Epistola ad Ephesios

Translated 2026-08-19, one translator, 6 chunks (5,309 Latin words), on Opus. Structure: ARGUMENTUM + CAPUT PRIMUM (chunk 0); CAPUT II (1); CAPUT III (2); CAPUT IV (3); CAPUT V (4); CAPUT VI (5).

Apparatus: 10 [var:] · 4 [sic:] · 2 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 14 [n:] (exact [n:] parity with the Latin twins). verify-english.mjs 8991: clean.

Marker parity verified after writing, chunk by chunk, with marker contents excluded from the italic count:

Every figure is identical to its Latin twin's. Colons and semicolons were compared band by band, not by total (brief test 3), and eight English marks were found and removed that Migne does not print — they are named in §7 so the method can be told from the claim. Zero guillemets and zero em-dashes in either language.

Conventions follow GLOSSA-NT-AGENT-BRIEF.md: VERS. n.-- passes through verbatim; the lemma is Englished; usque ad → as far as (shipped 8978/8970/8980 convention, one occurrence here, at 0589C, with no verb supplied); notes ride as [n:], contents untranslated.

Pattern 17. The second person in this work is overwhelmingly plural (vos, vobis, estis, scitis, sciatis, impleamini, redarguite, obedite) — the Ephesians as a body — and is "you". Six passages address a single hearer and take thou: 0588D distingue … intelligas; 0590A cognosce … audis … vides … tuo (Augustine's inner-man passage, where cognosce is singular and putetis … audite three words later is plural — Migne's own shift, reproduced); 0591C dicas … putes … promerui … emisti; 0594B facis; 0594C operaris … tribuis … perseveras … facias … amas; 0596D te maculat … consentis … redarguis (0598A). The number changes are Migne's, not ours, and none was smoothed.

§0 · PLATE STATUS — a SPOT-CHECKED zero, and it is not a proof

Per the launch: cols 587/588 (p. 299, the Galatians→Ephesians division) and 593/594 (p. 302, full width, 300 dpi) were read at the plate this session by the orchestrator. No foot-of-page apparatus and no asterisk-keyed EDIT. note. So no [cn:] is expected in this work and no conjecture of Migne's was available to settle anything below. The two [cj:] of §3 and the four [sic:] of §4 therefore stand on our reading alone.

⚑ It is a sample of two openings out of eight, not the whole work. Three of the sites below are exactly the kind an editor would have queried, and if a fuller plate read is ever cheap, these are the columns to read:

• 0595B, mancida — a hapax in the whole corpus (§4a). A nineteenth-century editor querying anything in this work would have queried this.
• 0600C, rectus for tectus (§3b) — the conjecture is forced by the gloss's own next clause, and it is the sort of one-letter slip Migne's Forte notes are made of.
• 0599C, nomo for homo (§4c).

§0a · A form that is suspiciously TIDY, raised per brief §4 and left unmarked

0596B · indolorii. Migne glosses Desperantes with Vel indolorii, quod ex Graeco trahitur — a transliteration of ἀνάλγητοι that exists nowhere else in this corpus and in no lexicon I can appeal to. It is not flagged as broken type: it is offered by the glossator as a Greek-derived coinage and immediately explained (id est de peccato non dolentes), so it is doing precisely what a technical term does. Carried into the English roman and untranslated, as Migne sets it roman. No [sic:] — the register rule for untranslatable technical Latin covers it, and marking it would say the plate is broken where the glossator is coining.

I record it here as a checked negative of the opposite kind to brief §4's three positives: nothing here reads too tidy for its context; the rough passages (0595B, 0599C, 0600C, 0598D) are rough on the page and stayed rough in the English.

§1 · THE LEMMA DIVERGENCE LIST — every italic span collated against the Clementine

197 italic spans across the six chunks: 196 lemmata (the launch's list, in order) plus the one usque ad at 0589C, which is Migne's abridgment formula and not a lemma. ⚠ The launch states 202 spans; the list it supplied contains 196, and the Latin files contain 197 italic spans in total. I report the count I can verify against the files. Every one of the 196 was read word by word against sources/vulgate/clementine-flat.txt before its English was written, and each rendering was then held fixed against every re-quotation of it in the gloss below.

