. 0458C Permittuntur apostoli velle, sed non ire. Host is ire. ⛔ This is the 2 Thessalonians @0624B shape — a sed non beside a verb of willing — and the temptation is to press the negative onto velle ("they are unwilling"). It is not on velle; they are permitted to will. English: "permitted to will, but not to go."

Comparatives checked separately (7a mechanism 4): 0456B Fuga apostolorum non plus fuit ex timore, quam amore verbi Dei → "was not more out of fear than out of love" — not reversed; 0456B altior erat sententia quam sua lingua barbara → "higher than"; 0452D minor est illis → "is less than they"; 0458A quorum munditia magis ad Deum pertinet, quam intestinorum → "pertains more to God than that of the entrails".

Every other negative carried on its printed host, including where it makes the sentence awkward: 0450B Non in nocte, non in somnis · 0450C Non Satanae · 0450D nec per somnium, nec in exstasi · 0451D ex parte non tangit · 0452B non quod quotidie … non est … non indigens · 0452C non assidue · 0452D etiamsi nulli datur · 0453C Nihil mutavit vel addidit · 0453D non servasse · non hoc eum improbare · sed non pro eis · non regnum Dei, sed temporalia · ista non cogitemus · 0454A nec aliis refulgebat · 0454C non exhorreret · 0455B Non veni nisi · 0455C non moritur · non dominabitur · Non credetis · non credere · 0455D Non quia scirent · 0456A Non ait · 0457A Non alterius · nullus apostolorum · nondum dicunt · 0457C non comedant · ne comedant · non est effusus · 0457D ne aliqua … relinquatur · nullum verbum ponitur · 0458A non est salus · necessaria non est · non vultis · ne faciatis · 0458B Non in hoc delinquebant · quae juste non fit · 0458C ne … videretur · quasi non necessariam · non imponere · non circumcidit · ne … acciperent · Non praedicabant · quod non feceris · non autem tanquam facturi · 0459A non enim angeli … indigent · 0459B non in facie, sed post eos · 0459D Nescit tarda molimina. Nothing was added that Migne does not print.

Test 2b — each anchor checked against the WORD it stands beside, not merely present. 39 anchors, all 39 verified in place, and four were repositioned in draft after this check rather than left where the English fell naturally: 0450B (moved to stand before illud, matching secundum [0450B] illud), 0453C (moved to stand before "not only to the Jews", matching civitates [0453C] non solum Judaeis), 0456A (moved to stand before "said", matching quod Christo [0456A] dicitur), 0457D (English reordered to "Of the apostles the [0457D] common title", matching Apostolorum [0457D] communis titulus). No anchor was split mid-construction; none needed it.

Test 3 — punctuation compared BY COLUMN BAND, never by total. Colons, semicolons, question marks, exclamation marks, guillemets and parentheses, band by band across the 39 anchor-bounded bands. ⛔ First pass returned FOUR mismatched bands in 0010 (0451A colon 3/2 and semi 0/1; 0451D semi 0/1; 0452B semi 0/1; 0452D colon 1/2 and semi 1/3), one in 0011 (0453B semi 0/1) and two in 0012 (0456B colon 1/0; and one supplied colon in the same sweep). Every one was mine: a colon of Migne's softened to a semicolon at 0451A, three semicolons supplied where he prints commas, an "As if:" colon invented at 0452D, and one colon of his dropped at 0456B. All repaired. Final state: 0010 0 mismatched · 0011 0 · 0012 0 · 0013 0. Commas were deliberately excluded from the comparison — English syntax requires different comma placement, and the corpus ruling closed comma→semicolon sweeping as a rate question.

Totals for the record, and note that they would have hidden three of the seven: Latin/English question marks 3/3 (0450C, 0455B, 0456A — all direct questions keeping their marks with no recasting; Pattern 8b was not needed in this range, there being no indirect question closed with ?), guillemets 0/0, em-dashes 0 in the English.

Structural parity, re-checked after every edit: column anchors 39/39 verbatim, in order and in place (12 + 6 + 9 + 12); [n:] notes 107/107 at parity and in order (31 + 12 + 32 + 32); paragraph counts 1:1 in every chunk (39 + 23 + 38 + 47); italic spans 1:1 — the English's asterisk surplus is exactly the internal italics of the [var:] markers, verified pair by pair (0010 +11 pairs, 0011 +3, 0012 +8, 0013 +11).

Notes pass through verbatim, contents untranslated. Zero were converted to [nt:] — Pattern 15 was tested against all 107 and every one is a citation or an attribution siglum ((RAB.) 44 · (ID.) 30 · (BEDA.) 11 · (BED.) 1 · (CHRYS.) 2 · (GREG.) 1 · (HIER.) 1 · (TERT.) 1 — 91 sigla — plus 16 scripture references). None is editorial prose. Judged by content, not by length; the longest, (Psal. XCVIII), is still a locator.

⭐ Test 4 — does the paragraph ASSERT what he asserts?

Run last, with the counts already clean. It returned four things nothing mechanical would have seen, and all four are conformations or mis-parses that read better than the truth:

1. ⭐⭐ 0459A Lydda → "Lydia". The familiar name arrived over the plate in draft. §2.4.
2. ⭐ 0455C dispergimini → "be destroyed". The Clementine's word had been carried into Migne's, flattening the very distinction the next six words gloss (ab Assyriis et Chaldaeis, novissime a Romanis). Now "be scattered", with the [var:].
3. ⭐ 0454D Salamina → "Salamis". The received English form of the city, in a sentence whose whole content is duo autem sunt nomina unius civitatis — the names are the subject.
4. 0451A quem … quaerere et salvare venit. quem is masculine and its antecedent is mundum, not the feminine salutem standing next to it: Jesus came to seek and save the world, not the salvation. A draft "which Jesus came to seek and to save" bound to the wrong noun; the morphology decided it.

Sections read clean under the fourth test and named as such: 0010's VERS. 6, 7, 9, 10, 11, 12, 13 (the sheet and the fourfooted beasts, 0450D–0451C) — dense allegory, no scripture re-quotation, nothing to conform; 0011's CAPUT XII entire (0453D–0454C, Peter's deliverance) — eight glosses, all narrative, the only trouble in it being the catenae carcer comma of §3.4; 0012's VERS. 46–51 and CAPUT XIV entire (0455D–0456D); 0013's VERS. 22–29 (0457C–0458A, the council's letter) and VERS. 25–34 (0459B–0459D, the jailer). ⚑ Named because "a flagged passage is not a checked passage" cuts both ways: a reader is entitled to know which stretches were read and found sound.

§P · MIGNE'S CITATIONS — one misprinted, left EXACTLY as printed

Identified from the quoted words, never from the numeral, per the standing rule. Not touched in the text; referred to citation-corrections.json here, for the merge to enter.

1. 0457C — [n: (Ezech. IV)] on Omne morticinum et captum a bestia de avibus et pecoribus, sacerdotes non comedant. The words are Ezech. 44:31, the priestly food law. Ezechiel IV is the siege-tile and the barley cake; its only contact with the quotation is 4:14's morticinum … non comedi, which is Ezechiel speaking of himself and contains no sacerdotes and no prohibition addressed to priests. The clinching words are sacerdotes non comedant, which stand only at 44:31. refDisplay keeps (Ezech. IV); refKey should resolve to Ezek.44.31.

The other fifteen scripture citations in this range were checked against their quoted words and are CORRECT: (Act. XI) ×2 (Act 11:3, 11:16) · (Matth. V) (5:7) · (Philipp. III) (3:20) · (Matth. XXVI) (26:29) · (Joan. III) (3:8) · (I Cor. XII) (12:13) · (Matth. VI) (6:34) · (Joel. II) (2:28) · (I Cor. XV) (15:9) · (Matth. XV) (15:24) · (Psal. I) (1:4) · (II Cor. XI) (11:25) · (Num. VI) (6:27) · (Psal. XCVIII) (98:6). ⚑ Recorded as checked negatives: the first stint found two misprints and the second two, so a range with only one is worth being able to tell apart from a range nobody read.

§8 · THE PLATE — a checked zero, and what a read would still settle

No [cn:] anywhere in this range, and that is a checked zero rather than a skip: PL 114's plate spot check for this work found no numbered Forte apparatus, and both mid-page work divisions were examined, which is what makes the zero cover these columns. [cn:] is Latin-only in any case and none is in the Latin twins.

Four questions a read of PL 114 pp. 228–232 would close, none of which blocks the work because none of them fired a marker (Addendum B):

1. p. 228, col. 0452D — is there a ? (or any mark) after potest prohibere? §3.1.
2. p. 228, col. 0451C — does the plate read Manducavi or Mundacavi? §2.1.
3. p. 231, col. 0458D — Hoc es or Hoc est? §2.5, where I have argued it is ours.
4. p. 232, col. 0459A — Lydda or Lydia? §2.4. ⚑ This one I expect to go Migne's way, on Addendum A's 5-for-5 record on real words, and I want the negative recorded either way.

Also worth one look while those pages are open, per Addendum C's rule that the normalization class is found four words from the word you went for: 0453A paritet (§2.2), 0455A e c. (§2.3), 0452A aperuit potestatis (§3.2) and 0450C vocantem cum (§3.3).

§9 · FOR THE MERGE

• [var:] marker form is the house form throughout — book, chapter and verse explicit in all 21. No re-run of 8964's two-form drift; the first two stints use the same form.
• One vocabulary seam is flagged and NOT harmonized: omnes gentes → "all nations" at the first stint's 0432D against my "the Gentiles" throughout 0450B–0459D. §V. I judged the two senses genuinely different and did not sweep a peer's file; the merge should rule.
• One displaced verse address, 0455B, reproduced as printed. §3.7.
• One misprinted citation for citation-corrections.json. §P.
• Five plate candidates, zero markers. §2. Three are non-words or split type (Addendum A's winning class); two are real words rendered as printed.
• Nothing outside 0010–0013 was touched, and cruces.md was not written.

