Liber Judith

Translated 2026-08-18, one translator, 3 chunks (2,959 Latin words), on Opus. verify-english.mjs clean: columns, notes and sections aligned, no duplicate paragraphs.

Apparatus: 9 [var:] · 7 [sic:] · 3 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 35 [n:] (9 / 9 / 17, exact parity with the Latin twins). Structure: PROLOGUS + PRAEFATIO + CAPUT PRIMUM → CAPUT XVI, eighteen ## heads, 1:1 with the Latin. Guillemets: 91 « / 89 », 1:1 with the Latin. Two quotations are UNCLOSED in Migne and one closing mark is ORPHANED (no opener); all three stay as printed (§3). Italics: 6 spans in chunk 0 (four usque ad + the two italic lemmata of CAPUT IV), 1 in chunk 1, 0 in chunk 2 — 1:1. Question marks: ZERO in the Latin, zero in the English. Column anchors: 12 / 9 / 9, verbatim and in order.

Tobit (8970) §1 governs this book unchanged and was not re-opened: guillemet lemmata are Englished inside marks that are the plate's; usque ad → as far as with no verb ever supplied; notes and column anchors sit inside guillemet spans where Migne puts them; sigla keep his internal spacing. What Tobit did not cover is in §7.

§1 · PLATE STATUS — READ, AND A CHECKED ZERO

Foot-of-page apparatus for cols 0731–0740: a CHECKED ZERO. Read by the orchestrator during this session, not by the translator, and merged here. Method: archive.org patrologiaecurs04migngoog (on disk at raw/scans/pl113/), PDF page = (column + 11) / 2, verified against the running head (p. 372 = cols 733/734, head GLOSSA ORDINARIA. — LIB. JUDITH.). Pages 371, 372, 373, 374 and 375 read at full page width across the whole foot. No foot rule, no numbered notes — the columns simply end.

⚑ Both of this work's boundaries fall MID-PAGE and both were read there — p. 371, where Tobit ends and LIBER JUDITH. begins below the rule, and p. 375, where Judith ends and LIBER ESTHER. begins below the rule. That is the case that lost three notes on PL 202.

Tier 3, coverage full, notes_recovered: 0. Third sampling of PL 113, so the volume's "no conjecture apparatus" finding now rests on fourteen pages across three books. Consequence for this file: the three [cj:] below are MINE, not conjectures Migne had already answered.

Corroborations supplied from the plate (not a substitute for the collation below, and each was independently checked against the twin): cols 733/734 confirm « et vermis eorum non morietur, et ignis non exstinguetur. » and VERS. 12 « Tunc ad Oziam congregati. »; cols 737/738 confirm « Confidit in ea cor viri sui, et spoliis non indigebit » and VERS. 14 « Porro autem. »; cols 739/740 confirm CAPUT XVI's VERS. 15 « Juvenibus. » Quia tunc neque nubent neque nubentur, the doubled neque … neque standing, and the closing VERS. 21 « In carnes eorum. running into the column where Esther begins.

§2 · THE DIGITIZATION'S OWN DEFECTS — logged, not silently repaired

2.1 ⚠ The column anchors jump 0731D → 0732D. Reproduced as given; NOT renumbered.

Chunk 0's anchor sequence is [0731D] [0732D] [0733A] …. Columns 0732A, 0732B and 0732C carry no anchor in our TEI, although the running text passes straight through them. This is either Migne's own setting or (far more likely) a loss in Corpus Corporum's transform. Nothing was supplied and nothing renumbered — hard rule 1, and the brief's "Migne's own printing is not yours to tidy". Flagged for the tei-patch queue rather than fixed here, because supplying an anchor invents a citation address (migne.app/pl/113/732a) that nothing on the plate warrants. ⚑ This is the one thing in the work I most want a plate reader to settle, because it is an address, not a word.

2.2 ## CAPUT VII prints with NO closing period — in the body AND the frontmatter

Every other head in the book carries one (CAPUT PRIMUM., CAPUT II. … CAPUT XVI.). Both the heads: field and the ## line lack it, so the two sides agree and there is no internal evidence of loss. The English head therefore reads ## CHAPTER VII without a period, and the frontmatter is copied verbatim as always.

⚠ This is the same shape as Tobit's CAPUT XIV, where the plate turned out to print the period and our TEI had dropped it. I could not read the plate myself and declined to assume the same answer: importing Tobit's finding into this book is exactly the "a rendering found in another BOOK is a hint, never an authority" failure. One glance at p. 373 closes it; if the plate prints CAPUT VII., this becomes a two-entry tei-patch (head + frontmatter) and the English head takes the period.

2.3 VERS. 19.--Et ecce, » etc. at 0738A — a missing space AND a missing opener

Chunk 2 prints VERS. 19.--Et ecce, » etc. where every other address in the work prints VERS. n.-- « …. Two defects at once: no space after the --, and no opening guillemet for a span that is duly closed. Reproduced exactly, VERS. 19.--And behold, » etc. — the orphan » is why the counts are 91/89 rather than 91/90 (§3). Same class as Ruth's VERS.9 missing-space patch.

§3 · UNCLOSED AND ORPHANED GUILLEMETS — Migne's, and they stay

Pattern 5's corollary. Three sites, and they are why the marks do not balance.

⚑ The 0733B case is the one a span-extractor gets wrong, and it did: the lemma-list file supplied for this work carries 86 spans against 91 opens, and Tunc Holofernes. is one of the five it could not see. The lemma is Tunc Holofernes. (Jdt 2:7); everything after the siglum is gloss. Same for Centum quinque, etc. (Jdt 16:28) at 0740A and for the Ascendam in coelum quotation at 0733A–B, both of which are in §4 below although the supplied list does not name them.

⚠ The 0737C site is the worst-behaved, because the unclosed « runs across a VERS. address, and the same verse (1 Pet 5:8) is quoted again eleven lines later at 0738A — that second time properly closed. Both are rendered; neither was balanced against the other.

§4 · THE LEMMA LIST — ALL 91 SPANS, CHECKED AGAINST THE VULGATE

Every guillemet and italic span in the work, with what the check found, including the ones that agree exactly. Judith's Latin is as unstable as Tobit's and this is where the value is. Divergences are grouped first; the clean ones are named, not summarized.

