Liber Tobiae

Translated 2026-08-18, one translator, 3 chunks (2,567 Latin words), on Opus. verify-english.mjs clean: columns, notes and sections aligned, no duplicate paragraphs.

Apparatus: 4 [var:] · 4 [sic:] · 2 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 57 [n:] (29 / 19 / 9, exact parity with the Latin twins). Structure: CAPUT PRIMUM → CAPUT XIV, all fourteen heads present. Guillemets: 76 « / 72 », 1:1 with the Latin. Four quotations are UNCLOSED in Migne and stay unclosed (§3). Italics: 17 usque ad + 2 etymological glosses, 1:1. Question marks: ZERO in the Latin, zero in the English.

§1 · CONVENTIONS THIS BOOK SETTLES — for 8948, Judith (8960), and the guillemet books to come

Ruth (8968) is the Glossa pilot and its §1 governs everything here unchanged: VERS. n.-- passes through verbatim; the lemma is Englished; a fragmentary lemma is rendered as the phrase it stands for; the attribution sigla are ordinary [n: …] notes, contents unanglicized. This work re-opened none of that. What follows is what Ruth did not have to decide, because Ruth's plate has no guillemets at all.

(a) The guillemet lemma — marks are the plate's, contents are Englished

Migne sets Tobit's lemmata in guillemets, not italics (Ruth's and the PL 114 epistles' are italic). This changes the typography and nothing else:

• The marks are sacred and 1:1 — every « and » stands where Migne sets it, with his spacing, and the counts match the Latin twin exactly (Pattern 6, verify-english check 7).
• The contents are Englished, exactly as an italic lemma is. Corpus precedent is shipped work 8930 (Commentaria in Cantica, PL 112), where a guilleted scripture quotation has its words translated and its marks and spacing untouched. Followed here without variation.
• A note inside a guillemet span stays inside it, in Migne's position — this book does it constantly: « Omnia primitiva [n: (ID., ibid.)]. », « Latum mandatum tuum nimis [n: (Psal. CXVIII)], ». So do column anchors: 0728A falls INSIDE the Matt. 3:2 quotation (appropinquavit enim [0728A] regnum coelorum) and is reproduced inside it.
• Migne's punctuation inside the marks rides as printed, including the odd ones: « Et factum est. » etc. sets a full stop inside the guillemets before etc., and the English does the same.
• « Nolite, pueri, effici, etc. »-type commas are his, not ours. Where Migne's commas around a word would change its syntax, render the construction the Latin bears and log it (§4, 0726B).

(b) usque ad → as far as — settled, not re-opened

Migne's abridgment formula (etc., usque ad … = quote the head of Bede's sentence, skip, resume at these closing words). Englished in its own italic span, per shipped work 8978 (cruces.md §6), which is the site that set it. 17 occurrences here — by far the heaviest in the Glossa block so far, and none was varied.

⚠ The brokenness is the formula working. Sic populus Israel, etc., usque ad unde: « Primogenitus meus Israel » is not a defect, and neither is Apostolus gentium eos qui ad fidem venerant, etc., usque ad …, whose main verb the abridgment has cut out. No verb was supplied to any of the seventeen and none should be in a later book. Where the resumption lands mid-clause, the English lands mid-clause.

(c) Tyconius vocabulary — fixed here for the OT Glossa books

0726B cites Tyconius by name and by number: secundum illam regulam Tyconii, qua tanquam de eodem agentes, aliquando propter bonos omnes laudamus, aliquando propter malos omnes vituperamus, quod est secunda regula.

• regula = "rule", throughout, never "canon" and never "principle". The Liber regularum is the Book of Rules in English and the ordinal is load-bearing here.
• secunda regula = "the second rule" — rendered as the ordinal Migne prints, not expanded to its traditional title De Domini corpore bipartito. Expanding it would supply text the plate does not carry.
• tanquam de eodem agentes = "as though treating of the same [subject]", with the supplied word in brackets (corpus practice for supplied English). eodem is masculine or neuter and the plate does not disambiguate; "the same body" is what the rule means and is precisely what Migne does not print, so it stays out of the text and lives here.

(d) Sigla with Migne's own internal spacing are reproduced character for character

(BEDA ubi supra. ), (BED., ubi supra. ), (BEDA, ex Isid. ), (BEDA, in Tobiam. ), (BEDA, in Esdram. ) — the space before the closing parenthesis is his and is carried. (BEDA ubi supra. ) at 0725C also lacks the comma its four siblings carry: reproduced bare.

⚠ Two sigla attribute a Tobit gloss to Bede ON ANOTHER BOOK: (BEDA, in Esdr., tom. II.) at 0728B and (BEDA, in Esdram. ) at 0732B. Reproduced verbatim, uncorrected. This recension's attributions are known to be unreliable and the work page says so; passing a siglum through is not endorsing it.

(e) The frontmatter is copied verbatim, including its own defect

Chunk 2's heads: field reads "CAPUT XIV" without the closing period. The plate prints CAPUT XIV. with the period (§2), so the English ## head reads ## CHAPTER XIV. like every other, while the frontmatter carries the TEI's reading unaltered — because "copy the Latin chunk's frontmatter VERBATIM" is an unconditional rule and the pending tei-patch will fix both sides at once.

