Liber Psalmorum

BAND A — chunks 0000–0038, cols 0841A–0918C, Psalms I–XLIX

Merged 2026-09-04 from eight stint files. Bands B (chunks 0039–0076, Ps 51–100) and C (0077–0105, Ps 101–150) are separate runs; the work ships only when all three are merged.

⛔ The merge is not the concatenation. §A below is the reconciliation — the conventions that now bind Bands B and C, the drift that was found, and how each was settled. The eight stint files follow verbatim and unabridged in §E; nothing in them has been cut.

A1. The tituli — ruled from the Glossa's own argument, not from preference

1. In finem → "Unto the end." ⛔ NOT "for the choirmaster" (the sense of the Hebrew lamnatseach). The Glossa argues from finis meaning END: Finis duobus dicitur modis, quando aliquid deficit, vel quando perficitur. A stint found further evidence at 0884B, where Migne writes Non moriar jam in finem in his own prose four lines from the titulus. Verified across Band A: 26 occurrences, zero rivals.
2. Diapsalma / Diapsalmus → unanglicized and ROMAN, never italic, never "Selah". ⚑ This was the one true convention conflict in Band A, and the fault was the launch brief's, not a stint's. The brief said "italic as Migne sets it" — and Migne sets it ROMAN. Two stints read the plate and overrode the brief; one obeyed it and italicised a single occurrence. They were not diverging from each other, they were following contradictory instructions. Settled on five plate reads (0849B, 0870D–0883A, 0887C, 0891D, incl. the masculine form); the single italic is reverted; the brief is corrected. Italicising it also breaks the Latin/English span parity every stint checks — the error had a second, silent cost.
3. Canticum graduum → "A canticle of degrees", and gradus → degree throughout. Load- bearing, not stylistic: the gloss numbers the gradual psalms as a series in its own voice (Quartus gradus, Septimus gradus, Decimus gradus). ⚑ Band A contains none of these — they begin at Ps 119, in Band C. The ruling is recorded here so Band C inherits it.
4. Psalmus David and Psalmus ipsi David are different lemmata and stay different. Several carry a ⚑ in the lemma brief because Migne's titulus and the Clementine's differ; three stints fired [var:] on such tituli, consistently.

A2. Psalm numbering — Vulgate, and the one apparent exception is not one

Vulgate/LXX numbering governs the whole book. Never convert a psalm number. Chunk 0010 carries two heads both numbered X — PSALMUS X.-- Secundum Hebraeos. (the Hebrew psalm 10, which our Clementine holds as the tail of Ps 9, its VERS. restarted at 1) and PSALMUS X. (the Vulgate Ps 10). Migne glosses the passage twice under one numeral, and that is his text: nothing was renumbered, merged or marked. The stint collated all nine ⚑ Ps 9:2x spans individually and every one agreed word for word.

⛔ The natural inference — that the Glossa follows the Hebrew division and runs one ahead thereafter — is false, and was tested rather than assumed: first-lemma probes for every psalm head in the work score 138 same-number / 0 offset / 3 undecided.

A3. Heads

PROTHEMATA IN PSALTERIUM. → PROTHEMATA ON THE PSALTER. · PSALMUS PRIMUS. → PSALM ONE. but PSALMUS II. → PSALM II. — Migne spells the first out and numbers the rest, and his inconsistency is followed, not regularised. Roman numerals stay Roman.

A4. The vocabulary axes — measured across all 39 chunks, not asserted

A mechanical sweep of the merged Band A English:

⚑ The archaism check needed a refinement, and it is a real finding for Bands B and C. The settled test for whether -eth/saith is licensed is italic membership — archaism is permitted only inside quoted scripture. A sweep flagged one saith as gloss-voice. It is not: it stands inside « guillemets », quoting Isa 45:1. Quoted scripture in this work is marked TWO ways — italic lemmata AND guillemet quotations — so an italic-only test mis-flags. Use both. Confirmed clean otherwise: 1 saith, 1 doth, both licensed.

A5. RULED BY WILSON 2026-09-04 — settled, inherit, do not re-open

• ⭐ gens tenebrarum @0887D → "people of darkness" (was "nation of darkness"; the English is changed at 0022.md). The Manichaean power of Non appropiant qui dicunt fatum facit peccare, vel fortuna … vel gens tenebrarum, quae Deo rebellat — third in a list of determinist excuses. The gentes axis does not reach it: it rules the ethnic/political plural, and this is a singular gens naming a cosmological principle. "Race" was offered and declined. The axis gains no sixth word — this is a named exception to it, and a singular gens in a non-ethnic metaphor is flagged, not folded into the axis.
• Dominus virtutum → "Lord of hosts" (the gloss glosses it by angels). Bands B and C will meet it again; recorded so they inherit it.
• sedes → "seat", not "throne", because "throne" collides with the order Throni.

A6. ⭐⭐ RULED 2026-09-04 — WHAT THE ENGLISH IS A TRANSLATION OF

The plate where the leaf has been read; our twin where it has not. Where a translator has opened the column and Migne's type differs from our TEI, the English renders what Migne printed, and the divergence goes in data/briefs/<idno>-PENDING-TEI-PATCHES.md so the Latin converges on the English when the patch lands. Where no leaf has been read, the twin is all there is and the twin is translated.

Why this and not "always translate the twin": at 0973B our twin prints the non-word ominus (qui non ominus jaculatur) and the plate reads eminus. Translating the twin there would have forced either untranslatable garble or a [sic:] publicly accusing Migne of Corpus Corporum's defect. The doctrine was already stated in the pending-patch file — on this work the plate is the text and our TEI is a witness to it — and this makes it govern the English too.

⛔ The gate is the plate read, not the translator's confidence. This licence exists only for a column somebody actually opened; it is never a warrant to emend from the Latin alone. And every such site MUST appear in the patch file — an English that silently outruns its Latin, with no pending patch to close the gap, is the drift this rule is designed to prevent.

A7. ⭐⭐ RULED 2026-09-04 — usque ad: THE etc. SPLICE TEST REPLACES ITALIC MEMBERSHIP

usque ad is Migne's abridgment formula ("as far as") only where it follows an etc. and names the point at which the quotation resumes. Everywhere else it is ordinary Latin and is rendered by sense.

Italic membership is WITHDRAWN as the test. It fails in both directions, and col 0953C carries both cases four words apart inside one italic run:

Occiduum tempus, etc., usque ad contra quos Dominus: Vobiscum sum usque ad consummationem saeculi, etc.

The first is the formula; the second is Matthew 28:20. No italic test can separate them; the etc. test does. Measured over the whole work: 777 occurrences, 733 formula, 44 not. Over the translated chunks 0000–0076: 483 formula, 24 not.

⚑ This re-grounds Band A's 209 renderings; it does not overturn them. Only five of the 24 non-formula sites used the formula's words, and four are legitimate sense-renderings that happen to reuse them (spread as far as the ends of the earth). One was repaired: Mic 4:10 @0984C, usque ad Babylonem venies, "as far as Babylon" → "even unto Babylon", to agree with 0073's already sense-rendered unto the consummation of the world — two quoted-scripture sites of one construction that had been rendered two ways.

⭐ Band C inherits this and needs it: the gradual psalms carry usque ad constantly.

46 columns read, every marker standing on a column somebody opened (plate-gate OK). Zero foot-of-page conjecture apparatus on any page read, in either layer.

⭐⭐ The measured case for the plate gate, and it is the strongest this project has produced.

• One stint had six [sic:] candidates and the plate withdrew five. Every withdrawn one read as a confident Migne defect from the Latin alone.
• Another had the exact mirror: it declined In sua in virtute @0890A reasoning that a doubled short word is the classic transcriber artifact — and at 500 dpi Migne sets it plainly. A [d: in] now stands there.

Neither "trust the text" nor "suspect the transcriber" is a rule. Only the leaf is.

⛔ The mod-4 clipping rule the launch brief once gave is WITHDRAWN. It said a left-hand column numbered 1 (mod 4) is clipped at the inner margin. Refuted: 841, 849, 893, 897, 1045 clipped against 853, 857, 861, 865, 869, 877, 881, 889, 901, 909 clean. A coin flip. The original claim rested on four consecutive leaves plus one — a single physical gathering — so the congruence was an artifact of contiguity. Clipping is real and sporadic; open the column, never infer. Its failure mode runs where nobody audits: applied blind it silently withholds good markers, and it had blocked three before a stint looked.

39 TEI patches applied in one pass after the band's translation was complete (data/tei-patches/8967.json; queue and classes in data/briefs/8967-PENDING-TEI-PATCHES.md). Our Latin is Corpus Corporum's transcription, and across the ~46 columns read it departs from Migne at roughly 1.5 sites per column — consistent with 9003's independent collation of the sibling PL 114 Psalter at ~3 per page. The class runs in three directions:

1. CC moves Migne ONTO the Clementine. Dominus conservet (0903D), audiebam (0899B), populis (0895A), Multiplicati sunt super numerum (0902C), sordescat (0847A). Invisible to any Vulgate collation by construction — the divergence is erased before an instrument looks.
2. CC REPAIRS Migne's real defects, destroying markers we are entitled to fire. Nocere (0908C, restored — [sic: Nocere] now stands), iniquitetem (0910C), and two erased dittographies (etsi. Etsi → etc. @0887B; doc-/ctus).
3. Our Latin carried words the plate lacks. 0904A, where CC replaced Migne's quasi, with sed Deus and inserted in. Verified on the leaf before patching.

Four English twins changed with the Latin: 0908C ([sic: Nocere] fired), 0904A, 0866D ("or the fat" → "the fat"), 0902C ("above number" → "in number").

⛔ Three reported sites were REFUSED, and each refusal is a rule:

• locutus → locutos — at 900 dpi the glyph is ambiguous, and locutus (perfect deponent, "and afterwards I myself spoke") is correct Latin where locutos is a solecism. A patch needs positive evidence.
• ⭐ possit → rossit, and its whole class. The plate's broken sorts are not Migne's errors. A damaged p that images as r is the condition of this copy's inking, not a reading of the edition. Compositor errors (wrong sorts SET) are restored; ink damage is not — that line is why iniquitetem and Nocere are in and rossit is out.
• Desiderabilia → Desiderabillia — deferred: two occurrences stand in col 0871D and the report does not say which. Re-read the leaf; do not guess.

⚑ And one site was confirmed against the editor's own doubt. Multiplicati sunt super numerum looked wrong to challenge, because the gloss beneath repeats "Super numerum". The plate settles it: the lemma reads numerum and the gloss's Super numerum is the gloss expanding its lemma, which is how this commentary works throughout.

## C1. Patched in the Latin, deliberately NOT marked 0903D Dominus → Dominum conservet. The English is unchanged. Convention 3 renders a fragmentary lemma as the phrase it stands for (Ps 40:3), and no [var:] is owed: the gloss does not argue from the word, and Dominum/Dominus is exactly the letter-level corruption the settled single-word [var:] rule excludes.

## C2. Markers the patch pass UNLOCKED, for a later pass

• Seb isti @0876B — our twin read Sed isti and did not construe; the plate reads Seb isti; the conjecture is Sedisti (Prov 23:1–2), supplying exactly the missing verb and person. No marker could reach it before the patch: [sic:] may wrap only type the twin carries and [cj:] needs the printed word present. The patch is a precondition for the apparatus, not a cleanup after it.
• 0902C a [var:] against Ps 39:6 is now available on the restored lemma.

1. The blind polarity read on all 39 chunks — required, and it is a reading pass, not a grep.
2. Bands B and C, then a three-band merge of this file.
3. Gallica, three sites this copy cannot answer: quae non est nisi [inquieta] mente @0885B (f443) · enumerantur/emunerantur @0898D (f449) · excrescentibus pilis @0905A (f453).
4. One In Hebraeo clause unread. Two of three are closed: 0868A prints (quod in Hebraeo) with no Hebrew type at all, and 0902B's In Hebraeol was CC debris for In Hebraeo.
5. The gens tenebrarum ruling (§A5).

Band A marker totals: 37 [var:] · 6 [sic:] · 2 [cj:] · 1 [d:].

Nothing below is abridged. Each stint's own reasoning, declined list and checks stand as written.

§E.0000 — stint file cruces-0000.md, verbatim

Range: src/latin/8967/0000.md–0004.md, cols 0841A–0850A. Content: the PROLOGUS BEATI HIERONYMI, the PROTHEMATA IN PSALTERIUM, and Psalms I–IV. Lemma brief: data/briefs/8967-lemmata-0000.txt.

⚠ This file is one stint's range. It must be merged into src/english/8967/cruces.md.

§0 · SPAN COUNT RECONCILIATION (the required check)

My brief's header states 6 spans for chunks 0000–0004. Counting … in my own Latin:

The totals AGREE (6 = 6), and the six spans are the same six, in the same order. No head-of-range shortfall, no tail spill. This range's brief was split correctly.

⚑ One discrepancy, and it is in the brief's column labels, not its membership. The brief files all six spans under [0841A]/[0842B]/[0842D]. In the Latin, nebel and organum. stand after the [0841B] anchor, not [0841A]. Membership is unaffected and nothing downstream reads those labels, but a later stint reconciling by band rather than by chunk would be misled. Reported rather than worked around.

⚑ Front matter carries few lemmata by nature — six spans across 3,921 Latin words — exactly as the launch message predicted. Four of the six are single words used as words (nebel, organum., visio,, videns.) and two are the name-etymologies forti manu / pacificus,. None of the six is a VERS.-addressed verse lemma; the verse lemmata in Psalms I–IV are set in guillemets, not italics.

The six spans, checked in place

§1 · MARKERS FIRED — three [var:], no [sic:], no [cj:], no [d:], no [ed:]

All three stand on columns I read at Migne's plate this session (see §2); plate-gate.mjs 8967 returns ✓ for all three.

1. [var:] @ 0844D (chunk 0002) — Gn 3:12. Migne prints « Mulier quam dedisti mihi me seduxit. » The Clementine reads Mulier, quam dedisti mihi sociam, dedit mihi de ligno, et comedi. Sense-changing: Adam's excuse in the Vulgate is that the woman gave him of the tree; Migne's makes her seduce him — importing the verb the Vulgate assigns to the serpent one verse later (Gn 3:13, serpens decepit me). The gloss is building the abiit / stetit / sedit triad on exactly this clause (sedit, dum se excusavit), so the wording is load-bearing. Plate-confirmed at 600 dpi, p. 427, mid-line, right-hand (clean) column. Rendered as printed; the marker sits beside the rendering.

1. [var:] @ 0848B (chunk 0003) — 1 Pt 3:19. Migne prints « Quando iis qui erant in carcere, spiritibus profectas praedicavit… » The Clementine reads In quo et his, qui in carcere erant, spiritibus veniens praedicavit. Plate-confirmed at 600 dpi, p. 429, right-hand (clean) column: the plate genuinely reads profectas. ⚑ profectas is a real word-form (fem. acc. pl. of profectus, -a, -um) with nothing in the clause to agree with; the sense the sentence needs is the masculine nominative profectus, "having gone forth" — parallel to the Vulgate's veniens. I rendered it for that sense ("having gone forth") and put the divergence and the conjecture in the marker rather than quarantining the word, per 7a⁗: quarantining it would have taken the participle that carries Christ's descent, which is the whole point of the citation. I did not fire a [sic:] beside it — see §4.2.

1. [var:] @ 0848D (chunk 0004) — Rm 5:3. Migne prints « Gaudemus in tribulationibus scientes quod tribulatio, etc. »; the Clementine reads gloriamur in tribulationibus. Plate-confirmed at 600 dpi, p. 429, right-hand (clean) column. Rejoicing for glorying — a lexical substitution in a proof-text adduced for consolation. Fired because the column is clean, read, and mid-line; the divergence is one word and I say so in the marker rather than arguing sense from it.

§2 · PLATE READS DONE THIS STINT — three new pages, recorded

data/plate-reads.json had 0841A–0842D and 0849A–0850D for this work. My range also covers 0843A–0848D, which was unread, and half my [var:] candidates fell there. I rendered and read pp. 427, 428 and 429 of raw/scans/pl113/patrologiaecurs04migngoog.pdf (map PDF page = (column + 11) / 2, corner numbers checked first on every page) and recorded all three reads. Cols 0841A–0850D of this work are now fully read.

• Zero foot-of-page conjecture apparatus on all three pages, of either the numbered or the asterisk layer. That is consistent with the five pages read at chunking; coverage stays spot.
• The mod-4 clipping rule held 6 of 6: 843 clean, 845 clipped, 847 clean.

⭐ 2a. A CORPUS CORPORUM DEPARTURE FROM THE PLATE — 0847A, needs a TEI patch, NOT a marker

The plate reads sordescet; our TEI reads sordescat (Apoc 22:11, in the Ps 2:9 gloss). Confirmed at 600 dpi. This is the class CLAUDE.md test 4 names: CC silently normalizing Migne toward a familiar text — Migne's future indicative pulled into the Clementine's present subjunctive, and it is invisible from the Latin alone.

⛔ No marker is owed and I fired none. A [sic:] may wrap only words carried verbatim from the plate, and our Latin column does not carry this one; a [var:] would assert that our printed Latin diverges from the Vulgate, which — as CC prints it — it does not. The correct repair is a data/tei-patches/8967.json entry restoring sordescet, after which the divergence becomes a real (small) [var:] candidate. My English renders our Latin twin (sordescat, "may be defiled still") and flags it here. Escalated: this needs an editor's decision, not a translator's.

2b. Two further CC-vs-plate departures, both cosmetic, both logged for the same patch pass

• 0845A — the plate sets one note, (AUG. CASS.). CC splits it into a bare running-text AUG. plus a separate [n: (CASS.)]. My English reproduces CC's shape exactly (an unanglicized bare AUG. in the running text) because marker parity is checked against the twin. Reported, not repaired.
• 0848D — the plate sets duxisti, v. « (comma, lowercase v.); CC gives duxisti. V. «. I reproduce CC's. ⚑ I could not settle what Migne's v. abbreviates — vide? versus? — and I did not guess: it passes through as printed.
• 0849D — the plate closes « In corde. »; CC drops the closing guillemet, which is why chunk 0004's guillemets run 29 « against 28 ». I reproduce CC's (sacred 1:1 against the twin). Same for chunk 0002 (25/24) and chunk 0003 (56/57); I did not "balance" any of them.

2c. What the clipped column cost me — one candidate withdrawn for want of a second copy

0845B, Jo 7:38. CC reads « Qui crediderit in me… » against the Clementine's Qui credit in me — a future perfect for a present, and a real candidate. On the plate the word breaks across the line exactly at the clipped inner margin (Qui cre- / [3–5 chars lost]…in me), and 845 is a left-hand column ≡ 1 (mod 4). Per §3 of the launch brief this does not ship on this copy. Marker withheld. Settleable in one Gallica IIIF request (ark:/12148/bpt6k5505319w, f = (column + 1) / 2 → f423); left for whoever runs the second-witness pass.

§3 · THE DIVERGENCE LIST — every scripture quotation in the range, including the clean ones

7a″ says the list is the deliverable. Every guillemeted quotation and every VERS. lemma in cols 0841A–0850A was collated against sources/vulgate/clementine-flat.txt. Findings first, then the checked-clean.

3.1 Divergences found, marker DECLINED, with the reason

3.2 Quotations collated and found to AGREE with the Clementine — the checked negatives

Recorded because a findings list alone cannot be told from a stint that never looked.

• 0842B Nolite tangere christos meos = Ps 104:15 verbatim (also 1Par 16:22).
• 0842D–0843A Haec dicit Dominus christo meo Cyro = Is 45:1 verbatim.
• 0843B In principio, creavit Deus coelum et terram = Gn 1:1 (Migne's comma after principio is his pointing; the Clementine spells caelum where CC prints coelum — an orthographic normalization of the flat file, per sources/vulgate/README.md, not a divergence).
• 0843B Ecce Virgo concipiet = Is 7:14 verbatim (twice).
• 0843C Quadraginta dies, et Ninive subvertetur = Jon 3:4 verbatim (Migne drops only the verse's opening Adhuc).
• 0844A Miserere mei Deus = Ps 50:3; Misericordiam et judicium cantabo tibi = Ps 100:1; Omnis spiritus laudet Dominum = Ps 150:6 — all three verbatim, and all three the correct hinge-verses of the three fifties the gloss is counting.
• 0844B Domine, quid multiplicati = Ps 3:2 verbatim.
• 0844C Tecum principium = Ps 109:3; Ego dormivi = Ps 3:6; Longe a salute mea = Ps 21:2 — all verbatim incipits.
• 0847A qui in sordibus est, sordescat adhuc = Apc 22:11 verbatim as our Latin prints it (see §2a for the plate).
• 0847D Quod facis, fac citius = Jo 13:27 verbatim.
• 0849B Sol non occidat super iracundiam vestram = Eph 4:26 verbatim.
• Every VERS. n.-- lemma of Psalms I, II, III and IV was checked against Ps 1–4 in the Clementine and every one matches its own verse, verbatim, at the printed number — see §3.3.

3.3 ⭐ THE VERSE NUMBERING IS RIGHT, AND I CHECKED IT RATHER THAN ASSUMING IT

Per §1 of the launch brief. All 30 VERS. addresses in my range resolve to the Vulgate verse of that number in their own psalm, with zero offset:

• Ps I — VERS. 1 Beatus vir qui non abiit (1:1) · 2 Sed in lege (1:2) · 3 Et erit (1:3) · 4 Non sic impii (1:4) · 5 Ideo non resurgent (1:5) · 6 Quoniam novit Dominus (1:6). ✓
• Ps II — 1 Quare fremuerunt gentes · 2 Reges · 3 Dirumpamus · 4 Irridebit · 5 Loquetur ad eos · 6 Ego autem · 7 Dominus dixit · 8 Postula · 9 Reges eos · 10 Et nunc, reges, intelligite · 12 Nequando irascatur · 13 Cum exarserit. ✓
• Ps III — 1 Psalmus (the titulus) through 8 Quoniam tu percussisti. ✓
• Ps IV — 1 In finem through 9 In pace in idipsum. ✓

⚑ VERS. 11 is missing from Psalm II, and it is a GAP, not a defect. Migne runs 10 → 12 → 13. Plate-confirmed, p. 429. Ps 2:11 is Servite Domino in timore, et exsultate ei cum tremore, and the recension simply carries no gloss on it: 12 and 13 sit on their own correct verses either side, so nothing has slipped. This is not the Jude-style out-of-sequence numbering the brief warns about and must not be "restored" — reproduced exactly as printed.

§4 · CANDIDATES REJECTED, AND WHY — the measured instrument

4.1 modo etiam apposuit @ 0841C — declined, and the copy cannot settle it

Migne prints Nam David composuit omnes psalmos: modo etiam apposuit, in quo multos auxiliatores habuit. apposuit has no object and modo is a bare ablative; the sentence wants something like modos etiam apposuit ("he supplied the melodies as well"), which is what the following in quo multos auxiliatores habuit is about. I did NOT emend. I rendered it "he moreover appended them, wherein he had many helpers", taking modo etiam adverbially, which keeps every printed word and invents no noun.

⛔ And the plate cannot arbitrate: the verb stands at a LINE OPENING in column 841, which is clipped (the read at chunking recorded exactly this). Our copy shows …uit,. Needs Gallica f421. Logged, no marker.

4.2 No [sic:] on profectas @ 0848B — deliberate

The obvious move is to stack a [sic:] on the [var:]. I declined, and the reason is the class prior, not a hunch: profectas is a real, well-formed Latin word — Pattern 12 marks type the plate got wrong, and nothing here shows the plate got anything wrong. It reads exactly as a father's loose participle would if his copy had profectus and a compositor set the wrong ending — but that is a guess about whose error it is, and 9004 measured translating agents' attribution guesses at 5 right against 9 wrong. The [var:] states both readings without claiming a culprit, which is what Pattern 14 is for. ⚑ On that same work 8 of 27 [sic:] fell at the plate and every fall was a single word; profectas is a single word.