4a · THE DIVERGENCES — nine [var:] fired

4b · DIVERGENCES FOUND, MARKER DECLINED — with the reason, because a blind reader will look

• *0733A · « Omnes uno animo contradixerunt, etc. » — Jdt 1:12 has Sed* omnes. A dropped opening connective inside an usque ad resumption. No marker.
• 0733D and 0740A · vermis eorum non morietur, et ignis non exstinguetur (Isa 66:24). The Vulgate has et ignis eorum non extinguetur. Migne drops the second possessive at both sites, identically — so it is his text, not a slip, and the assertion is unchanged. No marker at either. ⚑ Recorded because the consistency across two columns is the evidence.
• 0734A · Instauras testes contra me (Job 10:17). The Vulgate has Instauras testes tuos contra me. The English therefore reads "Thou renewest witnesses against me", not "thy witnesses" — deliberately. This is the 7a″ trap in its purest form: the familiar possessive arrives before the printed Latin is read. No marker, because dropping tuos does not change what Job asserts; logged so the absent possessive is visibly a decision.
• 0734B · « Porro filii Israel descendentes de Bethulia venerunt. » — Jdt 6:10 reads de Bethulia descendentes. Word order only. No marker.
• 0734D · « Erant tamen non longe a muris fontes. » — I believe the Vulgate has autem where Migne prints tamen, but I have no Vulgate text on disk and would not fire a [var:] from memory on a connective. Declined on a half-check, exactly as Tobit declined Exsurgens. ⚑ The blind reader also declined it, explicitly, as a stylistic inference it would not stand behind. Not a live defect either way — the English renders the printed tamen as "Yet", which is the right treatment of whatever Migne prints. Recorded as OPEN, for whoever has a Vulgate.
• 0735A · « Domine, Domine, cor autem eorum longe est a me » (Matth. VII, XV). A deliberate composite, and the note says so by printing both chapters: Domine, Domine is Matt 7:21, cor autem eorum longe est a me is Matt 15:8, quoted exactly. Not a divergence at all — the citation is doing its job. ⚑ Note the speaker shifts mid-quotation (the carnal men speak, then Christ); rendered as printed, nothing harmonized.
• 0735C · « Ecce nunc benedicite Domino. » (Ps 133:1). The Vulgate has Dominum, accusative. A case change with no English exponent — "bless ye the Lord" renders both — so this falls under Pattern 9's reasoning for gender mismatches: there is no emendation to make in English and a [var:] would announce a difference the reader cannot see. Declined and logged. ⚠ It is a real difference on the Latin side and a Latin-column reader will meet it.
• 0735C · « Ubi venit fides, jam non sumus sub paedagogo… » (Gal. III). Migne drops the Vulgate's opening At and elides quae est from fidem, quae est in Christo Jesu. Neither changes the assertion. No marker.
• 0735C · « Mortificati estis legi per corpus Christi Jesu… » (Rom. VII). Rom 7:4 reads per corpus Christi and opens et vos. An added Jesu; no marker.
• 0735D · « Caput enim Christi Deus » (cited II Cor. II). 1 Cor 11:3 has caput autem Christi Deus. A connective adapted to the new sentence frame — the citing author's, not a textual variant, and Tobit set that precedent at 0727A. No marker. The citation itself is wrong (§4d).
• 0737D · « Impius cum in profundum peccatorum venerit, contemnit » (Prov. XVIII). Prov 18:3 has cum in profundum venerit peccatorum. Word order inside a genitive. No marker.
• 0738B · « Venient ad te curvi filii eorum qui te humiliaverunt… » (Isa. LX). Isa 60:14 has qui humiliaverunt te and opens Et venient. Word order and a dropped opening connective; everything else word for word. No marker.
• *0740B · « Flevimus, dum recordaremur Sion » (Psal. CXXXVI). Ps 136:1 has cum* recordaremur. dum for cum with the same sense. No marker.

4c · CHECKED AND AGREEING — named, not summarized

Judith lemmata, chunk 0: Arphaxat. (1:1) · Quam appellavit. (1:2) · In gloria. · In campo magno. (1:6) · Tunc exaltatum est. · Et misit ad omnes. (1:11) · Tunc indignatus Nabuchodonosor. (1:12) · Tunc Holofernes. (2:7, the unclosed span the extractor missed) · Cumque pertransisset. (2:12) · Praedavitque omnes. · Descendit in campos. (2:17) · Tunc mise unt, (3:1 — type defect, §5.2, not a variant) · Memores estote. (4:13, italic) · Tunc Achior dux, (5:5) · Et fecit ei coenam magnam. (6:19) · Tunc ad Oziam congregati. (7:12, plate-confirmed) · Et hos quinque dies, (7:23).

Non-Judith scripture inside chunk 0's glosses — read word by word, since this is where the reach for a familiar English happens: Haec omnia tibi dabo, si cadens adoraveris me (Matt 4:9 — exact, word for word) · Hi in curribus et hi in equis (Ps 19:8, unguilleted in the plate and left unquoted in the English) · Instauras testes contra me, (Job 10:17, §4b, italic).

Judith lemmata, chunk 1: Et factum est cum, (8:1) · Filia Merari. (8:1) · Et vir ejus fuit Manasses, (8:2) · Cui vir suus reliquerat divitias multas. (8:7) · Non est iste sermo. (8:12) · Et dixerunt. (8:28) · Et dixit illis Judith. (8:30) · Quibus abscedentibus. (9:1) · Domine Deus. (9:2) · Respice castra. (9:9) · Abstulit a se cilicium, et exuit se vestimentis viduitatis suae, (10:2) · Ut incomparabili decore. (10:4) · Et panes et caseum. (10:5) · Dixeruntque, · Et dixit illi Judith: Sume verba, (11:4) · Missa sum. (11:17) · Et accepit. (12:19).

Non-Judith scripture inside chunk 1's glosses: Facis teipsum Deum (John 10:33 — exact) · Quia Filium Dei se fecit (John 19:7 — exact) · Astitit regina a dextris tuis in vestitu deaurato, circumdata varietate (Ps 44:10 — exact, word for word) · Fili Dei, fili Dei, sicut scis, et sicut vis, miserere mei (not scripture at all — the gloss says quidam patrum in oratione sua dixisse legitur; see §4d on its note) · Cum oraveris, intra in cubiculum tuum (Matt 6:6, unguilleted, and tu autem dropped).