§2 · PLATE STATUS — READ, AND WHAT IT SETTLED

The foot-of-page apparatus for these columns was read by the orchestrator during this session, not by the translator, and the result is merged here. Method: archive.org patrologiaecurs04migngoog (on disk at raw/scans/pl113/), PDF page = (column + 11) / 2, verified against the running head. Columns 0725–0732 are pp. 368–371, all four read.

1. Foot-of-page apparatus: CHECKED ZERO — tier 3, coverage full, notes_recovered: 0

No foot rule and no numbered notes on any of the four pages; the columns simply end. This includes both mid-page work divisions — the case that lost three notes on PL 202 — p. 368 carrying the end of Nehemiah above the rule and LIBER TOBIÆ. below it, and p. 371 carrying the end of Tobit and LIBER JUDITH. below it. Both were read at the division.

Write this down as a checked zero, not as a skip. It is the second sampling of PL 113 (Ruth's five pages were the first), so the volume's "no conjecture apparatus" finding now rests on nine pages across two books.

2. ⭐ « Nephtlali. » at 0725C is OUR DIGITIZATION'S defect, not Migne's plate

The plate reads Nephthali. Read at 450 dpi and again at 900 dpi. The scan clips line-initial characters down the lower left column of p. 368, so the word survives as hthali. » — but the surviving letters are decisive: h-t-h-a-l-i, not h-t-l-a-l-i. -hthali cannot yield the TEI's -htlali. Corpus Corporum has metathesized the h and the l.

So the English reads « Nephthali. », and this is not a silent normalization of a plate reading — it is the correction of a transcription defect against the plate, exactly Ruth's significaRuth class (8968 cruces §2, 0534A). A data/tei-patches/8970.json entry follows, so the Latin side of the published page shows Nephthali too and the defect does not read as ours. The plate is untouched and unemended. The gloss's own etymology — Latitudo, "Breadth", of the tribe Naphtali — corroborates it.

Two treatments were live before the plate read and both are now moot, recorded because a blind reader will look here: [sic: Nephtlali] (rejected — a [sic:] may only mark a defect Migne's plate HAS, and it does not have this one) and rendering the printed form with a [var: Vulg. Nephthali] beside it (rejected for the same reason).

3. CAPUT XIV — the plate prints CAPUT XIV. WITH the period

Read at 450 dpi on p. 371. Our TEI dropped it; every other CAPUT head in the book carries one. Same digitization class as Ruth's VERS.9 missing-space patch, not Migne setting his last head differently. The English head therefore reads ## CHAPTER XIV. like the other thirteen. Second tei-patch entry pending; see §1(e) for the frontmatter.

§3 · FOUR UNCLOSED QUOTATIONS — Migne's, and they stay unclosed

Pattern 5's corollary: where Migne opens « and never closes it, the English mirrors the plate and supplies nothing. Four here, which is why the guillemet counts are 76/72 and not 76/76. All four were checked against the Latin twin individually; none is a transcription loss of the kind §2.2 describes (each ends at a hard structural boundary the compositor evidently treated as a close).

⚑ The 0728A case is the one to watch: the unclosed « swallows the two italic etymological glosses and the two proof-texts that follow, so a span-extractor reports one 60-word "lemma" there. It is not one. The lemma is Ego sum, inquit, Azarias Ananiae magni filius, (Tob 5:18) and everything after the [n:] is gloss.

§4 · THE LEMMA LIST — ALL 73 SPANS, CHECKED AGAINST THE VULGATE

Every guillemet span in the work, with what the check found — including the ones that agree exactly. A list of findings alone cannot be told apart from a report by an agent that never looked. Tobit's Latin text is genuinely unstable and this is where the value is.

4a. THE DIVERGENCES — four [var:] fired

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

• 0727B · Qui parcit virgae, odit filium (Prov. XIII). The Vulgate reads odit filium suum. ⚠ The English therefore reads "hateth THE son", not the Douay's "hateth HIS son" — deliberately, and this is precisely the 7a″ trap the brief names: the familiar cadence arrives before the printed Latin is read. No [var:], because dropping a possessive does not change what the proverb asserts; recorded here so the un-Douay article is visibly a decision and not a slip.
• 0727A · Caecitas enim ex parte contigit in Israel (Rom. II). Vulgate Rom 11:25 has quia caecitas ex parte contigit in Israel. A connective swapped to fit the new sentence frame — the citing author's, not a textual variant. No marker. The citation itself is wrong** (Rom. II for Rom. XI); see §4d.
• 0728A · Et vidimus gloriam ejus, quasi unigeniti a Patre (Joan. I). John 1:14 reads gloriam ejus, gloriam quasi unigeniti a Patre. An elision of the repeated noun, not a different reading. No marker.
• 0728C · Pater, transfer calicem hunc a me (Marc. XIV). Mark 14:36 reads Abba Pater, omnia tibi possibilia sunt: transfer calicem hunc a me. Migne quotes the head and the tail and drops the middle — an abridgment of the same family as usque ad, not a variant. No marker. (Luke 22:42's Pater, si vis, transfer calicem istum a me is closer in shape but is not what Migne cites; the citation stands as printed.)
• 0732A · Aperiens autem Tobias os suum (Tob 13:1). The Vulgate has Tobias senior. A dropped epithet in a lemma the gloss does not build on. No marker.
• 0730A · Benedicimus te, Domine Deus (Tob 8:17). The Vulgate continues Israel; the lemma simply stops where Migne stops it. Not a divergence.