Agreements are listed as well as divergences (§1c), because a list of findings alone cannot be told apart from a report by someone who never looked.

1a · The ten that diverge — all marked [var: …]

1b · Divergences examined and deliberately NOT marked, with the reason

Real differences from the Clementine, each weighed against Pattern 14's bar and declined. A blind reader will meet all of them and must be able to tell a decision from an oversight.

• 0588C · Sanctis omnibus. — Eph. 1:1 omnibus sanctis. Word order only, and English has no exponent for it: "To all the saints" is the only natural rendering of either order. This is the Pattern 9 principle (a Latin feature with no English exponent, logged not marked), not a conformation — no word is altered, added or lost. ⚑ Same class, same decision: 0589C Omnia operatur for Eph. 1:11 operatur omnia.
• 0589A · In dilecto Filio. — Eph. 1:6 in dilecto Filio suo. Terminal clipping of a lemma; nothing at stake. Same class: 0589C In quo et nos (Vulg. in quo etiam et nos), 0589C In quo vos (Vulg. in quo et vos), 0592D ipse est pax nostra (Eph. 2:14 Ipse enim est pax nostra, a back-reference, not a proof-text), 0595D Et in aedificationem (Eph. 4:12, an added Et), 0597B Spiritum sanctum, (Eph. 4:30 Spiritum Sanctum Dei), 0599A Ego dico in Christo, (Eph. 5:32 ego autem dico). In none does the gloss touch the missing or added word. The precedent is 8990 @0611D, where an added et was weighed and declined on exactly this test.
• 0589C · Omnia, quae in coelis. — Eph. 1:10 instaurare omnia in Christo, quae in caelis. The lemma splices across an elided in Christo which the gloss supplies four words later in its own voice (In ipso. Id est per Christum). Abridgment, not variance.
• 0592A · vos qui gentes eratis, — this string is in no verse of Ephesians, and that is the point: it is the reading the glossator is reporting (In quibusdam codicibus reperitur scriptum esse: qui, ibi: vos qui gentes eratis). A [var:] here would report as a textual divergence the very thing the sentence exists to describe. Declined, and it is the clearest decline in the work.
• 0596C · In exspoliatione corporis carnis, — Col. 2:11 in expoliatione. Orthographic (exsp-/exp-), Pattern 9, and the identical call was made and declined in the Colossians work at 0612D (§6 below).
• 0597B · Et nolite contristare. — the Clementine on disk prints constristare at Eph. 4:30 (and again at 2 Rg. 13:21), which is the source's own oddity, not Migne's; the ordinary form contristare stands at Jer. 22:3 and Mk. 6:26 in the same file. A [var:] turning on this would be a claim about a typo in the comparand. Declined, and flagged here so a later collator does not re-raise it.
• 0599D · Contra spiritalia. — Eph. 6:12 contra spiritualia nequitiae. The syncopated spiritalia is attested medieval orthography, Pattern 9.
• Orthographic throughout, all Pattern 9, none marked: charitate/caritate, coelis/caelis, coelestibus/caelestibus, caetero/cetero, unquam/umquam, quodcunque/quodcumque, Tychicum (Migne's spelling here; 8990 has Tichicum).
• Capitalization is not variance. Migne capitalizes In dilecto Filio., Spiritum sanctum, and Mundi rectores. variously against the Clementine's lower case. English has no exponent.

1c · The eleven scripture quotations from OTHER books — checked first and hardest

This is where the reach for a remembered English Bible happens. Six agree with the Clementine exactly and are recorded as agreements; five diverge and are in §1a. Named individually:

• 0589C–D · Non vos me elegistis, sed ego vos, etc. (John 15:16) — agrees, with the verb gapped: the Vulgate has sed ego elegi vos, and Migne's etc. closes an abridged quotation of exactly the kind §1b covers. English gaps it too — but I you — which English syntax permits and which keeps the elision visible. No marker.
• 0591D · Surgam, et ibo ad patrem meum [n: (Luc. XV)] (Luke 15:18) — agrees verbatim.
• 0594A · Fili, dimittuntur tibi peccata tua [n: (Matth. II)] — ⭐ the words agree verbatim with the Clementine, but at MARK 2:5, not at the Matthew the note cites. Matt. 9:2 reads Confide fili, remittuntur tibi peccata tua. No [var:] was taken, and the reason is that Pattern 14 measures the printed words against the received text, and these words ARE the received text — of another Gospel. ⚑ And the note's own numeral is evidence for that reading: II is Mark's chapter, not Matthew's, so what is wrong on the plate is the book abbreviation, not the quotation. The note rides verbatim; the observation lives here. See §5.
• 0598C · Omnium me servum feci [n: (I Cor. IX)] (1 Cor. 9:19) — agrees verbatim.
• 0599D · Et mundus eum non cognovit [n: (Joan. I)] (John 1:10) — agrees verbatim, opening Et included.
• 0596C · In exspoliatione corporis carnis, (Col. 2:11) — agrees but for the orthography of §1b. ⚑ This is a quotation of the parallel epistle inside my own work; see §6.
• Diverging, in §1a: 1 Cor. 15:41 · John 14:2 · John 1:16 · Jer. 31:22 · Eccles. 7:16–17 · Matt. 10:34.

1d · Scripture the gloss uses WITHOUT italics — checked, and correctly unmarked

Roman-set scriptural matter is the glossator's prose, not a quotation, and Pattern 14 does not reach it (the ruling is 8990 §1d). Named so a blind reader can see they were noticed:

• 0599C apud quem non est iniquitas (Rom. 9:14 / 2 Chr. 19:7) — roman. ⚑ The identical phrase is roman in the Colossians work at 0610A and was left roman there.
• 0599B Non jam duo, sed una caro sunt — a recast of Matt. 19:6 (jam non sunt duo, sed una caro) with the negative and the verb both moved. Roman; rendered as the glossator recast it.
• 0599B Sicut sponso alligavit mihi mitram, et sicut sponsam ornavit me ornamento — offered as Isaias (61:10), where the Clementine reads quasi sponsum decoratum corona, et quasi sponsam ornatam monilibus suis. A very large difference, and still no [var:], because the span is roman: this is the glossator quoting from memory or from an Old Latin text, in his own prose. Rendered as printed. ⚠ If any single decline in this work deserves a second opinion it is this one — the divergence is bigger than several I did mark — and it is declined on the form of the span, not on the size of the difference.
• 0598D vestis ejus in monte tanquam nix dealbata effulsit (after Mark 9:2) — roman.
• 0592C Ipse enim est lapis angularis — the glossator's own prose, roman.

§2 · THE UNEXPRESSED SUBJECT — every supplied pronoun, named with its evidence

Sixteen places where English had to print a pronoun the Latin does not. Where two readings construe, both are set out and the one adopted is stated.

§3 · THE TWO [cj: …] — the printed word is REAL and the faithful English misleads

3a · 0590A, voluntas for vocatio

Migne prints, in the second construal of Eph. 1:18:

Et sensus est: Ut sciatis quae sit, scilicet voluntas ejus, id est quantum valet spes vitae aeternae ad quam vocat.

The lemma is Ut sciatis, and the verse it opens is ut sciatis quae sit spes vocationis ejus. The gloss is supplying the noun the lemma elides — and the noun the verse supplies is vocatio, not voluntas. The control is inside the same sentence: ad quam vocat picks up vocationis and answers nothing in voluntas. Rendered as printed ("namely his will") with the gloss beside it: [cj: voluntas; read vocatio, "calling"].

⚠ The rival considered and rejected: that voluntas is right and looks back to Eph. 1:9 (sacramentum voluntatis suae). It loses because the whole "Vel ita" is announced as a re-reading of spes (ut spes proprie accipiatur) and because ad quam vocat has no work to do under it. Naming the rival without adjudicating is what the brief forbids; this is the adjudication.