Stint: chunks 0014–0017, cols 0460A–0470A (3,553 Latin words), CAPUT XVII – CAPUT XXVIII. The Areopagus · Ephesus and Diana · the Miletus farewell · the arrest and the defences before Felix, Festus and Agrippa · the shipwreck · Rome. This is the tail of the book, and the Acts→Romans division at cols 469/470 falls inside it.

Three peers hold 0000–0004, 0005–0009 and 0010–0013. This is a stint file. It is NOT cruces.md; the merge is the launcher's. I wrote only my four chunks and this file.

Marker tally for this stint: 4 [var:] · 0 [sic:] · 0 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 0 [cn:] · 92 [n:] (exact parity with the Latin twins: 22 + 22 + 32 + 16). Zero guillemets in either language, all four chunks. Zero em-dashes in the English. EN/LA ratio 1.41 · 1.43 · 1.50 · 1.45.

verify-english.mjs reports OK on all 18 chunks of the work — the three peers are on disk and the whole book now verifies.

§0 · WHAT I INHERITED, AND FROM WHERE — read before writing a word

Both earlier stint files were on disk when I began and I read them first, as instructed. Nothing below was re-derived. ⟵ marks an inheritance; the few new decisions are marked new and declared rather than smuggled.

⚑ On that last row: my range has three non-words and they are RENDERED, not carried. The launcher ruled the peers' disagreement (§S2 of cruces-0005.md) and I followed the ruling rather than either peer. Sites in §W below.

Corpus controls run before inventing anything (src/latin/, src/english/): Atheniensis 2 : Athenienses 2 (both attested — see §W4) · Sceno 1 (mine alone) · Tyrrhenc 1 (mine alone) vs Tyrrheno 1 elsewhere · dithalassum 0 : Bithalassum 1. "by antiphrasis" attested (8976/0466C, peer's §X).

§S0 · THE SEAM WITH 0010–0013 — its cruces landed while I was drafting, and I checked it

cruces-0010.md was not on disk when I began (only cruces-0000.md and cruces-0005.md were, which is what the launch message named). It appeared while I was drafting, I read it before filing, and it found one real drift in my chunks, now repaired.

⛔ 0461A — Isaiae had been rendered "Isaiah". The third stint fixes the work's biblical name forms as Douay-shaped and consistent — Isaias · Ezechiel · Chanaan · Jonas · Judaea · Moses · Jerusalem — and records that it corrected its own Judea to the peers' Judaea before filing for exactly this reason. My Isaiah was the AV form, and the only occurrence of the name in my range. Repaired to "not of Isaias". ⚑ Recorded rather than silently conformed, per 8964 §M: a translator that quietly matches a peer's form leaves no trace that the two ever differed.

Checked across the seam and sound, needing nothing:

• gentes / gentilis → the Gentiles / gentile (its declared extension) holds through my five sites: 0460C de colentibus gentilibusque and gentiles permanserant · 0461C gentes sine circumcisione salvari · 0465D evangelizare gentibus · 0467C Philosophi gentium. Reached independently and identical.
• designat → designates, its third verb held off significat. No designat in my range, so the extension is untested here; significat occurs twice (0463C Sacrum … mundum et profanum significat → "signifies", 0465A ascensionis in castra significat → "It signifies") and both take signifies, per 8964 §M. Consistent.
• Eutichus → Eutychus (0463D) — I did not keep the form, and the ruling is the first stint's, not a lapse: Eutichus/Eutychus is the ordinary medieval i-for-y, the same class as Andraeas → Andrew (cruces-0000.md §V), not the class of Salamina, Lydda or Barnabas, which are different names. My §Q lists it among the Pattern 9 items.
• Its §2.7 records "the . [a-z] class the first stint found at 0435B — ZERO in my range." ⭐ Mine has one (§M4), which together with 0435B is the whole population of the work, and its zero is what makes that count trustworthy.
• Its §P has one misprinted citation, mine has two (§P), and all three are treated identically: left exactly as printed, identified from the quoted words, referred to citation-corrections.json. Four stints, six misprinted citations, one treatment.
• The [var:] marker form. Its 21 markers, the first stint's six, the second's five and my four all carry book, chapter and verse explicit. Thirty-six markers, one form, and no repeat of 8964's two-form drift.
• Its §2.5 declines a divergence because the evidence points at OUR file. ⚑ I met the same question at 0467B and came down the other way, with the reasoning written out — see §D1. The two decisions are not in conflict: its Hoc es is a single occurrence with a sound twin nearby, mine is printed twice against one sound occurrence.

§V · THE DIVERGENCE LIST — every italic span in cols 0460A–0470A walked

⛔ The agreements are listed too. A findings-only list cannot be told apart from no look at all. Every reading was fetched at its own verse in sources/vulgate/clementine-flat.txt, never matched "somewhere in the Clementine" and never from recollection.

V1 · The four that took a [var:]

All four [var:] contents were checked not to occur verbatim in their Latin twin (Pattern 14's mirror test, which verify-english.mjs enforces): Sopater, Nolite possidere aurum, detuli sententiam, dithalassum — none of the four marker strings is in the chunk it annotates.

V2 · Divergences found and a marker DECLINED, each with its reason

Pattern 14 fires on a divergence that changes the sense. These do not, and a [var:] is a public claim; the crux is always available and always safe (ship test B).

• 0460B · Quod ignorantes colitis, hoc annuntio. — Act 17:23 Quod ergo ignorantes colitis, hoc ego annuntio vobis. Three trimmed words, no sense change.
• 0460B · Nova quaedam infers, — Act 17:20 nova enim quaedam infers auribus nostris, and it is glossed under Migne's VERS. 22, 23. A trimmed connective; the verse displacement is §M3.
• 0460B · vivimus, movemur et sumus, — Act 17:28 vivimus, et movemur, et sumus. Trimmed conjunctions in a re-quotation whose own lemma (In ipso enim vivimus) agrees exactly.
• 0460A · Omnis qui se regem facit, contradicit Caesari — Jo 19:12 Omnis enim qui se regem facit. One connective. ⚑ The citation on it is misprinted — §P1.
• 0460C · Viri Atheniensis, — Act 17:22 Viri Athenienses. A morphological form, not a reading. §W4 gives the whole argument and why no marker fires.
• 0461C · Sceno factoriae artis. — Act 18:3 scenofactoriae artis: the same word, divided across a space. That is §W1's plate candidate, not a textual variant.
• 0461D · Apprehendentes omnes Graeci Sosthenem, — Act 18:17 Apprehendentes autem omnes Sosthenem. Migne announces it himself (In Graeco), so it is §V3's class.
• 0462B · Ego baptizo in aqua, ille baptizabit in Spiritu sancto. — a conflation: Ego baptizo in aqua is Jo 1:26 verbatim; ille … baptizabit in Spiritu sancto is the sense of Jo 1:33 / Mc 1:8. Unattributed (no [n:]), so there is no single host verse for a marker to name. Recorded here instead, exactly as the peer did with Gal 3:12/Lv 18:5 at 0443B.
• 0462C · Judicia Dei abyssus multa — Ps 35:7 judicia tua abyssus multa. Roman, not italic, so it is not a lemma and Pattern 14 does not reach it (peer's §1.5 class).
• 0460B · Si vobis seminavimus, etc. — 1 Cor 9:11 Si nos vobis spiritualia seminavimus. Roman, same class.
• 0463A · Capite nobis vulpeculas quae devastant vineas [n: (Cant. II)] — Ct 2:15 Capite nobis vulpes parvulas quae demoliuntur vineas. ⚠ The one I came closest to marking, because two content words differ rather than a particle. Declined: vulpecula is the ordinary diminutive of vulpes parvula and devastant/demoliuntur are the same act, so the proof-text proves exactly what it proves in the Clementine — the guardians of the vineyard catching the foxes. Nothing the gloss argues turns on either word. Named here so a reader can disagree with me from the evidence.
• 0463A · Multi autem ex eis qui curiosa fuerant sectati. — Act 19:19 qui fuerant curiosa sectati*. A transposition.
• 0463C · Vox facta est una omnium — Act 19:34 vox facta una est omnium. A transposition of the copula. ⚑ The numeral in the same lemma agrees: quasi per horas duas is Act 19:34's own reading.
• 0464C · Huic autem erant filiae. — Act 21:9 Huic autem erant quatuor filiae virgines prophetantes. A clipped lemma that drops a number. ⚠ Checked with care because a dropped numeral is the prize this brief names: the gloss does not reason from the four. It argues only whether these were the daughters of Philip the apostle or of Philip the deacon.
• *0464D · Cum autem septem dies, — Act 21:27 Dum* autem septem dies consummarentur. Cum for Dum, and the verb clipped — but the gloss's first words are Nondum consummati erant, which shows he had the clipped verb in view. Same sense.
• 0465B · Aemulator legis Dei existentis — announced In Graeco, §V3's class.
• 0465B · Testimonium illis perhibeo, quod aemulationem Dei habent [n: (Rom. X)] — Rom 10:2 Testimonium enim perhibeo illis quod. A transposition and a trimmed connective; the citation itself is correct and was checked (§P3).
• 0466C · Tremefactus, etc., et sperans. — tremefactus is Act 24:25, et sperans is Act 24:26 (simul et sperans quod pecunia ei daretur a Paulo). Both agree with the Clementine; the verse address is §M3.
• 0467D · Ut autem judicatum est eum navigare, — Act 27:1 Ut autem judicatum est navigare eum in Italiam. A transposition, and the gloss's whole point is the Greek's nos against the Latin's eum, which he states himself.
• 0468B · Eramus universae animae. — Act 27:37 Eramus vero universae animae in navi ducentae septuaginta sex. ⭐ The shipwreck's number is clipped away, and the gloss does not reason from it — it reasons from animae against viri (Non ait viri, quia fortasse erant et mulieres). A checked zero on the numeral the brief singled out. See §N.
• 0468D · Salvavit eum. — Act 28:8 agrees exactly; recorded here only because the neighbouring Dysenteria vel dysin is §W3's candidate.