Judith lemmata, chunk 2: In lecto. (13:4) · Caput Holofernis. (13:11) · Custodibus murorum. (13:13) · Aperite portas. (13:13) · Illa autem ascendens. (13:16) · Et interfecit. (13:18) · Et ecce, (13:19, the orphan-» span, §2.3) · Quia hodie nomen tuum, (13:25) · Videns autem Achior. (13:29) · Tunc ingressus. (14:13) · Videntes itaque, (15:3) · Omnis itaque. · Per dies autem. (15:13) · Porro autem. (15:14, plate-confirmed) · Juvenibus. (15:15, plate-confirmed) · Horruerunt. (16:12) · Adonai Domine. (16:16) · In carnes eorum. (16:21, plate-confirmed) · Et factum est. (16:22) · Centum quinque, etc. (16:28, the unclosed span the extractor missed) · Abram suam. (16:28) · Luxitque illam omnis populus diebus septem. (16:30).

Non-Judith scripture inside chunk 2's glosses: Sobrii estote et vigilate (1 Pet 5:8 — exact, twice) · Super muros tuos Jerusalem constitui custodes (Isa 62:6 — exact; the citation is wrong, §4d) · Beatus qui causam suam loquitur in aure audientis (Eccli. 25 — declined, §6) · Nemo ascendit in coelum nisi qui descendit de coelo (John 3:13 — exact, truncated before Filius hominis) · Ipsa conteret caput, etc. (Gen 3:15 — exact, and the feminine ipsa is the Vulgate's own reading, not Migne's; the citation is wrong, §4d) · Confidit in ea cor viri sui, et spoliis non indigebit (Prov 31:11 — exact, plate-confirmed) · Ecce isti de longe venient, et ecce illi ab Aquilone et mari (Isa 49:12 — exact; the capital Aquilone for the Vulgate's lower case is orthography, Pattern 9) · Vermis eorum non morietur, et ignis non exstinguetur (Isa 66:24, §4b) · Flevimus, dum recordaremur Sion (Ps 136:1, §4b).

⛔ « Abram suam. » at 0740A is a conformation trap and was checked twice. Abram is the accusative of Abra, Judith's maidservant (Jdt 16:28, Abram suam liberam dimisit) — not Abram/Abraham, whom this book never mentions. Rendered « Her Abra. », and the gloss's Abram carnalium conversationem likewise. A translator reading fast turns this into the patriarch and the whole paragraph with it.

4d · SIX CITATIONS ARE WRONG AS PRINTED — reported, NOT corrected

Runbook false positive 2 and the brief's rule: an ID., ibid. chain resolves against the printed sequence, so a corrected locator misdirects every note that follows. This book runs (RAB., ubi supra. ) twice, (RAB., ibid.), (ID., ibid.) and (Ibid. XIX). Nothing was corrected.

• 0734A · (Cap. XII) on Instauras testes contra me — Job X:17, not XII.
• 0735D · (II Cor. II) on Caput enim Christi Deus — I Cor. XI:3.
• 0737D · (Apoc. I) on Qui habet aures audiendi — Apoc. II (the formula recurs through II–III); Apoc. I has no such clause.
• 0737D · (Isa. LXI) on Super muros tuos Jerusalem constitui custodes — Isa. LXII:6.
• 0738A · (Gen. II) on Ipsa conteret caput — Gen. III:15.
• 0738A · (II Cor. II) on Timeo ne sicut serpens — II Cor. XI:3.

⭐ A seventh is NOT an error and is the most interesting note in the book: (Matth. VI) at 0736A is MISPLACED, not wrong. It is attached to Fili Dei, fili Dei… miserere mei, which is not scripture at all. But the very next sentence is Regnum tantum Dei quaerere debemus, et omnia adjicientur — Matt 6:33. So Migne's reference is correct for a clause it stands two lines above. Left exactly where he sets it. Correcting or moving it would destroy the evidence that the note travelled.

⚑ (Ibid. XIX) at 0735D is right: Ibid. resolves to John (from the preceding (Joan. X)), and John 19:7 is the verse. The one Ibid. in the work is sound, which is precisely why the six wrong locators around it must not be touched.

§5 · THE BOOK'S OWN CRUCES

5.1 ⭐ 0733B · et similis milis ero Altissimo — a catchword doubled across the column break

…et ponam sedem meam in Aquilonem, et similis [0733B] milis ero Altissimo.

The word similis is broken by the column break and its second half is re-set at the head of the next column: simi-lis / -milis. The plate therefore prints an intrusive milis that is not a word and has no slot. Carried under Pattern 10 as a whole run and marked:

and I will be like [0733B] [sic: milis] the Most High.

7a⁗ and test 1 both run explicitly. With the marker stripped the clause reads "and I will be like the Most High" — complete: verb (will be), predicate (like the Most High), no negation or conjunction swallowed, and the object/complement present, which is the fourth question Tobit's 0728C failure proves 7a⁗ does not ask. The rendering is beside the marker, not inside it. The anchor stands between the two fragments exactly as Migne sets it.

⚠ Why [sic:] and not [d:]. Pattern 11 is for a run set twice; here only a syllable is repeated, and milis is not a copy of any neighbouring word — it is half of similis. Pattern 11a's test ("mark the intrusive copy") has no purchase on a fragment that is not a copy. Logged because the two markers are genuinely close here and a later agent will meet the same shape.

5.2 Five split-type runs — [sic:] marked and rendered

Pattern 10 carries the whole broken run; 7a⁗ requires the sense to be said in plain words beside it. All five do both, and test 1 was run on each by deleting the bracket and reading aloud.

Each conjecture is near-certain from the immediate context: subjugaverit is the verb the imperio requires; miserunt is Jdt 3:1 word for word; sollicitudo… divitiarum suffocat verbum is Matt 13:22 verbatim; quilibet… deditus is the only reading that agrees with deditus; correcti… salvantur is answered by Achior… consociatus populo Dei in the same sentence.

5.3 0733D · fallaci divitiarum — a REAL word with no slot. [sic:], and the boundary is close.

Migne prints In his enim solli itudo hujus saeculi, et fallaci divitiarum suffocat verbum. Matt 13:22 reads sollicitudo saeculi istius et fallacia divitiarum suffocat verbum.

fallaci is a real form — dative or ablative singular of fallax — and Pattern 12's closing paragraph says a real word that is merely wrong takes no [sic:]. I marked it anyway, and here is the reasoning, because it is a genuinely near call. The test I applied is Pattern 12's test 1, the one Tobit used on astutius at 0729A: the form has no grammatical slot. It cannot govern divitiarum, it agrees with nothing in the clause, and it stands as the second of two coordinated subjects of suffocat — a position only a nominative can hold. A dative/ablative adjective floating as a subject is not "a real word that is wrong"; it is a word with a letter missing. Marked AND rendered: "the deceitfulness [sic: fallaci] of riches".