4.3 Candidates raised and rejected on inspection

• quem novit Deus @ 0849B — masculine relative for the neuter antecedent corde. A solecism, not a defect, and Pattern 9 territory; medieval Latin does this constantly. Rendered "which God knows". No marker.
• Non retropendet @ 0848C — retropendere is not a classical verb and reads oddly. But it is transparently formed (retro + pendere, "to weigh back"), it is plate-confirmed p. 429, and it stands in a clean column with normal spacing. Rendered literally, "He does not weigh back". Pattern 7, no marker. ⚑ Flagged so a blind reader does not think it was missed.
• in latitudine gaudii duxisti @ 0848D — ablative where the sense wants the accusative of motion (in latitudinem). Plate-confirmed. Rendered "into the breadth of joy" for the sense the verb duxisti forces, which is what the printed words allow. No marker; a case ending is not a Pattern 10 non-word.
• vel etiam diabolis et angelis ejus @ 0847D — a plural diabolis with a singular ejus hanging off it. Plate-confirmed p. 429. Rendered "by devils and his angels", carrying the mismatch. Per axis 3 the word is devil(s), never "demon". No marker.
• ut sitis aqua @ 0845B — sitis is the noun "thirst" (nominative), not the verb "you may be": qua reficitur interior homo, ut sitis aqua = the inner man is restored by wisdom "as thirst by water". ⚑ This is a live mis-parse trap of exactly the Judith ferias shape (a noun readable as a subjunctive) and I name it so no later pass silently "fixes" it.
• nomine, non numine @ 0844C — a nomen/numen jingle English cannot keep. Rendered "in name, not in the divine power", losing the sound and keeping the sense. Not a defect; recorded because the loss is real.
• PSAMUS CXXIV — outside my range and already settled at the plate; no action, per §3 of the launch brief.

§5 · RENDERING CALLS AND CONVENTIONS APPLIED

• In finem → "Unto the end." (launch brief §2.1) at 0848D, and the gloss immediately under it argues from finis meaning END (quando aliquid deficit, vel quando perficitur), which is the axis's own justification standing in my range. Rendered "The end is said in two ways, when something fails, or when it is completed."
• Diapsalma → diapsalma, unanglicized (brief §2.2), four occurrences at 0849A–0849C. ⚑ Migne sets it in ROMAN here, not italic — so my English sets it roman too. The brief says "italic as Migne sets it"; in this book, at these four sites, he does not. Italicizing it would also have broken span parity against the twin. Never "Selah", never "pause".
• gentes → nations throughout Psalm II (0846A Quare fremuerunt gentes, 0846C in omnibus gentibus, 0847A do tibi gentes haereditatem). Axis 4: this psalm sets gentes against populi, not against Judaei, and the gloss expounds it of Herod, Pilate and the spread of Christ's name through the world — plainly generic, so "nations". Stated here because the Douay gives "the Gentiles" at Ps 2:1 and Acts 4:25 and the familiar cadence pulls hard.
• mysterium → mystery (0842A plus mysteriorum, 0843C altioris mysterii, 0848D in magno mysterio). No sacramentum occurs in this range, so the 1:1 axis is untested here.
• usque ad — does not occur in my range, in either the formula or the temporal sense.
• Commentator's voice not archaised (axis 1): dicit/ait → "says", throughout. The only -eth forms in my English are inside guillemeted scripture (casteth, dwelleth, knoweth, showeth), which is where axis 1 permits them.
• (ID.) and (Ibid.) never expanded — 23 of them in this range, all passed through verbatim.
• AUG. at 0845A passes through as a bare unanglicized siglum in the running text, because that is what our Latin twin has. See §2b.
• Second person (Pattern 17): plural vos/vestris/dominemini → you/your (0847B, 0849A–B); singular cognosceris addressed to God at 0849D → thou. No reverence-thou anywhere.
• PSALMUS PRIMUS. → PSALM ONE. but PSALMUS II./III./IV. → PSALM II./III./IV. — numerals mirror the plate, per the shared brief's measured rule.
• PROTHEMATA IN PSALTERIUM. → PROTHEMATA ON THE PSALTER. ⚑ Prothemata is kept: it is a technical name for a class of prefatory material, it is current in English medieval studies, and "Preliminary Themes" would read as a description rather than a title. First use of this head in the corpus — flagged for the merge so the other seven stints and Bands B/C match it.
• PROLOGUS BEATI HIERONYMI IN LIBRUM PSALMORUM. → THE PROLOGUE OF BLESSED JEROME ON THE BOOK OF PSALMS. — the shipped corpus form (8946 Ecclesiastes, 8948 Esther).

§6 · THE THREE CONTAINMENT TESTS, PLUS THE FOURTH

1. [sic:] swallowing — n/a: I fired none. Both [var:] markers sit beside a complete rendering; stripping every marker from my English leaves four readable sentences with no stranded article, preposition or doubled mark.
2. Negation attachment, host named — all 34 printed negatives in the range walked, host word named, host located in the English. Two were on the 2a trigger population and are recorded: - 0847D qui gloriam majestatis non putant ad humilia posse descendere — the non is on putant, not on posse. Written "who do not think that the glory of majesty can descend", never "who think it cannot". - 0848C saeviunt illi, sed non dormirem, nisi vellem — non on dormirem, nisi on vellem, and the sed correlative confirms the scope. "they rage, but I would not sleep, unless I willed it." - Also carried without softening: nisi sequantur laudanda (0845A), nec otiosa (0845A), alii nec judicabunt, nec judicabuntur (0845D), ne mundus post eum abiret (0846B), qui nolunt jugum Christi sibi imponi (0846B), ut non temere dominemini (0847B), nihil patientiae (0847C), Aliter non sperarent (0848A), quae non Deus / quae non faciunt quod promittunt (0849A), non ei consentiat ratio (0849B), quae non ducitur ad effectum (0849C), non in multiplicitate saeculi (0850A).
3. Punctuation band by band — colons, semicolons, question marks, guillemets and exclamations compared inside every one of the 37 column bands in my range. All bands match 1:1 except one, and that one is licensed: [0842A] carries one extra colon in the English, from Pattern 8b — Migne closes the indirect question Solet quaeri cur… legantur? with a mark English cannot carry, so it is recast direct behind a colon: "It is commonly asked: why are the psalms read in the Church before other Scriptures?" The ? is preserved, not dropped. All three of the Latin's question marks are present in the English. ⚑ In the first pass this comparison returned 16 mismatched bands — every one a comma I had promoted to a semicolon or colon out of habit. Fifteen were reverted to the printed mark. The test earns its place.
4. "Does it ASSERT what he asserts?" — every paragraph read back against the Latin with the counts already clean. Two things this caught that nothing else would have: - 0846C — I had first written "they wish to subjugate us to themselves" with the unexpressed subject taken as the kings. It is the Lord and his Christ: the whole speech is the kings quoting their bonds (Dirumpamus… quibus nos sibi astringere volunt), and sibi points back through the lemma to adversus Dominum et adversus christum ejus. The antecedent is in the lemma above, exactly as the brief warns. - 0846B — Prior qui infantes occidit Herodes, posterior qui Pilato consensit: the posterior is also a Herod (Antipas), which the next sentence proves (Solus enim rex Herodes et princeps Pilatus). I did not supply the name, because Migne does not; the English keeps his bare "the latter, he who consented with Pilate".

§7 · SECTIONS READ CLEAN

Cols 0841A–0844C (the prothemata entire), 0844D–0846A (Psalm I), 0846A–0848C (Psalms II–III) and 0848D–0850A (Psalm IV) were each read line by line against the Latin twin, and against the plate for 0841A–0850D. Apart from the sites named above, no polarity defect, no dropped clause, no supplied or destroyed punctuation mark, and no unexpressed-subject ambiguity remained unresolved. verify-english.mjs 8967 is silent on chunks 0000–0004; plate-gate.mjs 8967 returns ✓ on all three of my markers.

§8 · HANDOFF — what the merge and the editor must act on

1. sordescat → sordescet needs a data/tei-patches/8967.json entry (§2a). Editor's call.
2. Two Gallica second-witness reads are owed and each costs one HTTP request (§2c, §4.1): f423 for crediderit @ 0845B, f421 for apposuit @ 0841C.
3. PROTHEMATA ON THE PSALTER. and PSALM ONE. / PSALM II. are first uses; the other seven Band A stints and Bands B and C must match them.
4. The lemma brief's column labels are off for chunks 0000–0004 (§0). Membership is correct; only the printed band label is wrong. Worth checking whether the same generator mislabels the other 21 ranges.
5. My three new plate reads are in data/plate-reads.json; cols 0841A–0850D of this work are now fully read and the mod-4 clipping rule stands at 6 of 6.

§E.0005 — stint file cruces-0005.md, verbatim

Band A, stint covering src/latin/8967/0005.md … 0009.md.

§0 · SPAN COUNT RECONCILIATION — the brief and the Latin AGREE

data/briefs/8967-lemmata-0005.txt declares 21 italic spans. Counted in my own Latin:

The identities match too, not merely the total: 12 × usque ad, plus (Isa. [0851A] XXVI), Cognoscetur., Convertantur peccatores., convertentur,, In infernum,, Patientia., Exsurge., exsurge., Sciant. — the nine named spans of the brief, in the brief's order.

⚠ The brief's COLUMN BANDS are unreliable, though its list is not. Several spans are labelled with the chunk's colContext rather than the band they actually stand under: e.g. all three of chunk 0005's are labelled [0850A] when they fall at 0850D, 0850D→0851A and 0851B; chunk 0009's mid-chunk spans are all labelled [0857A], when two of them stand at 0858C and 0859A; and the brief's [0852D] and [0855C] entries are really 0853A and 0855C-in-chunk-0008. The count and the identities are sound; do not use the brief's band labels to locate a span.

§1 · PLATE READS — I READ MY WHOLE RANGE AT THE PLATE, AND IT CHANGED THREE CALLS

At launch only cols 0849A–0850D of my range had been read. I rendered and read PDF pages 430–435 of raw/scans/pl113/patrologiaecurs04migngoog.pdf (map PDF page = (column + 11)/2, re-verified against the printed corner numbers on every page), covering cols 0849A–0859D, plus one 600-dpi crop. All five new reads are recorded in data/plate-reads.json.

Foot-of-page apparatus: ZERO on all six pages, of either layer (numbered Forte conjectures or asterisk cross-references). Full page width was rendered each time, so this is not a bottom-strip crop. Coverage for this work stays spot; a checked zero over ten more columns is worth having, but it is not a claim that the work has none.

⚑⚑ 1a. THE mod-4 CLIPPING RULE IS FALSIFIED — 853 AND 857 ARE CLEAN

data/plate-reads.json recorded, from five pages, that "a left-hand column whose number is congruent to 1 (mod 4) loses ~3–5 characters at the start of most lines; 3 (mod 4) is clean," and the launch brief's §3 turned that into a policy: a quarter of this work's columns cannot license a claim about line-opening type.

It does not hold. Read today at 200 dpi:

Two counter-examples in five pages. Against the earlier tally (841 clipped·1, 849 clipped·1, 1043 clean·3, 1045 clipped·1, 1047 clean·3) the rule now stands at 5 confirmations and 2 refutations, so congruence 1 (mod 4) is a predictor of risk, not a rule. The scan's clipping is a per-gathering accident of how tightly that copy was held open, not an arithmetic property of the column number. scan_defect in data/plate-reads.json has been amended to say so.

⭐ Operationally: never decline a reading on the strength of the column number. Render the page. It costs one pdftoppm call, and on this range it turned a policy-declined column into two shipped markers and one withdrawn accusation.

§2 · deeidam @0853B — NOT MIGNE'S. CORPUS CORPORUM'S. NO [sic:] OWED.

Our Latin at 0853B prints

« merito » (etsi non videatur hominibus) « deeidam ab inimicis meis inanis, »

a non-word, with the correct form Decidam printing eleven words later in the same paragraph (« Decidam merito ab inimicis. »). Every internal signal said "carried type, fire a [sic:]" — and CLAUDE.md says in as many words that internal corroboration proves an error exists, never whose it is.

The plate settles it: Migne prints de- at the end of one line and cidam at the head of the next. The word is decidam, correctly set, hyphenated across a line break. Corpus Corporum's OCR fused the break and misread c as e.

• English: rendered normally — « let me deservedly fall away from my enemies. » — because that is what Migne prints.
• No [sic:]. Firing one would have accused Migne of his transcriber's defect, exactly as the already-repaired PSAMUS CXXIV would have.
• ⭐ Recommended: a data/tei-patches/8967.json entry, deeidam → decidam at 0853B, on the same footing as the PSALMUS CXXIV repair. Left for the merge — I did not write the patch file, since a patch changes the Latin column for every downstream stint.

⚑ This is the single most instructive site in my range: it is the shape 9004 measured (confident our-file/Migne calls running 5 right and 9 wrong), and it went the way only the plate could decide.

§3 · MARKERS FIRED — 10, ALL ON COLUMNS I READ AT THE PLATE TODAY

node scripts/plate-gate.mjs 8967 returns ✓ (read: read) on all ten. Unread markers added by this stint: ZERO — the ratchet is not touched.

The one [sic:]

@0852C — [sic: poenitentes avertitur] · Narrat aerumnas quibus poenitentes avertitur. A plural subject against a singular passive verb. Plate-verified at 600 dpi (page 431, crop): Migne prints pœnitentes avertitur. This is Pattern 12's second test — English "penitents are turned away" is number-neutral in a way the Latin is not, so rendering it silently repairs the disagreement and leaves the reader nothing to see; it is the signa … ardeat case (11632 @0304A) exactly.

Per 7a⁗ the marker does not stand alone: the sense is rendered in plain words beside it ("the hardships by which penitents are turned away [sic: …]"), so stripping the marker leaves a clause with its verb intact.

⚠ The alternative parse, recorded because it is not absurd: if poenitentes were read as a singular participle agreeing with an unexpressed subject, the clause construes. It cannot be: the form is plural and Latin does not take an accusative object under a passive. But a reader who wants to re-adjudicate has the printed reading and this note.

The nine [var:] — Migne's scripture against the Clementine

Every one is plate-confirmed as Migne's own text, not our source's, and every one is cited from sources/vulgate/clementine-flat.txt.

§4 · pro occultis filiis @0857A — read at 600 dpi, and the gloss is the witness

Migne heads Psalm IX « In finem pro occultis filiis. » and glosses it at once: « Nec additur cujus, per excellentiam est unigenitus Dei » — nor is it added whose. The question "whose?" can only be asked of a genitive singular, which is what the Clementine prints (pro occultis filii). Migne's filiis is a dative/ablative plural and cannot govern cujus.

I re-cropped the line at 600 dpi (page 434) before touching it, precisely because a terminal -s is the class of reading a 200-dpi render gets wrong. The plate is unambiguous: filiis.

• English renders Migne: « Unto the end, for the hidden sons. » Pattern 7, no exception.
• [var:] beside it, naming the Clementine's filii and saying that this is what the gloss's cujus presupposes.
• Not a [cj:]. The reading is not a conjecture question: nothing is defective, and a second witness (the Clementine) settles what the received text says. Pattern 14 is the marker that carries a divergence between Migne's scripture and the received text, and this is exactly that.

§5 · THE RULED AXES — what my range did with each

• In finem → "Unto the end." Three occurrences: 0853A (In finem, psalmus), 0855C (In finem, pro torcularibus), 0857A (In finem pro occultis filiis). All rendered "Unto the end." No choirmaster anywhere.
• Diapsalma. ⚠ My range does NOT contain the explaining passage — the launch message says it does, but Diapsalma secundum Hieronymum semper, Hebraei Sela… stands at 0849B, in chunk 0004, the previous stint's range. What my range has is one genitive at 0858D–0859A: Hic est canticum diapsalmatis, rendered "Here is the canticle of the diapsalma" — the term kept, unanglicized, italic, per the ruling. Note that Migne does not italicize it here, so this is one italic span in the English that has no counterpart in the Latin; that is the register rule for untranslated technical Latin, and it is deliberate.
• usque ad. All 12 italic occurrences → "as far as", the splice formula, unvaried. ⚑ Three UNITALICISED occurrences are NOT the formula and were rendered by sense, per the axis's own exception clause: 0855D–0856A Scripturae usque ad infantes ductae ("led down as far as babes") and per eas usque ad robur nutriuntur ("nourished unto strength") — ordinary extent; and 0857C et qui modo occulte, usque ad quem aliquid valent hostis frameae — a relative ("up to which"), which "as far as" cannot render at all. A raw usque ad grep over my chunks returns 15, not 12; that is the instrument, not drift.
• gentes. One occurrence, 0850B, inside Ps 2:8 dabo tibi gentes, haereditatem tuam, rendered "Gentiles." The frame is the Judaei/gentes typology in its sharpest form — the Church against the aemula who is shut out of the inheritance — not the generic in omnes gentes.
• turba / multitudo / daemonium / mysterium / sacramentum. daemones at 0857C → demons; diabolus throughout → devil; mysterium at 0853A (futura mysteria, Mysterium incarnationis) → mystery. No turba, no multitudo, no sacramentum, no natio, no gentilis in my range.
• Sigla. 111 [n: …] across the five chunks, passed through verbatim. No (ID.) or (Ibid.) expanded — 32 (Ibid.) and 0 (ID.) stand exactly as printed. Migne's (CAS.) at 0852D and (CASS.) elsewhere are both reproduced as printed; they are not normalized to one another.
• Numerals in heads. PSALMUS V. → PSALM V., numerals kept, after 9003's shipped IN PSALMUM VI. → ON PSALM VI.
• Psalm numbering. Vulgate throughout, nothing converted. My range ends at 0859A, one column before the double-numbered head at 0859B; the ⚑ Ps 9:2x flags of §1 of the launch brief fall in chunk 0010, not here.
• Pattern 17. One live site: 0858B imo non est oblitus ut putatis → "as you think" (2nd person plural). Also 0850D non dicit intendite, sed intende — the gloss's whole point is the number, so it is rendered "he does not say attend ye, but attend."

§6 · CRUCES — passages rendered literally, with the difficulty named

1. @0850C — contra timorem mortis, periculi, utilitatis. The three fears of adversity are given as death, peril and utility, which is not a fear. Rendered as printed, "against the fear of death, of peril, of profit." Probably inutilitatis, or utilitatis in the sense of the loss of one's goods, but nothing on the plate decides it and the conjecture would be ours. Plate-confirmed (page 430) — Migne prints utilitatis. No marker: it is a real word in a real slot, and the English does not assert something false, only something odd.
2. @0850C — the closing sentence of the argumentum has no main verb. Ad postremum commemorans praemia beatorum, ut … converteret, et … concitarent. Rendered as the fragment it is. Note also the number: poena converteret (sg.) beside praemia concitarent (pl.), which is correct Latin, each verb with its own subject.
3. @0850C — Trina oratio, verba, oris psalmodiam, clamorem, cordis affectum … Ecce ter de eodem. The gloss announces three and names two (words = the mouth's psalmody; a cry = the heart's affection); the third is intellige in the psalm verse, which the gloss does not repeat. Rendered as printed. No emendation.
4. @0852B — cujus comparatione coactus est confiteri mala sua. The unexpressed subject is the rich man (Dives), not Lazarus: he is the subject of Confessus est dives two clauses back, and it is he who is compelled. ab his praemoneri — "forewarned by these" — is genuinely ambiguous between Moses-and-the-prophets (Lc 16:29) and Lazarus; rendered with the ambiguity kept, since English can keep it here.
5. @0854B — Superbus diabolus, quia per iniquitatem vincitur, subdit: An anacoluthon: the participial clause has no finite verb of its own and subdit has no expressed subject. Rendered flat — "The devil is proud, because he is overcome by iniquity, he adds:" — with the roughness visible.
6. @0854A — Et quia ablatio hominis, quia poena est diabolo, dicitur ira Dei contra eum. A second anacoluthon (Et quia … dicitur, with no apodosis). Rendered as printed.
7. @0854C — et earum fines dirigit justum, et non solum dirigit, sed adjuvat justum. justum is accusative where the sense wants justi ("their ends, the just man's"). Rendered literally, "directs also their ends, the just man". ⛔ Declined for [cj:]: the English is awkward, not false — the sentence still asserts that God directs and helps the just man, which is what the gloss asserts, so Pattern 18's bar ("the English then asserts something the author did not") is not met.
8. @0854D/0855A — « Nunquid irascitur. » and « Nunquid adducit iram per singulos dies. » Both are questions in the psalm and both are closed with a full stop on the plate (page 432, verified). Pattern 8: the marks ride as printed, so the English reads « Is he angry. » and « Doth he bring anger every day. » Not 8b — nothing is being recast, and English is not refusing the mark; Migne simply did not print one. Both chunks 0007 and 0009 carry zero question marks in Latin and zero in English.
9. @0855A — Alia littera: « Nunquid adducit iram per singulos dies. » and @0855A Alia littera: Frameam suam splendificavit. ⛔ No [var:] on either. Migne flags these himself as variant readings; marking our own would tell the reader twice and blur whose observation it is (the ( sic ) principle of Pattern 12).
10. @0855A — nec impar dictum est hoc, quia arcum suum tetendit, » — a closing guillemet with no opener. Reproduced. Guillemets across my range are 1:1 with the Latin and deliberately unbalanced in five places (0005 23«/22», 0006 24«/23», 0007 62«/61», 0008 29«/26», 0009 41«/36»), all plate-confirmed: Migne opens quotations he never closes (« Ut scuto., « In inferno., « Frameae., « Conculcet in terra vitam meam,, « Stulta mundi « elegit Deus). Do not balance these.
11. @0855C — superdescendent (0007, VERS. 17). A rare compound, not a defect; rendered "they shall come down upon". detrudent has an unexpressed object, supplied as "it" (the caput/anima of the same clause).
12. @0856A — cum tamen et illi fidem, et hi sapientiam veram impugnent. The pronoun binding is the trap: illi = the heretics (defenders of the faith), hi = the philosophers (defenders of wisdom), so each impugns the thing he defends. Rendered "the former … the latter" to fix the binding in English, which has no ille/hic proximity contrast.
13. @0856B — hoc bene convenit primitivis. primitivis is substantival with no noun; rendered "those of the first age" rather than supplying "[Christians]" in brackets, since the gloss's subject (the apostles seen in posteris) makes the reference plain.
14. @0852D — CC drops a comma the plate prints (Dat exemplum fideliter, orantes audiri, page 431). Noted, not marked: the Latin column is our source's, [sic:]/[var:] do not reach punctuation, and the sense is unaffected either way ("faithfully" attaching to orantes or to audiri). Rendered on the sense our Latin gives: "those who pray faithfully are heard."
15. Migne's own verse numbering skips, reproduced as printed, never reordered: chunk 0006 runs VERS. 3 → VERS. 5 (no 4); chunk 0007 runs VERS. 8 → VERS. 10 and VERS. 13 → VERS. 15; chunk 0008 runs VERS. 5 → VERS. 8; chunk 0009 runs VERS. 2 → VERS. 4 and VERS. 7 → VERS. 10. All plate-confirmed on pages 431–434.

§7 · DECLINED — divergences and defect candidates I checked and did NOT mark

⭐ On Luke, 12 of 17 corrections came out of declined lists. This is the list.