V3 · Readings MIGNE HIMSELF ATTRIBUTES to another witness — no marker, by construction

Pattern 14 is for a quotation that diverges without saying so. These announce the divergence in their own words, so a [var:] would tell the reader twice and blur whose observation it was — the same logic as Migne's own ( sic ), and the same treatment the opening stint gave its four.

• 0460A in Graeco: Jasonem et quosdam alios fratres* — Act 17:6 quosdam fratres. The gloss's conclusion (Unde intelligendum est Jasonem fratrem et fidelem fuisse) rests on the Greek's alios*, which is exactly why he cites it.
• 0460D In Graeco: ex uno sanguine — Act 17:26 ex uno omne genus hominum, no sanguine*. He is reporting the Greek (ἐξ ἑνὸς αἵματος) and glosses it as such.
• 0461D In Graeco: Apprehendentes omnes Graeci Sosthenem.
• 0465B In Graeco: Aemulator legis Dei existentis. Rendered "A zealot of the law of the God that is," — see §R2 for the parse and its rival.
• 0465C Quaedam exemplaria habent: Saul, Saul, against his own lemma Saule, Saule. See §M2: both forms are kept in the English*, because the gloss exists to distinguish them.
• 0461D / 0462A Alia translatio: / Alia editio: — 0462A Profuit multum credentibus (against Act 18:27 contulit multum his qui crediderant), 0467D ventus tempestuosus for Typhonicus, 0465D Ego multa summa's fuller edition, 0467B Occidendisque detulit sententiam, 0466C servus "as is had in other codices". Six announced variants, none marked.

V4 · Lemmata checked and found to AGREE with the Clementine exactly

Named, because the brief requires it, and each fetched at its own verse.

Act 17 (0014): Et de colentibus gentilibusque · Trahebant Jasonem · Regem alium · A Paulo · Profecti · Epicurei · Quid vult seminiverbius · Alii vero · Ad Areopagum · Ignoto Deo (both occurrences) · Deus, qui fecit mundum · Ex uno omne genus hominum · In ipso enim vivimus · Sicut et quidam vestrorum poetarum · Ipsius enim · Genus ergo · non debemus · Despiciens Deus · Fidem praebens omnibus · Dionysius. Act 18 (0014): Eo quod praecepisset Claudius discedere omnes · Sanguis vester · Titi Justi, colentis Deum · Si vero quaestiones sunt de verbo · et nominibus · Qui sibi totonderat · Caesaream · Contulit multum. Also inline Rom 11:36 Ex ipso, et per ipsum, et in ipso sunt omnia (verbatim — the citation on it is misprinted, §P2) and 1 Cor 3:6 Ego plantavi, Apollo rigavit (verbatim).

Act 19 (0015): Joannes baptizavit · Erant autem omnes · Quotidie disputans in schola · In schola tyranni · Per biennium · Semicinctia · Exorcistis · Septem filii · His autem expletis · Romam videre · In Macedoniam · Magna Diana Ephesiorum (both) · Alii autem aliud clamabant · Vox facta … quasi per horas duas clamantium (but for the copula, §V2) · Neque sacrilegos, neque blasphemantes deam vestram · Et cum haec dixisset. Act 20 (0015): Ad Graeciam · Una autem sabbati · Eutichus (Clem. Eutychus, orthographic) · Cum descendisset Paulus, incubuit · Anima enim ipsius in ipso est · Pentecosten · Memoria retinentes · Beatius est. Also inline Mt 9:24 Non est mortua puella, sed dormit (verbatim but for a trimmed enim).

Act 21 (0016): Cum autem · Coum, Rhodum, inde Pataram · Philippi evangelistae · Haec dicit Spiritus sanctus · Dicens non debere · Neque secundum · Quia quae de te · Et cum venisset ad gradus · Graece nosti? (his ? included) · A Tarso. Also Ex 20:17 Non concupisces. Act 22 (0016): Natus in Tarso · Aemulator legis · Saule, Saule · Audierunt · Et ibi · Saule frater, respice · Praeordinavit te · Fieri me in stupore mentis · Consentiebam · Ego multa summa · Civilitatem. Act 23 (0016): Percutiet te · Nesciebam, fratres · Quid si spiritus? · Ne discerperetur Paulus ab ipsis · Devoverunt se dicentes · Parate milites ducentos. Act 24 (0016): Patri et Deo meo · Nisi de una hac · De suis ministrare · Tremefactus · et sperans (at 24:26, §M3) · Reliquit Paulum vinctum. Act 25 (0016): Festus ergo cum venisset in provinciam · De quibus · Sua superstitione · Cum multa ambitione. Act 26 (0016): Quid incredibile? · Detuli sententiam · Puniens eos · et amplius insaniens · Contra stimulum · Insanis, Paule · Multae te · Ad insaniam. Also Ps 39:5 Non respexit in vanitates et insanias falsas — verbatim, and the citation (Psal. XXXIX) is correct, which is worth saying in a book with three misprinted ones.

Act 27 (0017): Adrumetinam · Venimus Lystram (⚑ checked with care: the Clementine also reads Lystram, not Myra — the gloss's Pro Lystra, in Graeco Smyrna is a note on the Greek, not a divergence from the Latin) · Jejunium · Portum · Ad Africum · Asson · Typhonicus · Euroaquilo · Syrtim · Oportebat quidem · In insulam autem quamdam oportet nos devenire (verbatim, in full) · In Adria · Sumere cibum. Act 28 (0017): Pyra · Dysenteria · Salvavit eum · Puteolos · Audissent fratres · Conductu · Sine prohibitione.

§P · MIGNE'S CITATIONS — two misprinted, both left EXACTLY as printed

Identified from the quoted words, never from the numeral, per the standing rule. Neither is touched in the text; both are referred to citation-corrections.json here. refDisplay keeps what Migne set; only refKey resolves. No Ibid. chain depends on either (runbook false positive 2 checked: there is no Ibid. anywhere in my four chunks).

1. 0460A — [n: (Joan. XVIII)] on Omnis qui se regem facit, contradicit Caesari. The words are Jo 19:12, verbatim but for a trimmed enim. John XVIII is the arrest and the first interrogation and contains no such sentence; the words stand in the trial before Pilate, one chapter later. → refKey John.19.12.
2. 0460D — [n: (Rom. II)] on Ex ipso, et per ipsum, et in ipso sunt omnia. The words are Rom 11:36, verbatim. Romans II is the judgment of the Gentiles and the inward Jew. → refKey Rom.11.36.

⛔ Neither is a plate candidate. The peers found the same class twice each in this book ((Matth. XXVIII) for Mc 16:19, (Zach. V) for Zach 9:11, (Matth. VI) for Mt 5:10, (II Tim. II, 9) for 2 Tim 2:19) and every one read at the plate this session proved Migne's own. An abbreviated book-name or numeral in an [n:] is exactly the class where the plate goes his way. Four in the book, and mine make six.

1. Checked and CORRECT, recorded so nobody re-raises them: (Rom. X) on Rom 10:2 ✓ · (Psal. XXXIX) on Ps 39:5 ✓ · (I Cor. III) on 1 Cor 3:6 ✓ · (Cant. II) on Ct 2:15 ✓.

One reference in Migne's PROSE, not in an [n:], that is loose — the author's own sentence, rendered as written, no apparatus: 0461C, Non fuit hic Titus discipulus Pauli … cum quo et Barnabas ascendens Jerosolymam, contulit Evangelium cum apostolis. The episode is Gal 2:1–2, where Paul, not Barnabas, is the subject and Titus is taken along. Migne makes Barnabas the nominative. Rendered as printed; see §S for the unexpressed subject that follows from it.

§W · PLATE CANDIDATES — RAISED, NEVER PATCHED, NO MARKER FIRED

Nothing in this section carries a marker, so my four chunks assert nothing about the plate. Addendum A's split governs: a non-word is worth raising and has been our digitization's error ten times out of ten; a real, well-formed word that reads oddly is MIGNE'S until a plate says otherwise, refuted five of five.

W1–W3 · Three NON-WORDS, all rendered for their evident sense per the launch ruling

⚠ None is patchable from the file; a data/tei-patches/8976.json entry needs the plate. ⚑ 7a⁗ was run on each: none of the three is a verb, a negation or a conjunction, and none governs anything. Nothing is quarantined, because nothing was carried — under the launcher's ruling the English says what Migne evidently meant and the crux says what he printed.

⛔ What a plate read would change, stated so a reader can act on it. If the plate prints the sound word in any of the three, the fix is a tei-patch and the English is already right. If the plate prints the broken run, a [sic:] becomes correct and should be added — and under cruces-0005.md's treatment the run would be carried instead. Either way nothing in the running text asserts anything false today. PL 114 = archive.org patrologiaecurs06saingoog, PDF page = (column + 11) / 2 — cols 0461, 0468 are pp. 236 and 239 (0468 covers both W2 and W3, so one crop settles two of the three).

W4 · Viri Atheniensis (0460C) — a REAL word, no marker, and the reason is Pattern 9

Migne sets VERS. 22, 23.-- Viri Atheniensis, etc. where Act 17:22 reads Viri Athenienses. Atheniensis is a perfectly good form — the corpus has it as a true singular at 7017/0000 (Solon Atheniensis) — so it is a real word in an odd slot, not a non-word, and the standing ruling is that it is Migne's until a plate says otherwise.