3b · 0600C, rectus for tectus

Marcus dicit apostolos calceatos sandaliis, vel soleis, ut neque pes sit rectus, nec nudus ad terram, id est nec occultetur Evangelium, nec terrenis commodis innitatur.

rectus is a perfectly good adjective, so Pattern 12 has nothing to wrap and Pattern 7 renders it — and the English then says the apostles were shod so that the foot should not be straight, which is not a hazard anyone was guarding against. The gloss decodes its own figure in the next clause, and the decoding is a strict pair: neque pes … [1] · nec nudus ad terram [2] ↔ nec occultetur Evangelium [1] · nec terrenis commodis innitatur [2]. Term [1] is concealment, which is tectus, "covered"; and tectus/nudus is the standing opposition that rectus/nudus is not. Rendered as printed with [cj: rectus; read tectus, "covered"].

⚠ The rival considered and rejected: rectus in the sense "upright, unbent", i.e. the foot not rigid. It loses on the decoding, which says nothing about rigidity and everything about hiding, and it leaves nec occultetur Evangelium answering no member of the pair.

⛔ Neither [cj:] removes anything. Migne's word keeps its English in both; the conjecture stands beside it, per Pattern 18's additive form, and both printed words appear verbatim in the Latin twin as the verifier requires.

§4 · THE FOUR [sic: …] — carried type, with the sense kept OUTSIDE the quarantine

Every one was closed against the brief's fourth question — does the clause still have the thing it acts on? — as well as 7a⁗'s three. The mechanical strip test was run on all six files: with every marker deleted, no stranded article, preposition or doubled mark remains, and every sentence can be read aloud.

4a · 0595B, mancida

Talis actio nec frangitur negotio, nec turbulenta est, nec mancida, nec audax, nec fugax, nec praeceps, nec jacens

mancida is a hapax in the entire corpus — one occurrence, this one — against a neighbouring family of 30-odd mancip- forms and no other mancid- anywhere. It is not a Latin word. Carried untranslated in italics, with its nec outside the marker: nor [sic: mancida]. The negation is one of seven in a list and it is present and on its own member; nothing is governed by the quarantined run, so nothing left the English with it.

Conjecture: mancipata, "enslaved" — the reading the mancip- family supports and the only one that fits a list of things an action can be. manca, "maimed", is the alternative and is palaeographically closer; it loses only because mancip- is what this corpus actually writes. Recorded as a conjecture in both directions rather than settled, because the plate would settle it and the plate for col. 595 was not read.

⚑ Note the paragraph ends with no terminal stop (nec jacens). Pattern 8: never supply one. The English stops where he stops. Same at 0596C (haec vetustas et novitas) and at 0599B, the work's last sentence of chunk 4 (ad litteram dicitur).

4b · 0598D, cas for eas

debent mori pro salvandis uxoribus, et verbo vitae eas instruere, ut exhibeant cas Deo.

A one-letter non-word, and — this is the point — it is the clause's OBJECT. This is exactly the shape that took four content words out of Tobit's English. Rendered that they may present them [sic: cas] to God: the object is in the English in plain words, the broken type is marked, and neither stands in for the other. The conjecture is eas, which the same sentence prints correctly eleven words earlier (eas instruere).

4c · 0599C, nomo for homo

ut nomo homini conditionis vinculo subderetur

One occurrence corpus-wide against 1,785 of homo, and it is the clause's subject. Rendered that man [sic: nomo] should be subjected to man by the bond of condition — subject in the English, type marked. Conjecture homo, near-certain, and confirmed by homini in the same four words.

4d · 0590D, ab ipse for ab ipso

Secundum deitatem: quia omnes ab ipse facti

A nominative governed by ab. This one sits on the Pattern 9 / Pattern 12 boundary and the call is argued rather than assumed: Pattern 9 exempts Latin features with no English exponent (gender, section labels), and English case after a preposition has none either — "by him" is the only rendering there is. But Pattern 12's second test decides it the other way: where the English has no exponent for the defect, rendering it silently repairs it, and a preposition governing a nominative is a fact about this printing, not a systemic feature of Latin. So: rendered ("because all were made by him") and marked ([sic: ab ipse]), which is the only treatment that leaves both the sense and the evidence on the page.