Scripture that AGREES with the Clementine — checked, verbatim, clean

Divergences that are REAL and were judged not material — logged, not marked

• 0852A, 1Pt 4:18 — Migne « Si justus vix salvabitur, impius ubi apparebit? »; Clementine et si justus vix salvabitur, impius et peccator ubi parebunt? An omitted et peccator and a singular for a plural. The proposition is unchanged and the gloss uses only the justus/impius contrast. Declined — Pattern 14's bar is "changes the sense."
• 0853D, Gn 3:14 — « Terram manducabis » for the Clementine's terram comedes. A synonym; the gloss turns on terram, which both texts have. Declined.
• 0854A, Ps 7:2 — « O Domine Deus meus, etc. » for Domine Deus meus, in te speravi. A supplied vocative O on an abridged citation. Declined.
• 0854D, Ps 6:3 — « Infirmus sum, sana me, Domine » for Miserere mei, Domine, quoniam infirmus sum; sana me, Domine. An abridgment stitched from the verse's own words, and one this very commentary glossed forty columns earlier at Ps VI. Declined.
• 0856A, 1Pt 2:2 — « quasi modo geniti » for sicut modo geniti infantes. Synonym plus abridgment. Declined.
• 0856B, Is 40:12 — « Qui appendit tribus digitis molem terrae » for quis appendit…. A relative for an interrogative in a phrase quoted as a description. Declined.
• 0857A, Is 14:13–14 — « Ascendam in coelum et exaltabo solium meum, et ero similis Altissimo » for In caelum conscendam, super astra Dei exaltabo solium meum … similis ero Altissimo. ascendam/conscendam is a near-synonym, the missing super astra Dei is the recension's ordinary abridging, and the word order is the only other difference. Declined — this is the one I came closest to firing, and the reason I did not is that no clause of the gloss rests on any of the three differences.
• 0858A, Ps 9:11 — lemma « Quoniam non reliquisti » for quoniam non dereliquisti. Simplex for compound, identical sense. Declined.

Defect candidates that were NOT defects

• deeidam @0853B — plate says de-cidam. Withdrawn before it was ever fired. See §2.
• praedicaris @0856A (et est, non solum a perfectis, sed et ab insipientibus praedicaris) — a second-person passive dropped into a third-person context. It is correct Latin in a change of address: the gloss turns to God, as thou hast perfected praise four lines later confirms. Not a defect; rendered "thou art proclaimed."
• quibusdam quies, quibusdam tormentum @0852C — verbless, but a normal Latin ellipsis. No marker.
• Arguuntur in judicio, qui sine fundamento. @0851D–0852A — verbless relative clause, the same ellipsis. No marker.
• superdescendent @0855C — a rare compound, not a non-word; Pattern 9. No marker.
• The [Gen. III] / (Gen. III) bracket difference @0853D — the plate sets square brackets where our TEI gives [n: (Gen. III)]. That is the chunker's note form, not a textual matter. Noted only.

§8 · CHECKS RUN, WITH RESULTS

• Column anchors: 36 across the five chunks, verbatim, in order, and each checked against the word it stands beside in the Latin (2b), not merely for presence. Two are deliberately mid-construction, matching Migne: (Isa. [0851A] XXVI) sits inside an italic span, and [0855D], [0851D], [0852B], [0853D], [0857C] open a line immediately before VERS., exactly as the Latin sets them.
• [n: …] notes: 111, identical strings in identical order. No [nt:] conversions — every note in my range is a bare attribution siglum or a scripture locator, none is editorial prose (Pattern 15).
• Guillemets: exact 1:1 per chunk, unbalanced where Migne is unbalanced. See §6.10.
• Question marks: 0/0, 2/2, 0/0, 1/1, 0/0. Exact.
• Punctuation by band (Test 3): semicolons now exact in all five chunks (1/1, 2/2, 12/12, 6/6, 3/3) — my first draft ran 10 against Migne's 1 in chunk 0005, which is precisely the comma→semicolon drift the test exists to catch, and it was corrected rather than swept. Colons run 31/30, 28/28, 56/56, 30/29, 47/47 after markers are stripped.
• Word ratios: 1.46 · 1.47 · 1.44 · 1.45 · 1.43. Tight, and within the band on every chunk.
• Em-dashes: zero in all five chunks.
• 7a⁗ strip test: every marker deleted and the English read back. Nothing stranded; the one [sic:] leaves a clause with its verb, its subject and its sense.
• Test 2 / 2a (negation attachment): all 88 printed negation words walked one at a time, each mapped to the single Latin word it attaches to. Five sites carry the 2a trigger shape (a negative on a verb with a correlative adjacent) and all five were resolved from the structure, not from what reads well: 0851D non vult a mortuis, sed a vivis rogari · 0854C non delectari in carne, sed in Domino · 0855A non in ipso est ira, sed in angelis irae · 0855A non propter justitiam qua non gaudent, sed propter malitiam (two negatives, two different hosts, one sentence) · 0856A qui nondum escam, sed lac sumere possunt.
• Test 5 (does it ASSERT what he asserts?): each finished paragraph read back against the Latin. The two that changed as a result are §6.12 (the illi/hi binding, which had been reversed in draft) and §6.4 (the subject of coactus est).
• node scripts/plate-gate.mjs 8967 — ✓ on all ten of my markers.
• node scripts/verify-english.mjs 8967 — zero errors and zero warnings on 0005–0009 (the run fails only on chunks other stints have not yet written).
• node scripts/scan-raw-markers.mjs — clean.

§9 · FOR THE MERGE

1. data/plate-reads.json was edited by this stint — five new reads (cols 0851A–0859D) and an amendment to scan_defect. The amendment is a substantive correction to a rule the launch brief relies on; it should be reflected in the launch brief's §3 before the remaining stints run, or they will keep declining markers on 1 (mod 4) columns that are in fact perfectly legible.
2. deeidam @0853B wants a TEI patch (data/tei-patches/8967.json, → decidam). Not written here because it changes the Latin column for everyone.
3. The lemma brief's band labels are wrong for at least chunks 0005 and 0009 (§0). The counts are right. If another stint reports "my spans are not at the bands the brief gives," that is this, not a mis-split.
4. (CAS.) @0852D vs (CASS.) elsewhere is Migne's own inconsistency, plate-confirmed, and is passed through. If a later pass normalizes sigla across the work, this is one of the sites.

§E.0010 — stint file cruces-0010.md, verbatim

Stint 5 of Band A. Psalms X (Hebrew division) · X · XI · XII · XIII · XIV · XV · XVI · XVII.

0. Span reconciliation — REQUIRED CHECK, and it AGREES

data/briefs/8967-lemmata-0010.txt declares 307 spans for chunks 0010–0014. Counted mechanically in my own Latin (re.findall(r'\[^]+\*') over the body of each chunk):

307 = 307. No discrepancy. The brief's first and last entries ([0859A] Secundum Hebraeos. and [0870B] Magnificans, salutes.) are also the first and last spans of my Latin, so the split is correct at both ends, not merely in total.

⚑ Re-checked against the corrected brief issued mid-stint (stale column labels repaired): the header still declares 307 and the file still carries 307 [0…] rows. Membership is by chunk, not by band, so nothing moved. The count stands.

English span counts run higher by exactly the italics inside our own markers: 0010 +4 (three [var:]), 0011 +3 (one [var:], one [cj:]), 0012 +0, 0013 +0, 0014 +2 (one [var:]). Every italic span of the Latin has an italic span in the English at the same place.

1. The two heads both numbered X — NOT a defect, and nothing was touched

Chunk 0010 carries ## PSALMUS X.-- Secundum Hebraeos. (0859B–0861A) and then ## PSALMUS X. (0861A–0861D). Both are reproduced as Migne prints them, in his order, with his VERS. numbers restarted at 1 in the second. Read at the plate (p. 435, cols 859/860 and p. 436, col 861): the heads print exactly so. No renumbering, no merge, no marker.

The nine ⚑ MATCHES ONLY ELSEWHERE — Ps 9:2x flags in my brief (0859B–0860D) are the checker reporting correctly — the Hebrew division is the tail of Clementine Ps 9 — and no [var:] was fired on any of them. Collated individually against sources/vulgate/clementine-flat.txt:

Ut quid. = Ps 9:22 · Dum superbit. = 9:23 · Incenditur pauper…cogitant. = 9:23 · Quoniam laudatur. = 9:24 · Irae suae. = 9:25 · Inquinatae. / In omni tempore. / omnium inimicorum suorum dominabitur. = 9:26 · Dixit, / sine malo, / Non movebor. = 9:27 · cujus maledictione os plenum, / sub lingua ejus, = 9:28 · In occultis, = 9:29 · Oculi ejus in pauperem, / Leo in spelunca, = 9:30 · Inclinabit. = 9:31 · In finem. = 9:32 (ne videat in finem) · Contere brachium. / Quaeretur. = 9:36 · Dominus regnabit…saeculi. = 9:37 · desiderium pauperum, / Praeparationem cordis. = 9:38 · Judicare pupillo et humili, = 9:39. All agree with the Clementine word for word except the three sites marked in §2.

Vulgate numbering governs from PSALMUS XI onward and it was checked, not assumed: every lemma of Psalms XI–XVII in chunks 0011–0014 resolves to Clementine Ps 11–17 at the same number (spot-verified across all 24 VERS. heads in the range). Nothing is offset.

2. Markers fired — five, every one on a column read at the plate this session

Plate: on-disk raw/scans/pl113/patrologiaecurs04migngoog.pdf, PDF page = (column + 11) / 2, re-calibrated against the printed corner numbers on every page opened (p. 435 = 859/860, 436 = 861/862, 437 = 863/864, 438 = 865/866, 439 = 867/868, 440 = 869/870). All six pages of my range were read; all six are recorded in data/plate-reads.json. plate-gate.mjs 8967 passes every marker below.

And one Pattern-18 conjecture note:

3. ⛔ SIX CORPUS CORPORUM DEPARTURES FROM THE PLATE — patch candidates, NOT [sic:]

Every one was found by reading the plate, none is detectable from the Latin alone, and none takes a marker, because a [sic:] would accuse Migne of his transcriber's defect. This is ~1 per page, against the standing 9003 estimate of ~3 per page, so the class is behaving as documented.

1. 0862C — a compound siglum split. The plate sets (AUG., CASS.). Our TEI gives bare running text AUG., followed by a separate [n: (CASS.)]. The English reproduces the twin.
2. 0863D — a siglum unencoded and its point dropped. The plate sets (CASS.); our TEI gives bare text (CASS) — no [n: ] wrapper, no full stop. English reproduces the twin.
3. 0864A — a dropped letter inside a siglum. The plate sets (CASS.); our TEI gives [n: (CAS.)]. ⚠ Do not sweep this: Migne's own (CAS.) at 0863A and 0863C was read at 400 dpi and IS his. Two forms, one of them ours, and only the plate tells them apart.
4. 0866D — an inserted word. The plate reads non humiles, adipem, id est, carnalem sensum retinuerunt; our TEI inserts vel before adipem (read twice at 400 dpi). This is the only one of the six that touches the sense.
5. 0870A — a full stop softened to a comma. The plate reads et dilatatio. et confirmatio (500 dpi); our TEI gives et dilatatio, et confirmatio. Pattern 8 class.
6. 0870C — an inserted comma. The plate reads Magnificans salutes.; our TEI gives Magnificans, salutes..

Plus a seventh which is not a word at all: 0867C captu+m. The plate reads captum cleanly (p. 439); the + is the only one in this work's TEI. The English renders Migne's word ("a man taken captive") and fires no marker — carrying an unrenderable CC artifact into the English under a [sic:] would blame Migne for it.

⛔ THE POLICY I APPLIED, stated so the merge can undo it in one pass. For (1)–(6) — all renderable — the English follows the Latin twin as it currently stands, so the parallel columns agree and a blind reader does not report a dropped or invented word that is really a pending patch. Only for captu+m, which cannot be rendered at all, does the English follow the plate instead. If data/tei-patches/8967.json is extended and the work re-chunked, item 4 requires one English edit (0013.md: drop the "or" before "the fat") and items 5–6 require none, being punctuation of the Latin only. I did not patch or re-chunk: seven other stints are writing chunks of this work in parallel and chunk-work.mjs rewrites every chunk from TEI.

4. Divergences from the Clementine CHECKED AND DECLINED — the declined list

Per 7a″ these are the deliverable as much as the fired markers. Each was collated against sources/vulgate/clementine-flat.txt and each is Migne's, confirmed at the plate.

Declined as immaterial (sense unchanged):

• 0861B quia quae perfecisti destruxerunt — Ps 10:4 has quoniam. A connective swapped for its synonym inside the gloss's own re-quotation; the lemma above it prints Quoniam correctly. No sense turns on it.
• 0863A Cupio dissolvi et esse cum Christo (Phil. I.) — Phil 1:23 reads desiderium habens dissolvi. Same sense; the cupio form is the ordinary patristic citation of this verse.
• 0863B Est peccatum ad mortem, non pro eo dico ut quis oret — 1 Jo 5:16 reads non pro illo dico ut roget quis. Pronoun and word order only; the negation and the sense are intact.
• 0863D Veniet ex Sion qui avertit captivitatem, vel impietatem a Jacob — Rom 11:26 (citing Is 59:20) reads qui eripiat, et avertat impietatem a Jacob. ⚑ Declined for a positive reason, not a weak one: Migne's own vel shows the Glossa is offering the two readings itself. A [var:] would tell the reader what the text already tells him, in our voice.
• 0864C Nec fecit proximo malum — Ps 14:3 has proximo suo malum. A possessive dropped in a lemma that is clipped anyway.
• 0864C Vade retro, Satanas — no Clementine verse reads this; Mt 16:23 has Vade post me Satana, Mk 8:33 Vade retro me Satana. This is the proverbial form, adduced by the gloss as a commonplace ("ut cum dixit"), not quoted as a proof-text. Crux, no marker.
• 0864C timentes Dominum glorificat — Ps 14:4 has timentes autem Dominum glorificat. Clipped, not divergent.
• 0864D Jam non dicam vos servos, sed amicos — Jo 15:15 reads Jam non dicam vos servos … vos autem dixi amicos. A compression of two clauses into one, with no change of assertion.
• 0865A Sanctis qui sunt in terra — Ps 15:3 has in terra ejus. Clipped.
• 0868D Ecce praevenerunt — Ps 17:19 has praevenerunt me in die afflictionis meae; the Ecce is the gloss's own pointing word, not a variant reading.
• 0869A/B tu, Domine, populum humilem salvum facies — Ps 17:28 has no vocative. An added Domine in a formula the gloss quotes three times; nothing turns on it.
• 0869A Habemus thesaurum istum in vasis fictilibus — 2 Cor 4:7 has Habemus autem. A connective dropped at the head of a quotation.
• 0870B Disperdidisti, Pater, — Ps 17:41 has odientes me disperdidisti and no vocative; the gloss introduces it as Supra ait and supplies the addressee its own argument needs (eadem operatio Patris et Filii). The gloss's construction, not a divergent text.
• 0870B Et exaltetur Deus salutis — Ps 17:47 has Deus salutis meae. Clipped, and the gloss immediately restores de Deo salutis meae.
• 0868A Virga furoris mei Assur, ipse vero non cognovit — a conflation of Is 10:5 and 10:7, freely quoted; neither half is being adduced as a text with a reading at stake.
• 0868A Clarificavi, et iterum clarificabo — Jo 12:28 exactly, with the Clementine's leading Et dropped.

Declined suspicions that the plate REFUTED (recorded because a checked negative is worth as much as a finding):

• 0866B doubled scilicet — Scilicet opera hominum: quas scilicet merentur opera hominum reads like a dittography and is not: the plate prints it exactly so (p. 438). Rendered as printed, no [d:].
• 0869A sordescat — checked at the coordinator's request, since the same form was a confirmed CC normalisation elsewhere in this work (the other site's plate reads sordescet). Here the plate agrees with our TEI: p. 440, read at 500 dpi, prints qui in sordibus est, sordescat adhuc. No patch is owed at this site, and the other site's finding must not be generalised to it.
• 0867B iustorum for justorum — our TEI's only i/j outlier in the range; the plate prints justorum. An orthographic normalisation of no consequence, exempt under Pattern 9. Recorded only so the next reader does not spend the page-fetch I already spent.

Declined because they are Migne's own apparatus, not ours to mark:

• 0868A aqua satians [ alias saliens] — Migne's own inline variant note, in his brackets, with his italic alias. Carried through verbatim, the reading untranslated (a reading needs no English), exactly as a numbered [cn:] would be. Not marked as ours.
• 0868A [n: (Matth. II)] — the reference is attached to non per speciem, which is 2 Cor 5:7, and it does not fit; the next paragraph cites (II Cor. V) correctly. Migne's citation, passed through verbatim under convention 4. Not expanded, not corrected — a reference correction is citation-corrections.json's business, not a translator's.
• (ID.) / (Ibid.) — 20 of them in the range, none expanded.

5. Cruces in the gloss prose

• 0859D vel, Deus eum, ut hic ulciscatur (Ps 9:25). The verb is elided; it must be carried over from non quaeret three clauses earlier, so the English reads "God will not seek him out." The negation therefore governs a verb the Latin does not print at this point. Rendered with the negation, per 7a; the alternative (an affirmative "God seeks him out to avenge himself here") would be exactly the silent repair 7a forbids.
• 0859D id est levior, lingua ejus, quae intus nocet: illa, scilicet poena extra. The comparative has no quam, and illa has no clear antecedent. Rendered as closely as the words allow ("that is, it is the lighter; his tongue, which harms within: that other, namely, is the punishment without"). Not smoothed; not marked — the words are all real and in their slots.
• 0861C usque ad qui tunc dicit sancta sacramenta, cum per sanctos dentur. The abridgment's resumption tail opens with a relative whose antecedent was cut away with the omitted text. Parsed as ordinary prose per the brief's warning and rendered "he then calls the sacraments holy"; the qui is left to do the work of a demonstrative, which is what the splice made of it.
• 0866B Quo sui judicatus videatur prodisse (Ps 16:2). sui is doing something the syntax does not license cleanly. Rendered "whereby he who was judged may be seen to have come forth as his own." Real words in real slots; crux, no marker.
• 0866C Mira, sunt quae nova sunt: haec sunt quae in carne cessit mirantibus Judaeis. cessit is intransitive and takes no quae. Rendered literally ("which he yielded in the flesh"). Crux, no marker.
• 0867B et hoc qui praeoccupavit mortalitas et possibilitas (Ps 17:8 gloss). qui does not agree with anything; the sentence does not construe. Rendered as closely as the printed words allow, "and this, which mortality and possibility forestalled." No conjecture is offered, because two are available (quod, quae) and nothing in the passage chooses between them.
• 0868A Dedit vocem: ut non dixit: Hic est Filius meus dilectus — flagged, examined, and found sound: at Jo 12:28 the voice from heaven did not say what it said at the Transfiguration, and Cassiodorus is drawing exactly that contrast. The plate prints non (p. 439) and the English carries it. Recorded here because it reads at first like a dropped cum and a later reader will stop on it.

6. Unexpressed subjects supplied, with the authority for each

Per the brief's dominant-defect warning, every pronoun the Latin does not print:

• 0859B Dum superbit. → "While the wicked man is proud." Subject impius from Ps 9:23, the verse the lemma clips.
• 0859B Quoniam laudatur. → "For the sinner is praised." Subject peccator from Ps 9:24.
• 0859C Inquinatae. → "Defiled." Its noun is viae illius (Ps 9:26), and the gloss's own next clause supplies it (et ideo viae ejus … sordent), so the English keeps the participle bare and lets the gloss govern.
• 0860C Vides enim quoniam tu laborem et dolorem. → "…that thou considerest labour and sorrow." The verb consideras is the next word of Ps 9:35 and the lemma stops before it; English cannot leave the accusatives without a verb. Expanded per convention 3.
• 0861B Quoniam quae perfecisti. → "For the things which thou hast perfected." Here I did not supply destruxerunt, because the same paragraph re-quotes the fuller quia quae perfecisti destruxerunt and expanding both would erase Migne's clipping.
• 0863D Dominum non, → "The Lord they have not called upon." invocaverunt from Ps 13:5.
• 0864D Et munera super innocentem. → "And gifts against the innocent." ⚑ I did NOT expand to the Vulgate's non accepit. Convention 3 would license it, but 7a forbids a negation in the English that the Latin does not print, and 7a wins: the gloss's own sed non contra innocentes carries the sense two clauses later.
• 0867B Frameam. → "The sword," not "thy sword": Ps 16:13 has frameam tuam, and the gloss makes the sword the soul, not God's.
• 0867A Populus catulus leonis — "the people is the young of a lion." The predicate is unexpressed; Populus is the printed subject.
• 0865B sed ab ipso habet omnia — subject of habet supplied as Christ the man, not God: "but Christ the man has all things from him." Added by the polarity pass (chunks 0010–0019). The English had carried Deus forward as the subject ("but has all things from him"), which made the sentence say God receives all things from a creature — the opposite of the anti-Pelagian conclusion the same sentence draws (destruitur arrogantia hominum et haeresis Pelagianorum). §7a mechanism 5 (agency mis-bound); the noun is the sentence's own.

7. Vocabulary axes — every decision in the range

• In finem → "Unto the end" at 0860A, 0861A, 0862D. ⚑ The 0860A occurrence glosses Ps 9:32's ne videat in finem, so the rendering is not merely the titulus convention here — the gloss (Negant videre Deum) argues from the end of seeing. No exception taken.
• Diapsalma at 0859B: unanglicized, and set roman, because Migne sets it roman there (p. 435). The launch brief says "italic as Migne sets it"; he does not, so neither do we, and the span counts stay 1:1.
• gentes family, all seven occurrences decided from the context, not by rule:
• 0859D Omnes gentes occupabo (Antichrist speaking) → nations — generic.
• 0860D ab hoc regno peribitis gentes → nations — the psalm's own vocative (Ps 9:37), generic, no Judaei in view.
• 0862C Inopes, gentiles … pauperes, Judaei → Gentiles — the typological opposition in its purest form, printed in one clause.
• 0863D plenitudo gentium (Rom 11:25) → Gentiles — same opposition, the Jews named in the same sentence.
• 0865C Haereditas enim ad gentes → Gentiles — Christ's inheritance against the cup of the passion, in a psalm the gloss reads ut sit rex Judaeorum.
• 0867B Gentiles in inferno → Gentiles (gentilis per the 2026-09-04 addition).
• 0868C transitus ad gentes and 0870B exaltatio gentium → Gentiles — both stand against Judaei in their own clause (depressio Judaeorum, et exaltatio gentium).
• 0869D non dii gentium → nations — generic, and set against paganos in the next clause, which is rendered "pagans".
• singular gens at 0864A ut pro gente et loco → nation. This is the Jo 11:48 formula (tollent locum nostrum et gentem) and it is the axis's own worked example.
• usque ad. 25 italic occurrences, all the abridgment formula, all → "as far as". ⚑ Three look-alikes were decided the other way and are named here per addendum D: 0863D qui sunt usque ad secundam divisionem (unitalicised, spatial → "as far as the second division"); 0863D usque ad unum is not in my range but its neighbour usque ad: Timor Dei ante oculos IS the formula with Migne's colon carried; and 0865D Usque ad noctem, which is Ps 15:7's own words and renders "Even till night" — a lemma, not a splice.
• turba / multitudo / daemonium: the range contains no turba and no multitudo (grep run and checked against the brief's instrument warning — the natio hits in this range are all examinatio-, conclusione, and the like, i.e. the documented false positive). One daemonibus at 0868C → demons; diabolus at 0864C, 0866D, 0867A, 0870A → devil.
• mysterium / sacramentum: no mysterium in the range; sacramenta at 0861C and sacramentorum at 0868A → sacraments.
• thou / you (Pattern 17): every second-person singular is thou — the psalm's addressee throughout, and the gloss's own ut justos accipias (0861D) and Hinc trahe exemplum (0865D), both singular imperatives to the reader. No plural second person occurs in the range.