⚑ And it takes no defect marker for a second, independent reason: English has no exponent for it. "Men of Athens" and "Athenian men" are both number-neutral in the adjective, so a rendering cannot show the disagreement either way — Pattern 9's gender/number class exactly. Rendered "Men of Athens,", no italic Latin, no crux beyond this line.

W5 · Dysenteria vel dysin (0468D) — kept as printed, and it is a NAME not a word

dysin is offered by Migne as an alternative name for the affliction, alongside Dysenteria, and both are Greek in Latin dress. It is kept as printed and italic (Migne italicizes it), on the same footing as Adrumetis, Bithalassum, Euroaquilo and Asson — transliterated names are carried, not Englished. Not raised as a non-word candidate, because a name has no "evident sense" for the launcher's rendering ruling to recover; if a plate reader is at col. 0468 for W2/W3, a glance at band D would settle whether he prints dysin or dysenteria a second time.

W6 · ⭐ THE SUSPICIOUSLY TIDY FORMS — a checked list, and it came back EMPTY

Brief §4 and addendum §C: the digitization silently corrects as well as corrupts, and the only handle is a form that reads too tidy in a rough neighbourhood. I read the whole line around each of W1–W3 and around all 92 [n:] sites (addendum §C's discipline: examine the neighbours in the same look), and found nothing of that shape in 3,553 words. Recorded as a negative, because the class is only credible when the zeros are reported.

The two forms I looked hardest at and cleared:

• 0467B interrogando in etc., [0467B] interrogando, quasi dicat — a bare gerund after an abridgment tail. It construes ("by way of questioning") and it is the ordinary ablative of manner. Not tidy, just terse.
• 0466A civilitatem — a late but well-attested noun which the gloss then defines at length (Id est socialem inter cives conversationem). The definition is the control: he knew the word.

§D · WHERE THE EVIDENCE POINTS AT OUR FILE RATHER THAN HIS PLATE — one candidate, weighed

The brief invites this reasoning explicitly, so here is the one place it was live, and why I came down the other way rather than declining.

D1 · 0467B, detulit for detuli — the OUR-FILE reading considered and REJECTED

The peer's model case is a divergence declined because Migne's own next clause printed the word correctly, making the dropped letter likely ours. That shape is present here: four lines below the lemma Migne prints Detuli sententiam as his own second lemma, in the first person, and glosses it Judicavi illos occidendos.

It loses, on two counts, and both are on the page.

1. detulit is printed TWICE, not once — in the lemma and again in the Alia editio he quotes (Occidendisque detulit sententiam). A digitization that adds the same letter to the same word twice, eight words apart, and leaves the third occurrence alone, is a much worse account of the page than an editor printing what his copy had.
2. The gloss's own work is to reconcile the two persons. He quotes a third-person text, then paraphrases it in the first (qualiter occiderentur ipse sententiam dedi), then re-lemmatizes the first-person form. That is a discussion of the variant, not a slip: the paraphrase exists precisely because the printed form does not say what Paul says.

So the [var:] fires on the Clementine, which is a true and checkable claim, and nothing in the English or the apparatus claims the plate is defective. ⚑ The rival is named because a reader never sees the reading I rejected: if a plate read shows detuli in the lemma, the fix is a tei-patch and this [var:] comes out with it.

§N · THE NUMBERS — every one checked, and they all hold

The brief named the shipwreck's numerals and place-names as the highest-value target in this range. The result is a checked zero, and here is the whole of it rather than the conclusion.

• 0468B, Act 27:37 — the two hundred and seventy-six. ⭐ Migne clips the number away (Eramus universae animae, no ducentae septuaginta sex) and the gloss does not reason from it: it reasons from animae against viri. There is nothing to diverge and nothing to mark.
• 0462C — the twelve and the hundred and twenty, with the arithmetic stated. in duodecim discipulos, et supra in centum et viginti (qui numerus perficitur duodenario decies ducto). 12 × 10 = 120. ✓ The gloss reasons from the multiplication and the multiplication is right. Both numbers agree with Act 19:7 and Act 1:15.
• 0462C — the hours of Paul's day. Ab hora quinta usque ad horam nonam et decimam … horas operi quinque, alias quinque doctrinae, duas dabat cibo et orationi. 5 + 5 + 2 = 12, and the teaching hours (the fifth to the tenth) are the five the sentence claims. ✓ Internally sound, and Migne himself flags that it rests on nothing (quod nulla tamen auctoritate firmatur).
• 0463A — the seven sons of Sceva ✓ Act 19:14. 0463C — the two hours' shouting ✓ Act 19:34. 0462D — the two years at Ephesus ✓ Act 19:10. 0466B — the two hundred soldiers ✓ Act 23:23. 0464D — the fourfold synod, and four are actually enumerated ✓.
• 0464C, Act 21:9 — the four daughters, clipped out of the lemma and not reasoned from (§V2).
• 0468C — the four virtues of Bede's allegory (sapientiae, fortitudini, temperantiae et justitiae): four named ✓.
• 0468D — the fasts. jejunium septimum … in quo occisus erat Godolias ✓ (Gedaliah is killed in the seventh month, 4 Rg 25:25 / Jer 41:1) and dies jejunii decimi in Januario, quando cognoverunt in Babylone captivi templum esse destructum ✓ (the tenth-month fast, and the news reaching the exiles, Ez 33:21). Both attach to the right month. Nothing to mark.
• 0461B — Josephus's nono anno Claudii for the expulsion of the Jews. Not resolvable from anything on disk, and the gloss does not compute with it. Rendered as printed, raised here.
• ⚠ 0470A — Post passionem Domini vigesimo quinto anno, the LAST gloss in the book. It sits in an abridgment head (etc., usque ad Petrus et Paulus martyrio coronati sunt), so the words that would have reasoned from it are the ones the formula removed. Rendered as printed, and flagged: this is the one numeral in my range I cannot check against anything, and it is the one a reader is most likely to query, since the received chronology puts the two martyrdoms more than thirty years after the passion. No marker: there is no witness to compare it to, and [var:] is for scripture.

Place-names of the voyage, form for form as Migne sets them, none normalized: Adrumetis (not Adrumetum) · Lystram (not Myra — and it is the Clementine's own word) · Smyrna and Myrrham, the two names the gloss reports from the Greek and from Jerome · Asson · Syrtis · Adria · Puteoli · Coum, Rhodum, Pataram → Cos, Rhodes, Patara · Phoenice · Creta. Wind-names kept Latin — Africus, Corus, Euroaquilo, Boreas, aquilo — see §M6.

⚠ Two geographical claims of Migne's that are simply wrong are rendered as printed and marked nowhere, because they are what he asserts and the fourth test's question is whether the English asserts it too: 0461D Caesarea metropolis est Cappadociae (Act 18:22's Caesarea is the maritime one, in Palestine) and 0469A Puteoli locus est ultra Romam (Puteoli is south of Rome, not beyond it). 0464C Cous et Patara insulae sunt is a third: Patara is a Lycian city, not an island. None is a textual defect and none takes a marker.

§M · THE HOUSE CONVENTIONS AS APPLIED

M1 · The mirror rule — its ruled scope, and the two words that tested it

• Ecclesia stands capitalized at all three of its occurrences in my range (0461B rexit Ecclesiam, 0463A custodibus Ecclesiae, 0466B Christi et Ecclesiae) and all three render the Church. There is no lower-case ecclesia in cols 0460A–0470A. Nothing to adjudicate — recorded as a checked negative.
• ⭐ spiritum at 0462C is LOWER-CASE and is mirrored. In quo ostenditur illos tantum … spiritum replere, qui sunt in catholica unitate — against Spiritus sanctus capitalized two lines above in the same paragraph. Rendered "the spirit fills only those … who are in catholic unity". ⚠ It reads oddly and it is deliberate, on cruces-0000.md §V's ruling (which found the same alternation at 0430A and 0427A) and cruces-0005.md §S1's (0444D). My range has exactly this one lower-case site, against ten capitalized Spiritus; a later sweep that capitalizes it would destroy something the plate makes. Also lower-case and mirrored: 0466A–B sadducaeos spiritum non confiteri and Pharisaei spiritum nominant → "a spirit", where English would take the indefinite article anyway.
• apostolus NOT mirrored, following cruces-0005.md §M1 and on fresh evidence of my own: my range prints lower-case apostoli (0460A, Apostoli vero contradicebant is sentence-initial), apostolis (0461C, 0464B), apostolos (0463A) against capitalized Apostolus as Paul's title at 0462A and 0466A (sed inspiratione divina dixit Apostolus). The title is capitalized and rendered "the Apostle"; the plural is not. That is the same distinction the peer drew, reached independently here.
• Mirrored where the case is semantic and his practice is consistent: Propheta (0467C) → the Prophet · Scripturarum (0463D) → the Scriptures · Judaei, Graeci, Christiani, Sadducaei, Pharisaei → capitalized as printed · Testamenta (0463D), Novi Testamenti (0464D) → the Testaments, the New Testament · Evangelio (0460A, 0464B) → the Gospel.
• NOT mirrored where English has no lower-case option: Spiritu sancto → the Holy Spirit, Deum, Dominus, Trinitatis.

M2 · ⭐ Saule, Saule — the ONE lemma left in Latin, and why the convention requires it

VERS. 7.-- Saule, Saule. … Quaedam exemplaria habent: Saul, Saul, ut ostendatur de nomine Saulis Saulum nominatum esse. The gloss exists to distinguish two forms of one name: the declinable Saulus (vocative Saule) against the indeclinable Hebrew Saul. Englishing the lemma collapses both into "Saul, Saul" and the sentence explains nothing.