4c. CHECKED AND AGREEING WITH THE VULGATE — named, not summarized

Tobit lemmata, chunk 0: Tobias. (1:1) · Cum irent, etc. (1:5) · Supra Naasson. (1:1) · Ita ut omnia, etc. (1:3) · Nihil tamen puerile. (1:4) · Omnia primitiva (1:6) · Cum vero factus esset vir (1:9) · Quem ab infantia timere, (1:10) · In conspectu Salmanasar, (1:13) · Cum autem venisset, etc. (1:16) · Sepulturam exhibebat, etc., · Denique cum reversus esset, etc. · Contigit, (2:10) · Sed immobilis, (2:13) · Insultabant reges, · Ut Sara, etc. (3:7) · Corpus meum sepeli. (4:3) · Ex filiis Israel. (5:7) · Tunc ingressus, (5:10) · Et dixit ei angelus. · Ego sum, inquit, Azarias Ananiae magni filius, (5:18) · Tunc paratis. (5:22).

Tobit lemmata, chunk 1: Profectus est, (6:1) · Et ecce piscis, (6:2) · Exivit. · Quem expavescens, (6:3) · Apprehende branchiam, (6:4) · Quod cum fecisset. · Exentera hunc piscem. (6:5) · Sunt enim haec necessaria. · Assavit carnes ejus, (6:6) · Et oportet. (6:12) · Quo audito verbo Raguel expavit, (7:11) · Postquam vero coenaverunt, (8:1) · Religavit. (8:3) · Et factum est. (8:11) · Duas quoque, (8:22) · Quatuor arietes. · Occidi fecit. · Et adjuravit Raguel Tobiam. (8:23) · Tamen obsecro te. (9:3) · At vero Raguel, (10:8) · Tradidit, (10:10).

⚑ Religavit. at 0729D was a live [var:] candidate and is not one. The active religavit looked like a divergence from a passive religatus est, and the gloss's own Angelus… in deserto ligavit seemed to be repairing it. The Clementine reads religavit too (Tob 8:3, Raphael angelus apprehendit daemonium, et religavit illud in deserto superioris Aegypti). Checked, agreeing, no marker. Recorded because the shape of the near-miss is the shape of a false positive a later agent will meet again.

Tobit lemmata, chunk 2: Praecedamus. (11:3) · Blandimento. (11:9) · Exsurgens. (11:10) · Tunc sumens. (11:13) · Albugo. (11:14) · Ingressa est, (11:18) · Et nunc, (12:14) · Et postquam illuminatus est Tobias, (14:1) · Quinquaginta namque et sex annorum lumen oculorum amisit, (14:3, word for word) · In hora autem mortis suae vocavit ad se Tobiam. (14:5) · Tobias abscessit ex Ninive cum uxore sua, et filiis et filiorum filiis, (14:14) · Sepelierunt. (14:16).

⚑ Exsurgens. at 0731A — CLOSED at the blind read: clean, no action, no [var:]. The two verbs are synonymous in this context and "Rising up" serves either, so nothing was conformed. The open question as I left it is kept below, because the shape of it is worth a later agent's time. The Clementine at Tob 11:10 has consurgens (consurgens caecus pater ejus coepit offendens pedibus currere); Migne's lemma prints Exsurgens, and the gloss picks it up as exsurgit populus Judaeorum. The prefix change is real but the two verbs are near-synonyms, the Tobit tradition is unstable exactly here, and I do not have a Vulgate text on disk to confirm the Clementine reading against my memory of it. Declined rather than fired on a half-check, and named so a reader with the text could close it in one look — which is what happened. ⭐ Declining to fire on an unverifiable half-check was right and produced the right outcome: had I fired a [var:] from memory, the work would carry a marker asserting a divergence that is not one.

Non-Tobit scripture quoted inside the glosses — the brief names these as exactly where the reach for a familiar English Bible happens, so each was read word by word: Latum mandatum tuum nimis (Ps 118:96 — exact) · Non potest civitas abscondi supra montem posita (Matt 5:14 — exact, word for word, Migne's Latin, not the Douay's word order) · Nolite, pueri, effici (1 Cor 14:20, minus sensibus, which etc. carries) · Surge qui dormis et exsurge a mortuis (Eph 5:14 — exact) · A sanctuario meo incipite (Ezek 9:6, minus the opening et) · Charitas patiens est, benigna est (1 Cor 13:4 — exact) · Poenitentiam agite, appropinquavit enim regnum coelorum (Matt 3:2 — exact; coelorum for caelorum is orthography, Pattern 9, not a defect) · Ego sum, inquit, lux mundi, qui sequitur me non ambulat in tenebris (John 8:12 — exact, the inquit interpolated per Pattern 5 and kept inside the guillemets) · Adjutor meus et liberator meus es tu, Domine, ne moreris (Ps 69:6 — exact) · Serpens erat callidior cunctis animantibus terrae (Gen 3:1, minus the opening Sed et).