⚠ The reading I rejected, recorded because it construes if you squint: taking fallaci as an ablative of means with suffocat ("chokes the word by the deceit of riches") — rejected because et has then nothing to coordinate and sollicitudo is left as the sole subject of a verb the Gospel gives two.

5.4 ⭐ 0738D · ad malitiam Domini veniunt — the faithful English asserts something false. [cj:] FIRED.

Ad praedicationem verbi Dei ex singulis gentibus et provinciis probatae personae et viribus accinctae ad malitiam Domini veniunt.

Rendered literally: "persons approved and girded with strength come unto the MALICE of the Lord." The Lord has no malice, the sentence is praising the preachers, and the whole paragraph is a military figure — viribus accinctae, "girded with strength", and the preceding gloss on Gedeon's trumpets, pugnant praedicationis voce. The printed English asserts something the author did not, which is Pattern 18's bar exactly, and 18a's clarification that the test is falsity and not merely a better word available.

The conjecture is militiam — one letter — giving "unto the warfare of the Lord", the standard phrase, and the one the girding and the fighting both demand. Marked additively so both readings stand:

unto the malice [cj: malitiam; read militiam, "unto the warfare"] of the Lord

⚠ This is the site I most want a blind reader to open, and I fired rather than declined because Tobit's notionum taught the cost of the other choice: a declined marker leaves a reader meeting "the malice of the Lord" with nothing at all telling him the page is broken, which is worse than either available answer.

5.5 0735A · corporale vitium solatium — [cj:] FIRED, and the marker does not swallow the word

…qui corporale vitium solatium auditoribus indiscrete promittit…

corporale and solatium agree (neuter singular); vitium stands between them agreeing with neither and governed by nothing. And the same paragraph answers it four lines later: si enim praesentis vitae negatur solatium, subditos deserunt — the identical noun pair, with vitae where this clause has vitium. So the conjecture is vitae, "the bodily solace of life", and the intruded vitium is a compositor reading a familiar word for an unfamiliar one.

⭐ CONFIRMED at the blind read on exactly this evidence, and it is decisive: the conjectured phrase corporale vitae solatium is reproduced VERBATIM four lines later in the same paragraph. The conjecture does not have to be imagined — Migne prints it himself, a few words on.

Rendered so that both the printed word and the conjecture are on the page:

promises his hearers a bodily vice [cj: vitium; read vitae, "the bodily solace of life"] for a solace

⚑ Test 1 run: strip the bracket and the clause reads "promises his hearers a bodily vice for a solace" — odd, which is the point, but complete, with both its objects present. The failure mode Tobit measured (the marker standing instead of a content word) does not occur here because vitium is rendered as "a bodily vice" and solatium as "for a solace" in separate breath.

5.6 ⛔ 0735B · quindecim generatione progenita — [cj:] FIRED, then OVERTURNED at the blind read. The rival I named was right.

quindecim is indeclinable and cannot agree with an ablative singular generatione; the phrase does not construe. The gloss's own arithmetic settles the intent: per hebdoadem legis et ogdoadem evangelii — seven plus eight is fifteen, and Jdt 8:1 gives Judith fifteen forebears.

The marker now in the text: brought forth by fifteen generation [cj: quindecim; read quindecima, "in the fifteenth generation"]

What I filed first, and why it was wrong. My first marker glossed the other word — [cj: generatione; read generationibus, "by fifteen generations"] — reading the phrase as descent through a count of generations. I called it the weakest of the three markers and named quindecima as the rival. The rival is the reading. Three arguments converge on it and none of them is mine:

• Palaeography. QUINDECIMA GENERATIONE → QUINDECIM GENERATIONE is textbook haplography of the final -A before GENERATIONE. It explains the printed ablative singular exactly, and it leaves the printed word untouched. My version required emending the word that is not corrupt while preserving the one that is — which is the diagnostic that a conjecture is aimed wrong.
• Idiom. progenitus + ordinal + ablative of time is the standard phrase. quindecim generationibus progenita would want post quindecim generationes.
• Sense. Seven plus eight is the fifteenth, not fifteen separate things; the gradual psalms the gloss reaches for immediately after are fifteen in number and the topos is the fifteenth step. My own cited evidence pointed at the reading I did not take.

⭐ The lesson, and it is not "I was unsure and should have declined." I identified the correct conjecture, wrote it into the file by name, ranked my own marker weakest — and still fired the other one. Naming a rival in the cruces is not the same as adjudicating between them, and the apparatus records only what the marker says. Where a crux names two conjectures, the file must say which one the marker took and why the other lost, in those words; anything softer is an undecided site wearing a decision's clothes.

⚑ Body English is unchanged and stays literal on the printed text — "brought forth by fifteen generation" — per Pattern 18's additive rule. Only the bracket moved.

5.7 THE UNEXPRESSED SUBJECT — every supplied pronoun, and what decided it

This book runs Judith = the Church, Holofernes and Nabuchodonosor = the devil, Achior = the heretics, and the referent moves between the literal figure and the allegory mid-sentence. Six sites needed a decision. Five took a bracketed supply rather than a bare pronoun, on Ruth's 0538C lesson that a bare "she" binds silently to the nearest noun rather than to the right one.

(i) 0733C · resistentes occidit in ore gladii. Rendered "he [the devil] slays them that resist with the edge of the sword." The lemma is Praedavitque omnes — Holofernes plundering — but the immediately preceding paragraph names its subject outright: ex quibus diabolus vindicat sibi aliquam partem, nec pugnae formidat difficultatem. The whole run from 0733A (Cum diabolus perditorum multitudinem…) has the devil as its grammatical subject and never changes it. The literal Holofernes is present only as the figure being read. Devil, supplied.

(ii) 0734A · Per exempla sanctorum corroborat. Rendered "By the examples of the saints [Eliachim] strengthens them." The lemma is Memores estote — Jdt 4:12, where Eliachim the high priest writes to all Israel and says exactly those words. So the subject is the speaker of the lemma, which is the brief's own rule that the antecedent is very often in the verse the lemma comes from rather than in the sentence. ⚠ The rival reading, recorded: corroborat of Scripture or of the sacred writer, which would make the supply wrong but the sense nearly identical. I supplied the named person because the verse names one; a reader who prefers the impersonal has the argument here.

(iii) 0737A · Sub dominatione saecularium principum libere manere permiserunt. Rendered "Under the dominion of secular princes they permitted [the Church] to remain free."

The substance was right first time and the bracketing was backwards; repaired at the blind read. The verb is plural and the sentence prints neither subject nor object. The two clauses that follow supply the subject and are the whole evidence: Trajan tempering his edict, Hadrian writing to the proconsul of Asia — emperors, granting toleration; and saecularium principum in this very sentence names the class.