So the lemma is rendered "Saule, Saule.", the variant "Saul, Saul,", and the conclusion "that it may be shown that Saulus was named from the name of Saul". This is the Song-of-Songs case 2 exactly (where the gloss IS the etymology, an English word sharing no root makes the gloss explain nothing), and it overrides the corpus control the peer ran (Saul 187 : Saulus 1) for this paragraph only — every other Saulus in my range is Saul. ⚠ Declared, not smuggled: if the merge prefers a fully Englished lemma here, the cost is the paragraph.

M3 · The verse addresses, reproduced as printed

• Every VERS. number in my four chunks is in ascending sequence, chapter by chapter: XVII 4·6·7·13·15·18·19·22,23·24·26·28·29·30·31·34 · XVIII 2·3·6·7·15·18·22·27 · XIX 4·7·9·10·12·13·14·19·21·22·28·32·34·37·40 · XX 2·4·7·9·10·16·31·35 · XXI 1·8·9·11·21·24·27·35·37·39 · XXII 3·7·9·10·13·14·17·20·28 · XXIII 3·5·9·10·12·23 · XXIV 14·21·23·25·27 · XXV 1·18·19·23 · XXVI 8·10·14·24 · XXVII 1·2·5·9·12·13·14·17·21·26·27·34·37·41 · XXVIII 2·8·13·15·30·31. No displaced number in this range — a checked negative, since a sibling stint found several.
• VERS. 22, 23.-- at 0460C is a double address, and both lemmata under it are correct (Viri Atheniensis is 17:22, Quod ignorantes colitis is 17:23). The third lemma in that paragraph, Nova quaedam infers, is Act 17:20 — a mid-gloss second lemma reaching back, which convention does not address by its own VERS. Not a mis-numbering; recorded so the merge does not read it as one.
• VERS. 25 at 0466C covers Tremefactus (24:25) and et sperans (24:26), and Migne's own gloss joins them deliberately (Sic junge: tremefactus et sperans). Same class.
• VERS. 4 at 0463D, Sosipater, is Act 20:4 ✓. VERS. 30 at 0470A, Conductu, is Act 28:30 ✓ — and it is the last verse address in the book.
• Every VERS. is well formed (VERS. n.-- ), with none of the malformed VERS 34.-- the peer found at 0443A.

M4 · ⭐⭐ THE SITE THE OPENING STINT PREDICTED — found, mirrored, and the plate has since answered its twin, which changes what I think mine is

⚠ cruces-0000.md §2.4 raised a full stop followed by a lower-case word at 0435B, mirrored it, and recorded that it is "the only . [a-z] in all eighteen chunks of this work except one at 0015." ⭐ That one is here, and this is it:

0462A–B — sicut circumcisio in patriarchis signaculum fidei quam habebant fuit. ita et hoc lavacrum populus poenitens quasi signaculum suae devotionis accepit. Mirrored: "…was a seal of the faith which they had. so also the penitent people received this washing as a seal of their devotion."

⛔⛔ AND THE TWIN HAS NOW BEEN READ AT THE PLATE, AND IT WAS NOT WHAT ANYONE GUESSED. data/tei-patches/8976.json (commit ae9ea49, written this session and deliberately not yet applied because agents were reading src/latin/8976) records the answer at 0435B: Migne prints a COLON there, not a full stop. A colon before a lower-case word is ordinary, so the sentence was never his defect — the mark is OURS, and the same read turned up a second lost colon (become a spaced hyphen) and a normalization (Chistus tidied to Christus) in the same line.

⚑ The consequence for my site, stated as a prediction so it can be checked cheaply: fuit. ita is very probably the identical class — a colon of Migne's that our transcription turned into a full stop — and not a lower-case sentence opening at all. Three things point that way, and none of them is the resemblance alone:

1. A colon is exactly what the sentence wants. sicut circumcisio … fuit : ita et hoc lavacrum … accepit is a sicut … ita correlative, and Migne's colon before the answering member is his ordinary practice — my range prints 22 colons of his in 0014–0015 alone.
2. His habit of continuing lower-case after a mark is documented right here, in M4a below: eight ? [a-z] sites in this work, six of them mine. Lower case after a mark is unremarkable in this book; a full stop before one is the thing that is odd, which is why the odd item is the mark and not the capital.
3. The two sites are the entire population of . [a-z] in eighteen chunks. A defect class with exactly two members, both at a sicut … ita-shaped hinge, in one transcription, is one mechanism, not two coincidences.

So: the English stays mirrored and NOTHING is repaired here — I have no plate, and a translator patching a Latin chunk on a resemblance is precisely what the patch file's own "prove it before patching" rule forbids. What I am handing over is a one-crop question with the answer already half-known: PL 114 p. 237 = cols 461/462 (re-calibrated the same way the existing patch was, page = (column + 11) / 2, and the patch file's own p. 223 = cols 435/436 confirms the map). If the plate prints fuit : ita, the fix is a second entry in data/tei-patches/8976.json:

`` "find": "quam habebant fuit. ita et hoc lavacrum", "replace": "quam habebant fuit : ita et hoc lavacrum", ``

…and my English full stop must become a colon in the same commit, which is a one-character edit to src/english/8976/0015.md. ⚠ Note the ordering constraint the patch file already states: applying a patch means re-chunking, so this belongs with the existing 0435B entry, after the last Acts stint lands — and I am the last Acts stint, so that gate is now open.

⚑ And addendum C says to read the neighbours in the same look. When someone crops col. 0462 for this, the same page carries §W1's Sceno factoriae (col. 0461) and §R1's aurum ex argentum (col. 0461) — three of my open questions on one crop — and the 0435B read is the reason to expect a fourth nobody has flagged: I searched my four chunks for the spaced-hyphen tell that read turned up and there is none (zero - in cols 0460A–0470A), which is a checked negative worth having.

⛔ M4a · AND THE ? [a-z] CASES ARE NOT THE SAME CLASS — I nearly filed them as one

My first draft of this section named a second site, 0469A Christiani Romae habitantes unde? patet Paulum non primo praedicasse Romanis, as a companion anomaly. It is not one, and the control is what shows it.

⚑ A control has to be the same class as the thing it controls. Scanning the work for ? [a-z] returns eight sites, six of them in my own four chunks: 0015 baptizaretur? cum · 0016 te esse? ego · apud vos? etc. · quid dicis? incredibile · 0017 relinquat? nisi · unde? patet, plus two in the peer's range at 0005. Continuing lower-case after a question mark is Migne's ordinary practice in this book, not a defect — and against it, . [a-z] stands at two sites in eighteen chunks, which is what makes that one an anomaly worth a crop.

All eight are mirrored in the English anyway, because Pattern 8 governs the marks and nothing licenses recasting the sentence that follows one. But they are recorded here as a checked NON-defect, so that nobody spends a plate read on them and nobody reads the mirroring as our slip. ⚑ Had I filed them beside the . [a-z] case, I would have manufactured an inconsistency that is not there — the 2026-08-19 addendum's exact failure, and three readers made it in one batch.

M5 · Two mid-italic oddities, reproduced

0460A — Trahebant Jasonem. [n: (BED.)] in Graeco: …. The note sits inside the italic run and the sentence that follows it opens with a lower-case in. Reproduced exactly: "They dragged Jason. [n: (BED.)] in the Greek: …". Same class as M4, smaller — and here too the lower-case is mirrored rather than tidied. 0462D — Semicinctia [n: (RAB.)], with no period before the note where every other lemma in the range has one. Reproduced exactly.

M6 · usque ad — CENSUSED PER OCCURRENCE, and the census is 16 tokens, 15 formula, 1 not

Decided from what stands on either side, never by rule (addendum §D).

The abridgment formula (15), rendered as far as: 0460C · 0460D · 0461C (0014, three) · 0463A ×2 · 0463D ×3 · 0463D/0464A (0015, five) · 0464D · 0465A · 0465B · 0466A · 0467B (0016, five) · 0468B · 0470A (0017, two).

⛔ NOT the formula (1), and it is the trap: 0462C, Ab hora quinta usque ad horam nonam et decimam — an ordinary prepositional phrase in roman, inside a gloss about how Paul divided his day, and the sentence goes on to compute from it. Rendered "From the fifth hour until the ninth and the tenth". ⚑ Confirmed mechanically: grep "as far as" returns 15 across the four English chunks and grep "usque ad" returns 16 across the four Latin ones.

⚑ No verb was ever supplied to complete a formula tail, and each tail was parsed as strictly as ordinary prose (the ferias discipline). Three needed it:

• 0461C …etc., usque ad et a ventis insidiarum verbo et opere defendit. The verb has no expressed object; the abridgment took it. Rendered "and by word and deed protects from the winds of snares" — an intransitive English construction, so nothing is supplied**.
• 0460D …usque ad quia quid non ex ipso de quo dicitur: Ex ipso… omnia? The tail keeps its question mark, and it is the tail's own. See §T3.
• 0470A …usque ad Petrus et Paulus martyrio coronati sunt. A complete clause; the head (Post passionem Domini vigesimo quinto anno) is the fragment. Rendered as it stands, abrupt.

M7 · Lemma expansion — both classes stated, so the contrast is not misread

Per the 2026-08-19 addendum, a control has to be the same class as the thing it controls.