⚠ The rival I rejected: leaving it to a crux line with no marker, on the Pattern 9 analogy. It loses because a reader of the English would have no way to know anything was wrong, and the facing Latin — which is what makes normalization safe for gender and section labels — shows only two letters that a reader would take for our transcription slip rather than Migne's.

§5 · MIGNE'S OWN PRINTING — reproduced, not tidied

• 0590D · (Joan. X) on a quotation of John 1:16. Note verbatim, unaltered.
• 0594A · (Matth. II) on words that are Mark 2:5 (§1c). Note verbatim, unaltered.
• 0595D · septemque filiae Philippi — Acts 21:9 gives Philip four daughters. Migne prints septem. Rendered the seven daughters of Philip. Roman prose, so no [var:] reaches it; and it is a real word in its slot, so no [sic:] either. Pattern 7, straight: the English says seven because the plate says seven.
• 0598D · quae significabat. Ecclesiam omni macula peccati mundatam — a full stop breaking the middle of a construction, leaving Ecclesiam an accusative with no governing verb. Pattern 8's founding class (11534 @1050D). The stop is preserved and the accusative is rendered as it stands: which signified. The Church cleansed from every spot of sin, which… ⛔ Merging the two into one sentence is precisely the deletion Pattern 8 forbids, and it is the single most tempting repair in this work.
• 0594C · quare isti dat, non vero illi and, eleven lines later, quare autem illi det, isti non det — Migne reverses the demonstratives between the two statements of the same question. Both rendered as printed (why he gives to this man, but not indeed to that / why he should give to that man, and not give to this). Not a defect of ours to fix; the reader can see the reversal only if we carry it.
• 0600B · Recte lorica justitiae comparatur followed by Aliter: Justitia loricae potest comparari. Migne compares the breastplate to justice first and justice to the breastplate second — the terms are genuinely swapped, and the Aliter announces that something has changed. Rendered in Migne's direction both times. Reversing the first to match the second would be the tidy reading and would delete the point of Aliter.
• 0593D · Hujus rei gratia. twice, the second re-quoting the first inside the gloss. Reproduced twice; not a dittography (Pattern 11 does not reach a deliberate re-quotation), so no [d:].
• 0597A · Sol non occidat. twice — two separate glosses on the same lemma, as Migne sets them. Not a doubling.

§5a · One word rendered two ways, declared

0593A–B · investigabiles. The lemma is Eph. 3:8's investigabiles divitias, where the prefix is privative: unsearchable. The gloss then writes sed per gratiam et revelationem sancti Spiritus investigabiles sunt fidelibus — and under the privative sense that sentence denies what its own sed asserts. Latin investigabilis genuinely carries both senses, and the sed fixes the second as the positive one. So: lemma Unsearchable, gloss searchable.

⚑ This is declared here rather than buried because rendering one word two ways is the shape a silent repair takes. It is not one: no word is translated as its opposite to rescue anything — the Latin word is lexically ambiguous, the correlative disambiguates it, and the alternative (rendering both "unsearchable") would make Migne assert that grace leaves the riches exactly as unsearchable as nature does, which is the reverse of the sentence.

§6 · THE COLOSSIANS PARALLEL — where it was consulted, and what it decided

The launch names 8990 as the nearest precedent and warns that the parallel is a hazard as well as a help. Four consultations, all recorded:

1. ⭐ 0596D · Eph. 4:24, Qui secundum Deum creatus est. The Colossians work carries, at 0614C, a gloss that cites this book by name: The renewal of the mind is said to be according to God, as is read in the epistle to the Ephesians, or secundum imaginem Dei, as is read here. So the Colossians glossator's cross-reference is a witness to what Ephesians ought to print — and Migne prints exactly secundum Deum. The two books agree, and the check is worth more as an agreement than a divergence would have been: it confirms both the Colossians [var:] at 0614C and this book's lemma at once. No marker taken here; none needed.
2. 0596C · In exspoliatione corporis carnis (Col. 2:11), quoted inside this book's gloss on Eph. 4:23. The Colossians cruces declined a [var:] on the identical exsp-/exp- orthography at 0612D. The same call is made here, for the same reason, rather than re-derived.
3. 0597D · Eph. 5:5, Avarus — the covetousness-is-idolatry gloss, which runs parallel to Colossians 0614B (Et avaritia). ⚠ The two glosses are NOT the same text and were not allowed to lend each other wording. Colossians argues from solus (ne solus nomen deitatis habeat … ut solus usurpet) — the site whose negation attachment is the brief's worked example. Ephesians has no solus anywhere in its version and argues instead from praefert: the covetous man prefers God's goods before God, sicut idololatria Dei honorificentiam. Rendered from the Ephesians text alone.
4. 0599C · Eph. 6:5, Servi, obedite ‖ Col. 3:22. Checked because the household code is where the pull toward familiar wording is strongest. Migne's Ephesians lemma stops at obedite and the gloss is a servitude-and-sin argument with no Colossians counterpart. Nothing imported.