8. Sections read clean, and how

• Question marks: 0010 five, 0011–0014 one each; English matches Latin exactly, chunk by chunk. The five in 0010 are propter quid? / quid proficit…? (0860B), Cur dicunt transmigra? / …sagittare rectos? (0861B), Justus autem, quid fecit? (0861C) — all read on the plate. No mark was supplied and none deleted; no 8b recast was needed, because Migne prints no ? on an indirect question anywhere in this range.
• Column anchors: all 26 verified present, in order, and against the word they stand beside in the Latin (test 2b), not merely present. Two sit mid-construction and were split to match: et exaltetur [0860B] manus → "and let [0860B] thy hand be exalted", and De absconditis [0867B] tuis. → "Of the hidden things [0867B] that are thine." — two spans in the Latin, two in the English, the anchor between them. ⚠ One anchor cannot be placed exactly: Bonorum [0865B] meorum reverses in English, so it stands as "Of my [0865B] goods"; the slot inside the two-word phrase is the same, the word order is not.
• Notes: 146 [n: …] across the five chunks, identical in count, order and content to the Latin. No [n:] was converted to [nt:] — every one is a bare attribution siglum or a scripture reference, i.e. a citation, never prose.
• Negation, checked by attachment and not by count (test 2/2a). The trigger population — a negative near a quantifier or correlative — fires at five sites in the range, and each host was proved from the sentence's structure:
• 0859C non solum non metuitur reprehensor, sed etiam auditur laudator — the first non is on solum, the second on metuitur; the sed etiam clause proves it.
• 0864C malignum non esse, nisi qui a Creatore ad creata convertitur — the non is on the predicate and the nisi restores the exception; "no one is malignant save he who…".
• 0861D sed non nisi his qui peccatores sunt — non nisi is the correlative, rendered "but not save upon those who are sinners", both particles carried.
• 0867A nemo non prius in se quam in alium peccat — a double negative, carried as one ("there is no one who does not sin against himself first"), not collapsed into a positive.
• 0863B Non solum propter me illuminare, sed ne inimicus gaudeat — non on solum, and the second clause's ne kept as its own negative.
• 7a⁗ (a marker swallowing the sense): not applicable — no [sic:] is fired in this range, so nothing is quarantined and nothing can hide inside a bracket. Stripping every marker from the English leaves five readable chunks with no stranded article or preposition.
• Punctuation by band: colons and semicolons compared band by band against the Latin twin across all 26 anchors; zero mismatched bands. (The two plate-vs-TEI punctuation departures at 0870A and 0870C are §3 items — differences between the plate and our Latin, not between our Latin and our English.)
• Foot-of-page apparatus: six pages read full width (435–440) and ZERO notes of either layer — no numbered Forte conjecture and no asterisk cross-reference. The work's coverage stays spot; this is six more checked zeros, not a claim that the work has none.
• The mod-4 clipping rule fails on my pages. Columns 861, 865 and 869 are all ≡ 1 (mod 4) and all three are CLEAN in this copy. That is three more counterexamples on top of the ones stints 0015–0019 and 0025–0028 recorded the same day, and it means no marker in this range needed the Gallica second witness: every column of 0859–0870 is legible in the on-disk scan. The rule should be treated as withdrawn, not merely qualified.
• In Hebraeo clauses: one of the work's three outstanding ones falls in my range, at 0868A (quod in Hebraeo), and it is now read: the plate sets no Hebrew type there at all, so nothing was dropped and none is to be supplied. Two remain unread elsewhere in the work.
• The fourth test (does the English ASSERT what he asserts?): run last, paragraph by paragraph against the Latin, with the counts already known clean. It changed three renderings: the elided quaeret at 0859D (§5), the refusal to supply non accepit at 0864D (§6), and the scope of non in labiis dolosis at 0866A, where the negative belongs on the prepositional phrase and not on prodeat — "let my judgment come forth from thy countenance not in deceitful lips", never "let not my judgment come forth".

§E.0015 — stint file cruces-0015.md, verbatim

Psalms XVIII–XXVIII. Stint 0015 of Band A.

⚑ Every column label in this file was derived from the [NNNNx] anchors in my own Latin chunks, not from the lemma brief. The brief's column labels were stale when I received it (corrected on disk 2026-09-04); membership was never affected, so §0's reconciliation stands unchanged, but any label here that disagrees with an older copy of the brief is mine and is the one to trust.

0. SPAN COUNT RECONCILIATION — the brief is CORRECT

data/briefs/8967-lemmata-0015.txt declares 330 spans. Counted in my own Latin:

Exact agreement. No mis-split at either end of my range: 0015 opens on VERS. 1.-- In finem. at 0870C, which is the brief's first line, and 0019 closes on Diluvium. at 0883A, which is its last.

English span counts run over the Latin by exactly the number of italic runs inside my own markers (0015 +1, 0016 +1, 0017 +0, 0018 +0, 0019 +5). Anchors, [n:] notes and question marks are at 1:1 parity in all five chunks; verify-english.mjs reports no error on any of them.

## 1. ⛔⛔ THE HEADLINE, AND IT IS NOT A TRANSLATION QUESTION: OUR LATIN TWIN HAS BEEN SILENTLY ## CORRECTED, AT ROUGHLY ONE SITE PER COLUMN

I read six plates for this range (archive.org patrologiaecurs04migngoog, PDF page = (column + 11) / 2, re-verified against the printed corner numbers on p. 441, which reads 871 in the left corner). Eleven places where our chunked Latin does not say what Migne printed — and the direction is overwhelmingly one way: our twin is tidier than the plate, and tidier in the direction of the Vulgate and of ordinary Latin.

⚑ Two further sites are recorded with lower confidence because the italic a of this face at this scan quality is not always separable from u: 0878D Oculi supientis in capite (twin: sapientis) and 0878D De necessitutibus (twin: necessitatibus). Both would be normalizations if real. I record them as unsettled, not as findings — a second witness (Gallica ark:/12148/bpt6k5505319w, f = (column + 1) / 2) settles either in one HTTP request. ⭐ The control that makes them worth raising at all rather than dismissing as a scan artifact: unambiguous italic a's on the very same column (Anima, usque ad, Semen) image cleanly.

⚑ What follows, and it binds any later reader of these chunks:

1. This is the same class 9003's stints found in the sibling Glossa Psalter (PL 114), at the same order of density. It is not a local accident and it is not confined to one volume.
2. No [sic:] or [cj:] may be fired anywhere in this range on the strength of the Latin twin alone. The twin has already been cleaned of most of the plate's defects; a marker keyed to it would either miss the real defect or, worse, blame Migne for a reading he does not carry. My one [sic:] and my one [cj:] are safe only because I read their plate lines.
3. ⛔ Five [sic:] candidates were WITHDRAWN by the plate — ocus, commmendando, spiriritus, lilexi, genus. Every one of them read as a confident Migne defect from the Latin alone, and every one is ours. That ratio (5 withdrawn, 1 confirmed) is the measured answer to CLAUDE.md 8a's warning that a translating agent's attribution guess is not evidence.
4. The fix is data/tei-patches/8967.json + a re-chunk, which is outside a translating agent's remit and must not be done mid-batch. I am escalating it, not doing it.
5. My English follows the TWIN, not the plate, on the 9003 precedent — our English sits in a parallel column beside the twin — except at the four sites where the twin is a non-word and there is nothing to render: ocus → "a place", commmendando → "commending", spiriritus → "spirit", lilexi → "I have loved". Each takes the plate's reading, and each is listed here so the call can be overruled.

1a. ⭐ THE mod-4 CLIPPING RULE IS REFUTED — two counter-examples in my range

The launch brief's §3 stated that a left-hand column congruent to 1 (mod 4) loses 3–5 characters at the inner margin, verified 5 of 5. That rule is now WITHDRAWN work-wide (the coordinator's correction of 2026-09-04, on counterexamples from three stints). In my range:

⛔ Two of the three predicted-clipped columns in my range are whole. Other stints found the same at cols 853, 857 and 901. Work-wide tally after my six pages: clipped 841, 849, 873, 893, 897, 1045 · clean 853, 857, 871, 875, 877, 879, 881, 895, 899, 901, 1043, 1047. The clipping is real but it is per leaf, not arithmetic — the rule was derived from five pages and generalized past its evidence.

⚑ The failure mode runs in the direction nobody audits. Applied blind, the rule does not produce false markers; it silently withholds true ones on columns this copy answers perfectly well, and a withheld marker leaves no trace anywhere. My [var:] at 0881C is exactly such a case: it stands on a line-opening word in a column the rule would have disqualified, and the page shows the line whole. Predict with the rule if you like; never decide with it. Where a column genuinely IS clipped and the marker rests on a line opening, the second witness (Gallica) is still required — that part stands.

2. MARKERS FIRED — six, every one on a column I read at the plate

node scripts/plate-gate.mjs 8967 reports ✓ on all six. My nine plate reads are recorded in data/plate-reads.json with honest band-level from/to: I read parts of six pages, not six whole pages, and the entries say which parts.

⚠ Per Pattern 14 none of these say why the readings differ. Migne's compositor, his copy-text, an Old Latin reading and the glossator's memory are all live, and telling them apart is research.

3. 7a″ — THE VULGATE COLLATION. This is the deliverable, and it includes the clean ones.

Every lemma in my range was read word by word against sources/vulgate/clementine-flat.txt before its English was written. Six divergences took a marker (§2). The rest are below, INCLUDING the ones that agree exactly — a list of findings alone cannot be told apart from a report by an agent that never looked.

3a. Divergences found, marker DECLINED, with the reason

3b. Lemmata checked and found to AGREE with the Clementine exactly

Recorded because a checked negative is worth having. Ps 18: Coeli enarrant (18:2), Dies diei (18:3), In sole, Tabernaculum suum, De thalamo, Gigas ad currendam (18:6), Nec est qui se abscondat (18:7), Lex Domini (18:8), Justitiae Domini (18:9), Timor Domini (18:10), Desiderabilia super aurum (18:11 — the twin's spelling; see §1), Ab occultis (18:13). Ps 19: Exaudiat, In die tribulationis, Dei Jacob (19:2), Mittat/Auxilium (19:3), Memor, Sacrificii (19:4), Tribuat, Omne (19:5), Laetabimur (19:6), In potentatibus, Salus dexterae (19:7). Ps 20: Domine in virtute tua laetabitur (20:2), Quoniam praevenisti, Coronam (20:4), In salutari tuo: gloriam et magnum decorem, super eum (20:6), In gaudio cum vultu tuo, Quoniam rex sperat in Domino (20:7), Inveniatur manus tua (20:9), Fructum eorum de terra perdes, et semen eorum (20:11), Quoniam pones (20:13), Exaltare (20:14). Ps 21: Deus, Deus, Respice in me, Quare me dereliquisti, Longe a salute mea (21:2), In sancto habitas, Laus Israel (21:4), In te speraverunt patres nostri, speraverunt (21:5), Ego autem, Vermis et non homo, opprobrium (21:7), Ab uberibus, De ventre (21:10–11), Ne discesseris a me, Tu extraxisti, Spes mea ab uberibus, In te projectus sum (21:11–12), Circumdederunt (21:13), Aperuerunt, Sicut leo (21:14), Sicut aqua, In medio ventris mei (21:15), Aruit … testa virtus mea, Lingua mea (21:16), Foderunt (21:17), Dinumeraverunt omnia ossa mea, Inspexerunt me (21:18), Diviserunt sibi vestimenta mea (21:19), Unicornium, humilitatem meam (21:22), Qui timetis Dominum, laudate eum (21:24), Timeat eum omne semen Israel, quoniam non sprevit, Pauperis (21:25), Apud te laus mea in Ecclesia magna (21:26), Laudabunt Dominum qui requirunt eum: vivent + Corda eorum … convertentur ad Dominum (21:27–28, continuous across the verse boundary and exact on both sides), Manducaverunt et adoraverunt, In conspectu ejus cadent (21:30), Annuntiabitur (21:32). Ps 22: Psalmus David, Dominus regit me (22:1), In loco pascuae, Super aquam (22:2), Animam meam, Deduxit, Super semitas (22:3), Mecum es, Virga tua (22:4), Parasti in conspectu, In conspectu meo mensam, Et calix meus (22:5), Et misericordia (22:6), Et ut inhabitem (22:7). Ps 23: Domini est terra (23:1), Et super flumina (23:2), Quis ascendet (23:3), Quis stabit, Innocens, Qui non accepit, In vano (23:4), Attollite, Et elevamini, Portae aeternales (23:7), Quis est (23:8), Dominus fortis, Fortis, Potens, Potens in praelio, Virtutes (23:10). Ps 24: In finem, Ad te, Domine, levavi (24:1), Confido, Neque irrideant me inimici, Supervacue, Vias tuas, Domine, demonstra (24:2–4), Doce, Deus salvator meus (24:5), Reminiscere, Miserationum, Quae a saeculo (24:6), Delicta, Juventutis, Ignorantias, Memento mei, tu, Secundum misericordiam (24:7), Dulcis, Rectus, Propter hoc, In via (24:8), In judicio (24:9), Misericordia et veritas, Testamentum (24:10), Propter nomen tuum, Domine (24:11), Quis est homo (24:12), Anima, Semen, Firmamentum est (24:13–14), Unicus (24:16), De necessitatibus (24:17), Respice (24:19), Innocentes et recti adhaeserunt (24:21), Israel (24:22). Ps 25: Judica, In innocentia (25:1), Proba me, Ure renes (25:2), Misericordia, Complacui in veritate (25:3), Non sedi, Odivi, Vani (25:4–5), Malignantium (25:5), Lavabo, Et altare, Inter innocentes (25:6), Vocem laudis (25:7), Domine, Dilexi decorem (25:8), Ne perdas, cum impiis (25:9), Dextera eorum (25:10), Pes, stetit, in directo (25:12). Ps 26: Dominus illuminatio, Dominus protector vitae meae, A quo trepidabo (26:1), Super me nocentes, edant, carnes, Infirmati sunt (26:2), Castra, In hoc ego (26:3), Unam petii, Omnibus diebus vitae meae, Ut videam (26:4), In abscondito (26:5), Circumivi (26:6), Exaudi, Domine, vocem meam (26:7), Tibi dixit, Cor (26:8), Ne avertas, Ne declines, Salutaris meus (26:9), Quoniam pater, patrem, Matrem (26:10), Legem pone mihi (26:11), Tribulantium (26:12), Credo videre, Bona (26:13), Exspecta (26:14). Ps 27: Ad te, Domine, Ne sileas, et assimilabor (27:1), Ad templum (27:2), Da illis (27:4), Quoniam non, Destrues, Benedictus, Dominus (27:5–7), Salvum, Et rege, Extolle (27:9). Ps 28: Afferte Domino, Filios arietum, Vox Domini (28:1–3), Libani, Cedri Libani (28:5), Desertum (28:7), Dominus (28:10), Diluvium (28:10).

⚑ Two spans in the brief carry a ⚠ NOT in Clementine verbatim that is answered by Migne himself, not by us. At 0880D the gloss prints Alia littera: Ut contempler delectationem Domini — Migne is announcing his own variant against Ps 26:4's ut videam voluptatem Domini, so a [var:] would be re-asserting what the text has just said. And at 0880D avertat vultum is the glossator paraphrasing Ps 26:9's Ne avertas faciem tuam, inside his own sentence, and italicized as a catchword rather than quoted. Neither is a divergence in Migne's scripture.

4. THE HARD READINGS — where I chose, and what decided it

4a. 0876B Sed isti ad mensam divitis — the most interesting crux in the range, and it is DECLINED

Our twin prints De hac dicitur: Sed isti ad mensam divitis, id est, Christi, scito et te similia debere praeparare. That does not construe: Sed isti has no verb and no antecedent, and the sentence then swings to a second-person imperative.

The plate (p. 443) reads Seb isti — a non-word in italic, and the b is unmistakable at 300 dpi.

⚑ The conjecture I believe is right, and the reason it is in the crux and not in the text: Sedisti — "thou hast sat at the table of a rich man; know that thou too must prepare the like," which is the well-attested school form of Prov 23:1–2 and supplies exactly the verb and the person the sentence is missing. The plate's Seb + isti is then a broken d plus a word division at the wrong place — Pattern 10's split-run shape.

⛔ I did not fire anything. A [sic:] may only wrap type carried verbatim from the plate, and our twin does not carry Seb; a [cj:] requires Migne's printed word to appear verbatim in the twin, and Sedisti appears in neither. The honest state is: the twin is wrong, the plate is broken, the conjecture is strong, and the repair belongs in data/tei-patches/8967.json, after which a [sic: Seb isti] with the conjecture beside it becomes both possible and correct. My English renders the twin as it stands — "But these at the table of the rich man," — which is deliberately as rough as the Latin.

⭐ The rival I weighed and rejected: reading isti as a dative ("but to that man, at the table of a rich man…"). It loses because it still leaves the clause verbless and it does not explain the plate's b.

4b. usque ad — the formula, and one look-alike

The italic usque ad splice formula occurs 44 times in my range (45 occurrences of the words in all) and is rendered "as far as" in every one — the English carries exactly 44 — including the three that carry Migne's own punctuation into the span (usque ad, at 0871D, usque ad: Dirige, at 0877C, usque ad Dulcis, at 0878A — the punctuation and the extra word ride as printed).

⚑ One occurrence is NOT the formula and is not italicized, and I checked it rather than assuming: 0870D haec doctrina diebus et noctibus continuatur usque ad posteros — this sits inside the resumption tail of a formula that has already fired, it is unitalicized, and its sense is extent in time. Rendered "down to those who come after." Per axis 6, the axis governs the formula, not the words wherever they occur.

4c. Diapsalma / Diapsalmus — unanglicized, and ROMAN, and here is why

Three occurrences: Diapsalma at 0872C (Ps 19:5) and 0872D (Ps 20:4), Diapsalmus at 0877A (Ps 23:7). All three pass through unanglicized per launch brief §2.2 — never "Selah," never "pause."

⚑ But all three are set ROMAN in my Latin, not italic, and the plate agrees (p. 444 confirms Attollite. Diapsalmus. with the term in roman while Attollite is italic). The brief's phrasing is "italic as Migne sets it"; Migne does not set it italic here. I have kept them roman, which also preserves the sacred 1:1 italic-span parity — italicizing them would have introduced three English spans with no Latin twin. ⚑ Flagging it for the merge: if the other stints in this band italicized theirs, the band is inconsistent and mine is the one that follows the plate.

⚑ Note also that Migne prints two forms of the word, Diapsalma and Diapsalmus, within four columns. Both ride as printed; neither is normalized to the other.

4d. In finem — 5 occurrences, all "Unto the end."

0870C (Ps 18), 0872B (Ps 19), 0872D (Ps 20), 0877B (Ps 24), 0880A (Ps 26). ⚑ Two further occurrences are lowercase and not the lemma and are rendered as ordinary prose: 0870C respicit in finem, id est, Christum and 0882B per totum hoc saeculum usque in finem. Never "for the choirmaster." ⭐ Psalm 18's own gloss in my very first line is the argument in miniature: hic propheta respicit in finem, id est, Christum, vel in ultima tempora — the commentary glosses the lemma as END and as Christ, and a "choirmaster" rendering would leave that sentence expounding a word not on the page.

4e. The gentes axis — two occurrences, decided opposite ways

• 0871A qui omnes gentes in unum erat congregaturus — Pentecost, all tongues, no Jew/Gentile opposition in view. Generic → "all nations."
• 0882C spiritum timoris dedit Judaeis et gentibus — the opposition is printed in the sentence itself. → "to the Jews and to the Gentiles."

⚠ I checked the instrument before reporting either. A raw natio grep over my range returns zero true hits; \bgentes catches nothing else. There is no natio, no gentilis, no gentilitas, no ethnicus, no turba, no multitudo, and no daemonium anywhere in cols 0870C–0883A. diabolus occurs twice (0876C a principe diabolo, 0881B saeculum pater vel diabolus) and is "devil" in both. mysterium occurs twice (0874C legis mysteria, 0877B novo mysterio) and sacramentum five times (0871A, 0872A, 0874C, 0876B, 0880A); they are 1:1 throughout, and 0874C carries both in adjacent sentences — sacramentum ablutio peccatorum and legis mysteria comparat cerae — which is where the axis is settled mechanically.

4f. The unexpressed subject — every supplied pronoun, and what decided it

4g. Negation fidelity — the sites where carrying the printed word makes the sentence odd

⭐ 0870D–0871A is the one that would have been repaired silently. Migne prints:

Planius diceret, in quibus non audiantur voces eorum, vel, omnium sermonum audiantur voces eorum: illud idem facit per relativum.

The first limb has the non; the second does not, and the two limbs are offered as alternatives (vel). Read straight, the gloss offers "in which their voices are not heard" and "the voices of all their words are heard" as the same thing, which they are not. I carried both exactly as printed and did not supply a non in the second limb. Column 871A was read at the plate (p. 441) and it prints exactly this. It is Migne's, and it is the kind of thing this edition exists to show.

⚑ Test 2a sites — a negative resting on a verb with a correlative nearby — fired twice, and both were re-hosted from the sentence's structure rather than from what read well:

• 0879D quia non eis, sed concordat domui Dei. The non looks as though it sits on concordat. The sed correlative shows it governs eis. Rendered "because it is not with them, but with the house of God, that he agrees," not "because he does not agree with them."
• 0880B ut maledico dente non me, sed carnalia desideria consumant. Same shape, same resolution: the negative is on me, and the English puts it there — "so that with their cursing tooth they consume not me, but my carnal desires."

Other negatives carried without incident and checked individually: 0871C qui non docuit quod non fecit (two), non serviliter · 0871D non foras mittitur · 0872A Non est majus delictum · 0873C non justitiae, sed delictorum · 0873D non exaudies (twice), non ad insipientiam, nullae in eo tenebrae, non auditur · 0874A Vermis et non homo · 0874B non vir, non in peccata hominis, Nec conceptio sine Deo, nec in peccato · 0874C non ad sapientiam reseratum · 0874D Non ut fenum · 0875A ne in majorem numerum crescerent · 0875B sed non tunica desuper, quam nemo potest dividere · 0875C Non ejus qui, non tumentis, non praesumi · 0875D non prohibeat, non pro instructione · 0876C qui non permanent · 0876D non deputavit animam rebus non permanentibus (two), non juravit in dolo, nullis eam fecibus · 0877D non erubescam, in nullo sustinente patienter fallit · 0878A Non pro aetate, quae mala esse non scimus, nemo bonus nisi Deus · 0878B non ex crudelitate · 0879B Non angelorum · 0879C non merita, sed misericordiam, ne quid delicti remaneat, non asperitas, non sedi · 0879D non est locus veri sacrificii extra · 0880A Ne perdas (twice), quibus hic non communico · 0881A non aliud quam doces · 0881A non nisi unam, non putat iram, nisi si avertat vultum · 0881B Non in ira declinat · 0882A non optare, non secundum quod · 0882B nec etiam visibilia, ut nihil mihi noceat, nec post, ne laedantur, Ne declinent.

⚑ Every one of these is present in the English, on the word it attaches to in the Latin — the test run was Test 2's, not "are all the negatives present."

⭐ And it was also checked mechanically, band by band, which is the shared brief's test 3 applied to negation rather than to punctuation: 105 negative tokens in the Latin, partitioned by the column anchors (which are verified 1:1), against the English's. No band anywhere in my range has Latin negatives and zero English negatives — the direction that would betray a lost negation. ⚠ Five bands flagged the other way (0870C, 0872B, 0872D, 0876C, 0882D: Latin 0, English >0) and all five are the instrument, not the text — an un\w+ pattern catching unto, unicorn, unheard and unlettered (the last rendering illiterati, whose negative is real and Latin). Checked before reporting, per the shared brief's standing warning that every one of these greps manufactures a number that reads like a violation.