• Clipped lemmata left as legible stumps (nothing elided that English needs; the Latin stands opposite): A Paulo. → "By Paul." · Profecti. → "They departed." · Alii vero. → "But others." · Genus ergo, etc. → "Being therefore, etc." · Ipsius enim, etc. → "For his, etc." · Erant autem omnes. → "And they were all." · Contulit multum. → "He helped much." · De quibus, etc. → "Of whom, etc." · Neque secundum. → "Neither according to." · Oportebat quidem, etc. → "It behoved indeed, etc." · Portum, etc. → "A haven, etc." · Cum autem, etc. → "And when it came to pass, etc."
• Clipped lemmata expanded, because the stump is not English at all: Quia quae de te, etc. → "That the things which they have heard of thee, etc." (a bare relative with no verb) · Multae te. → "Much learning doth turn thee." (a bare adjective and pronoun; the verb supplied is Act 26:24's own convertunt, which is what convention 3 means by "the phrase it stands for") · Ad Areopagum. → "To the Areopagus." · Sine prohibitione. → "Without prohibition."

⛔ Contrasting the two lists manufactures an inconsistency that is not there — the first group has nothing elided that English needs, the second does. Ruth's convention 3 governs both.

M8 · Four glosses where the Latin word had to survive into the English

The Song-of-Songs case 2 (where the gloss IS the etymology, an English word sharing no root makes the gloss explain nothing). Each adds one italic span the Latin does not have, and each is declared here rather than left to look like drift:

• 0460B, sermologus. vel seminiverbius, id est sermologus — a Greek-formed coinage offered as a synonym for seminiverbius. Kept Latin. The lemma itself is rendered: seminiverbius → "word-sower", because the gloss's own definition (qui verbum seminat) carries into English intact and Douay's "babbler" would delete it. ⚑ That is a conformation refused, in miniature.
• 0461B, pagus. Areopagus … nomen trahens a Marte, qui Graece Arios dicitur, pagus, villa. The word being decomposed cannot be its own translation; villa is rendered "a village". So: "…who in Greek is called Arios, pagus, a village."
• 0463C, sacrum. Sacrilegi dicuntur a profana diligentia. Sacrum enim mundum et profanum significat. Kept, because the whole sentence is about the Latin word's double sense.
• 0467A, ambitio / phantasia. Pro ambitione in Graeco ponitur phantasia. Both kept — a statement about which word each language sets.

⚑ And the four places where the Latin did NOT need to survive, recorded so the treatment does not look arbitrary: Typhonicus → "Typhonic." with ventus tempestuosus → "a tempestuous wind" (the second is a description, not an etymology) · Syrtim → "Syrtis" with angustia sive tribulatio Englished · amens/fortunatus (0463D, glossing Eutichus) → "out of his mind" / "fortunate" · caedens/occidens (0465D) → "he that strikes" / "him that kills", because English carries that distinction with no help from us.

M9 · Three lemmata rendered AGAINST the familiar English Bible, deliberately

The 11551 stylus failure in miniature, three times. Each is a place where Douay's word would have made the gloss explain nothing.

1. 0462D, Semicinctia → "Half-girdles", not Douay's "aprons". The gloss's second definition is vel zonae — "or girdles" — and only a rendering that keeps the girdle in the lemma lets the alternative be an alternative.
2. 0467A, Cum multa ambitione → "With much ambition", not Douay's "with great pomp". The gloss defines it as multa turba ambiente eos, "a great crowd going about them", which is the ambi- root doing the work; and pompa is a different word in the same sentence (pompa regalis officii), which "pomp" would have collided with.
3. 0470A, Conductu → kept as "Conductu", not "in his hired lodging". Migne glosses it in libera custodia per quam conducebatur, reading the noun from conducere "to lead" — and "hired" shares nothing with "conducted". The one lemma besides Saule, Saule that stays Latin, and for the same reason.

⚑ Also resisted, and recorded because the familiar English arrived first in my own reading before the collation caught it: 0460B Quid vult seminiverbius wanted Douay's "babbler" (see M8); 0468B Eramus universae animae wanted "we were in all two hundred threescore and sixteen souls", which is a number Migne does not print (§N).

§S · THE UNEXPRESSED SUBJECT — every supplied pronoun, named

The measured dominant defect class. Each site is a "he/she/it/they" the Latin does not print. Where two readings both construe I say so rather than choosing silently.

• 0460A · Apostoli vero contradicebant Christum regem esse Judaeorum → "But the apostles spoke against it, that Christ was king of the Jews". ⚠ Both the object and the verb's force are open, and I did not resolve them. (a) contradico + acc.-inf. normally means "to deny that", which would make the apostles deny Christ's kingship — and the appositive id est confitentium then explains nothing and the closing unde et Judaei ipsum Christum accusabant has nothing to follow from. (b) Migne is playing on the contradicit Caesari he has just quoted, and the apostles are answering back, asserting a kingship over the confessing rather than over the nation — which is what makes the Jews' accusation of Christ follow. I supplied a bare "it" and the neutral "spoke against", which carry (b) without excluding (a).
• 0460D · docet esse → "he teaches" — Paul, from Nota ordinem hujus disputationis.
• 0461A · loquitur, conservat et confirmat, ne nova videatur inferre → "he speaks", "he preserves and confirms", "lest he should seem" — Paul throughout; the paragraph is the Areopagus speech and Sicut et quidam vestrorum poetarum is his sentence.
• 0461A · statuit judicare (0461B) → "he has appointed to judge" — God, from Despiciens Deus two lemmata above.
• 0461C · contulit Evangelium cum apostolis, inquirens … an recte praedicaret → "Barnabas … conferred the Gospel with the apostles, inquiring … whether he preached rightly". ⚠ Migne makes Barnabas the nominative, so contulit is his; but Gal 2:1–2, which he is summarizing, makes Paul the one who goes up and lays out his gospel, and praedicaret has no expressed subject at all. Both construe; I followed the printed nominative and left praedicaret to take it, and did not resolve the second. See §P's closing note.
• 0461D · Nondum enim in Syriam Phoenicis venerat → "he had not yet come" — Paul.
• 0462A · Hanc prius non viderat forsitan (0463B) → "Perhaps he had not seen it" — Paul, and Hanc is Rome, from the lemma Romam videre.
• 0462B · quia in fide Trinitatis → "because it is in the faith of the Trinity". The Latin has no verb; English cannot leave the clause verbless. Named because it is a supply.
• 0463A · ostendit de quibus custodibus Ecclesiae praecipitur → "he shows" — Bede, whose siglum opens the note.
• 0463B · Transire volebat → "He wished to pass" — Paul, from In Macedoniam.
• 0463C · Hoc loquitur quasi Judaeus … aut quasi Christianus locutus est → "He speaks this as a Jew … or he spoke as a Christian" — the town clerk of Act 19:35, who is the subject of Et cum haec dixisset in the lemma. ⚠ Not Paul: the whole point is that the speaker is the Ephesian scriba.
• 0463D · transit autem a Macedonia et Achaia ire Jerusalem → "he passes" — Paul.
• 0464A · redivivus adducitur → "he is brought back alive" — Eutychus, from the lemma.
• 0464B · Non illis … praeponit … sed illos maxime glorificat → "He does not prefer … but he chiefly glorifies" — Paul, whose Beatius est is the lemma; the note's own closing address (Vos igitur episcopi) is a second voice inside the same abridged extract and is left as printed.
• 0464C · meruit vocari → "he deserved to be called" — Philip, the lemma.
• 0465A · Non enim ita reprobas … si autem ideo celebraret … non hoc diceret Jacobus → "thou dost not reject … but if he celebrated … James would not say this". ⚠ The person shifts inside the sentence and it is Migne's: reprobas is second singular (James speaking to Paul, which is the situation of Act 21:24), celebraret third. Rendered with the shift intact; see §T4.
• 0465B · Quo captus cum parentibus commigravit → "Whither, having been taken captive, he migrated with his parents" — Paul.
• 0465D · notat quia non occidit → "he notes that he did not kill" — Paul, the speaker of Consentiebam, in both places.
• 0466A · stipabatur → "she was thronged on every side" (0467A) — Bernice, from illam in the same ablative absolute and from Act 25:23's Agrippa et Bernice. ⚠ illam is accusative, so the passive's subject is genuinely unexpressed; the feminine of illam and ambiente is what fixes it.
• 0466B · quia hoc noverant contrarium esse illis → "because they knew this to be contrary to them" — the Pharisees, the nearest nominative.
• 0466C · dimisit eum liberae custodiae → "he dismissed him" — Felix.
• 0467B · Suscitavit quidem ante mortuos … vel suscitat → "he raised up the dead before … or he raises them up" — God, from si Deus mortuos suscitat in Act 26:8, which is the clause the lemma clips.
• 0468B · quod supra hortabatur eos Paulus — no pronoun supplied; Migne names Paul.
• 0468C · portum salutis consequitur, quaerat → "he attains … he may seek" — the qui of nisi qui pane vitae pascitur, carried through the whole period.
• 0469A · destruxerunt, et aedificia mutilaverunt → "the men of Salerno … destroyed" — Migne names them (Salernitani).
• 0470A · contra Judaeos disputabat → "he disputed" — Paul; the book's last clause.

§R · REAL WORDS THAT READ ODDLY — rendered as printed, every marker declined

Per the standing rule (refuted five of five), a real, well-formed word that reads oddly is Migne's until a plate says otherwise.

R1 · 0461A · aurum ex argentum — the one that came nearest a [cj:], and it is declined

Homo in medio est, habens Deum superiorem, aurum ex argentum, cum caeteris hujusmodi inferiora, quae non debent ei comparari.

ex is a real preposition and it governs the ablative; argentum is accusative, so the phrase cannot construe as printed. The conjecture is et — "gold and silver" — and it is near-certain from the sense (man in the middle, God above, metals below).