⚑ The general point. Every one of the four came back "no import needed" — which is the result a check should usually produce, and is the reason to record it. The one place the parallel actually did work was (1), and there it worked as confirmation, not as a source of wording.

§7 · THE TESTS FROM THE BRIEF, RUN — including what they caught

• Test 1 ([sic:] must not swallow). Run mechanically on all six files. Four markers, all four with their content word or particle present in plain English outside the bracket (§4). Zero stranded articles or prepositions.
• Test 2 / 2a (negation attachment). Every printed non · nec · neque · nisi · ne · nemo · nihil · nullus · nunquam · nusquam · non solum was listed with the single Latin word it attaches to, and the host was proved from the sentence's structure — the sed correlative, the sicut … ita pivot — never from what read well. The trigger population (a negative resting on a verb with a quantifier, correlative or predicate nearby) fired at these sites, and at each one the host is a phrase, not the verb, which is where a fluent English rendering would have put it:
• 0589C Sors … quia non est hominis electio, sed voluntatis Dei → "not the choice of man, but of the will of God" (host: hominis electio, not est).
• 0589D Ad impetrandum non obtendo merita, sed quod Deus et Pater Christi → "I put forward not merits, but that he is God…" (host: merita, the object the sed replaces).
• 0590B Nulla enim forma corporis Deus finitur → "by no form of body is God bounded" (host: the quantified ablative; putting it on the verb would say God is not bounded, full stop).
• 0591C Gratia praecessit meritum tuum: non gratia ex merito, sed meritum ex gratia → rendered as the bare correlative pair, "not grace from merit, but merit from grace", because the sed replaces the whole phrase and any verb we supplied would move the negative onto it.
• 0591C emisti, non gratis accepisti → host gratis, contrasted with emisti.
• 0591D Non hoc ideo dicit, quia … sed quia → host ideo, the reason, not dicit.
• ⭐ 0592B Non enim regionibus longe est quisquam a Deo, sed affectibus — the sharpest site in the work, and the Esther-class trap. quisquam is a negative-polarity item and the fluent English is "no one is far from God by regions" — which puts the negative on the quantifier. The sed affectibus is an elliptical ablative and shows the scope is on regionibus. Rendered "For it is not by regions that anyone is far from God, but by affections", which keeps quisquam as "anyone" and the negative where the correlative puts it.
• 0593A Non enim est meum sed Dei → host meum; and beside it nullae mihi gratiae reddendae, sed Deo, where nullae quantifies gratiae but the sed pivots on mihi/Deo — the two coincide in English ("no thanks are to be rendered to me, but to God") and the coincidence is stated rather than assumed.
• 0593B easque implet, non illo opere providentiae, quo … sed eo quo → host the ablative phrase, not implet.
• 0594D fides nec in solo Patre integra est, nec in Filio solo perfecta — two negatives, both on phrases containing solo, which is the pivot. "neither entire in the Father alone, nor perfect in the Son alone". Attaching either to the verb would deny that faith is entire in the Father at all.
• 0594D Non ut sint (plenus Deus) … sed ut perfecte sint pleni Deo → host the first ut clause; the sed supplies a second one.
• 0598C Non solum auditores praelatis, sed etiam praelati subditis → host the whole first phrase. Textbook non solum … sed etiam; recorded because it is the exact construction the brief's 8987 case mishandled.
• 0599A non corpus, sed corruptiones et pondus ejus odit → host corpus, the object.
• 0599D non timore subdolo, sed fideli dilectione · 0599D Non ideo sic appellantur … sed · 0600A Non dominantur daemones mundo … sed peccatoribus · 0600A non pro parva re, sed pro coelesti haereditate · 0600B non simulatorie, sed in veritate · 0600D non casu, sed de proposito — all hosts are the phrase the sed replaces, and none is the verb.
• Test 2b (anchor position, not merely presence). All 54 anchors were placed against the word they stand beside in the Latin, not merely in order. Three were moved during the pass after checking: 0588C (stands before describit, so before "describes"), 0589D (stands before me inside the John 15 lemma, so You have not chosen [0589D] me), 0590B (stands before aperit, so before "discloses"). 0597A splits a phrase mid-construction to match Migne — be indignant with your [0597A] own selves for indignamini vobis [0597A] ipsis — which is deliberate and is the 8995 @0605B precedent.
• Test 3 (punctuation by band). Colons and semicolons compared band by band against each Latin twin. Eight English marks were found that Migne does not print and were removed: 0589B (a supplied ; at non reciperent Christum, where he prints a comma), 0591A and 0591B (two supplied : after "Or", where he prints Vel bare), 0592A (a supplied ; at sub peccato fuissent), 0593A (a supplied : after Scilicet, ibi), 0594D (a supplied ; at Existit vero ex occulto illa charitas, where he prints a comma), 0595B (a supplied ; at gratiae attribuenda), 0598A (a supplied : after Sicut et Judas). Every band now matches at zero. ⚑ This is the check that earns its keep: the totals for three of those chunks would have been off by exactly one, and two of the eight sat in chunks whose other bands were already clean.
• Test 4 (does it ASSERT what he asserts). Run last, with the counts known clean, paragraph by paragraph. It produced three changes that no containment check would have seen: 1. 0591C, ut dona Dei consequantur merita mea. dona Dei and merita mea are both neuter plural and either can be the subject. "that the gifts of God may follow my merits" is the fluent reading and is the exact reverse of the paragraph's thesis, stated in the previous sentence: Omnia merita praecedit gratia. Rendered "so that my merits may follow the gifts of God". Nothing was missing in the wrong version; only the sense was. 2. 0597D, justa et vera convenienter loquamini. loquamini is second plural, not first — the paragraph is an address, and "let us speak" would have quietly moved the imperative onto the glossator. Rendered "do you speak". 3. 0593A, investigabiles (§5a), where the smooth reading is the one that contradicts the sed. ⚑ And the habit that made it bite: 0597B's Quod non debetis in quo vos quasi cera signati estis reads roughly in English and was left rough. A sentence that reads smoothly where the Latin is rough is a suspect; this one is not smoothed.