⚠ What was wrong with my first version — "[the rulers of this world] permitted them to remain free." Two faults, and they compound:

• The bracket restated saecularium principum, four words earlier in the same sentence, so the page read as though two different bodies were in play — the secular princes, and some further "rulers of this world" acting under them.
• "them" was the genuinely supplied word and it carried no bracket at all. The nearest plural noun in the English was "secular princes", so a reader parses the rulers permitted the princes to remain free — the permitters and the permitted collapsed into one.

The rule this yields: bracket what the Latin does not print, and only that. A subject recoverable from the sentence's own words is carried over with a bare pronoun; the object, which is nowhere on the page, is the thing that needs the bracket. Here that is [the Church] — she is what remains free under pagan dominion, and she is the subject of the entire surrounding gloss.

(iv) 0737B · Non inquinatur cibis gentilium aut superstitione. Rendered "[Judith] is not defiled by the foods of the Gentiles or by superstition." The lemma is Et accepit (Jdt 12:19, Judith eating in Holofernes' tent), so the literal figure is the subject; the very next sentence then switches to the allegory and names it — Ecclesia enim inter gentes habitans, idololatria non polluitur. Supplying "the Church" in the first sentence would erase the switch Migne makes. Judith in the first, Church in the second, as printed.

(v) 0737D · Recordationem confecti belli sollicite jubet memorari. Rendered "[Judith] carefully bids the remembrance of the finished war be called to mind." The lemma is Caput Holofernis (Jdt 13:11), where Judith shows the head over the wall and speaks. No other candidate in the sentence is singular and animate.

(vi) ⭐ 0740A · Abram… perducit. Rendered "[The Church], by leading and by exhorting, brings Abra, the manner of life of carnal men, which is always subject to fear, through into the liberty of grace."

⚑ This is the book's worst exposure and it took the longest. Abram is accusative — the object, not the subject — and so is carnalium conversationem, in apposition to it. The verb perducit is third singular with no expressed subject at all. The candidates:

• The Church. The lemma is Jdt 16:28, Abram suam liberam dimisit — Judith freeing her maid — and this work has stated the equation twice in so many words: Hanc conventionem nostra Judith, id est Ecclesia, respuit (0735B) and Judith, id est Ecclesia, commendat praesbyteris portam (0736B). Judith frees Abra; the Church brings carnal life into the liberty of grace. The abstract nouns ducendo et exhortando are the Church's own offices throughout this book (doctores Ecclesiae, magistri Ecclesiae, praedicationis obsequium).
• Abra herself as subject, i.e. reading Abram as a nominative slip. Rejected: suam in the lemma and the apposition with conversationem (unambiguously accusative) both fix the case, and the clause already has an object without it.

Church, supplied in brackets. The final clause ut non timore, sed amore serviat is left with its own singular subject unsupplied, because there it is Abra / the carnal manner of life — the thing brought into liberty — and that referent is two words away rather than absent.

⛔ A flagged passage is not a checked passage, so: this one is decided, not offered. If a blind reader disagrees, the disagreement is with "the Church", not with an undecided site.

5.8 The one two-sided call I could NOT close

0735B · subditos deserunt, ut cedentes persecutionibus, corporale devitent supplicium. Rendered "they forsake their subjects, that, yielding to persecutions, they may avoid bodily punishment."

Who yields and who escapes? Two readings, both construe:

• The teachers (subject continuity with deserunt, which is the reading in the text): the sluggish teachers abandon their flock so as to escape persecution themselves. This suits the paragraph, which is an indictment of doctor iners.
• The subjects (subditos, the nearest accusative): the teachers abandon them so that they, the flock, may yield and escape torment — a bleaker but coherent reading.

I took subject continuity and left the English's "they" ambiguous in exactly the way the Latin is ambiguous, rather than forcing a bracketed supply onto a site where the plate genuinely does not decide. ⚠ This is the one place in the work where I have left a pronoun unresolved on purpose, and it is named here so it does not read as an oversight.

⚑ A number shift runs through the same paragraph and is preserved: doctor iners expetit — singular — then subditos deserunt — plural. Migne makes the shift; smoothing it would erase it.

5.9 Sentences left broken, and punctuation the plate prints

• 0733A · unde sequitur. A paragraph ending "whence follows." with a full stop and nothing following. Reproduced; nothing supplied. (Compare 0735A, unde sequitur: — with a colon, and something does follow. Migne distinguishes them and so does the English.)
• 0735C · Et vir Manasses, Christus Ecclesiae sponsus. No verb. Rendered verbless: "And the husband Manasses, Christ the spouse of the Church."
• 0735C · cui Judith conjugio copulata quasi decalogo legis… obnoxia. No verb. Rendered with the copula ellipsis English cannot avoid; ellipsis of est is ordinary Latin, not a defect.
• 0738B · et sequaces suos errorem relinquere et fide liberi sese sociare hortantur. sese sociare has no dative — one joins oneself to something, and the something is not printed. The sense wants Ecclesiae from the clause before. [cj:] DECLINED, on Tobit 0730C's reasoning: the marker's form requires a printed WORD to gloss, and what is missing here is a complement, not a wrong word — there is nothing to put in the italics. Rendered as printed: "to join themselves", with no object supplied.
• 0738D · sed contra spiritualia nequitiae in coelestibus, etc, — Migne closes the abbreviation with a comma, not a period. Reproduced as etc,.
• 0733D · Migne's comma inside mortis periculum et voluptatis, detrimentum evadere possint. The comma falls between voluptatis and the noun it depends on. English has no exponent for a comma inside a genitive phrase, so the construction the Latin bears is rendered — "the peril of death and the loss of pleasure" — and the comma count of the clause is unchanged. Tobit's Nolite, pueri reasoning (its §5.10), applied.
• 0739B · ubicunque bis ponitur, Dominus: primum nomen… — a colon after a bare quoted word. Kept where he sets it.
• 0735D · cum mitteret apostolos suos… : venit aestus persecutionis — a colon between a cum-clause and its main verb. Kept.