⛔ Declined, and here is the bar it fails. Pattern 18a's test is whether the printed word makes the sentence assert something the author is arguing against. aurum ex argentum asserts nothing at all: it fails to construe, which 18a explicitly leaves outside the ruling ("mood, voice, number and person anomalies that merely fail to construe stay declined", 11064 @1154C the named parallel). A preposition governing the wrong case is that class, not the negation-or-correlative class the 2026-08-15 extension admits. Rendered as closely as the printed words allow: "gold from silver, with the rest of this kind as lower". ⚑ The rival, named because a reader never sees what I rejected: the other reading is that our digitization lost the t of et — a tei-patch, not a conjecture — and it loses only because I cannot read the plate. Col. 0461 is page 236, the same crop W1 wants.

R2 · 0465B · Aemulator legis Dei existentis — rendered, marker declined, rival named

existentis is a genitive singular participle and the only genitives near it are legis and Dei, so it must agree with one: "a zealot of the law of the God that is". The rival is that it is a calque of the Greek's nominative ὑπάρχων ("being a zealot for the law of God"), which the Latin's case forbids but which the Greek behind Migne's own In Graeco would support. It loses because the case is printed and the case decides; rendering it as a nominative would be reading through the Latin to a Greek the page does not print. Rendered as printed, no marker.

R3 · Rendered literally, no marker, reasons given

• 0466A, Non enim tribunus … civis esse non posset, sed… A double negative that construes perfectly ("it is not that he could not be"), and both negatives are carried on their own hosts. Not a defect; recorded because a count-based check will flag the pair.
• 0468D, hoc autem non indigebant.* A plural verb with hoc accusative, where indigere* usually takes the genitive or ablative. Attested with the accusative in later Latin; rendered "but these did not need this". Pattern 9 (attested medieval usage), no crux beyond this line.
• 0461D, Nunc nimirum declarato et narrato, id est Evangelio. An ablative absolute with no expressed noun; the antecedent is verbo, carried over from the lemma de verbo four words earlier. Rendered as elliptically as it stands ("Now indeed declared and narrated, that is, the Gospel"), nothing supplied. ⚑ The alternative — supplying "the word" — was rejected as the kind of quiet completion the abridged books invite.
• 0466C, jussit non prohibere quemquam ministrare ei de suis rebus. quemquam is the object of prohibere, not its subject; rendered "he commanded that no one be forbidden to minister to him", which keeps the negation on prohibere where Migne prints it.
• 0463D, ut sit praeparatio virtutis faciendae et dicendae. virtus here is the moral one, so "virtue", per cruces-0000.md §V's per-occurrence rule. (The other three virtus-family words in my range: 0463A signa faciunt is not one; there is no virtus meaning "miracle" in cols 0460A–0470A. A checked negative on the one term this work varies.)

§T · THE TESTS, AND WHAT EACH RETURNED

T1 · A [sic:] must not swallow the word it marks

Vacuously clean: zero [sic:] in the range, and zero [cj:], [d:], [ed:], [nt:]. The English was nevertheless read once with every marker stripped, which is the mechanical form of the test: no stranded article or preposition sitting on punctuation, no doubled mark, every sentence readable aloud. ⚑ The four [var:] were checked against the fourth question 7a⁗ does not ask — does the clause still have the thing it acts on? — and all four sit beside complete renderings (Sosipater, Do not receive gold, he brought the sentence, Bithalassum), none inside one.

T2 + T2a · Every printed negative named to its single Latin host

Eighty-six negative particles across the four chunks (0014 ×18 · 0015 ×18 · 0016 ×41 · 0017 ×9, counted per occurrence and not per line, taking non · nec · neque · nisi · nullus · nihil · ne · nondum · nemo), extracted mechanically with context and then each located in the English and checked to be ON its host, not merely present.

The 2a trigger population — a negative resting on a verb with a quantifier, correlative or predicate adjacent — fired twelve times. Each host was proved from the sentence's own structure, never from what reads well:

⚑ 0469A is the one that would have beaten test 2 on its own. The plausible wrong reading is "Paul did not preach to the Romans at first", which is grammatical, reads better, and asserts something the paragraph denies: the gloss's whole work is to explain how there were Christians at Rome already. The host is proved by the question, which is structure, not sense. ⚑ 0464A is the Esther shape — a negative correctly attached and still liable to land in the wrong place in English word order. "It wholly did not depart" and "it did not wholly depart" are both English and only one of them is what Anima enim ipsius in ipso est says.

T2b · Every anchor checked against the WORD it stands beside

All 39 anchors (9 + 9 + 13 + 8) were printed with their Latin context and the matching English context and compared one by one. Zero displaced. Seven are deliberate mid-construction splits matching Migne's, and one was repaired during drafting:

• 0460C vel Judaei sunt [0460C] qui hoc dicunt → "or it is the Jews [0460C] who say this".
• 0460D unum Deum auctorem [0460D] mundi omnium → "one God is the author [0460D] of the world, of all things".
• 0461B fidem praebens omnibus quod [0461B] resurgent → "giving faith to all that [0461B] they shall rise again".
• 0462C Judicia [0462C] Dei abyssus multa → "The judgments [0462C] of God are a great deep".
• 0465D audit ex ore [0465D] ipsius → "hears from his own [0465D] mouth". ⚠ English reverses the two words, so the anchor sits inside the same two-word construction but on the other side of it. Named because it is the only one where the order could not be preserved.
• 0466B spiritum non confiteri [0466B] et angelos → "do not confess a spirit [0466B] and angels".
• 0469A contra [0469A] quam, scilicet valerent aegritudinem → "against [0469A] which sickness, namely, they availed".
• ⛔ 0461A, repaired. Migne splits loquitur [0461A] testimonium. My first draft had fronted "a testimony" and left the anchor eleven words downstream, at versum Arati ponens — the 8995 failure shape in miniature. The sentence was restructured so the anchor stands where he puts it: "he speaks the [0461A] testimony, not of Moses…". Found by this test and by nothing else.

T3 · Punctuation compared BY COLUMN BAND, never by total

A band-partitioning script compared colons, semicolons, question marks, exclamations, guillemets and parentheses across all 41 anchor-delimited bands. ⛔ First pass returned ten mismatched bands, and every one was mine, and nine of the ten were a supplied semicolon — precisely the class the brief names.

After repair: 41 bands, and every band matches on every mark counted — with ONE exception, which is deliberate.

⚑ 0015, band [0462B]: one colon in the English that Migne does not print, and it is Pattern 8b. Quaestio est, utrum ille qui per ignorantiam forte a non baptizato … debeat rebaptizari? is an indirect question closed with ?, and English cannot end that construction with a mark. Recast direct behind a colon: "The question is: ought he who through ignorance has perchance been baptized by one not baptized, but nevertheless of right faith, to be rebaptized?" The mark is kept, nothing is deleted, and the extra colon is the convention's own cost. ⛔ This is the only band mismatch in the stint, it is intentional, and the merge should not "fix" it. (The peer declared the identical thing at 0443C; two 8b recasts now stand in this book.)

⛔ No comma→semicolon sweep was performed. Migne prints 33 semicolons of his own in my range (0014 ×9 · 0015 ×8 · 0016 ×10 · 0017 ×6), so Tobit's narrow exception — which rests on a book printing zero — does not apply and was not invoked.

Question-mark parity: 12 in the Latin, 12 in the English (0014 ×2 · 0015 ×2 · 0016 ×6 · 0017 ×2), each on the sentence Migne puts it on. Five are worth naming because they look like errors and are not:

• 0460D …usque ad quia quid non ex ipso de quo dicitur: Ex ipso… sunt omnia? — a ? closing an abridgment tail. Kept exactly there.
• 0461D fuere tamen Paulus vel Aquila; vel timore Judaeorum, vel aviditate convertendi eos? — a ? on what reads as a statement with alternatives. Kept (Pattern 8).
• 0465B Graece nosti? and 0466A Quid si spiritus? and 0467B Quid incredibile? — question marks inside italic lemmata, carried into the English lemmata unchanged.
• 0465D tam facile dicis civem Romanum te esse? — a direct question, and Pattern 17 makes it singular ("dost thou so easily say that thou art").

T4 · The fourth test — does the paragraph ASSERT what he asserts

Run last, with the counts already clean, on all 134 paragraphs. It is what produced §V1's four divergences, §M9's three refused conformations, §M2's Saule, Saule decision and §R1's decline. Four paragraphs were read twice under it and the reasoning is recorded, because a blind reader will stop at each:

1. ⛔ 0464B, Act 20:35 — repaired. My draft rendered Nolite accipere aurum as "Do not possess gold", from the familiar Mt 10:9 and from this work's own 0434B. It read perfectly and it was false: the paragraph proves beatius est dare quam accipere, and accipere is the word it proves it with. Repaired to "Do not receive gold" and marked. Nothing was missing, no count changed, the band matched. This is the conformation the brief warned would arrive over the plate, and it did.
2. ⛔ 0463C, Act 19:40 — the speaker. Hoc loquitur quasi Judaeus, nolens Judaeos ab idolorum cultoribus occidi. My first reading took the subject to be Paul; it is the town clerk of Act 19:35, who is the subject of the lemma Et cum haec dixisset and the only person in the chapter who could speak "as a Jew … or as a Christian" about protecting the Jews. Repaired before filing; §S records it.
3. 0465A, Act 21:24 — the person shift, left standing. Non enim ita reprobas legalia … si autem ideo celebraret, ut actione simulata suam occultaret sententiam, non hoc diceret Jacobus. Second person to third inside one sentence. It asserts what he asserts: James is addressing Paul (reprobas) and then speaking of him in the third person as the subject of James's own hypothetical. Rendered with the shift intact, no marker, no repair.
4. 0468C, Act 27:34 — a long period that reads smoothly, and it is Bede's. Nemo saeculi tempestates evadit, nisi qui pane vitae pascitur, et qui … innititur, auxilio Domini coruscante portum salutis consequitur, ita ut saeculo expeditus, flammam dilectionis … quaerat. ⚑ A sentence that reads smoothly where the Latin is rough is a suspect, not a success, so it was re-parsed: consequitur is the main verb of the second qui-clause, not a third parallel, and auxilio Domini coruscante is an ablative absolute, not an instrument of innititur. The English reproduces that structure. Left as written.