§8 · READ CLEAN — named, so a blind reader can tell checked ground from unexamined

• 0587D, the ARGUMENTUM — one sentence, 27 words, the standard Marcionite prologue. Nothing at stake; Tychicum diaconem rendered "Tychicus the deacon" (Migne's spelling with y; 8990 spells it Tichicum and neither was made to follow the other).
• 0591A–0592C, CAPUT II entire — 34 lemmata, zero [var:], zero [sic:], zero [cj:]. This is the cleanest chapter in the work and the collation is a genuine string of agreements, not an unexamined stretch: every one of the 34 was checked, including Surgam, et ibo ad patrem meum (Luke 15:18, verbatim). The only thing in it worth a second look is the hospes ~ ostii pes etymology at 0592A, which is set roman in Migne and therefore carries no added italics in our English either.
• 0592D–0595A, CAPUT III entire — 29 lemmata, zero [var:]. Its trouble is prose, not apparatus: the investigabiles pair (§5a), the four-dimensions passage at 0594B–D with its reversed demonstratives (§5), and the (Matth. II) note (§1c).
• The four-dimensions passage, 0594B–0594D — read three times, because it is the longest continuous argument in the work and the launch named continuous close-packed argument as this book's real difficulty. sublimitas and altitudo are Migne's two words for the same dimension; rendered "sublimity" and "height" respectively, kept distinct so a reader can see that he used two, with the identification left where he leaves it (habes latitudinem … altitudinem non habet).
• Chapters V and VI, the household code and the armour of God — the launch's named high-risk ground, and it produced four of the ten [var:] (0598B, 0598D, 0600D, 0600D) plus both remaining [sic:]. Every lemma in the armour passage (0600B–0600D) was collated individually: Succincti lumbos vestros · Induti loricam justitiae · Calceati · Pedes · Scutum fidei · Et galeam · Et gladium spiritus — all seven agree with the Clementine exactly, and are recorded as agreements. The divergence in that stretch is not in the armour lemmata at all but in the Matthew quotation attached to the sword (§1a).