4d. THREE CITATIONS ARE WRONG AS PRINTED — reported, NOT corrected

Runbook false positive 2: a following Ibid./ID., ibid. chain resolves against the printed sequence, so correcting a reference makes the Ibid. inherit the corrected key and resolve wrong. This book runs eleven (ID., ibid.) / (BED., ibid.) notes. Nothing was corrected.

• 0725C · (Matth. XV) on Non potest civitas abscondi supra montem posita — the Sermon on the Mount is Matt. V. The brief flags this one by name.
• 0727A · (Rom. II) on Caecitas enim ex parte contigit in Israel — Rom. XI:25.
• 0727D · (Deut. LII) on Constitui terminos populorum — Deuteronomy has 34 chapters, so LII is impossible; the verse is Deut. XXXII:8. This is the clearest of the three and still gets no correction in the note.

A fourth is a near-miss and is not an error: (II Cor. XIII) at 0727B on Charitas patiens est is I Cor. 13:4 — printed as II, reproduced as II.

§5 · THE BOOK'S OWN CRUCES

5.1 ⭐ 0727A · fidei lumen non admittit — the sentence says the opposite of its own argument. [cj:] FIRED.

Fatigatus a sepultura est caecatus. Qui enim infatigabilis in bonis operibus persistit, fidei lumen non admittit.

Rendered literally: "For he who persists UNWEARIED in good works does NOT admit the light of faith." The preceding sentence has just said that Tobit was blinded because he was wearied from the burial. So the printed clause asserts that the man who does not weary is the man who does not receive faith's light — the exact reverse of the paragraph.

Both negations are carried (infatigabilis, non), and admittit is rendered as what it means. No word was translated as its opposite to rescue the sentence.

The conjecture is amittit — "does not lose the light of faith" — one letter away, and it yields precisely what the argument requires: the unwearied man does not lose what the wearied man lost. This is the Pattern 18 shape exactly: admittit is a real Latin word in its correct person and slot, so Pattern 12 has nothing to wrap and Pattern 7 must render it, and the English then asserts something the author did not. Marked additively, so both readings stand on the page:

does not admit [cj: admittit; read amittit, "does not lose"] the light of faith

⚠ This is the one site in the work where the printed reading and the argument are in open contradiction, and it is the site a blind reader should open first.

5.2 ⭐ 0729A · Num ignoramus astutius ejus — a broken word inside a divergent quotation

Migne prints, for 2 Cor 2:11: « Num ignoramus astutius ejus [n: (II Cor. II)]. » The Vulgate has non enim ignoramus cogitationes ejus. Two independent things are wrong and each takes its own marker, per Pattern 14's clause that a [var:] may sit beside a [sic:]:

1. astutius cannot be the object of ignoramus. It is a real word — the neuter comparative of astutus, or the comparative adverb — and it has no grammatical slot here, which is Pattern 12's test 1 (the Pecte/ama/sternis class from 11632). The required accusative is astutias, "his crafty designs". So it is carried untranslated in italics inside [sic: astutius], not rendered.
2. Num for the Vulgate's non enim inverts the sentence frame from an assertion into an interrogative — and Migne prints a full stop, not a question mark. Pattern 8: the mark rides as printed, and this work prints zero question marks, so the English prints zero.

Rendering: « Surely we are not ignorant of his cunning [sic: astutius] [n: (II Cor. II)]. »

("cunning" was missing from my first version — "of his ___" is not English. See §7.1.)

⚑ On the "not": num is an interrogative particle that expects the answer no, and "surely… not" is its own English exponent — the same lexical fact as nonne → "is it not". No negation particle was added and none was dropped. The alternative renderings were both worse and both are recorded: "Are we ignorant of his…" (drops num entirely and asserts the opposite of the gloss's whole point, that Scripture makes the enemy's craft known) and a recast question (supplies a ? Migne does not print, breaking Pattern 8).

7a⁗ check, run explicitly: with the marker's contents deleted the clause still has its verb (ignoramus) and its interrogative particle (num) — the quarantine takes the object, not the force.

5.3 0728C · di bolum — split type, and what it takes down with it

Apprehendit Christus di bolum qui eum capere in cruce voluit, et moriendo vicit.

Pattern 10: word-division defects are carried as whole runs, spacing included, and marked [sic:]. [sic: di bolum] for diabolum, near-certain — the same clause's relative (qui eum capere in cruce voluit) and the sentence before it (Domino in cruce passo diabolus… advenit) both name the devil, and the word appears correctly four more times in the same paragraph.