⚑ Two further smooth-where-rough suspects were re-examined and left rough: 0461D's verbless ablative absolute (§R3) and 0470A's headless abridgment (§M6).

§Z · WHAT I DID NOT DO, AND WHY

1. No [sic:]. The three non-words (§W1–W3) are rendered, per the launcher's ruling on the peers' split, and a [sic:] on any of them would publicly blame Migne's plate for what addendum A says is our digitization's error ten times out of ten. Ship test B.
2. No [cj:]. One candidate, aurum ex argentum at 0461A, declined with the bar stated and the rival named (§R1). A preposition governing the wrong case is 18a's declined class.
3. No [ed:]. Nothing in my four chunks is a hole in the digitization. The three non-words are corruptions of text, not losses of text, and Pattern 13 is for the second.
4. No [nt:] — and I did not force one. All 92 notes in my range are attribution sigla ((RAB.) ×37, (ID.) ×26, (BED.) ×19, (AUG.) ×2, (CHRYS.) ×2) or scripture locators ((Joan. XVIII), (Rom. II), (I Cor. III), (Cant. II), (Rom. X), (Psal. XXXIX)). Not one is editorial prose, judged by content and not by length; the longest is thirteen characters. Pattern 15 never fires, exactly as the peer found for its 115.
5. No [f:] — runbook false positive 4. Lemma-and-gloss commentary, not a florilegium.
6. No [d:] — no run is set twice anywhere in the four chunks. Ignoto Deo appears twice at 0460C but they are lemma and re-quotation, twenty words apart, with usque ad between them; Magna Diana Ephesiorum twice at 0463B–C in two different verses' lemmata. Neither is a compositor's doubling, and Pattern 11a's "which copy is the intrusion" question does not arise because neither copy is intrusive.
7. No [cn:], and none is owed. PL 114's spot check for this work is a recorded checked zero — no numbered Forte apparatus, both mid-page work divisions examined, and one of them is the Acts→Romans division at cols 469/470, which is inside my range and is the reason the zero covers this tail. That zero licenses shipping; it does not license the claim that the work has no notes, and I make none.
8. No plate was read by me, and none blocks the range. ⭐ But one of my questions is now half-answered by a read someone else already did — the 0435B crop behind data/tei-patches/8976.json — and §M4 turns that into a prediction with a written patch. If the merge will spend crops, spend two, in this order: (a) cols 461/462, p. 237 — the highest-yield page in the range, and it closes three entries at once: §M4's fuit. ita (predicted to be a lost colon, exactly as 0435B was), §W1's Sceno factoriae, and §R1's aurum ex argentum. (b) col. 0468 band B/D, p. 239 — settles §W2 Tyrrhenc and §W3 Thelasson together, with §W5's dysin on the same page. Two crops close six entries. ⚠ And the patch file's own ordering gate is now open: it says the 0435B patch is held back until the last Acts stint lands, because applying it re-chunks src/latin/8976 under any agent still reading. I am the last Acts stint and I am done reading.
9. Column-band arithmetic not reported — runbook false positive 1. My bands skip letters freely (0466D → 0467A → 0467B → 0467C, then 0467D → 0468A) and 0469A → 0470A jumps a whole column at the work division. Normal, and the 469/470 jump is the book ending.
10. (cont.) handling: not applicable. No ## head repeats across my chunk boundaries — 0014 ends inside CAPUT XVIII and 0015 opens CAPUT XIX, 0015 ends inside CAPUT XX and 0016 opens CAPUT XXI, 0016 ends inside CAPUT XXVI and 0017 opens CAPUT XXVII. Checked, not assumed.
11. I did not write cruces.md, and I touched no file below 0014.

§Q · READ CLEAN — the sections where I found nothing

Named, because a findings list without its negatives is not a report.

• Chunk 0017, cols 0468D–0470A (CAPUT XXVIII, the last chapter of the book) — six verse addresses, thirteen lemmata, eight notes. The only entries anywhere in it are §N's vigesimo quinto anno, §W5's dysin, §M4's unde? patet, and §N's note that Puteoli is not beyond Rome. The lemmata all agree with the Clementine.
• CAPUT XXIII and CAPUT XXV entire (0466A–0467A) — eleven lemmata, all agreeing; two supplied pronouns, both named in §S; no divergence, no candidate, no marker.
• Ecclesia: three occurrences, all capitalized, all rendered "the Church." Checked at the source, not taken on trust.
• Guillemets: zero in both languages, all four chunks. Patterns 5 and 6 have nothing to govern.
• Migne's own ( sic ): none in this range, so nothing to stack ours on (runbook false positive 5).
• Pattern 9 items met and left alone: coelum/coelis for caelum throughout · tanquam · nunquam/Nunquid · Jerosolymam · sylva-class spellings · Eutichus for Eutychus · Ismaelites-class name forms. None marked, none italicized, per runbook false positive 6.

All 18 chunks read against the Latin twin, fenced from this file until findings were formed. Every printed negation in cols 0425C–0470A is carried — roughly 200 sites across non, nec/neque, ne, nihil, nemo, nullus, nunquam, nisi, nondum, the non solum … sed etiam frames, and three double negatives (0446D Non quod ignoret Petrus eum poenitentiam non acturum, 0466A Non … tribunus … civis esse non posset, 0448C nisi prius excaecatus). Not one dropped, not one supplied. ⭐ Zero polarity defects. Chunks read clean: 0000–0007, 0009–0017.

The site — ours, and it is a relative, not a negation

⭐ 0008 · 0447C · ut ex hoc Gentibus is praedicetur quem prius nesciebant — was "that from this he might be preached to the Gentiles whom before they knew not". ⛔ quem is masculine accusative singular and cannot agree with Gentibus (dat./abl. pl.); its only possible antecedent is is, Christ. English binds a relative to the nearest noun, so our page said the Gentiles were the unknown party — which is also incoherent, since the sentence is about Philip opening the Church's mouth so that Christ may be preached to them. Now "that from this he whom before they knew not might be preached to the Gentiles." ⚑ Same class as 8964 @0648C, found earlier in this batch: a relative that construes one way in Latin by morphology and another way in English by proximity.

The strongest near-miss in the book, checked and CORRECT

0439A · qui nec territi minis cessant praedicare → "who, terrified by threats, cease not to preach." Read first as a dropped nec. ⭐ It is not. If nec attached to territi, cessant praedicare would stand affirmative and say the apostles stop preaching — the opposite of what the sentence praises them for. The nec negates cessant, and the English keeps it on "cease not"; territi minis is concessive. Check the rule before acting on the fact.

Candidates raised blind and found ALREADY LOGGED — the file's coverage held

Each of these was reached independently before this file was opened, and each was already recorded with its reasoning: 0448B Dominus jubetur (§R2), 0452D domum for donum (§3.5 — which names the same control I found, Ita enim datur sicut donum Dei), 0459A Lydda for Lydia (§2.4), 0429D Matthaeus for Matthias (§205), and all five non-words — posili, ite rum, credeban, quisbusdam carried, Mundacavi normalized (§2.1, §1035–1077). ⚑ The differential treatment of the five reads as inconsistent from the page and is reasoned in the file; nothing to reopen, but it is the one thing in this work a reader will notice unaided.

Candidates raised and REJECTED, with reasons

• 0432D Non iterum movear — "Let me not again be moved" reads as a wish beside indicative neighbours (Scio me semper esse victurum, iterum non moriar), but movear is subjunctive in form and sits naturally as the purpose of the lemma Non derelinques. Underdetermined; left.
• 0460A Apostoli vero contradicebant Christum regem esse Judaeorum — contradico + acc.-inf. normally means "deny that", which would have the apostles denying Christ's kingship; but the gloss then redefines Judaei as confitentes, which makes a political denial coherent. The Latin is genuinely ambiguous and the English preserves the same ambiguity rather than resolving it wrongly.
• 0454A quia princeps erat, non ipse Herodes, sicut et Jacobus — the appositive may belong to the killing rather than to princeps. Compressed past deciding; left as printed.
• 0461A aurum ex argentum (for et) and 0461B episcopus Corinthiorum (against the Areopagite's own surname, which the next clause derives from where he presided) — both Migne's, both carried, both worth a line in §2 if the file is revised; neither is ours.

The fourth stint predicted a second instance of the 0435B class here — fuit. ita at a sicut … ita hinge, which is exactly where a colon belongs — wrote the patch text, and did not patch. ✅ Read.

PL 114 p. 236 (patrologiaecurs06saingoog, PDF page = (column + 11) / 2, corner numbers re-read before the crop: 461 and 462), rendered at 600 dpi. The plate prints:

… signaculum fidei quam habebant fuit : ita et hoc lavacrum …

⭐ A COLON. So this sentence, like 0435B, was never Migne's defect — the mark is ours, and the English's apparent mid-sentence full stop before a lower-case "so" went with it. Both are repaired: the TEI patch restores his colon (data/tei-patches/8976.json, third entry) and the English now reads "…was a seal of the faith which they had: so also the penitent people…". ⚑ Unlike the two 0435B patches, the TEI does not break this line, so the find string carries no newline.

⭐⭐ The prediction was made from the file alone and proved right at the plate. That is worth recording as a method result: a stint that flags a punctuation anomaly at a named rhetorical hinge is making a testable claim, and this one tested true. The class now stands at two confirmed in this book, both ours, both invisible to every automated check.