4h. Punctuation ridden as printed

• 0873A the gloss breaks off: Ideo super eum, quoniam dabis . . . . . — five spaced points, reproduced verbatim. Migne's ellipsis, not ours, and not a lost passage.
• 0879C vel ipsa ustio est misericordia, non asperitas — our twin prints no closing period, and the English carries none. ⚑ The plate (p. 445) DOES print one; this is a twelfth CC departure, of the punctuation class, and it is listed here rather than in §1 because it changes no word.
• 0874D [n: (CASS.)]: foderunt. — the colon stands after the note. Carried in place.
• 0873A / 0880A [n: (ID.)]. and [n: (AUG.)], — the note's own trailing mark rides with it.
• 0882A [n: ( Ibid. )] — Migne's internal spacing inside the siglum, carried verbatim.
• Question marks: 3 / 1 / 5 / 1 / 3 across the five chunks, at parity with the Latin, and none supplied. ⚑ The one place I wanted to supply one and did not: VERS. 13.-- Delicta quis. at 0872A is the fragment of Ps 18:13 Delicta quis intelligit?, and convention 3 expands a fragment to the phrase it stands for — but Migne prints a period, so the English reads "Who understandeth sins." with a period. Four lines later the same words return as a full quotation with Migne's own ?, and there the English has one.

4i. (ID.) and (Ibid.) never expanded

38 (ID.) and 13 (Ibid.) in my five chunks, every one passed through verbatim and unexpanded, per launch brief §4. Note that the sigla vary in case and form across my range — (CAS.) at 0871B, (CASS.) everywhere else; (Hier.), (HIER.) and (Hier., AUG.) all within two columns; (AUG., CASS.) and (CASS., AUG.) both occur. All ride exactly as our twin sets them. Migne's attributions in this recension are unreliable and passing one through is not endorsing it.

5. WHAT I READ CLEAN, AND WHERE A BLIND READER SHOULD GO FIRST

Read clean: the whole of Ps 19 (0872B–0872D) and Ps 20 (0872D–0873B) — short, formulaic, and every lemma agreeing with the Clementine; Ps 22's ten-graces sequence (0875D–0876C) apart from the Seb isti crux; Ps 24's Reminiscere/Delicta juventutis stretch (0877D–0878B).

Go here first:

1. §1's table. It is the finding with consequences beyond this range, and eleven of its twelve rows can be confirmed or refuted from one Gallica request each.
2. 0876B Sed isti (§4a) — the one place where our Latin column prints a sentence that cannot be construed and the English is deliberately rough.
3. 0880B corruptionem (§2) — the only [cj:], and the only place a reader could believe the Glossa asserts the opposite of what it means.
4. 0870D–0871A, the non that appears in one limb and not its twin (§4g). If someone later "repairs" it, the repair will read better than the truth, which is the signature of the class.
5. 0870D et ipsi eidem facti nox, tenebrosis scientiam quae de humanis — an abridgment resumption tail that I rendered as closely as the words allow ("and these same are made night to him, giving to the darkened the knowledge which is of things human") and do not claim to have fully construed. scientiam is an accusative with no governing verb in the surviving text. ⚑ Per the shared brief's §5: the splice removes words, it does not suspend grammar — the tail is genuinely defective here, not merely elliptical, and I am naming it rather than smoothing it.
6. 0877C usque ad qui non per omnes, minus perfectos — a second tail of the same kind, rendered flat ("who not through all, the less perfect") because no reading of it is secure.

6. WHAT I DID NOT DO

• No data/tei-patches/8967.json. Twelve entries are owed (§1) and re-chunking mid-batch would rewrite every stint's Latin. Escalated.
• No second-witness check at Gallica for the two low-confidence italic a/u sites (§1) or for the eleven confirmed ones. One request each; it is the cheapest open item in this range.
• No expansion of (ID.)/(Ibid.), no reordering of anything, no Hebrew invented — the one In Hebraeo clause in my range (0881D, Minus est in Hebraeo) asks for nothing.

§E.0020 — stint file cruces-0020.md, verbatim

Band A, stint covering Psalms XXIX–XXXIII (Vulgate numbering). Latin src/latin/8967/0020.md–0024.md.

§0. Span reconciliation — the count agrees

data/briefs/8967-lemmata-0020.txt declares 267 spans. Counting … runs in the body of my own five Latin chunks (frontmatter excluded):

267 = 267. The brief is not mis-split at this range. Head and tail also agree by inspection: the brief's first line is [0883A] VERS. 1.-- Psalmus cantici in dedic., which is the first span of 0020; its last is [0892B]-band, matching the tail of 0024.

Mechanical parity, all five chunks, Latin ↔ English: column anchors 5/10/6/9/8 present, verbatim and in order; [n: …] notes 22/42/22/31/24, contents identical and in order; italic spans 29/74/44/63/57 after stripping [var: …]; question marks 1/1/1/4/3. Punctuation compared band by band (Test 3), ? ; : — zero mismatched bands across all 38 bands.

§1. THE PLATE WAS READ — five pages, cols 0883–0892, all recorded

⛔ Nobody had read any column in my range. Under the corrected §3 of the launch brief (the mod-4 clipping rule is withdrawn — the plate is the arbiter, a rule about the plate is not) I opened all five leaves before firing anything: PDF pages 447–451 of raw/scans/pl113/patrologiaecurs04migngoog.pdf, at 200 dpi full-page with 450–500 dpi crops, corner numbers read first on every page. Map PDF page = (column + 11) / 2 verified — p. 447 carries 883/884 in the corners. All five reads are recorded in data/plate-reads.json.

Clipping, observed rather than predicted:

⭐ Col 889 is 1 (mod 4) and is perfectly clean — a further counterexample to the withdrawn rule, alongside 853, 857 and 901. Had I decided by the rule rather than by the plate I would have withheld nothing here (no marker of mine falls in 889), but I would have withheld the [d:] at 0890A on the belief that its neighbourhood was unreliable, and that marker turns out to be licensed. The failure mode is exactly the one the correction names.

⚑ Zero foot-of-page conjecture apparatus of either layer on all five pages — no numbered Forte notes, no asterisk layer. Recorded as five more spot reads; it does not license the claim that the work has none.

§1a. Markers fired — six

Five [var: …] and one [d: …]. Every [var:] rests on sources/vulgate/clementine-flat.txt, cited book chapter:verse, and every one now also rests on the plate: I confirmed that Migne prints what our TEI says he prints before making a public claim about his text. No [sic:] and no [ed:] anywhere in the range — see §3.

⚑ node scripts/plate-gate.mjs 8967 exits 0: every marker in this range stands on a column that was read.

⚑ On the three tituli (0886D, 0888B/C, 0890C): brief §2.4 names Migne-vs-Clementine titulus divergence as "a real finding, not a checker fault," and the band is already doing this — the 0019 stint fired [var:] on the tituli of Ps 26 and Ps 27 (0880A, 0881C) and the 0026/0027 stint on Ps 35 and Ps 36. Mine are consistent with that practice, and all three are plate-confirmed.

§2. In-force conventions — where each bit in my range

• In finem → "Unto the end." Two occurrences: 0884C (Ps 30 titulus, In finem psalmus. → "Unto the end, a psalm.") and 0888B/C (Ps 32 titulus). ⭐ The range supplies fresh evidence for the ruling: at 0884B, glossing Non compungar, Migne writes Non moriar jam in finem — in finem in the commentator's own prose, plainly temporal, "unto the end." A "for the choirmaster" reading of the titulus would make the same two words mean two unrelated things four lines apart.
• Diapsalma — four occurrences, all in-range: 0887C, 0887C, 0887D (Diapsalma) and 0891D (Diapsalmus, masculine). All four are carried unanglicized. ⚠ Not italicized, because Migne does not italicize them — ✅ checked at the plate on all four, pp. 449 and 451: they are set in roman, and none of the four is inside a … span, so adding italics would also create English spans with no Latin twin and break the 267-span parity. ✅ Diapsalmus at 0891D is confirmed masculine at the plate, so the form beside the work's 37 Diapsalma is Migne's. The launch message's "unanglicized and italic" is followed on the first half and yielded to the plate on the second, per the shared brief's rule that every italic span in the Latin is an italic span in the English in the same place. Flagging for the merge in case the band wants it otherwise.
• Psalmus David / Psalmus ipsi David — one occurrence, 0890C, rendered "A psalm of David" (see §1).
• Sigla — 141 notes across the five chunks, all passed through verbatim. No (ID.) or (Ibid.) expanded; 0883A carries a compound (AUG., CASS.), likewise verbatim.
• Vulgate numbering — untouched. Heads printed PSALM XXIX–XXXIII; every first-lemma probe I made against the Clementine resolved to the same-numbered psalm (Ps 29:1, 30:1, 31:1, 32:1, 33:1). No offset anywhere in this range, consistent with the brief's 138/0/3 measurement.

§3. The three type-defect candidates — ALL THREE SETTLED AT THE PLATE

⭐ Every one of these was declined while the columns were unread, and the plate then settled all three — two against my instinct and one for it. That is the measured lesson of 9004 with a case attached: frequency and internal corroboration prove an error EXISTS, never whose it is.

### 3a. 0884B — in aeter num laudabo te → CC's. No marker owed. Our Latin prints in aeter num laudabo te. The plate reads in æter-num laudabo te, hyphenated across a line break (450 dpi crop, right column, band B). Corpus Corporum dropped the hyphen and left the space. Our corruption, exactly the class the coordinator's de-cidam→deeidam names.

• Rendered "For ever I will praise thee." No [sic:], and none is owed: firing one would have accused Migne of CC's defect.
• ➜ TEI-patch candidate: PL 113 col. 0884B, PDF p. 447 right column. aeter num → aeternum.

### 3b. 0890A — In sua in virtute nullus salvatur → MIGNE'S. [d: in] FIRED. ⛔ This is the one I got wrong while the column was unread. I had reasoned that a doubled short word is the classic CC artifact and declined the marker under Ship Test B. The plate sets In sua in virtute plainly, at 500 dpi, with normal spacing on both sides — Migne's own dittography, and col. 890 is clean, so nothing about the copy explains it away.

• Marker now fired, per Pattern 11a on the intrusive copy: In his own [d: in] strength no one is saved. Both occurrences are rendered, as Pattern 11 requires.
• ⚑ The general point, and it is the reason this crux is worth its length: my confidence pointed the wrong way, and the only thing that corrected it was opening the page. The withdrawn mod-4 rule would have reinforced the error rather than caught it.

### 3c. 0886B — --VERS. 21. → CC's. No marker owed. Our Latin reads [0886B]--VERS. 21. A conturbatione., the only such form in 2,035 verse addresses in this work. The plate reads the ordinary VERS. 21. — A conturbatione (p. 448, right column, band B). CC transposed the dashes to the wrong side of the address across the column break.

• Reproduced verbatim in the English, since the address passes through untranslated. No marker.
• ➜ TEI-patch candidate: PL 113 col. 0886B, PDF p. 448 right column.

§3a. ⭐ FIVE MORE CORPUS CORPORUM DEPARTURES, found by reading the whole line

These are the addendum-C class — every one was found while reading the plate for a different word a few words away, and not one is detectable from our files, because in each case CC's reading is the tidier one. They are TEI-patch candidates per the coordinator's ruling, not markers: a [sic:]/[d:] must wrap text that is verbatim in the Latin twin, and in each case the defective text is precisely what our twin does not contain.

⚑ The consequence for the English, stated plainly: my translation renders the Latin twin as shipped, so it currently reads "moved from this good," "etc.", and "in a ring." If these five patches land, three of my sentences must change with them — the etsi one materially, since Migne's doubled word wants a [d: etsi]. Flagging it here rather than pre-empting a patch that has not been made.

### 3d. Real words that read oddly — logged, no marker, per addendum A ### 3d. Real words that read oddly — logged, no marker, per addendum A Addendum A of the shared brief: a real, well-formed word that reads oddly is Migne's until a plate says otherwise. Four sites, none marked:

1. 0883C — quod morte solvistis (2 pl.). "the first sin, which by death you have loosed." The surrounding discourse is 3 sg. (vindicavit in nos). ✅ Plate-confirmed as Migne's. Rendered as printed; the address to a plural "you" may be Augustine turning to his hearers, or may be for solvisti. No marker.
2. 0883D — Vespere facit. Rendered "He acts at evening." Two words, no object, standing between the ad vesperum / ad matutinum gloss and Adam's speech. Reads like an abridgment stub. ✅ Plate-confirmed as Migne's, so nothing is missing from our side. Rendered literally.
3. 0887C — sicut supra de rectione inveteratio (Ps 31:5). Rendered "as above the growing old from governance." ⚑ rectione makes no sense here; the conjecture is tectione, "from the covering," which restores the chain Migne himself builds two columns earlier — Ps 31:1 tecta sunt peccata and 0887A's Tegit, ne in judicio reveletur. ✅ Read at the plate: Migne prints de rectione inveteratio. So this is a conjecture about Migne's text, not about CC's, and it stays a conjecture: rendered as printed (Pattern 7), no [sic:] (a real Latin word), no [cj:] (the crux is where the conjecture lives).
4. 0891A — et si in primis in jucundum accedite. Rendered "and if among the first, come ye into what is pleasant." The doubled in … in construction here is grammatical (unlike 3b) but the clause is obscure. Rendered as printed.

§4. The negation and scope work — 7a, 7a″, Tests 2 and 2a

### 4a. ⭐ The site that would have been silently repaired: 0885B ira per zelum turbat: quia et haec, concussa mente, contemplationem dissipat, quae non est nisi inquieta mente; sed inde post paululum plus proficit ad videndum.

Read straight, the printed words say contemplation does not exist except with an unquiet mind — the opposite of what the sentence is arguing, since the whole point is that a shaken mind scatters contemplation. The obvious repair (quieta mente, or moving the non) is exactly the silent repair 7a forbids.

• Rendered as printed: "…scatters contemplation, which is not except with an unquiet mind; but thereafter, after a little, it profits the more toward seeing."
• Two parses both construe, and I am naming both rather than choosing silently: (i) est is existential — "which does not exist save with an unquiet mind," which contradicts the argument and points at a plate defect; (ii) est carries an ellipsis of dissipata from the preceding clause — "which is not [scattered] save by an unquiet mind," which construes and agrees with the argument. The English takes neither exclusively: "which is not except with an unquiet mind" leaves the ellipsis open exactly as the Latin does.
• ⛔ Read at the plate and NOT settled — this is the one place in my range the copy fails. Col. 885 is clipped, badly, and the plate's lines run contemplatione dissipat, quæ non est nisi / […]mente ; sed inde post paululum plus proficit. The word our TEI gives as inquieta falls exactly at a clipped line opening, so this copy cannot confirm that Migne prints it, nor rule out that he prints quieta — which is the reading that would dissolve the whole crux. ➜ Needs the second witness: Gallica ark:/12148/bpt6k5505319w, f = (column + 1) / 2 = f443. One IIIF request settles it. Recorded as an open plate read, and no marker fired in the meantime.

### 4b. ⭐ Test 2a fired, host proved from structure: 0886D si nulli hoc faceret, non videretur vigilare divina providentia, si omnibus non servaretur divina patientia.

• Rival reading: "if the divine patience were not kept for all" — i.e. omnibus inside the negated clause. Rejected, because it makes the second limb a non-sequitur and destroys the si nulli … si omnibus parallel that the sentence is built on.
• Taken: si omnibus is elliptical for si omnibus hoc faceret, and the non stays on servaretur — "if he did this to none, the divine providence would not seem to be watchful; if to all, the divine patience would not be preserved." The negative did not move; only the parallel identifies where the scope of the condition ends. This is the Colossians/Esther hole exactly. Migne prints commas throughout and the English keeps them (band parity holds).

### 4c. ⭐ Test 2a examined and NOT repaired: 0890D Mansueti, non de sua intelligentia superbientes; non jugum Dei, indomitorum more, ferentes.

• Taken: the surface scope, non on ferentes — "not bearing the yoke of God after the manner of the untamed." This is what the parallel member forces: the first limb puts non on the participle (non … superbientes), so the second does too.
• Rival: non scoping the manner-phrase indomitorum more — "bearing the yoke of God, but not after the manner of the untamed" — which is theologically far smoother and is why it must be named and refused: reading well is what the wrong host gives you.
• The English wording preserves the same order the Latin has and so preserves the same ambiguity.

### 4d. ⛔ A question mark Migne does NOT print: 0891C Pascebat Dominus contemnentem se, et deseret timentem. Promittit autem non superflua, sed necessaria. The a fortiori sense obviously wants a question — he fed the man who despised him, and will he forsake the man who fears him? — and every English translation of Augustine here supplies one. Pattern 8 forbids it: a ? Migne does not print is never supplied. ✅ Read at the plate (p. 451, left column, band C): Migne prints a full stop. The statement form is his, not our source's, so the decision is now evidence-backed rather than merely rule-backed. Rendered as he prints it: "The Lord was feeding him that despised him, and he will forsake him that fears him." The next sentence (Promittit autem non superflua) contradicts it, which is the tell 7a always gives and is precisely what this edition exists to show. No marker; logged.

### 4e. Negatives checked clean, host named (Test 2 run on each) non tantum … sed etiam 0883A (host tantum) · non terreat 0883A · non sinit 0884A · nemini … nullum 0884A · nunquam sine ea 0884A · non viderent 0884B · Non compungar / Non moriar 0884B · non differas 0884D · nec parcitur 0885B (host parcitur, with the et parcitur morti pivot proving it) · nec … nec … ulla 0885B · ne … videatur 0886A · nullam … non turbantur 0886B · Ne … turbaret 0886B (host turbaret, tantum on promissio) · ne noceantur 0886B · non sum confusus 0886C · non amant … neque 0886C · non diligere 0886C · Non in quibus non invenit 0887A (host of the first non is the whole in quibus phrase, proved by the sed quibus remitit correlative) · non sit 0887A · ne … reveletur / non imputabitur / Non est dolus 0887A · non vult confiteri 0887B · non imputavit / non imputat / non justos se praedicant 0887B (host justos, proved by sed se agnoscentes) · Etsi non sit sibi conscius … non tamen justus 0887B · non appropiant 0887D · ne putetur 0887D · Nolite fieri / non approximant / non recedunt 0888A · nisi … non possumus 0888B · non sufficit, nisi abundet 0888C · Ne quis putet / non nisi 0888D · non cognoscens / quis non dabit 0888D · non indigent 0889A · Non terroribus … non nisi a Deo 0889B · Ne habeant effectum 0889C · non peccata 0889D (host is the object peccata, not vidit — the sed vidit omnes filios correlative proves it) · non ex Adam 0889D · Non salvatur … nec facti … Non modo … sed et … Nemo … nisi … nec … nec 0889D–0890A (host of Non modo is modo, proved by sed et exteriora) · non vult sola facere 0891A (host examined: non on vult with sola predicative; unde vocat alios shows the force lands on sola, and the English "does not wish to do alone" carries both) · Non possunt … non aliud quam 0891A · Non confundantur / non erubescit / non accessit 0891A · nec erubescet 0891B · Nullus … non securitatem praesumat 0891C · non superflua, sed necessaria 0891C · Non est inopia 0891C · fidem non habent / non satiantur / Deum non habent 0891C · ne nomen timoris horreatur 0891D · non capietur 0892A · Nec … diffidas 0892A · ut nec mentio 0892A · Non corporaliter, sed 0892B · ut nec patientia pereat 0892B · Non modo mala … tamen pessima 0892B (host modo, proved by the tamen pivot) · Non omnium … sed eorum 0892B (host omnium, proved by sed eorum) · non reconciliatur 0892C. Every printed negative is in the English, on the word the Latin attaches it to, and none is supplied.

§5. 7a″ — the lemma inventory walked. Divergences found and NOT marked

The declined list is the deliverable. Each of these was collated against sources/vulgate/clementine-flat.txt; each is a real difference from the received text; none rose to Pattern 14's "changes the sense" threshold, and each is recorded so a later reader need not re-derive it.

### Lemmata collated and found to agree with the Clementine exactly So the list above cannot be mistaken for a report by an agent who only checked the flagged ones. Ps 29: 29:3, 29:4 (in lacum), 29:6 (Quoniam ira, ad vesperum, et ad matutinum laetitia, fletus), 29:7 (Ego autem), 29:8 (voluntate tua praestitisti decori meo virtutem, avertisti faciem tuam), 29:9 (Ad te, Domine, clamabo), 29:10 (Quae utilitas in sanguine meo, dum descendo in corruptionem), 29:11, 29:12, 29:13 (non compungar). Ps 30: 30:2, 30:3 (accelera, esto mihi), 30:5 (educes, de laqueo … absconderunt), 30:6 (in manus), 30:8 (exsultabo et laetabor, humilitatem meam, necessitatibus), 30:9 (in loco spatioso, statuisti in loco spatioso), 30:10 (miserere mei Domine, tribulor, conturbatus, in ira oculus), 30:11 (paupertate), 30:12 (super omnes inimicos meos, vicinis, notis), 30:13 (qui videbant me foras, vas perditum), 30:14 (vituperationem multorum, in circuitu, in eo dum convenirent simul), 30:16 (sortes meae, eripe), 30:17 (illustra, salvum me fac in misericordia), 30:18, 30:19 (all three), 30:20, 30:21 (a conturbatione, proteges), 30:22 (benedictus, mirificavit), 30:23 (in excessu mentis, a facie oculorum tuorum, ideo exaudisti…), 30:24 (all four), 30:25. Ps 31: 31:1 (beati quorum remissae sunt, iniquitates, peccatum), 31:2 (beatus vir), 31:3 (tacui, quoniam tacui, dum clamarem tota die), 31:4 (configitur spina, spina, aerumna), 31:5 (delictum meum, dixi), 31:6 (pro hac orabit, tempore, verumtamen in diluvio), 31:7 (tu es refugium, erue), 31:8 (intellectum tibi dabo, ×3), 31:9 (all six), 31:10. Ps 32: 32:1 (exsultate), 32:2 (cithara, in psalterio), 32:3 (cantate ei, canticum novum), 32:4 (quia rectum est, et omnia opera, in fide), 32:5 (terra), 32:6 (verbo, coeli, omnis virtus eorum), 32:7 (sicut in utre, abyssos), 32:8 (ab eo, commoveantur), 32:9, 32:10 (all four), 32:11 (all three), 32:12 (beata gens), 32:13 (de coelo), 32:15 (finxit, sigillatim ×2), 32:16 (non salvatur rex, gigas), 32:17 (fallax equus, equus), 32:18 (ecce oculi, oculi, metuentes eum), 32:19, 32:20, 32:21 (quia in eo, speravimus ×2), 32:22 (fiat misericordia tua, in te). Ps 33: 33:2, 33:3 (in Domino laudabitur anima mea, mansueti), 33:4 (magnificate Dominum, magnificate ×2), 33:5 (exquisivi Dominum, ex omnibus), 33:6 (accedite ×3), 33:7 (pauper, iste), 33:9 (gustate et videte, gustate, suavis est Dominus), 33:10 (omnes sancti ejus, quoniam, non est inopia), 33:11 (divites eguerunt et esurierunt, esurierunt), 33:12 (venite filii audite me, docebo vos), 33:13 (dies videre bonos), 33:14 (prohibe ×2), 33:15 (diverte a malo), 33:16 (oculi Domini ×3, in preces), 33:17 (memoriam), 33:19 (juxta est Dominus, juxta est), 33:20 (multae tribulationes justorum, multae, liberabit), 33:21 (ossa, et unum), 33:22 (pessima, qui oderunt), 33:23 (redimet animas).

⚑ One conflation, logged and not marked: 0892B Sicut et in Novo Testamento, juxta est, et salvabit,. Both words stand in Ps 33:19 but not adjacent (Juxta est Dominus iis qui tribulato sunt corde, et humiles spiritu salvabit). The Glossa is compressing its own psalm's verse while saying "as also in the New Testament," which is itself odd. Rendered as printed; no received text to collate a conflation against.