7a⁗-b applies and was worked through: the quarantine is the size of a whole lexeme, and here that lexeme is the object of apprehendit. Checked against the three questions — the clause keeps its verb (apprehendit), carries no negation, and carries no conjunction or interrogative. Nothing governing was swallowed — but the OBJECT was, and that alone broke the sentence: my first version read "Christ laid hold of the [sic: di bolum]", a dangling article with no noun. Corrected to "Christ laid hold of the devil [sic: di bolum]"; see §7.1. ⚠ Note precisely where the reasoning failed: I ran 7a⁗'s three questions (verb? negation? conjunction?), got three clean answers, and concluded the marker could stand alone. 7a⁗'s three questions do not ask whether the clause still has an OBJECT.

5.4 0727A · spiritualer and 0727B · incantus — two more carried runs

• spiritualer for spiritualiter, in Ita spiritualer fatigatus dormit. A non-word; carried in [sic: spiritualer], and rendered "spiritually" in the running text beside it — see §7.1. My first version marked it INSTEAD of rendering it, which deleted the literal/allegorical hinge of the sentence and left the clause merely repeating "wearied".
• incantus for incautus, in qui levitati lasciviae, ac superbiae se incantus subjicit. The classic u↔n turned-type confusion. The [sic:] decision was close and the reasoning is recorded: incantus, -us is an attested late-Latin noun ("enchantment"), so the word is not simply non-existent — but as a nominative singular agreeing with qui it has no slot, which is again Pattern 12 test 1, and the conjecture incautus ("unwary") makes the clause construe at once. Marked AND rendered — "who incautiously [sic: incantus] subjects himself" — after the same correction; see §7.1.

5.5 The unexpressed subject — every supplied pronoun, and what decided it

The brief names this as the dominant defect class, and this book runs the Tobit-father / Tobit-son / Israel / Church / Gentiles allegory with the referent switching mid-sentence. Three sites needed a decision; all three are recorded, two took a bracketed supply rather than a bare pronoun.

(i) 0726C · quod nunquam peccatum faceret, sed ipsum timoris Domini spiritu impleret. Rendered "that [the boy] would never commit sin, but that [God] would fill him with the spirit of the fear of the Lord."

⚠ REVISED at the blind read (§7.2). My first version put [God] on BOTH verbs and is superseded. Both readings genuinely construe, so this entry keeps the losing argument in full rather than quietly replacing it — a reader who thinks I chose wrong should have the material to say so.

The reading now in the text: faceret of the SON, impleret of the LORD. It follows the lemma being glossed — Quem ab infantia timere, Tob 1:10, the son taught to fear God and to abstain from all sin — so the two limbs of the sed are the two halves of the verse, and the subject of impleret is named on the spot by timoris Domini. The decisive objection to my first version: peccatum facere is the New Testament idiom for committing sin (1 John 3:8), so "confessing that God would never commit sin" is a strange article for a father to teach his son.

The argument I first took, kept because it is not worthless: it turns on ipsum, which is not reflexive — a reflexive would be se — so the person it names is not the subject of impleret. That still rules out the reading English reaches for first ("that he would never sin, but would fill himself…"). The subject of impleret is therefore God and ipsum is the boy, which the shipped reading also holds. Where the two part is only whether faceret shares that subject: joined by sed under one quod, two subjunctives normally do. Against that stand the idiom and the lemma, and they carry the more weight.

Both verbs take a bracketed supply rather than a bare pronoun, because a bare "he" binds silently in a sentence whose referent changes mid-clause — Ruth's 0538C failure in this book's clothing.

(ii) 0728D · molitus est adhuc electis suis persecutionem commovere. Rendered "he [the devil] contrived still to stir up persecution against his elect."

Two readings construe and both are set out here, but I have taken one and it is in the text. The cum-clause subject is Dominus; the main clause has none expressed. For the Lord: suis is reflexive to the main subject, and "his own elect" then belongs to whoever is contriving. For the devil: molior is a scheming verb — one contrives a persecution, one does not contrive one for one's own people — and the allegory the whole paragraph runs is that the exposed enemy keeps striking at the feet (the elect) by which Christ walks on earth, which the very next clause states. The devil is the subject. The looser use of suus for a non-subject possessor is ordinary medieval Latin and is the weaker of the two arguments; the verb is the stronger. Supplied in brackets so the reader sees that the choice was made.

(iii) 0731A · tandem, scilicet, se Ecclesiae de gentibus… conjungat. Rendered "at last, that is to say, [the Jewish people] may join itself to the Church."

⚠ The bracketed supply was added at the blind read. A bare "it" was right as to NUMBER, and for that reason I thought the site was finished — but the nearest singular noun in the English sentence was "divine grace", so the pronoun bound to the wrong one anyway. Carrying a number faithfully is not the same as carrying a referent, and this is the same lesson as §7.1 in a different dress: what is correct in the analysis has still to arrive in the sentence a reader actually reads.

⚑ A number shift decides this and it is easy to miss. The same sentence runs ut… in libris suis cognoscant… multum gaudeant — plural — and then conjungat — singular. The plurals are rendered plural ("they may know", "they may rejoice") and the singular singular ("[the Jewish people] may join itself"), because that singular is populus Judaeorum, the collective the paragraph opened with. Smoothing the singular into a plural to match its neighbours would erase a shift the plate actually makes. The other candidate subject, gratia divina, is singular too but cannot take se conjungat here — grace joining itself to the Church is not what the passage is about; the Jews' incorporation is.