5.10 Candidates considered and DECLINED — with reasons

• 0732D · per vastum tumoris et elationis agitur. fastum ("pride") is one letter from vastum and would suit tumoris et elationis well. Declined: vastum as a noun ("waste, vast expanse") construes as printed, "through the vastness of swelling and of elation" asserts nothing false, and 18a's bar is falsity, not improvement.
• 0736C · Simonis in Bene in oratione actum Simonis commemorat. Jdt 9:2 names Simeon, and the deed described (with Levi, avenging their sister on strangers, Gen 34) is Simeon's. Rendered "Simon" as printed, no marker — Simon/Simeon is one Greek name in two Latin spellings, both attested, which is Pattern 9 territory. ⚑ Logged because a translator "fixing" it to Simeon would be conforming the plate to the verse, which is the 7a″ mechanism.
• 0736C · purae ascensionis incensum. orationis would give "the incense of pure prayer", which is what Ps 140:2 and the Matt 6:6 quotation in the same sentence would license. Declined: ascensio is the word this book has just built on (ad coelestem gratiam ascensura, futurae ad coelos ascensionis figurativus, 0735B–C), so the printed reading is the one the argument has prepared.
• 0737A · Munitium Fundanum. The proconsul is Minucius Fundanus. Rendered as printed, no marker: a proper name misspelt is not a word with no slot, and Pattern 12's boundary puts a real-but-wrong form under Pattern 7. Same treatment as Carmin for Charmi at 0734C.
• 0734C · Carmin. Jdt 6:11 has Charmi. Rendered "Carmin" as printed. See above.
• 0740A · Centum quinque with the gloss centenarius numerus a laeva transit in dextram. The gloss counts only the hundred and drops the five. Not a defect — the finger-reckoning figure (a hundred passes from the left hand to the right) is about the hundred. No marker.

5.11 Negation — 43 particles, every one carried, every one on its own word

Test 2 run as the brief specifies: for each printed negative I named the single Latin word it attaches to, then found that word in the English and checked the negative was on it. None was dropped, none was added, and none was relocated.

⚑ Two sites are the 2 Thessalonians shape exactly and are the ones to check first:

• 0734D · non solum divinam, sed et humanam Christianis interdicunt philosophiam. The non attaches to solum, not to interdicunt. Julian and the pagan princes DO forbid; what they do not do is stop at divine philosophy. Rendered "forbid Christians not only divine but also human philosophy" — the verb affirmative, the negative on "only". Pressing that non into the verb, which is what 8987 @0624B did with velint, would have produced smoother English saying the opposite.
• 0740B · in dolore et gemitu non affliguntur ex desperatione, sed futurae vitae recordatione. The non attaches to ex desperatione through the non… sed correlative. The elect ARE afflicted — that is the sentence's whole claim — but not out of despair. Rendered "are afflicted in grief and groaning not out of despair, but out of the remembrance of the life to come." "Are not afflicted… but" would be the relocation, and it reads better.

Three more where the negative sits on an OBJECT and not on the verb, and English can easily move it:

• 0734C · nec persecutorum rabiem, nec mortem pertimescere docent — "teach their catechumens to dread neither the rage of persecutors nor death." Not "do not teach them to dread."
• 0736A · nec tempus, nec modum praescribere Domino debemus — "we ought to prescribe to the Lord neither the time nor the manner." Not "we ought not to prescribe."
• 0738C · non hastam, non clypeum, sed tubas et lampades tulit — "took not a spear, not a shield, but trumpets and lamps." The verb is affirmative for the third member.

The remainder, checked one by one: nisi forte (0731D) · non divinae permissioni (0733A) · nec pugnae formidat (0733B) · fructum non facit (0733D) · non morietur / non exstinguetur (0733D, and again 0740A) · perfecte non didicerit (0734A) · renatus non sit (0734A) · viam veritatis non teneant (0734A) · omnino non celant (0734A — scope repaired at the blind read, §9.5) · latere non possent (0734B) · ne… timeatur (0734B) · quos flectere… nequit (0733C) · ne faciant (0734D) · non longe a muris (0734D) · non meretur (0734C) · jam non sumus (0735C) · nec erroribus… dignatur (0735D) · non novit + non subito fiat (0737A) · ut nemini liceret + sine criminis objectione (0737A) · idololatria non polluitur (0737B) · Non inquinatur (0737B) · Nemo ascendit… nisi (0737D) · Timeo ne (0738A) · non recedet (0738B) · Non est… non est… nec consilium (0738C) · Non fugiet… non salvabitur (0738C) · non est colluctatio (0738C) · neque nubent neque nubentur (0739A, plate-confirmed) · spoliis non indigebit (0738D) · non resistit (0740A) · ut non timore, sed amore serviat (0740B).

5.12 Read clean, and named so

CAPUT IX (0736C), CAPUT X (0736D–0737A), CAPUT XI (0737B), CAPUT XII (0737B–C), CAPUT XIV (0738C) and CAPUT XVI's VERS. 16 and VERS. 22 (0739A–0740A) carry no defect marker of any kind: lemmata checked, guillemets balanced, negations carried, no supplied subject needed. Chunk 1's only apparatus is one [cj:] and one [var:]; chunk 2's is seven [var:] and one [cj:] and no [sic:] at all. All five split-type runs and the doubled catchword are in chunk 0, which is where the plate is worst.

Verse numbering ascends within every chapter — no out-of-sequence address anywhere in this book. Checked, and recorded as checked. CAPUT VII opens at VERS. 7 and CAPUT XVI at VERS. 12; those are Migne's starting points, not gaps, and nothing was supplied.

5.13 ⭐⭐ 0737B · sed sopor tibi quem ferias praeparavit — a SUBJUNCTIVE VERB demoted to a noun. Repaired at the blind read.

[n: (AUG., serm. 229 de temp.)] Species custodit quae blanditur, etc., usque ad sed sopor tibi quem ferias praeparavit.

Now reads: "The beauty which allures keeps guard, etc., as far as but slumber hath prepared for thee [him] whom thou mayest strike."

What I filed first: "Beauty guards that which flatters, etc., as far as but the slumber which he prepared for thee for holidays." Both halves were wrong, and they are one clause-pair.

(a) ferias is ferio, not feriae. It is the second person singular present subjunctive of ferio, "that thou mayest strike" — not the accusative plural of feriae, "holidays". The construction is sopor (nominative, subject) praeparavit tibi [eum] quem ferias: slumber has prepared for thee the man thou art to strike. That is the point of an Augustine sermon on Judith 13 — the drunken sleep delivers up the man she is to behead. My version made quem the object of praeparavit, invented an unexpressed "he" for a sentence that has a perfectly good subject in sopor, and left ferias doing no grammatical work at all.

(b) quae is nominative, agreeing with species. "The beauty which allures keeps guard" — Judith's adornment is itself her protection, which is precisely why this excerpt is drawn in here, two paragraphs after CAPUT X's ut incomparabili decore videretur gloss. My "that which flatters" took quae as the object of custodit and left the verb with nothing stated to guard.