§6. Vocabulary axes — every call in the range

• Axis 1, the commentator's voice is not archaised. ait 0886D → "says"; inquit 0888B → "says"; dicit throughout → "says"/"means". -eth forms appear only inside italic scripture spans (In the Lord shall my soul be praised, Standeth for ever, That hate, The rich have wanted), never in gloss prose. Checked by the italic-membership test.
• Axis 2, turba/multitudo. ⚑ Instrument checked first, per the brief. Neither word occurs in my 267 spans or in the gloss prose of chunks 0020–0024. Nothing to report — and nothing to mis-report, since a raw \bturba grep here returns turbatus, conturbatus, turbantur, perturbati, turbat and perturbationis, all of them turbare, none of them turba.
• Axis 3, daemonium/diabolus. daemonum 0889C → "of demons". diabolus 0883D, 0884D → "the devil"; a diabolo 0889B, 0892B → "by the devil". The two are never merged. ⚑ Note the brief's warning holds here: \bdaemoni misses daemonum, which is the only occurrence in the range.
• Axis 4, the gentes family — three calls, all stated:
• 0889C Gentium. → "Of the Gentiles." This is the Judaei/gentes typology at its plainest: Migne glosses Gentium. Dum idololatriam destruxit and immediately opposite Populorum. Judaeorum. The gloss itself supplies the opposition.
• 0889C in quibus gentes conversae → "in which the Gentiles are converted." Same paragraph group, same opposition, Christian times against the Jews of v. 10.
• 0889C Beata gens. → "Blessed is the nation." Singular gens, and the axis's nations branch: the lemma is the Vulgate's own generic "blessed is the nation whose God is the Lord." ⚑ The two calls sit five lines apart and are deliberately different — gens the blessed people in the lemma, gentes the converted Gentiles in the gloss. Naming it here so the merge does not read it as drift.
• 0888A gentilis indomitus → "the untamed Gentile." The 2026-09-04 addition, applied; and the context is the sharpest possible warrant, since the very next clause reads mulus, Judaeus piger ad credendum.
• 0884C persecutione omnium gentium → "the persecution of all the nations." Plainly generic (omnium gentium, the axis's own cited pattern), the whole world persecuting the Church, no Jew/Gentile opposition in view. Stated because it is the one gentes in the range rendered "nations".
• ⭐ A word the axis does not rule on: gens tenebrarum (0887D, the Manichaean "race of darkness"). Rendered "the nation of darkness" — i.e. held to the axis's nation rather than invented as a sixth word. Flagging rather than deciding silently, which is the behaviour the axis section rewards. If the series wants "race" here, this is the only site in my range.
• natio does not occur. ⚑ Instrument checked: a raw natio grep in my chunks returns dedicatione, incarnatione, tribulatione, desperatione, damnatio-shaped forms — zero real natio.
• Axis 5, mysterium/sacramentum. mysterium does not occur. sacramento 0891B → "sacrament"; sacramentis 0891D → "sacraments". 1:1.
• Axis 6, usque ad → "as far as". ⚑ Decided per occurrence (addendum D). 44 occurrences in the range. 43 are the abridgment formula and take "as far as". One is not: 0883D, inside the italic span Tristis est anima mea usque ad mortem — Mt 26:38's own words, rendered "even unto death." Taking the formula reading there would have swallowed a lemma's words, which is the trap the addendum names.
• Additional in-range chains held 1:1, stated because the gloss argues from them: confiteor → confess everywhere (0883B Confitemini "Confess ye", 0884A confiteamur peccata / confiteamur laudem, Confessio gemina est, 0884B confitebor tibi). The gloss at 0884A builds its whole point on the word being one word — a "give thanks" for the lemma and a "confess" for the exposition would break it. exsulto → exult, laetor → be glad, gaudeo → rejoice, forced by 0885A's own definition (Exsultare est … gaudere: laetari, temperata affectione). fingo → fashion against formo → form, forced by 0889D (Finxit. Manu gratiae formavit).

§7. Anchors, subjects, and the sections read clean

Anchor positions (2b — presence and order are verified mechanically; position is not): every one of the 38 anchors was checked against the word it stands beside in the Latin, not merely that it is present. Four sit mid-construction and were deliberately split to match Migne:

• 0883C at quod morte [0883C] solvistis → "which by death you have [0883C] loosed."
• 0884B at tempore [0884B] tristitiae → "in a time of [0884B] sadness".
• 0886C at vocem orationis [0886C] meae → "the voice of my [0886C] prayer." ⚑ The one case where the split point could not be kept identical: Latin splits orationis | meae, English inverts to my | prayer. The anchor stays inside the same two-word unit, which is the closest available position; 8995's precedent at 0605B is the model.
• 0888D at id est, [0888D] remunerat → "that is, [0888D] rewards".

Supplied subjects and elisions (the dominant defect class — every one named):

• 0883C Demorabitur → "It shall abide", subject = fletus, the immediately preceding lemma Fletus., not the laetitia of the following clause.
• 0883D Deus a Christo faciem → "God turned away his face from Christ." The verb is elided and lifted from the lemma directly above (Avertisti). English cannot leave a clause verbless; the verb is Migne's own, one span away, not a conjecture.
• 0889B sed quod a Deo utiliter, ideo ab eo → "but what is moved by God is moved profitably" — commovetur supplied twice from the clause immediately preceding, where Migne prints it.
• 0885C Foras, Christum → "Without, Christ" — the verb (videbant) is in the lemma above; the ellipsis is carried into the English rather than filled out.
• 0886C si omnibus → "if to all" — see §4b for the proof.
• 0885B quot et quanta ea perturbant → "how many and how great are the things that disturb": ea is nominative plural neuter and the verb has no expressed object. Rendered without supplying one.
• 0885B Statuisti in loco spatioso, expanded to "Thou hast set my feet in a spacious place" (convention 3; the lemma carries etc. and Ps 30:9 supplies pedes meos). This is expansion of a fragment, not conformation — the printed words are not altered, and Migne's text agrees with the Clementine here.
• 0887A Beati quorum remissae sunt, expanded to "Blessed are they whose iniquities are forgiven" — same class, and the gloss's very next lemma is Iniquitates., which names the supplied word.
• ⚑ By contrast 0885B In ira oculus. was NOT expanded → "In anger the eye." Migne is glossing the two words as a pair and immediately treats in ira on its own (In ira persecutorum), so an expansion to "my eye is troubled with anger" would import a verb the gloss does not use. The two treatments are different classes (fragment-with-etc. vs. a two-word gloss target), named here so a later reader does not manufacture an inconsistency between them.

Read clean, no findings: 0884A (Ps 29:10–11, the twofold-confession passage), 0884D–0885A (Ps 30:3–8), 0888C (Ps 32:2–3, harp and psaltery), 0889A (Ps 32:6, the Apostles as heavens), 0891D–0892A (Ps 33:12–16). These were read line by line against the Latin under Test 5 — does the English assert what he asserts — and produced nothing.

Em-dashes: two in the range, both carrying real parenthesis (0890C Cum immutavit vultum — Christ, that is, the sacrifice — coram; 0890B Because in him — lest anyone murmur about patience — he adds the gift). Not a defect class; not marked; noted only so the count is on record.

§E.0025 — stint file cruces-0025.md, verbatim

Band A, stint covering Psalms XXXIV–XXXVIII. Latin src/latin/8967/0025.md–0029.md.

0. SPAN COUNT RECONCILIATION — the brief is CORRECT

data/briefs/8967-lemmata-0025.txt declares 243 spans. Counted mechanically over the five chunks' Latin bodies (\(?!\)(.+?)\*, DOTALL, frontmatter excluded):

243 = 243, and the brief's own line count (^\[0) is also 243. No mis-split. The head of the range (VERS. 1.-- Judica, Domine.) and the tail (Remitte mihi.) are both present, so the file is not shifted at either end. English span parity is 1:1 in every chunk after markers are stripped.

1. PLATE READS — SIX PAGES, ALL SIX COLUMNS-PAIRS OF THIS RANGE

None of cols 0892–0901 had ever been read. All six pages were read for this stint from the on-disk PL 113 scan (raw/scans/pl113/patrologiaecurs04migngoog.pdf, PDF page = (column + 11) / 2), corner numbers checked first, full page width, with 500 dpi crops on every candidate. Recorded in data/plate-reads.json. Zero foot-of-page conjecture apparatus of either layer on any of the six pages (consistent with the work's existing spot coverage; still a sample, not a proof).

⚠⚠ THE MOD-4 CLIPPING RULE HAS A COUNTEREXAMPLE IN THIS RANGE

The launch brief §3 states the rule as regular, 5 of 5: a left-hand column congruent to 1 (mod 4) loses 3–5 characters at the inner margin. In my range 893 and 897 are clipped exactly as predicted (VERS. prints as s.), and 895 and 899 are clean as predicted (3 mod 4). But column 901 is 1 (mod 4) and is CLEAN — page 456's left column opens Verumtamen. (AUG.) Diapsalmus. with the V whole, and every line beneath it is entire.

So the rule is a strong tendency, not an invariant. Render the page before invoking it. Had I applied it blind, I would have refused to settle three candidates in col 901 that the copy in hand answers cleanly (hahet, Iter, Diapsalmus).

2. MARKERS FIRED — six, every one on a column read at the plate

⚑ Class split for this range: 4 [var:] and 2 [sic:], and both [sic:] rest on a plate-visible defect rather than on a corpus argument. Every one of the eight suspicions I raised against a real, well-formed word was refuted or declined; every one I raised against a non-word was settled at the plate — and five of those seven non-words turned out to be Corpus Corporum's, not Migne's (§3). That is the reverse of what a corpus-frequency argument would have predicted, and it is the whole reason for reading.

3. ⛔⛔ SIX SITES WHERE OUR LATIN IS NOT MIGNE'S — no marker owed, TEI patches wanted

This is the highest-value finding of the stint. In every one of these, firing the marker the Latin file invites would have publicly blamed Migne for Corpus Corporum's defect. Recommend data/tei-patches/8967.json entries; I have not written them, because a patch rewrites the chunk under a translation that has already been filed.

3a. ⭐ TWO NEW MEMBERS OF THE NORMALIZATION CLASS (brief §4), and neither is detectable from the Latin

Both were found by reading the whole line for a different word — exactly the mechanism addendum C describes.

• 0899B, VERS. 14. Migne prints Tanquam surdus non audiebant, et sicut mutus non aperiens os suum. Our TEI prints audiebam — which is the Clementine's reading at Ps 37:14. The digitization has silently conformed Migne's lemma to the Vulgate. This is the fourth instance of the class in the series and the first where the tidied form is the received biblical text itself, which is the worst possible case for 7a″: a divergence that a Vulgate collation cannot see, because the collation was run against the text the corruption produced. Our English renders our Latin ("I heard not"); the plate wants "they heard not".
• 0895A, VERS. 28. Migne prints quando Novum Testamentum populus praedicavit — an ungrammatical nominative, since the subject is lingua ejus. Our TEI prints populis, the dative that makes the clause construe. Our English renders our Latin ("to the peoples"). Migne's page says "when the people preached the New Testament".

⚠ Weaker, NOT settled: 0898D enumerantur. The plate at Narratio, in qua … miseriae appears to read emunerantur (m/n transposed), a non-word, where our TEI reads enumerantur. The inking on that line is heavy and I could not separate m from nu with confidence at 500 dpi. Recorded as unsettled, not as a finding. A single Gallica IIIF crop (ark:/12148/bpt6k5505319w, f = (column + 1) / 2, so f449) would close it.

4. DECLINED — every candidate raised and not marked, with the reason

⚑ Listed because on Luke 12 of 17 corrections came from declined lists rather than from fired markers. Nothing below is a marker; several are places a later reader should look.

4a. Declined SCRIPTURE divergences (checked against sources/vulgate/clementine-flat.txt)

1. 0892D Tanquam pulvis ante faciem venti. vs Ps 34:5 tamquam. Orthographic only (tanquam/tamquam); Migne uses the -n- form throughout. Plate-confirmed as Migne's. Pattern 9. No marker.
2. 0893C Pax vestra ad vos revertetur (Migne cites Matth. X) vs Mt 10:13 pax vestra revertetur ad vos. Word order only — the verb agrees exactly. Pattern 14 requires a change of sense; the gloss's point is the returning, which both orders give. Plate read (col 893 is clipped, but the divergent words sit mid-line). No marker.
3. 0894D Devoramus. vs Ps 34:25 Devoravimus eum. Present for perfect, plate-confirmed (p.452). Declined as tense-only: the psalm asserts the same thing either way, and Migne's own gloss is present-tense too (Devorantur incorporati), so the tense is not idle. ⚑ This is the closest call in the range; a later reader who thinks Pattern 14's bar is lower than I have taken it should look here first.
4. 0896B in tuo lumine vs Ps 35:10 in lumine tuo. Word order only, plate-confirmed. No marker.
5. 0896D Miseretur et retribuet vs Ps 36:21 miseretur et tribuet. A real divergence (retribuet, "shall repay", for tribuet, "shall give") and arguably sense-changing — but it stands at a LINE OPENING in column 897, which is 1 (mod 4) and is clipped on this copy (the plate shows Miseretur et ending one line and the clipped …et, opening the next). §3 forbids the marker on this witness. ⛔ Owed: one Gallica IIIF crop at f449. Rendered as our Latin has it; no marker fired.
6. 0898A Psalmus David in rememoratione de sabbato. vs Ps 37:1 in rememorationem. Ablative for accusative, plate-confirmed. Case only; the titulus says the same thing. Declined. (Contrast the two tituli that were marked, at 0895A and 0896B, where a different word is at stake, not a different case of the same word.)
7. 0898C livores mei; vs Ps 37:6. Migne himself flags it — alia littera — so it is his own apparatus, not an unannounced divergence. Pattern 18a's ❌ class ("anything the edition already glosses"). No marker.
8. 0898C Currentes cum sponso post odorem unguentorum. vs Ct 1:3. Declined because it is not offered as a quotation: the gloss says gemamus: and then weaves the verse into its own sentence (the Clementine has Trahe me, post te curremus in odorem unguentorum tuorum; there is no sponso and no currentes in it). A [var:] here would assert a divergence in a citation that was never made.
9. 0899B Quia dixi: Ne quando supergaudeant… vs Ps 37:17 Nequando. Word-division of an orthographic compound; the Clementine's own printing of nequando varies. Pattern 9. ⚑ Deliberately NOT treated as the 0894C class: there the split produces super gaudeant, two independently meaningful words that destroy the verb, and there was a fused control in the same column. Here nothing is destroyed and there is no control.
10. 0899D Ut quid me dereliquisti?, referred by Migne to supra, i.e. Ps 21:2 quare me dereliquisti. Ut quid and quare both mean "why"; the form Migne gives is the Old-Latin/Mt 27:46 one. No change of sense. Declined; recorded because a reader hunting Old-Latin survivals in this book will want it.
11. 0901B Et tabescere fecisti sicut araneam animam. vs Ps 38:12 …animam ejus. Migne drops ejus. A clipped lemma, not a divergence — convention 3 territory. No marker.
12. 0900D Verum tamen. split vs Verumtamen. fused, both plate-confirmed on p.455/456 within a few lines of each other. Declined: verum tamen as two words is perfectly good Latin ("but yet"), so there is no non-word and nothing for [sic:] to wrap. Rendered identically ("But yet") in both places, deliberately.
13. Verse numbering. Migne addresses Ps 36:19 as VERS. 18. (0896D) and Ps 38:11/12/13/14 as VERS. 10./11./12./14. (0901B–D). His numbering, reproduced verbatim, never corrected. The psalm numbers themselves are the Vulgate's throughout and none was touched.

4b. Declined CONJECTURES ([cj:])

1. 0901C Iter mala saeculi petit sibi adesse. Plate-confirmed at 500 dpi: Migne prints Iter, not Inter. Inter mala saeculi ("amid the evils of the world") is the near-certain reading and is what the surrounding gloss requires. The [cj:] is DECLINED and Inter is the conjecture that lost only in the sense that no marker carries it: Pattern 18a's bar is that the printed word must make the sentence assert something the author is arguing against, and a bare nominative Iter with no verb does not assert anything — it fails to construe, which 18a explicitly keeps in the decline class. The English renders the printed words as they stand ("A journey. He asks for himself the evils of the world to be at hand…"), which reads as strangely as the Latin does.
2. 0894A Intende … ut ego intensus immerito. Elliptical: the apodosis is missing. Rendered literally ("that I, undeservedly stretched out") with nothing supplied. The pun is on Intende/intensus and the stretching is the cross. Not a defect; Migne's abridgment. No marker.
3. 0901D Hic sibi petit dimitti, ut certus dijudicium perveniat. perveniat wants ad; rendered "may come to the judgment" without supplying a preposition in the Latin's voice. Real word, plate-confirmed. Crux only.

4c. Not defects at all — recorded so a blind reader does not re-raise them

1. VERS 5.-- without the period, twice (0025 at Ps 34:5, 0029 at Ps 38:5). The plate prints VERS. 5. correctly in both places (pp. 451, 455). CC's dropped period. Reproduced as our Latin twin has it, per convention 1's verbatim rule; no [sic:], since the defect is not Migne's.
2. [0901A]-- Verumtamen. — a bare -- where a VERS. n.-- address would stand. Plate-confirmed: Migne runs the paragraph on without a verse number. Reproduced.
3. Siglum spelling varies: (CASS.), (CAS.), (Cass.), (ID.), (Ibid.), (HIER.). Passed through verbatim and unanglicized, and no (ID.) or (Ibid.) was expanded (launch brief §4). 148 notes across the five chunks (0025: 42 · 0026: 14 · 0027: 28 · 0028: 30 · 0029: 34), order and count 1:1.
4. (ID). at 0895C is bare text, not a [n: …] note — CC failed to tag that one attribution. Left exactly as it stands in the Latin twin; not our error to fix in English.
5. Diapsalmus. at 0901A and 0901C. Migne sets it roman, not italic, and in the masculine Diapsalmus rather than Diapsalma. Kept unanglicized and roman, mirroring the plate — launch brief §2.2 rules the word untranslatable ("a word the text explains cannot be replaced by its explanation"); it says italic as Migne sets it, and here he does not.
6. In finem → "Unto the end" at 0895A and 0900A, per launch brief §2.1. No exception was needed or taken.
7. usque ad → "as far as", 29 occurrences of the abridgment splice formula (0025:1 · 0026:5 · 0027:8 · 0028:4 · 0029:11), all rendered as far as. ⚑ Checked per occurrence against addendum D: none of the 29 is a lemma's own words in this range — every one stands between an etc. and a resumption tail, and every tail was parsed as ordinary prose. The nearest look-alike, Usque ad nubes. at 0895D, is a lemma (Ps 35:6) and is rendered "Even to the clouds", not with the formula.
8. Non supergaudeant. (0894A) and Non super gaudeant. (0894C) both render "Let them not rejoice over me". Identical English by design; the [sic:] is what distinguishes them, and the launch brief's rule on bare lemmata (expand the stump, never contract its twin) forbids differentiating them by mutilating one.

5. THE FOUR TESTS, run on the finished files

• Anchor position (2b). All 37 column anchors present, verbatim, in order — and each was checked against the word it stands beside in the Latin, not merely for presence. Three sit mid-construction and were split identically in English: Loosed [0893A] by God, pressed him [0898D] that he should not now be raised up, who was crucified [0900A] with him. The two anchors that fall inside an italic span in the Latin (Amat judicium et non derelinquet [0897C] sanctos suos.) keep the anchor inside the span in English.
• Punctuation by band (test 3). Colons, semicolons, question marks and guillemets compared band by band across all 42 bands of the five chunks, with our own markers stripped. Zero mismatched bands. Five sites were repaired to reach that: an invented colon at 0893D (Alii interpretes, takes Migne's comma), an invented semicolon at 0895A, an invented colon at 0896D, an invented semicolon at 0900A, and an invented colon at 0900C. All five were mine, all were smoothing, and the totals would have hidden two of them.
• Question marks. Latin 1, English 1 in 0025 (o Domine, quis? etc.); Latin 1, English 1 in 0028 (Ut quid me dereliquisti?). No mark supplied and none deleted. ⚑ Two places invite a mark Migne does not print and did not get one: Domine, quando respicies. (0894A) and Quando videbimus vindictam de eis qui nobis insultant. (0894A) both end in a full stop on the plate, and both end in a full stop in English, however interrogative they read.
• Negation attachment (tests 2 / 2a). Every printed negative was named to its host word. The four that fire 2a's trigger condition (a negative on a verb with a quantifier or correlative nearby) and how each was resolved from structure, not from what reads well:
• 0894A Dissoluti … nec tamen prae duritia compuncti. — the nec governs compuncti, not dissoluti; the tamen is the pivot. Kept as two italic spans in order, with the negative on the participle.
• 0894A nec tamen compuncti sunt, id est, poenituerunt. — the appositive poenituerunt glosses the verb, and therefore stands under the negative. It is rendered as such, and it was tempting to render "that is, they did not repent", which would have moved Migne's negation from his word onto mine.
• 0895C Non quod non potuit; perniciosius agit contra seipsum qui quod potest non vult. Three negatives, three hosts: quod, potuit, vult. All three carried. ⚠ The sed-less correlative here is potest … non vult, and the negative belongs on vult, not on potest — "does not will what he can", never "cannot will".
• 0895C Sunt enim quaedam quae, etsi volunt, non intelligere homines possunt … et ut tamen bene agerent, intelligere possunt. The non governs possunt and the etsi volunt is concessive; the second clause is affirmative and had to stay so.
• 0894C Nemo enim malus, qui non prius sibi noceat. — nemo + non, both carried, producing the double negative Migne prints.
• Test 5 (does it ASSERT what he asserts). Read back paragraph by paragraph with the counts already clean. It moved three renderings:
• 0892D ne habeant quid oblatrent had been smoothed to "that they may have nothing to bark at", which quietly moves the negative onto quid. Restored to "that they may not have what to bark at".
• 0893B nec defecit, qui omnia fecit had "nor did he fail"; the perfect is a standing state here (nihil suavius … ipsum ergo pete), so "nor has he failed".
• 0901C Iter mala saeculi petit sibi adesse had been smoothed into "he asks that the evils of the world be present to him", which reads fluently and is false — it makes the psalmist petition for the world's evils. Rebuilt as the printed words stand (§4b.14).

6. UNEXPRESSED SUBJECTS — every supplied pronoun, and what fixed it

The dominant defect class. Sites where two readings both construe:

• 0894C Tuam, Domine Deus. Quae in se est, vel quam ei dedit. Quae is justitia. Both the subject of dedit and the referent of ei are unexpressed and neither is recoverable from the sentence. Rendered "Which is in himself, or which he gave to him" — deliberately as underdetermined in English as in Latin. The likeliest reading is God's justice, given to Christ, but Ps 34:24 (Judica me secundum justitiam tuam) does not settle it and I have not settled it silently.
• 0893A quos divina gratia non humectat — quos are the dissoluti, from the lemma above (Fiant tanquam pulvis), not from the sentence. "they whom".
• 0894D ut his illa curent — his = pudor and reverentia, illa = audacia and impudentia, both from the two preceding clauses. Rendered "that by these they may cure those", keeping the demonstratives rather than resolving them.
• 0894D Illi dicunt, euge animae suae quae laetitia fugit — quae may be nominative with laetitia ("from which gladness flees") or agree with laetitia. Rendered "from which gladness flees"; the contrast with isti … suam laetitiam ponunt … quae est perpetua is what decided it, since the point is that their gladness does not last. Flagged rather than settled.
• 0897C Vel, non permittit damnari justum cum judicabitur ille ab impio; vel, illi impio, scilicet ad malum ejus. ille is the just man in the first limb and the impious man in the second; ejus in the second is the impious man's. Rendered so.
• 0900A Vox est veteris hominis qui cum eo crucifixus, hic autem in persona propria loquitur. hic is Christ, not the old man — the sentence turns on the contrast. Rendered "but here he speaks in his own person".