5.6 0732A · Tobias remanet apud Patrem — Migne's capital is doing the work

The literal figure is Tobias staying with his father Tobit; Migne prints Patrem capitalized, in a sentence whose other half is Angelus redit ad Deum and whose sequel is Christus… divinitate Patri aequalis. The capital is the allegory surfacing, and it is reproduced: "Tobias remains with the Father." Lower-casing it to fit the story would be tidying the plate; the reader can see both senses because the ambiguity is Migne's.

5.7 Punctuation the plate prints and the English keeps

• 0729D · a full stop mid-sentence, followed by a lower-case quia: quae a Deo deserta. quia ejus habitatione indigna. Reproduced, lower case and all — Pattern 8 forbids deleting a printed mark, and 11542 @1041D is the precedent for exactly this shape.
• 0726D · a paragraph that simply stops. et ante faciem ejus non subsistit — no terminal stop. None supplied.
• 0729B · a second one. tanto certius declinamus — no terminal stop. None supplied.
• 0731B · a semicolon inside a quoted prayer: dicunt, Domine Deus, illumina tenebras meas; a Domino illuminantur. Kept where he sets it.
• 0729A · a colon after a verb of saying: Aiunt enim physici: quia calore…. Kept.
• 0730A · « Et factum est. » etc. — a full stop inside the guillemets, before etc. Kept inside.

5.8 Verse and chapter ordering as printed

• CAPUT II opens at VERS. 10 and CAPUT V at VERS. 7; CAPUT III has a single gloss at VERS. 7. Migne's ordering, reproduced, no reordering, nothing supplied.
• Within chapters the verse numbers ascend throughout — no out-of-sequence verse mark anywhere in this book, unlike Jude. Checked, and recorded as checked.

5.9 Sentences left broken because the plate breaks them

• 0731A · sed offendens gressibus operum, donec renatus instructus in Christo… percipiat has no finite verb after sed. Rendered as printed, participle dangling: "but stumbling with the steps of its works, until… it may receive the light of faith and of working." Nothing supplied.
• 0730C · qui litteram legis septuaginta interpretes acceperunt. septuaginta interpretes is accusative and stands as a second object of acceperunt, so the printed sentence says they received "the letter of the law, the seventy interpreters". The sense plainly wants per septuaginta interpretes, "through the Seventy". [cj:] DECLINED — the marker's form requires a printed WORD to gloss, and what is missing here is a preposition, not a wrong word; there is nothing to put in the italics. Rendered as printed, and the conjecture lives here. ⚑ The blind reader could not decide between "through the Seventy" and "received the Seventy interpreters" either; recorded, on his instruction, as UNDECIDABLE rather than as a declined conjecture.** ⚠ My first draft of this clause read "received the letter of the law FROM the seventy interpreters" — a silent repair, caught and undone before the file was written. Recorded because it is the exact mechanism 7a describes, and because it is the one instance of that mechanism I caught myself.
• 0732A · Nos autem omnes hujus vitae labores debemus sustinere. omnes agrees equally with nos and with labores. Rendered "we all ought to endure the labours of this life." ⚑ The blind reader notes that word order and the quadragenarius numerus laboriosus argument mildly favour the other binding — "all the labours of this life" — and declined to churn a moderate-confidence ambiguity. Both readings recorded; the English stands as written.
• 0730A · Cantus pullorum rendered "The singing of the young birds", not "cock-crow". Pullus is the young of any bird and Migne's word for a cock is not on this page; the phrase is the head of an abridged Bede sentence (usque ad follows immediately), so the context that would license "cock" is precisely what the abridgment removed.

5.10 [cj:] candidates considered and DECLINED, with reasons

Pattern 18a closed the corpus census and set a high bar: the test is whether the faithful English asserts something false, not whether a better word exists.

• 0727C · Turbam notionum signat (of Sara). ⭐ DECLINED FIRST, REVERSED AT THE BLIND READ — and the reversal, not the reading, is the lesson. Nationum is one letter away and is what the book's allegory runs on everywhere else (Raguel populum gentium signat; Ecclesiam de gentibus; plenitudo gentium), Sara being the Church from the Gentiles throughout. I declined it because notio is a real word in its correct case and because the usque ad abridgment removes the body of the sentence, so the referent cannot be recovered from the printed text. That reasoning was sound about the conjecture and wrong about the consequence. Declining the marker left the site with no apparatus at all, so a reader met "the multitude of notions" with nothing to tell him the page was broken — which is worse than either available answer. Now [cj: notionum; read nationum, "of the nations"], exactly as admittit is treated two columns earlier. The reportable defect was the absence of a marker, not the choice of reading.
• 0726C · alibi dicitur Salmanasar. Reads oddly, since the lemma being glossed is In conspectu Salmanasar — one expects alibi dicitur Sennacherib. Declined: the printed sentence construes as it stands ("it is no contradiction that he is elsewhere called Salmanasar, who is also Sennacherib"), and the double-name claim it goes on to make is coherent either way. Rendered as printed.
• 0726B · Migne's commas in « Nolite, pueri, effici, etc. ». The commas make pueri look vocative ("be not, O children, made…"). Rendered "Be ye not made children" — the predicate reading — because 1 Cor 14:20 is nolite pueri effici sensibus with pueri predicate nominative, and because the gloss adduces the verse for a boy who did nothing childish, which needs the predicate. The commas are Migne's punctuation of a construction, not a change to it, and English has no exponent for them; Pattern 9's reasoning. Logged, not marked.