§9 · What a blind reader should hit hardest

1. 0599B, the Isaias quotation (§1d). Declined for a [var:] on the ground that the span is roman, while the divergence from Isa. 61:10 is larger than several I did mark. If the roman/ italic rule is the wrong test for a quotation this far from its source, this is the site that shows it.
2. 0598B, Omne enim quod manifestantur. I took a [var:] and rendered the number disagreement into English ("which are made manifest"). The alternative was a [sic:], on the ground that the disagreement is a fact about this printing — the same argument I used to take the marker at 0590D (§4d). I did not take it here because manifestantur renders in place and English does carry the disagreement; but the two calls are close enough that they should be checked together.
3. 0590A, the voluntas/vocatio conjecture (§3a). A [cj:] on a word that construes.
4. 0594C–D, the four-dimensions passage. Long, dense, and where a relocated negative would be hardest to see.
5. 0598D, the full stop mid-construction (§5). Preserved. It reads like our error and is his.

§Adjudication — blind polarity read (2026-08-19)

Blind reader, fenced from this file, all six chunks read: 3 sites, 2 repaired, 1 declined. Its negation-host audit ran on every negative in the work and found the Colossians failure pattern does not recur here — including the hard case at 0592B, Non enim regionibus longe est quisquam a Deo, sed affectibus, where the negative is correctly clefted onto the phrase and not parked on the verb.

1. REPAIRED · 0595B · [sic: mancida] HAD SWALLOWED THE WORD IT MARKS. The marker stood in the slot where the rendering belongs, so one member of a seven-term asyndetic list vanished: nec frangitur negotio, nec turbulenta est, nec mancida, nec audax… read "nor is it turbulent, nor , nor bold." Now "nor languid [sic: mancida]". ⚑ The reader proved it structurally rather than by taste: every other [sic:] in this work — and ~40 sampled corpus-wide — stands AFTER an English rendering; only this one did not. That is the right way to establish the class, and it is the third consecutive batch in which it has appeared. marcida ("languid, withered") is the form suspected under the printed letters; it is NOT asserted — the printed mancida stays in the marker.
2. REPAIRED · 0599C · novit — rendered "knew", now "knows". The perfect-form present is coordinated with the present non est, so "knew" put God's discernment in the past beside a present predicate about him. Small, and no count could see it.
3. DECLINED · 0594A · the missing [var:] against Matt. 9:2 — THIS IS A PHANTOM, and a predicted one. Migne's note reads [n: (Matth. II)]; the words are verbatim Mark 2:5, confirmed against clementine-flat.txt: Fili, dimittuntur tibi peccata tua. Matt. 9:2 reads Confide fili, remittuntur…, so a reader collating against Matthew manufactures a divergence that does not exist. No [var:] is owed; what is wrong is Migne's book abbreviation, not his text. ⭐⭐ This is exactly the trap 8980's translator logged deliberately on the same day (its 0680D note cites Matthew over a sentence of Luke's, and it recorded the fact "so a blind reader collating against Matthew doesn't report a phantom"). The precaution worked in one work and the absence of it cost a finding here. ⚑ Standing rule for the next brief: when a note's book and its words disagree, collate against the WORDS, and say in the cruces which book they came from.

⚑ Plate: cols 587/588 and 593/594 carry no foot-of-page apparatus and no asterisk-keyed EDIT. note; coverage: "spot". Three sites still want a fuller read and the cruces names them: 0595B mancida, 0600C rectus, 0599C nomo.