7. VOCABULARY AXES — instrument checked before reporting

Greps run with the brief's warnings applied (the natio/damnatio- trap, \bdaemoni missing oblique cases, the raw usque ad overcount):

• turba / multitudo: zero occurrences of either noun. ⚑ The instrument fires anyway: \bturb returns turbata at 0899B (ratio nuper turbata), which is the verb, exactly the trap the brief names. Nothing to report.
• daemonium / diabolus / satanas: two sites, 0898D sed et diabolus, qui corpus affligit and 0901D Non in diaboli civitate → devil in both. No daemoni- and no satan- of any case in the range.
• natio / gentes / gens / gentilis / ethnicus: one site, 0897B quia per omnes gentes dilatata, of the Church spread through the whole world — generic, not the Judaei/gentes opposition, so "nations" per axis 4's second limb. (The raw grep also returns damnari, dilatata noise; checked by eye.)
• mysterium / sacramentum: zero of either.
• Commentator's voice not archaised: dicit/ait → "says"; the -eth forms in this stint appear only inside italic lemmata (He loveth judgment, He sheweth mercy, He hath said, Thou that preachest), which is the italic-membership test axis 1 names.
• gradus: zero (no gradual psalms in this range).

8. SECTIONS READ CLEAN

• 0025 (Ps 34) — read clean apart from the sites named above; the VERS. 12 retribuebant antithesis (ego … ipsi …, four limbs) checked limb by limb for a reversed comparative.
• 0026 (Ps 35) — clean; the coelum/nubes/montes chain at 0895D was checked for a dropped non at quae post, de eis et si non ab eis and it is present.
• 0027 (Ps 36) — clean; 8 of the chunk's 44 spans are abridgment splices and every tail was parsed as ordinary prose per addendum D's warning.
• 0028 (Ps 37) — the densest negation block in the range (Non est sanitas / Non est pax / non est absconditus / non aperiens, ten printed negatives across 0898C–0899C). All ten named to a host and all ten present. Clean.
• 0029 (Ps 38) — clean apart from 0901C.

§E.0030 — stint file cruces-0030.md, verbatim

Psalms XXXIX–XLIV. Band A, stint covering chunks 0030–0034.

0. Span reconciliation — MY LATIN AGREES WITH THE BRIEF

data/briefs/8967-lemmata-0030.txt heads its inventory 267 spans. Counted directly off my own Latin (body only, frontmatter excluded): 0030 = 47 · 0031 = 41 · 0032 = 63 · 0033 = 53 · 0034 = 63 → 267. Exact agreement, chunk by chunk and in total. No missing head-of-range spans, no borrowed tail. The English carries 267 spans in the same order.

⚑ The brief's column labels were stale when I began (a line carrying anchors but no spans never advanced the band, so spans inherited the chunk's opening column) — e.g. VERS. 7 and VERS. 8 of Psalm XXXIX were filed under 0902C when they stand under 0902D. The coordinator corrected the file mid-stint. Membership was never affected, and I located every span from my own Latin, not from the labels.

Other parity, all 1:1 against the Latin twin: column anchors (7 · 5 · 9 · 9 · 9, verbatim and in place beside the same word), [n:] notes (35 · 28 · 39 · 34 · 34, byte-identical and in order), question marks (0 · 1 · 7 · 1 · 0). Zero em-dashes supplied.

1. PLATE READS — FOUR NEW PAGES, cols 0903A–0910D

Nobody had read 0903–0910. I read all four leaves before firing anything. Recorded in data/plate-reads.json; PDF page = (column + 11) / 2, corners checked first on each.

⭐ A third counterexample to the withdrawn mod-4 rule. Col 909 is congruent to 1 (mod 4) and its line-openings are whole (Obdormis. Negligere videris.). Col 905, also 1 (mod 4), is clipped. The congruence predicted the risk on one leaf and mispredicted it on the other, in the same range. The coordinator's ruling — predict with it, never decide with it — is confirmed once more from a leaf it would have wrongly written off.

Page 456 (cols 901/902) was already recorded as read; I re-opened it anyway at 500 dpi, because the recorded read carried no findings against my own chunk's words and my chunk opens under 0901D.

2. MARKERS FIRED — five [var:], zero [sic:]

Every one stands mid-line in a column I opened at the plate. Each is a divergence between what Migne PRINTS (plate and our TEI agreeing) and the Clementine, cited from sources/vulgate/clementine-flat.txt, never from recollection.

1. 0904D · In fine intellectus. (Ps XLI titulus) — Clementine Ps 41:1 In finem. Intellectus filiis Core. Migne has the ablative fine for the accusative finem, and drops filiis Core. Material: §2 of the launch brief settles In finem as "Unto the end" precisely because the Glossa argues from finis as a terminus reached; the ablative says "at the end," not "unto." Plate p. 457, col 904D, mid-line, clean column.
2. 0907D · Si vos persecuti fuerint in una civitate, fugite in aliam. — Clementine Mt 10:23 Cum autem persequentur vos in civitate ista, fugite in aliam. Wholly different protasis (conditional perfect subjunctive against a temporal future, una civitate against civitate ista). Plate p. 459, col 907, clean.
3. 0908B · In similitudinem gentium. [n: (Ibid.)] — Clementine Ps 43:15 in similitudinem gentibus, dative. ⭐ The decisive evidence is internal: the gloss's own paraphrase two words later reads gentibus sunt appositi — it expounds the DATIVE while the lemma above it prints the genitive. That is 7a″'s pattern with the halves reversed, and it makes the divergence material. Plate p. 459, col 908, clean.
4. 0910B · Pro his qui commutabuntur ex filiis Core. — Clementine Ps 44:1 pro iis qui commutabuntur. Filiis Core. Migne's added ex turns the sons of Core from the addressees of the psalm into the source out of which the change is made, and the gloss beside it is arguing exactly about who is changed (multi de suis inimicis supponent se sibi). his/iis alone would not have earned a marker; the preposition does. Plate p. 460, col 910B, clean.
5. 0911B · Virgines et proximae. — Clementine Ps 44:15 virgines post eam ; proximae ejus, with a semicolon and no connective. The gloss turns on the connective: post enumeratio partium est distincta, cum dicitur: Virgines et proximae — the et is what makes the enumeration "distinct," so this is the gloss arguing from the word. Plate p. 461 (read by another stint, col 911 clean), confirmed against my own re-read of the page.

No [sic:] is owed anywhere in this range. Every non-word and every unconstruable form in my Latin turned out at the plate to be Corpus Corporum's, not Migne's — see §3. Conversely, every defect Migne actually prints in these ten columns has already been tidied away by CC before it reached us, so there is nothing left in our Latin to carry.

3. ⛔ CORPUS CORPORUM NORMALIZATION — 28 SITES IN TEN COLUMNS, AND THE CLASS RUNS BOTH WAYS

This is the largest finding of the stint and it is not a translation matter. Reading the plate word for word against our TEI over cols 0902–0910 turned up 28 places where our Latin is not what Migne prints. None is a marker; all are TEI-patch candidates in the sense of translation-style.md § "Source patches" and of the sordescat→sordescet patch already in data/tei-patches/8967.json. I have NOT patched them: eight stints are running on this work concurrently and chunk-work.mjs rewrites every chunk, so the patch must be applied once, after the band merges, and the chunks re-cut then.

⚠ Our English mirrors the Latin twin in every one of these, not the plate — except the two marked ⭐ below, where our Latin holds no word at all to render. When the patch lands, these sentences must be re-read against it.

3a. CC changed a word Migne prints (15)

⛔ Two of these are undetectable by any check this edition runs, and they are the reason this section is long. Dominum→Dominus (0903D) and the VERS. 17 dicunt→dicant (0903C) both move Migne's text onto the Clementine's reading — so a Vulgate collation sees agreement where the plate has divergence, exactly the judicabit→judicavit mechanism recorded in CLAUDE.md 8a. A [var:] on either was made impossible, not merely wrong, before I ever saw the file.

⭐ And the class runs in both directions on one leaf-pair: at 0902C CC manufactured a concord Migne does not print (miro→mira), and at 0908D it repaired a real corruption (Nocere→ Noscere). Neither direction is detectable from our Latin, because both leave a well-formed word.

3b. CC changed the words Migne sets, not just their spelling (2)

• 0904A — Migne: haec bona temporalia : quasi, facit haec doloribus plena, ne eis spes habeatur. Our TEI: haec bona temporalia: sed Deus facit haec doloribus plena, ne in eis spes habeatur. CC has replaced Migne's second quasi, with sed Deus and inserted in. This is much larger than the ordinary letter-level class and I flag it as the one entry here that deserves a second eye before patching: it is the only site in the ten columns where our Latin carries words the plate does not carry at all. The English follows our twin ("but God makes them full of sorrows"); on the plate the clause is verbless and elliptical.
• 0902D — Migne: Multiplicati sunt numerum. Our TEI: Multiplicati sunt super numerum. CC has added super, which is the Clementine's reading at Ps 39:6. Another member of the invisible class: the addition creates agreement with the Vulgate where the plate diverges.

3c. CC changed Migne's punctuation (9)

0903B bona dicit malo, animo (comma dropped) · 0903C professio in aeternum. gaudentes (full stop dropped) · 0903D the head PSALMUS XL. with a final period, dropped, so our head reads PSALMUS XL and my English head reads PSALM XL to match · 0904C malus ingressus. malus egressus. (full stop → comma) · 0906A nemo. confitetur (Migne's own mid-sentence full stop, dropped) · 0907B Deus auribus. (comma inserted, giving Deus, auribus.) · 0910A vel de populo ad populum. (terminal stop dropped, so my English paragraph also ends without one) · 0911B idolorum cultoribus. (full stop → comma) · 0904D Hieronymum.. (double stop reduced to one).

⚑ The PSALMUS XL. case is worth naming separately: a head is not a spelling. Our Latin prints ## PSALMUS XL and Migne prints PSALMUS XL., and the heads array in the frontmatter carries the period-less form. I have mirrored our twin rather than the plate, because heads are checked against it, but the plate reading is the one to restore.

3d. ⭐ Two sites where our Latin holds nothing to render — the English follows the PLATE

These are the only two places where I have not mirrored the twin, and I say so here because a silent departure is exactly what this file exists to prevent.

• 0902B — In Hebraeol. Our TEI prints a non-word. The plate prints In Hebraeo, clean, mid-line, in a clean column (p. 456). Pattern 7's non-word rule would carry it in italics under a [sic:], but a [sic:] accuses Migne's type and I have seen that his type is sound: firing one would be a false public claim of exactly the kind CLAUDE.md 8a exists to stop. The English reads "In the Hebrew." ⚑ This is one of the three unread In Hebraeo clauses named in launch brief §3; it is now read, and no Hebrew type stands at this site on the plate — Migne prints the Latin phrase only, so there is nothing for CC to have dropped and nothing to recover.
• 0902D — ἀποπομπάιohgr;ι. Our TEI carries an unconverted SGML entity: &ohgr; is Greek small omega, so CC's rendering of the single character ῳ (omega with iota subscript) has broken into ohgr;ι. The plate prints ἀποπομπάιῳ, read at 500 dpi on p. 456. The word is untranslated Greek in both columns either way, so nothing is gained by printing markup debris at the reader; the English carries the plate's ἀποπομπάιῳ. It is the dative of ἀποπομπαῖος, the LXX's word for the scapegoat at Lev 16, which is precisely what Migne's own gloss says next (id est, emissario).

4. RENDERINGS AND WHY — the calls a reader will want argued

4a. The settled axes, and the instrument check before reporting anything

gentes family: 9 occurrences, all rendered Gentiles. 0902A gentes idola colentes · 0908A traditi gentibus in servitutem · 0908B gentium / gentibus sunt appositi · 0908D fides tua gentibus · 0908D sed gentibus, quae putant mortuum · 0910D gentes designantur, Tyri, gentium, gentes scilicet. Every one stands in a plain Judaei/gentes opposition — three of them name the Jews or Judaea in the same sentence — so none is the generic "nations" case. Singular gens at 0906D (Gens non sancta) → nation, per the axis. pagani at 0904D → pagans, a fourth word.

⚑ Instrument checked before reporting, as the shared brief requires. A raw \bturb grep returns 3 hits in my range and not one is turba: they are turbatur (×2) and turbari, the verb turbare — the exact false positive the brief names. Zero turba, zero multitudo, zero natio (and zero of its -natio look-alikes), zero daemoni-, zero mysterium/sacramentum. One diabolo (0908C) → devil.

usque ad: 23 occurrences, and I decided each one from what stands on either side. All 23 follow etc., and open an italic span — every one is the abridgment formula, rendered "as far as". There is no Sap-8:1-style look-alike in this range (addendum D checked, not assumed). Two carry Migne's comma inside the span (usque ad, at 0905A and 0905C) and are rendered as far as,, comma inside the italics.

In finem → "Unto the end" (0901D, 0903C, 0906C, 0909B), never "for the choirmaster". In fine at 0904D → "At the end", because Migne prints the ablative; see marker 1. Diapsalmus at 0907D → "Diapsalmus", roman and unanglicized. ⚑ Note the FORM: Migne prints the Latinised nominative Diapsalmus, not Diapsalma, and prints it in roman, not italic. The launch brief's rule ("italic as Migne sets it") therefore yields roman here, and italicising it would have added a 268th span. The stint at 0911A–0912D independently confirmed the same form one column later. Psalmus David at 0906C → "a psalm of David"; no Psalmus ipsi David falls in this range.

4b. Unexpressed subjects — every supplied pronoun, named

• 0905B dilatavi super seipsam — "I have spread it out above itself." It = animam meam, from the lemma Effudi above (Ps 41:5 effudi in me animam meam), not from the sentence. seipsam is feminine and fixes it.
• 0906A Qua ad me turbatur anima — "Whereby the soul is troubled toward me." Rendered literally on Qua. The obvious conjecture is Quia ("because"), and the plate prints Qua (p. 458, col 906A, clean), so this is Migne's and not ours. I did not take the conjecture: it would be a silent repair, and the sentence construes after a fashion as printed.
• 0908C quod a circumstantibus irrogat — active singular verb with no expressed subject. Rendered "which one inflicts from among those standing round." The subject is the vicinus of the previous clause. Flagged rather than smoothed.
• 0910A Et quia deducet te inoffense — "And because it shall lead thee on without stumbling." It = dextera tua, from Ps 44:5 (deducet te mirabiliter dextera tua), which is the verse the whole Regna gloss is expounding. Nothing in the sentence names it.
• 0908B quando eos comparibus poenis affecerunt — "when they afflicted them." They = the gentes of the lemma; eos = the martyrs.
• 0910B supponent se sibi — "shall put themselves under him," sibi reflexive to the psalm's te, the bridegroom, not to the enemies.

4c. Negation — each printed negative, and the word it sits on

Test 2 run word by word; the trigger population of 2a (a negative on a verb with a quantifier or correlative nearby) fires at three sites and all three are resolved from the sentence's own structure, not from what reads well:

• 0909C Non sono, qui transit, comparat Verbum: sed scripto, quod manet. The non sits on sono, not on comparat — proved by the sed scripto correlative, which supplies the alternative term of the comparison, not an alternative verb. English: "He compares the Word not to sound, which passes away, but to writing." Had it floated onto the verb ("he does not compare") the sed clause would have nothing to attach to.
• 0909C quia non aliud post aliud, sed omnia simul in uno Verbo aeterno — non on aliud post aliud, again fixed by sed omnia simul.
• 0910B Nemo sibi multum de misericordia praesumat, blandiatur — Migne prints no negative on blandiatur; Nemo governs both jussives asyndetically. English: "Let no one presume much to himself concerning mercy, or flatter himself." The "or" carries Nemo's scope across; it is not a supplied negation, and writing "let him not flatter himself" would have added a non Migne does not print.

Not in the trigger population, resolved as printed: 0909D quae non sub cruciatu humano cogitat quod ... divulgat — I kept the printed order ("which does not under human torment devise what it divulges") rather than resolving the scope, because there is no correlative to settle whether the non governs cogitat or the prepositional phrase. The English preserves the ambiguity that the Latin has; naming it here is the honest treatment. Carried without smoothing elsewhere: non hominem (0901D) · non me (0901D) · non turpe, non theatrale (0902B) · qui non temporalia sperat (0902B) · non est quod putatur (0902C) · Nulli auctores (0902C) · noluisti / noluit / non audit (0902C–D) · non petisti (0902D) · non parva / non prohibebit / non sim piger (0903A) · Non abscondi (0903A) · non diffidens (0903B) · non sufficit / Ne tardaveris / Nisi breviati (0903C) · non petentibus (0903C) · Ne putes / ne ... superetur / ne ... spes habeatur / ne ... deserti (0904A) · Non solum aperti inimici (0904B) · Non audent (0904B) · nonne habet (0904C) · nondum ad damnationem (0904C) · Non amaritudo (0905B) · non extra (0905B) · Deum non videam (0905B) · nihil est in mundo / si non modo (0905C) · non obsistam (0905D) · nisi vitia carnis essent (0905D) · Nemo currit ... nisi ... nisi / nemo confitetur nisi (0906A) · non in superbis (0906A) · Non petit peccata sua (0906C) · Gens non sancta (0906D) · ne ab eis corrumpatur (0906D) · non curat (0906C) · et non patitur (0907A) · non amentur (0907B) · ne putetur ambigua (0907B) · Non juvenes (0907B) · auctor non mutatur (0907C) · Non enim in arcu meo sperabo / gladius meus non salvabit me / ne nos traherent (0907C) · non tamen obliti Dei (0907D) · nullus acquirit / vel nullus pro eis (0908A) · quia non transitorie (0908A) · non erant in nobis (0908A) · Non recessit (0908A) · non permisisti (0908C) · non obliviscuntur (0908C) · Non pro criminibus (0908D) · non dormire / non sibi / non persequerentur (0908D) · Hactenus non obliti (0908D) · etsi nesciamus (0909A) · non das spiritum (0909A) · non habet ultra (0909B) · nihil sit in opere quod non prius fuit (0909C) · non aliud (0909C) · Non deauratis (0911B) · Vel non in circumcisione (0911A). Every one is on the page, and none was added.

4d. Lemmata I checked against the Clementine and found in AGREEMENT — the negative list

A findings list alone cannot be told from a report by an agent that never looked, so: these are the multi-word lemmata I collated verbatim and found identical to the Clementine, spelling variation apart. Exspectans exspectavi (Ps 39:2) · Mirabilia tua (39:6) · Multiplicati sunt super numerum (39:6, but see §3b — our text agrees only because CC made it agree) · Sacrificium et oblationem noluisti (39:7) · Aures autem perfecisti mihi (39:7) · In capite libri scriptum est de me (39:8) · Domine, tu scisti (39:10) · Non abscondi (39:11) · Misericordia tua (39:12) · Qui volunt mihi mala (39:15) · Ne tardaveris (39:18) · Intellectus filiis Core (41:1) · Qui intelligit (40:2) · Dominus conservet (40:3 — agrees only because CC made it agree, §3a) · Et vivificet (40:3) · Dominus opem (40:4) · Inimici mei dixerunt (40:6) · Et si ingrediebatur (40:7) · Egrediebatur foras (40:7) · In idipsum (40:8) · Et retribuam eis (40:11) · Quemadmodum desiderat (41:2) · Quando veniam (41:3) · Fuerunt mihi lacrymae meae panes (41:4) · Haec recordatus sum (41:5) · In locum tabernaculi (41:5) · In voce exsultationis (41:5) · Quare tristis / Quare conturbas (41:6) · Ad meipsum (41:7) · Propterea memor (41:7) · De terra Jordanis (41:7) · Abyssus abyssum (41:8) · Apud me oratio (41:9) · Quare oblitus (41:10) · Dum confringuntur ossa (41:11) · Judica me / Et discerne (42:1) · Quare tristis incedo (42:2) · Emitte lucem tuam (42:3) · Et introibo (42:4) · Confitebor tibi in cithara (42:4) · Salutare vultus mei (42:5) · Ad intellectum (43:1) · Deus auribus (43:2) · Sed dextera tua (43:4) · Tu es ipse rex meus et Deus meus (43:5) · Non enim in arcu meo sperabo / gladius meus non salvabit me (43:7) · Nunc autem repulisti (43:10) · Vendidisti populum tuum sine pretio (43:13) · Opprobrium vicinis (43:14) · Venerunt super nos (43:18) · et declinasti / a via tua (43:19) · humiliasti nos (43:20) · Si obliti sumus nomen Dei (43:21) · Propter te mortificamur (43:22) · Et tribulationis (43:24) · In terra venter (43:25) · Speciosus forma (44:3) · Populi sub te cadent (44:6) · dilexisti justitiam / Et odisti (44:8) · Myrrha et gutta (44:9) · Filiae regum (44:10) · Astitit regina (44:10) · Audi, filia (44:11) · Et concupiscet rex (44:12) · Omnis gloria / Ab intus (44:14) · Adducentur regi (44:15) · Pro patribus tuis nati sunt tibi filii (44:17). Outside the Psalter: nomen quod est super omne nomen (Phlp 2:9) · Semel locutus est Deus (Ps 61:12) · Quod factum est in ipso vita erat (Jo 1:3–4) · eructat verbum (Ps 18:3) — all four verbatim.

4e. Divergences from the Clementine that I DECLINED to mark, with the reason

⛔ Twelve of Luke's seventeen corrections came off a declined list, so this section is not filler.

1. 0902C Non est qui similis tibi. — Clementine Ps 39:6 non est qui similis sit tibi. Declined: an ellipsis in a clipped lemma, not a change of reading. The Glossa cuts lemmata mid-phrase constantly (Lacu., Carmen., Nomen.), and nothing in the gloss under it turns on the copula. Marking it would put a [var:] on the abridgment's habit.
2. 0908A Derisum his qui in circuitu. — Clementine Ps 43:14 derisum his qui sunt in circuitu nostro. Declined for the same reason: the lemma simply stops. The gloss expounds derisus and contemptus, neither of which is at issue.
3. 0909A Quare oblivisceris? — Clementine Ps 43:24 runs Quare faciem tuam avertis? oblivisceris inopiae nostrae et tribulationis nostrae? Migne splices two clauses into one. Declined: a conflation of adjacent words of the same verse is the abridgment cutting, and the gloss immediately treats Et tribulationis. as its own lemma, exactly as the Clementine divides. Nothing is asserted about another text.
4. 0908D Exsurge, quia propter te mortificamur. — a conflation of Ps 43:23 and 43:22, both verbatim. Declined: the gloss says outright that it is joining them (et ut perseveremus:).
5. 0908A Dedisti nos tanquam oves escarum. — Clementine tamquam. Declined: tanquam is an attested orthography of the same word, Pattern 9's exempt class. Same for lacrymae/ lacrimae at 0905B, coelum/caelum, charitas/caritas.
6. 0903C Qui dicunt mihi: Euge, euge. — Clementine Ps 39:16 closes Euge, euge ! with an exclamation. Declined: Pattern 8 governs the plate's own marks and Migne prints a full stop; a difference in a received text's punctuation is not a divergence of reading.
7. 0910B Pro his ... — the his/iis half. Fired the marker on ex, declined it on his. his for iis is a routine orthography of the period; ex is a preposition that changes the relation. The marker names only what it can stand behind.
8. 0906C Modicum quidem contristavi illum, attributed to Isaias. Not in the Clementine verbatim under that name (nearest is Is 57:17, which shares no wording). Declined: this is the Glossa quoting a father's paraphrase, and [var:] compares Migne's scripture with the received text of the verse he cites — here I cannot identify the verse he cites, and Pattern 14 forbids speculating. Crux, not marker.
9. 0904D vere, vere, as Aquila's rendering. Not in the Clementine, and should not be: it is a report of a translator's version, not a citation.
10. 0902D Ecce venio (Apc 3:11 etc.). Verbatim in several places; the psalm's own Ps 39:8 also reads ecce venio. No divergence.
11. 0905A Filii Core, glossed filii calvi. An etymology, not a citation.
12. 0907D Salvasti,, 0910A Regna., and the other 96 single-word spans. Checked in place against the verse each stands in; all are the verse's own word.