5.11 Negation — 19 particles, every one carried

Checked one by one against the Latin. non solum… sed (0725D) · non abstulit (0725D) · Nihil tamen puerile (0725D) · sine querela (0726B) · non poterat accedere (0726C) · nunquam peccatum faceret (0726C) · Non obloquitur (0726C) · ut nec memoria… remaneret (0726D) · non subsistit (0726D) · si… non liceret + nunquam scriberetur (0727A) · nihilominus (0727A) · non admittit (0727A — see §5.1) · Qui non corripitur non est filius (0727B) · non ambulat in tenebris (0728A) · ne moreris (0728A) · cui nil debebat (0728B) · nec in eo… quidquam habuit (0728B) · non diabolum, sed mortem (0728C) · nec ipsos tantum laedere permittitur (0729D) · nondum… susceperunt (0730C) · nemo prohibere potest, quin… (0730C) · quos diu non viderant (0731A) · nec ad sexagenarium… pervenerat (0732B) · animi levitatem non deserit (0732B).

None was dropped and none was added. Two are worth naming because carrying them produces strain the English keeps:

• 0727A, §5.1 — the site where carrying the negation makes the sentence contradict its own paragraph. Carried anyway.
• 0732B · diu vivens animi levitatem non deserit — the closing words of an usque ad, "living long, he does not forsake lightness of mind." Read out of its (absent) context this sounds like a criticism where the passage is praising. Rendered as printed; the sentence it belongs to is not on this page.

5.12 Read clean, and named so

Chunk 2 in its entirety carries no defect marker: 13/13 guillemets balanced, nine notes, six anchors, three usque ad, every lemma checked (§4c) with one declined near-miss (Exsurgens, §4c). The heavy apparatus in this book is all in chunks 0 and 1. CAPUT IV (0727C), CAPUT VII (0729B–C), CAPUT IX (0730B–C) and CAPUT XIII (0732A) are each a single gloss and each read clean.

§6 · WHAT A BLIND READER SHOULD OPEN FIRST

Rewritten after the blind read (§7); items 1–3 of the original list have been adjudicated and are now closed. Kept in place so the record shows what was offered and what came back.

1. ✅ 0727A, non admittit (§5.1) — offered as the call I most wanted a second eye on. Endorsed: correct and correctly formed, amittit certainly right, since the preceding sentence makes weariness the cause of the blinding.
2. ✅ 0729A, Num ignoramus (§5.2) — offered with my reasoning exposed for testing. Tested and holds: num carries its own negative expectation, no particle was invented, and II Cor. II requires the affirmative-knowledge sense, so both Num and a possible underlying Non yield the same English. No [cj:].
3. ✅ 0728D, subject of molitus est (§5.5 ii) — offered as devil-or-Lord. Devil endorsed, on evidence I had not cited: the lemma is Tob. VI, where the fish palpitates before Tobias's feet, and the gloss's next clause supplies exactly that (hi enim sunt pedes ejus). The Lord contriving persecution against his own elect is not available.
4. ✅ 0727C, notionum / nationum (§5.10) — reversed: now carries a [cj:].
5. ✅ 0731A, Exsurgens (§4c) — closed clean, no [var:].

What is still open, and it is short:

• 0726C, §5.5(i) — the one genuinely two-sided call left in the work. The text now reads faceret of the son and impleret of the Lord; both readings are set out in full and a reader who prefers the other has everything needed to argue it.
• 0730C, §5.9 — recorded as undecidable by two readers.
• 0732A, §5.9 — omnes with nos or with labores; both recorded, English unchanged.

§7 · THE BLIND POLARITY READ — one systematic defect, and it was invisible

A blind reader who never saw this file read all three chunks. On the classes this book was most exposed to it came back clean: no scripture conformation anywhere, and he named the places the pull was resisted — Migne's omission of suum at Prov. XIII preserved against Douay's "his son", Marc. XIV not conformed to Douay's "chalice", Domine Deus left without meus, and the anacoluthon at 0731A reproduced broken with no main verb supplied. Guillemets 1:1 in all three chunks including the four Migne leaves unclosed; zero question marks; 17/17 on usque ad with no verb ever supplied.

Nine repairs were required. All nine are applied. Seven were single sites (§5.5 i, §5.5 iii, §5.10, and the four below); two were systematic.

7.1 ⭐⭐⭐ THE MECHANISM — a [sic:] does not stand in for the translation. 4 of 4 sites.