⚑⚑ This is the one site in the work where a real defect hid behind SMOOTH-LOOKING ENGLISH, and that is the whole lesson. "for thee for holidays" reads as an odd but harmless Latinism — the sort of thing a reader forgives in a nineteenth-century gloss — so nothing in it looks broken. No automated check can see a subjunctive demoted to a noun: the word count is right, the italics are right, the marker parity is right, the negation count is right, and the punctuation bands agree. Only a parse catches it.

⚠ And note what it is NOT. It is not the usque ad formula leaving a tail unfinished. All the printed words construe, and nothing is to be supplied beyond the antecedent of quem, which is bracketed. I had the abridgment rule loudly in mind from Tobit §1(b) and it made an unparsed clause look expected. The rule "leave usque ad tails unfinished" is about supplying words the plate lacks; it is not a licence to leave a printed clause unparsed.

§6 · WHAT A BLIND READER SHOULD OPEN FIRST

1. ⭐ 0738D, ad malitiam Domini (§5.4) — the one [cj:] I am confident in and the one that changes what the page asserts. If militiam is wrong, the English currently accuses the Lord of malice with a marker beside it; if it is right, this is the book's best find.
2. ⭐ 0740A, the subject of perducit (§5.7 vi) — decided as "the Church", with the reasoning laid out. This is the site with the doctrinal weight, and it is the shape of Ruth's 0538C.
3. 0735B, quindecim generatione (§5.6) — the weakest of the three [cj:], with a named rival conjecture. I would drop this one first.
4. 0733D, fallaci (§5.3) — a [sic:] on a real word, on the "no grammatical slot" test. Pattern 12's own closing paragraph pushes the other way and the call is close.
5. 0735B, subditos deserunt… devitent (§5.8) — left deliberately ambiguous. A reader who thinks the Latin does decide should say which way.
6. ✅ 0737D, Beatus qui causam suam loquitur in aure audientis (Eccl. XXV) — CLOSED, and the answer is that the English is right and NO marker belongs there. The blind reader settled it from its own memory of Sirach 25, with no text on disk: Migne's wording is not the Vulgate's, and my English gives Migne's words exactly, with no conformation. ⛔ A [var:] was nonetheless refused, deliberately, and the reasoning is the important part: a [var:] is a public claim about ANOTHER text, and we cannot check this one. One agent's recollection is not a witness. So the divergence is recorded here as prose and nowhere else: Migne's wording appears to diverge from the Vulgate at this verse; no Vulgate text is on disk in this repo; the claim is therefore UNVERIFIED and is left for whoever has the text. That is honest, and it is a different act from marking the page.
7. §2.1, the missing 0732A–C anchors — an address question, not a translation question, and the only item here a plate reader can settle outright.

§7 · WHAT TOBIT'S CONVENTIONS DID NOT COVER — for Esther and the remaining OT books

Tobit §1 settled the guillemet lemma, usque ad, the sigla spacing and Tyconius, and none of that needed re-deciding. Five things came up here that it does not answer, recorded because Esther (8948) and the rest of the guillemet books inherit from this file too.

7.1 ⭐ usque ad is NOT always the abridgment formula — and the italics tell you which

Tobit met seventeen of them and every one was Migne's abridgment marker, set in italics. This book prints the phrase twice in ordinary running prose, unitalicized, meaning simply "as far as" or "until", and it is not a formula at all:

• 0735A · Sunt enim in sagena Domini et boni et mali pisces, usque ad littus futuri judicii — "even to the shore of the judgement to come."
• 0735D · jejunio, et orationi operam dat, usque ad consummationem saeculi — "even unto the consummation of the age."

Both take a plain accusative object and complete their own sentences. The four formulaic ones in this work are all italic in the plate and all follow etc., — that pair of cues is the test. ⛔ Italicizing these two, or rendering them as as far as, would invent an abridgment Migne did not make and would leave two complete sentences looking broken. A later agent working from Tobit's §1(b) alone would very plausibly do it.

7.2 A lemma may be ITALIC in a guillemet book — CAPUT IV prints both markers, on one page

Every lemma in this work is guilleted except the two in CAPUT IV at 0734A: Memores estote. and Instauras testes contra me, are set in italics, like Ruth's and the PL 114 epistles'. The chapter has one gloss and it uses the other convention throughout. Reproduced as printed — italic lemmata in italics, guillemet lemmata in guillemets, in the same file. The books are not cleanly divided into "italic books" and "guillemet books"; the page decides.

7.3 A guillemet span may be CLOSED WITHOUT BEING OPENED

Tobit had four unclosed opens. This book has two unclosed opens and, at 0738A, a closing » with no opener at all (§2.3). The rule is the same in both directions — reproduce the plate — but the consequence differs: an orphan close makes the running counts go negative inside a band, which will look like a translator's error to any tool that walks the marks in sequence rather than counting them. Counting « and » separately, as this file does, is the check that survives it.

7.4 A note may be MISPLACED rather than wrong, and the two are treated differently

Tobit's three wrong locators were all wrong about the passage they sit on. 0736A here is a different animal: (Matth. VI) is a correct reference sitting two lines above the clause it belongs to (§4d). It is not a candidate for correction or for a "wrong citation" report — moving it would be as destructive as changing it, and calling it wrong would be false. Report the displacement, leave the note.

7.5 An unattributed scripture quotation

« Non est enim prudentia, non est sapientia nec consilium contra Deum; » at 0738C is quoted with no [n:] at all, unlike every other proof-text in the work. Identified as Prov 21:30 for the [var:] and for the index; no note was supplied, because supplying one would put a citation on the page that Migne does not print. ⚑ Worth knowing for the scripture index: a divergent quotation can be present with nothing to key it by except its own words.

7.6 The usque ad rule governs what is SUPPLIED, not what is PARSED

⚑ Added from this book's one hidden defect (§5.13). Tobit §1(b) is emphatic that an abridged sentence is left unfinished and no verb is ever supplied to complete it — correct, and it held for all six formula sites here. But the rule created a blind spot: a resumption tail that looked ungainly read as expected ungainliness, so sed sopor tibi quem ferias praeparavit was rendered without ever being parsed, and a subjunctive verb (ferias, "thou mayest strike") was demoted to a noun ("for holidays").

The distinction to carry into Esther and the rest: the abridgment removes WORDS, it does not suspend GRAMMAR. Parse every resumption tail as strictly as ordinary prose; only when the parse shows a genuine hole does the "supply nothing" rule come into play. A tail that reads oddly is not evidence the abridgment made it odd.