Every [sic:] in this work swallowed the word it marked. The convention is that a corrupt word is Englished for its evident sense AND marked — the marker annotates provenance, the running text still has to say what the sentence says. I used the marker instead of the rendering, so four content words vanished from the English:

⛔ Why this is the most damaging shape of error available here: it is invisible to every automated check we have. verify-english.mjs sees a well-formed marker whose contents are verbatim in the Latin twin, and passes. Column, note, guillemet and question-mark parity are all untouched. Nothing counts words. The four sites survived my own marker audit precisely because each marker was individually correct — the defect was in what stood beside it, and there is nothing to grep for an absence.

⚠ It was mine, not a house ambiguity: peers in this same batch applied the rule correctly (8982 wrote "will not desire" beside [sic: de iderabitis]; 8985 supplied the sense beside monebu).

⭐ THE TEST, IN ONE LINE, FOR 8948 AND JUDITH (8960):

A [sic:] marks the plate's defect; it does not stand in for the translation. If the English sentence cannot be read aloud without the bracket, the rendering is missing.

Run it on every marker before filing a chunk — delete the bracket and read the clause. It takes seconds and it is the only thing that finds this class. (Mechanically: after stripping all markers, an article or preposition left stranded on punctuation, or a doubled mark, is the signature. Both patterns now return zero across all three chunks.)

⚑ A second finding, worth carrying forward: 7a⁗'s three questions are not sufficient. At 0728C I ran them — verb? negation? conjunction? — got three clean answers, and concluded the marker could stand alone. The quarantine had taken the object. 7a⁗ asks what governs; nothing in it asks whether the clause still has the thing it acts on.

7.2 The other repairs

• ⭐⭐⭐ 0728C · et moriendo vicit had been attached to the wrong subject. It is a main clause resuming Christus, not a second limb of the qui relative; my word order put it inside the relative, so the page said the devil conquered by dying — the inversion of the paragraph's whole argument. Now "…who wished to seize him on the cross, and by dying [Christ] overcame him." The swallowed di bolum above compounded it by removing the noun that would have anchored a reader.
• ⭐⭐ 0729C · quos is PLURAL and resumes fidelium. "he restrained him, whom, although he is sometimes permitted to tempt…" put "whom" straight after singular "him" (the devil), so the surface reading was that someone is permitted to tempt the devil. Latin morphology carried the antecedent; English has to carry it by position. Now "he restrained him by ceasing from the hurting of the faithful, whom, although he is sometimes permitted to tempt for a trial, he is nevertheless forbidden to overcome."
• ⭐⭐ 0727C · notionum stood with no apparatus at all — see §5.10.
• ⭐⭐ 0726C · the supplied [God] made God the one who might commit sin — see §5.5(i).
• ⭐ 0731A · "it may join itself" bound to "divine grace" — see §5.5(iii).
• ⭐ 0728D · "against his elect" made the elect the devil's. Suis is loose for the Lord's, as the next clause proves (hi enim sunt pedes ejus… qui per omnia regnat in coelo). Now "against [the Lord's] elect".
• ⭐ 0731A · statim sits on adventuram, not on edunt — the grace is to come straightway, the declaring is not done straightway. Now "they declare that the grace of the Holy Spirit is straightway to come."

7.3 ⭐⭐ Punctuation — and why the totals could not have found it

Five of Migne's marks had been lost or downgraded, and six of ours supplied. Restored: 0726D pertimescit: · 0729A aperuit: · 0730B hostia viva: · 0732B annuntiant: · 0731C acceperit;. Removed: the supplied colon at 0725C (he prints a comma) and eight supplied semicolons (0725C, 0726B, 0726C, 0728A, 0728C, 0728D, 0730B, 0730C).

⛔ The trap, and it is the transferable part: in chunk 0 the colon totals CANCELLED EXACTLY, 19 against 19. One of his destroyed at 0726D, one of ours invented at 0725C. A count-based check reports that chunk clean. Only a positional check finds it — and this is the 8989 defect from the previous batch, repeating.

So the check is now run per column band, not per chunk. Splitting both texts on the column anchors (which are verified 1:1 and therefore partition the twins identically) and comparing colons, semicolons and guillemet-opens band by band: 0 mismatched bands across all three chunks. That is the check that cannot be faked by cancellation, and it is what a later guillemet book should run instead of a total.

⚠ On the semicolon removals specifically — the corpus-wide ruling declining this class elsewhere in this batch holds that it is a rate question, not an absence, because Migne uses semicolons freely. That ruling does not reach this book: he prints ZERO in chunk 0 and ONE in chunk 1. Against that, four in my chunk 0 was an absence. The exception is narrow, rests on his own practice in this book, and should not be generalized to 8948 or Judith without counting their plates first.

7.4 Post-repair state

verify-english.mjs clean. Per-band parity: columns 12/10/6, notes 29/19/9, guillemets 37+34 / 26+25 / 13+13, question marks 0/0/0, prose colons 19/14/3, prose semicolons 0/1/5 — every one matching the Latin twin, and matching band by band, not merely in total.