§8 · THE THREE TESTS, RUN ON MY OWN FILES

Test 1 — strip every marker and read what is left. Run mechanically on all three chunks: all seven [sic:], all three [cj:] and all nine [var:] removed, and the residue read aloud looking for a stranded article or preposition sitting on punctuation. All ten defect markers have their rendering beside them, itemized in §5.1–5.6. The [var:] markers all follow a completed sentence and take nothing with them. ⚑ The fourth question — does the clause still have its OBJECT? — was asked explicitly at 0733B (the Most High), 0735A (a bodily vice… for a solace) and 0733D (the care… the deceitfulness… the word), which is where Tobit's four failures would have landed.

Test 2 — name the word each negative attaches to. §5.11. 43 particles, all carried, all on their own word. Two non… sed correlatives and three object-negations are named individually because they are the shapes that relocate.

Test 3 — punctuation compared BY COLUMN BAND, never by total. Built over the 30 verified anchor bands (13 + 10 + 10, including a HEAD band before the first anchor), stripping all [x: …] markers from both sides so the marker colons do not pollute the count. Colons, semicolons, question marks, « and » compared band by band. Result: all 30 bands agree on all five marks.

⚑ This test earned its keep. The first pass found six bands where I had supplied a semicolon Migne does not print — 0733A, 0733D, 0734A, 0734C, 0735A (three of them) and 0736B — every one a comma or a full stop in the plate that I had "improved" into a semicolon while joining clauses. The totals would have hidden none of them here, but the site-by-site fix is only possible band by band. All eight are now commas, as printed.

⚠ The comma→semicolon sweep was NOT invoked, and the brief's warning is why: this book's plate prints 14 semicolons of its own (3 + 4 + 7), so Tobit's narrow zero-semicolon exception does not apply. What was fixed here is the opposite defect — marks I supplied — which Pattern 8 forbids outright and which has nothing to do with the closed rate question.

Question marks: the Latin prints zero and the English prints zero, verified per band. ⚑ Two constructions invited one and were refused: Quaeritur quo tempore quibusve regibus historia Judith fuerit at 0731D and Notare debet lector, utrum opinio vera sit at 0739A are both indirect questions, and Migne closes neither with a mark, so Pattern 8b's colon recast does not apply — it is bounded to the case where his own mark falls on an indirect question. Recast without a mark, as printed.

§9 · THE BLIND POLARITY READ (2026-08-18) — five repairs, one conjecture overturned

A blind reader that never saw this file read all three chunks against the Latin. Recorded in full, wins and losses together, because the shape of what it caught is worth more than the score.

What held — the classes this book was most exposed to

• ⭐ No silent scripture conformation anywhere. Every quoted verse was checked, and the two sites the brief warned would go were both named as correct: 0734A Job 10:17, where Migne prints Instauras testes contra me against the Vulgate's testes tuos and the English does not restore the possessive; and 0733D / 0740A Isa 66:24, where Migne's missing second eorum is kept at both occurrences. This was Ruth's failure class — four silent repairs — and it is clean here.
• ⭐ All seven [sic:] pass the swallow test, with complete English standing beside each. This was Tobit's failure class — four for four — and it is clean here.
• ⭐ 0740B (VERS. 29) called the sharpest negation-relocation test in the work. non affliguntur ex desperatione, sed… — the non stands before the verb and governs ex desperatione. The English puts it on the phrase. Floating it onto affliguntur would have inverted the sentence and read perfectly well — which is 8987's velint exactly.

Confirmed, on evidence the file did not cite — do not churn

• 0738D [cj: malitiam; read militiam] — confirmed. The added evidence: militia Christi / Dei is the settled phrase, malitia Domini is theologically impossible, a/i confusion is among the commonest in Migne's setting, and the military saturation runs further back than I traced (Gedeon's battle and Eph 6's armour in the preceding paragraph, not merely viribus accinctae in this one). Literal body + bracketed conjecture confirmed as the correct house treatment.
• 0740A, subject of perducit supplied [The Church] — confirmed and it holds. Added support: the literal verse is Jdt 16:28, where Judith frees her maid, so the agent is the mistress. Christ was the rival worth weighing and rejecting him was right: ut… serviat has Abra for its subject and the manumission is Judith's act.
• 0735A [cj: vitium; read vitae] — confirmed, on the internal parallel now written into §5.5: corporale vitae solatium is reproduced verbatim four lines later.
• Also confirmed and not to be re-raised: the anchor jumps (§2.1); the three unpaired guillemets (§3); ## CAPUT VII without its period (§2.2); the six wrong or displaced locators (§4d); and both plain-prose usque ad instances correctly distinguished from the six formula ones (§7.1).

The five repairs, all applied

Repair 5 in full, because it is a pure English fault with no Latin error behind it. veritatem omnino non celant. I wrote "they conceal not the truth altogether." omnino non is a fixed intensifier — "not at all, by no means" — but that English word order reads in modern usage as the opposite scope: they do not wholly conceal it, i.e. they conceal it in part. The context demands the strong reading: beaten in argument, the heretics do not hide the truth at all. ⚑ Nothing was added or dropped — the negative was present and on the right verb — so test 2 passed it, and so did the band counts. The defect is scope, and only a reader asking what does this English actually say finds it.

⚑ The project gap, recorded as a gap and not as a task

Two calls in this work were declined for want of a Vulgate text on disk (§4b 0734D, §6 item 6), and the orchestrator notes these are now the third and fourth such decline in two batches. It is recorded here as a project-level gap, not as work for anyone to route around: a Vulgate on disk would have closed both in minutes and would have made §4 stronger throughout.

What the blind read teaches that the three tests do not

Three of the five repairs were invisible to every check this edition runs, and they failed in the same way rather than in three ways:

1. §5.13 (ferias, quae) — a mis-parse that produced plausible English. Word count, italics, markers, negations and punctuation bands were all correct.
2. §5.6 (quindecim) — a conjecture aimed at the wrong word, where I had already written the right answer into the file as a rival and fired anyway.
3. Repair 5 (omnino non) — correct negation, correct attachment, inverted scope.

⭐ The common thread: every one of them is a question about what the English SAYS, and every test in §8 is a question about what the English CONTAINS. Parity, counts, attachment and bands are containment checks; they cannot see a right word in a wrong scope, a right conjecture on a wrong lemma, or a right vocabulary in a wrong parse. A fourth test belongs beside them, and it has no mechanical form: read each finished paragraph back against the Latin asking only "does this assert what he asserts", with the counts already known to be clean.