4f. Passages read clean, and hard calls where two readings both construe

• 0902D vel, possunt esse duae dictiones pro peccato. Two readings: (a) "pro peccato may be two words" — the phrase just discussed is a pair, not a compound; (b) "there may be two words for sin." I took (a), because the sentence's whole context is the lemma Pro peccato. and its relation to oblationem two clauses above, and because Migne prints pro peccato unitalicized here where he italicised it as a lemma four words earlier — which is why the English could not set it in italics without adding a 268th span. (b) loses because nothing in the passage is counting synonyms. Recording the rival because the apparatus never shows the reader what was weighed.
• 0903B Et qui bona dicit malo animo, et qui mala dolose laudando, uterque malus est. The second member has a gerund where a finite verb is wanted (laudando for laudat). The plate prints it so (p. 457, col 903B, clean). Rendered with the anacoluthon visible — "and he who evil things by praising them deceitfully" — rather than repaired to "and he who praises evil things deceitfully," which would smooth away a fact about this printing. No [sic:]: laudando is a real word in its own grammatical slot, which Pattern 12 explicitly exempts.
• 0907D Enumerat mala quae hic fecerunt boni, non tamen obliti Dei. fecerunt — "which the good did here." The sense the passage wants is plainly "suffered" (the whole psalm is the martyrs' complaint), and pertulerunt or passi sunt is the obvious conjecture. The plate prints fecerunt (p. 459, col 907D, clean column, mid-line), so it is Migne's, and Pattern 7 forbids reading the conjecture into the English. Rendered literally, conjecture recorded here.
• 0905C Qui, si investigas, nihil est in mundo quare hoc debeas. Qui stands where Quia is wanted. Rendered literally ("Who, if thou searchest, there is nothing..."). Real word, occupying a slot, so no marker; the conjecture is Quia.
• 0903B Complaceat, Trinitatem personarum; tibi, singulariter unitatem essentiae signat. ⚑ The argument does not survive translation and the reader should know it. Cassiodorus reads the Trinity out of the com- prefix of complaceat and the unity out of the singular tibi. English "May it be pleasing" carries no prefix, so the first half of the argument is invisible in our column and legible only in the Latin one opposite. I have not glossed it in the text — that would be an editorial act — but it is the clearest case in this range of an argument that lives in the Latin word.
• 0902B Haec falsa, sed opera tua mirabilia, coelum, terra, elidit superstitiones. Singular elidit against a plural subject. Rendered with the singular kept ("dashes superstitions to pieces"), because English has an exponent for verb number and normalising it would hide a fact about the plate. No marker: this is not carried type, it is our English rendering his reading.
• 0904A the Universum enim stratum, pair. Two italic spans with enim between them. English word order wants "for" first, which would leave the two spans adjacent and read as one run. I kept a word between them — "The whole indeed couch," — so the two lemmata stay visibly distinct. lectum → bed and stratum → couch throughout, kept apart because Ps 40:4 sets them in one verse and the gloss glosses each separately.
• 0906A modico. Jordanis. [n: (Ibid.)] — one italic run in our TEI containing two lemmata and a note. The plate sets in modico Jordanis as a single italic phrase with no interior stops (§3a). Our English mirrors the twin's structure, so it reads "in the little one. Jordan.", which is odd; when the patch lands it should become one phrase.
• 0904C the stray full stop after [n: (AUG.)]. Reproduced verbatim. Migne's compositor, not ours.
• Migne's verse numbering skips, all reproduced as printed and none reordered: Psalm XXXIX runs 14 → 16 (no VERS. 15) and prints VERS. 12. without the -- every other address carries; Psalm XL skips 5 and 12; Psalm XLIII skips 3, 6, 9, 16 and 25; Psalm XLIV skips 16.
• 0903C ## PSALMUS XL — the only head in the range without a final period. Mirrored as ## PSALM XL; the plate has the period (§3c).

5. UNREADABLE ON THIS COPY — one site, logged for the second witness

0905A excrescentibus pilis et cornibus. Column 905 is clipped at the inner margin, and the line break falls exactly at the clip: the plate shows excrescentibus ending one line and t cornibus opening the next, so whether Migne sets pilis cannot be settled here. Our TEI's reading is probably right — the same sentence ends cornua et pilos deponit — but "probably right" is not a plate read, and no marker rests on it either way. Second witness: Gallica ark:/12148/bpt6k5505319w, f = (column + 1) / 2 → f453, one IIIF request.

6. NOT DEFECTS — raised, checked, and closed

• amicabus at 0911B (laudatur sponsa ab amicabus). Looks like a slip for amicis. The plate prints amicabus (p. 461, col 911, clean) — the archaic feminine dative-ablative plural on the filiabus/deabus pattern, and exactly right here, since the friends of the bride are women. Migne's, correct Latin, no marker, no conjecture.
• Diapsalmus for Diapsalma at 0907D — plate-confirmed, and confirmed again by the neighbouring stint at 0912A. Migne's form, not a corruption.
• non declinasti (alia littera) at 0908C — Migne's own variant note, printed inline in his voice. Reproduced as printed; no marker of ours is stacked on it, per the shared brief's rule about Migne's own apparatus.
• In labiis, Dico opera, Lingua, calamus at 0909C — flagged by the pre-scan as not in the Clementine. All three are fragments of Ps 44:2–3 cut at a word boundary (diffusa est gratia in labiis tuis; dico ego opera mea regi; lingua mea calamus scribae). No divergence.
• (ID.) and (Ibid.) — 39 of the 175 sigla in this range. Passed through verbatim and not expanded to the names they point at, per launch brief §4.

§E.0035 — stint file cruces-0035.md, verbatim

Band A, final stint, Psalms XLV–XLIX. Latin src/latin/8967/0035.md–0038.md.

0. SPAN COUNT RECONCILIATION — the brief is CORRECT

data/briefs/8967-lemmata-0035.txt declares 222 spans. Counted mechanically over the four chunks' Latin bodies (frontmatter excluded, \[^]\):

222 = 222, and the brief's own inventory-line count (^\[0) is 222 as well. No mis-split. The head of the range (VERS. 1.-- In finem. at 0911C) and the tail (Salutare Dei. at 0918B) are both present, so the file is shifted at neither end.

English span parity is 1:1 in every chunk with markers stripped, as are notes (31 / 19 / 26 / 39), column anchors (identical lists, in order and in place), question marks (0 / 1 / 1 / 2), colons and semicolons band for band.

1. PLATE READS — FOUR PAGES, ALL EIGHT COLUMNS OF THIS RANGE, ALL PREVIOUSLY UNREAD

Cols 0911–0918 had never been read. All four pages (PDF 461–464 by the work's map, PDF page = (column + 11) / 2, on-disk raw/scans/pl113/patrologiaecurs04migngoog.pdf) were opened for this stint, corner numbers read before the text (911/912, 913/914, 915/916, 917/918), full page width at 300 dpi with 600 dpi crops on every candidate. Recorded in data/plate-reads.json. Zero foot-of-page conjecture apparatus of either layer on any of the four pages — consistent with the work's existing spot coverage, and still a sample.

⚑ All four left-hand columns (911, 913, 915, 917) are CLEAN. None is 1 (mod 4), so this range adds nothing to that tally either way; it is recorded only so a later stint need not re-open the leaves. Every marker below rests on a word read mid-line in a clean column.

2. MARKERS FIRED — three, each licensed by a column read at the plate

2.1 [sic: sucipientis] — 0914B, chunk 0036

Migne prints Id est, populi non sucipientis misericordiam. Read at 600 dpi, mid-line, in a clean column: the plate itself sets sucipientis, and Corpus Corporum transcribed it faithfully. A non-word (for suscipientis), and brief §A's rule — a suspicion aimed at a non-word has been right every time — is here settled by the plate rather than by frequency. The marker does not swallow the word: the English reads "of a people not receiving [sic: sucipientis] mercy", and with the bracket deleted the clause still has its participle, its negation (non) and its object (misericordiam).

2.2 [var: Vulg. Ps 48:2 Audite haec, omnes gentes] — 0915A, chunk 0037

Migne heads Psalm XLVIII's second verse Odite haec, omnes gentes. — "Hate these things" — where the Clementine has Audite. Plate-confirmed at 300 dpi in a clean column. This is the sharpest thing in the range, because the gloss under it argues from the Vulgate reading and not from Migne's: Cassiodorus answers Nullum excipio ("I except none") and the next verse's gloss Nulli sermonem subtraho ("I withhold my speech from none") — both are glosses on an invitation to hear, and neither has any purchase on a command to hate. Pattern 7 therefore governs the English: it reads "Hate these things, all ye nations", and the [var:] carries the divergence. ⛔ Conforming the lemma to Audite would have been the exact 7a″ silent repair, and it would have concealed the one place in this range where Migne's plate and his own commentary come apart.

omnes gentes here → nations, not Gentiles: the gloss reads the address as universal (Nullum excipio, Nulli sermonem subtraho, audiat haeres terrae … justus), with no Judaei/gentes opposition anywhere in the psalm.

2.3 [sic: Tempesta] — 0916D, chunk 0038

Migne prints Tempesta valida. where the Clementine (Ps 49:3) has tempestas valida. Plate-confirmed at 600 dpi. Why [sic:] and not [var:], which was the rival and lost: a [var:] asserts that Migne's text diverges from a witness, which implies a reading some tradition holds. No witness reads tempesta; the nominative does not exist for this noun, and tempesta valida will not construe. What the plate shows is a dropped final letter — a defect in the type — so [sic:] is the honest instrument and [var:] would have dressed a broken sort as a textual variant. The English still says "A mighty tempest": the marker records the type, not the sense.

3. CANDIDATES DECLINED — with reasons, and this list is the deliverable

Plate-confirmed as Migne's, and deliberately left unmarked:

1. 0916D in circumitu ejus (Clementine Ps 49:3 in circuitu ejus). Plate-confirmed, so it is Migne's and not the digitization's. Declined because circumitus and circuitus are the same lexeme and the sense is untouched — Pattern 14 marks a divergence when it is material, and an orthographic doublet is not. Recorded here so the negative is on file.
2. 0913B Deus sedet super sedem sanctam. (Clementine Ps 46:9 sedem sanctam suam). Plate-confirmed. Declined as a truncated citation, not a variant: the gloss cuts the quotation precisely at the words it is about to take up as its own next lemma (Sedem sanctam. follows immediately). Marking the omission would convict Migne of a reading where he was only quoting as far as he needed.
3. 0917A Deus est judex justus. Neither Ps 49:6 (quoniam Deus judex est) nor Ps 7:12 (Deus judex justus, fortis) matches it verbatim, and it stands inside an usque ad resumption tail, where the recension itself re-sets the words. Declined because I cannot say which verse the span is a divergence FROM, and a [var:] that names the wrong verse is worse than none. The crux is the safe instrument here and this is it.
4. 0916D usque ad: exibit de Sion lex, et Verbum Domini de Jerusalem. (Isa 2:3 Clementine: de Sion exibit lex, et verbum Domini de Jerusalem). Word order only, inside an abridgment tail. Immaterial; declined.
5. 0914D Dilectio fortis est ut mors. (Cant 8:6 Clementine: fortis est ut mors dilectio). Same words, different order, in the gloss's own voice under whence Solomon. Immaterial; declined. ⚑ Noted because a reversal of this shape was a real defect at 7914 @1079B — but there the reversal was ours, against a plate that printed the order straight. Here the plate prints the reordering, so there is nothing of ours to undo.
6. 0913D ab aquilone pandetur omne malum (Jer 1:14: Ab aquilone pandetur malum super omnes habitatores terrae). An adapted citation inside an usque ad tail — omne has been pulled forward out of omnes habitatores. Declined as adaptation, not divergence.
7. 0913D In Bethlehem Judae secundum Scripturas. (Mt 2:5: In Bethlehem Judae: sic enim scriptum est per prophetam). The gloss's own summary of the answer given to Herod, not a quotation offered as one. Declined.
8. 0913C Psalmus vel laus cantici filiis Core secunda, (Clementine Ps 47:1: Psalmus cantici, filiis Core, secunda sabbati). The vel laus is the glossator offering an alternative rendering of the titulus, marked as such by vel; it is not a claim about the Psalter's text. Declined.
9. 0918C quem novit, qui aliquid alii quam gratiae Dei attribuit. The sense plainly wants a negative — Augustine's point is that the man who ascribes anything to something other than God's grace does not know Christ. Migne prints no negative, confirmed at 600 dpi, and Pattern 7a forbids supplying one. The English carries the printed reading and contradicts itself, which is the evidence. (See §4 for what our TEI does to the same clause.)

Real words that read oddly, rendered as printed and NOT raised as plate candidates (brief §A: a real word reading oddly is Migne's): 0914C qui multis: (the clause has no verb and the colon hangs — rendered literally, "who to many:"); 0917D Nostra tribulatio proprie est timore salutis (ablative where a genitive is expected); 0918A In his omnia tonat vitia; 0916A quasi felicitas nominatur impiorum. All four are on read pages and all four print exactly as our Latin has them.

4. CC-NORMALIZATION CANDIDATES — SEVEN, all TEI-patch material, NO markers owed

Every one was found by reading the whole line for a different word, exactly as brief §C predicts. ⛔ None of these may carry a marker in either direction: firing a [sic:] on one would accuse Migne of his transcriber's defect. They are logged for data/tei-patches, with the leaf named.

⚑ Three further CC artifacts of a lower order, listed so nobody re-derives them: 0917C our TEI sets VERS, 11.-- with a comma where the plate sets VERS. 11.; 0915A our TEI sets [n: ((AUG.)] with a doubled parenthesis where the plate sets a single (AUG.); and at 0918B the plate's own line-break dittography doc-/ctus is silently repaired to doctus in our TEI. The English reproduces our TEI verbatim in the first two cases, because a VERS. address and a note both pass through unaltered.

⚑ And a habit rather than a defect, reported once as a class: CC supplies a full stop before a siglum where the plate prints none (0914D Furem, currebas (Ibid.), 0918B Tui similis (CASS.), 0918B Rapiat (CASS.), 0916C In progenies patrum (Ibid.)), and supplies the -- after VERS. n. where Migne omits the dash (0915D VERS. 12, 0916A VERS. 15). Nothing turns on either; no patch is worth filing for them individually.

5. VOCABULARY CALLS — the axes, and the two the axes did not rule on

gentes, eleven occurrences, split deliberately:

• → Gentiles wherever the Judaei/gentes typology is the gloss's own argument: 0911C (quod caecitas contigit in Israel, quod Apostoli transierunt ad gentes), 0911D, 0911D amaritudine gentium, 0912A Immanitas gentium, 0912A Contradicentes, gentes, 0912C–D Psalm XLVI's Omnes gentes and the two glosses under it (vos omnes gentes ad quas transivit gratia), 0913B (expulsis Judaeis, gentes in filios Abrahae succedunt — the opposition stated in as many words).
• → nations where the sense is plainly generic: 0912A cui gentes et regna subjiciuntur — the phrase is Ps 45:7's own (conturbatae sunt gentes, et inclinata sunt regna), a pair of political words with no typology attached; and 0915A Odite haec, omnes gentes with its gloss (see §2.2).
• gens singular of the Jewish people → nation: 0914C Per Judam, gens earum ostenditur.
• ⚑ I checked the instrument before reporting any of this: natio occurs zero times in these four chunks (every apparent hit is damnationis / incarnationis / significatio), and there is no gentilis or gentilitas in the range.

usque ad, thirty-four occurrences, decided per occurrence per brief §D. ⛔ A raw grep returns 34 and the formula is 32 of them — the trap brief §D names, and it fires twice in this range. The two look-alikes are both A solis ortu usque ad occasum: at 0916D inside the lemma span (Ps 49:1's own words) and at 0917A unitalicised in the gloss that re-quotes it. Both are temporal and both are rendered "from the rising of the sun unto the going down", identically in both places. That is the Lc 16:16 exception from the shared brief, met in the wild. The other 32 are the abridgment formula → "as far as".

Diapsalma / Diapsalmus — both forms occur and both stand. Migne sets Diapsalmus at 0912A, 0912B, 0916A and 0916B, and Diapsalma at 0913A. Neither is anglicized and neither is replaced by its explanation; each is carried in the form the plate prints, roman where Migne sets it roman (he does, in all five places — so no italics were added, which would have broken span parity as well as fidelity).

Dominus virtutum → "the Lord of hosts" (0912B, and twice at 0914A). ⚑ This is a departure from this work's other renderings of virtutum ("of the virtues" — 0866C, 0885B, 0901C) and it is taken because the gloss argues from the word: at 0914A the comment on In civitate Domini virtutum is quia ibi custodes angelos deputat, "because there he appoints angels as guards", and at 0913B the gloss on Sedem sanctam is Throni, Virtutes, et sancti — the angelic orders. Migne's virtutes in this title are the hosts, on his own showing. Saying so here rather than silently, per the brief: this is the axis a Band B or C stint will meet next (Deus virtutum recurs at 0947A, 0975B, 0978D, 0979B, 0981A).

sedes → "seat", never "throne" (0913B, twice). Forced by the same sentence: the gloss explains sedem sanctam as Throni, Virtutes, et sancti, so "throne" in the lemma would collide with the angelic order Thrones three words later and make the gloss circular.

saeculum → "the world" (0911D, 0916B), matching the work's 22 prior uses. diabolus → devil (0916A), daemonibus → demons (0917D) — axis 3 kept apart; there is no daemonium/diabolus collision in the range.

6. FRAGMENTARY LEMMATA — expanded per convention 3, and the ones left as they stand

Expanded, because the governed word is elided and the stump would say nothing:

• 0913A Omnis. [n: (Ibid.)] → "Of all the earth." (Ps 46:8 rex omnis terrae Deus). The gloss's whole argument is the universality — Non ut Jupiter Cretae, Mars Thraciae — so a bare "All." would strand it.
• 0918B Statuam contra tuam. → "I will set against thy face." (Ps 49:21 statuam contra faciem tuam). Faciem is elided between contra and tuam; the gloss's own next words are ante faciem ponam.

Left as they stand, because nothing is elided (the class distinction the shared brief insists on): Mane., Vocem., Noster., Posuit., Prodigia., Excelsus., Terribilis., Dextera., Gloria., Domus., Stultus., Asaph,, Sum. and the rest of the one-word run. Contrasting these with the two above would manufacture an inconsistency that is not there.

⚑ One I could not expand and did not: 0913D Quoniam reges contra. Ps 47:5 reads Quoniam ecce reges terrae congregati sunt, convenerunt in unum, which contains no contra. Read at the plate: Migne sets contra inside the italic run, and it governs nothing. The likeliest reading is that it answers the gloss's own question two words earlier (Tutandam, et quid opus est? — guarded against what? reges contra, the kings against her), but that is a conjecture and supplying "her" would put it in the text. Rendered literally, "For the kings against." and logged.

7. UNEXPRESSED SUBJECTS SUPPLIED — named, with what licenses each

• 0911D Nos enim fugiebamus eas — "we were fleeing them": eas is feminine and its antecedent is not in the sentence. It is tribulationibus from Ps 45:2 (adjutor in tribulationibus quae invenerunt nos nimis), which is also what supplies the subject of invenerunt in the lemma above. Both stand or fall together, and the verse settles both.
• 0913A in qua speciosus erit fulgens ut sol, qui nunc est abjectus — the subject of erit is the Christian people of the clause before, and qui nunc est abjectus attaches to it, not to a new person. Rendered "in which he shall be comely, shining as the sun, who now is abject", which keeps the Latin's own ambiguity rather than resolving it; flagged rather than decided.
• 0912D ut nondum credentes sub pedibus habeat — the subject is Ecclesia, named two words earlier (Sic est exaltata Ecclesia).
• 0918A cui charitate jungimur — the antecedent is Ecclesiae, which is the gloss's expansion of matris in the lemma Et adversus filium matris; the Church, not the mother's son.
• 0916C ut malus a malo quiescat — malus is the man, malo the evil; rendered "that the evil man may rest from evil" and not "from the evil man".

8. NEGATION — every printed negative named to its host

Sixty-three negatives across the four chunks (non, nec, neque, nisi, ne, nullus, nemo, nihil, nunquam), all carried. The ones where the host is not the adjacent verb, which is the 2a trigger population — a negative resting near a quantifier, correlative or predicate:

• 0912B non enim illius tantum est Deus — the negative belongs on tantum, not on est. The preceding clause is Qui tanta nobis dedit quanta dedit Jacob: the point is not that God fails to be Jacob's God, but that he is not Jacob's alone. English: "for he is not the God of him alone."
• 0912B Non nisi filiorum crucis — both particles carried ("Not save of the sons of the cross"); rendering it "Only of the sons of the cross" would have dropped two negatives and passed every count.
• 0915C pro qua tantum timebitur with quasi non est quod timeam before it — tantum is the restrictive, and the English puts it there: "for that alone shall there be fear."
• 0918A non casu alii derogat, sed studio — a non … sed correlative, so the negative is on casu and not on derogat: "he detracts from another not by chance, but of set purpose." Putting it on the verb would have said he does not detract, which the sentence is denying.
• 0918A non cum voluntate facit — the negative is on the phrase cum voluntate, not on the verb; the sentence concedes that he does the thing, and denies that he wills it.
• 0917D non ideo quod Deus inde delectaretur, sed ne daemonibus offerentur — a non … sed with a second negative (ne) inside the sed clause; both stand, and the ne clause is a purpose clause, not a second denial of the first.
• 0913B Nemo sapienter facit quod non intelligit — two negatives, two hosts, both kept.
• 0914C quia nullus est, et si nesciat cultum, qui huic nomini non se subjiciat — three negative words in one sentence, each on its own host; the nullus … qui non is a double negative that English must keep as a double negative, not flatten to "everyone submits".

9. THE FOURTH TEST — read back for what the English ASSERTS

Run last, with every count already clean. Two paragraphs read smoothly where the Latin is rough and were re-checked for that reason:

• 0914C populi non sucipientis misericordiam, qui multis: ut nulla videantur esse grana, sed tamen nonnulla — the smooth reading would supply a verb for qui multis and make the grains a simile of the temple. Nothing was supplied; the colon and the dangling multis stand.
• 0918C Christus, qui cum gratia venit, quem novit, qui aliquid alii quam gratiae Dei attribuit — see §3.9. The English asserts what Migne asserts, which is the opposite of what Augustine meant, and the crux is where that belongs.

10. GATE STATE — read this before shipping the work

node scripts/plate-gate.mjs 8967 passes all three of my markers (0914B, 0915A, 0916D) and fails on one marker outside my range: 0904D 0032.md [var: Vulg. Ps 41:1 …], chunk 0032, which belongs to the stint on 0030–0034. Col 0904 is unread. Either that page is read at PDF 458 and recorded, or the marker is withdrawn — it cannot ship as it stands.

Eight Opus stints, 38 chunks, 39,328 Latin words. verify-english clean, plate-gate OK, patch-audit OK. This section is the reconciliation, not the concatenation; the eight stint files stand verbatim in §G.

F1. ⭐ EIGHT FOR EIGHT ON THE SPAN COUNT

384 · 294 · 373 · 367 · 352 · 272 · 302 · 258 — every stint's own count of the italic spans in its Latin matched its lemma brief exactly. That count is the only detector a mis-split brief has (verify-english cannot see one), so the Band B split is sound everywhere it was tested. Two stints also checked that the brief's first and last entries matched their range's first and last spans.

F2. ⛔⛔ THE HEADLINE, AND IT CORRECTS WHAT THE BAND LOOKED LIKE BEFORE THE PATCH PASS

Before the patch pass Band B stood at 106 [var:] against 4 [sic:], and that ratio was read — by me, in this session — as evidence that the defects in these columns are overwhelmingly Corpus Corporum's rather than Migne's. That reading was wrong, and the error is instructive.

The [sic:] count was suppressed by patch ordering, not by Migne's cleanliness. A [sic:] may wrap only type our twin carries; CC had already repaired dozens of Migne's real defects, so the markers were unfireable while the twin disagreed with the plate. One stint reported zero [sic:] on a range containing twelve Migne defects, and another found that its twin carried none of the six standing on its leaves.

After the patch pass restored them, Band B stands at:

⭐ The lesson is general and belongs to the series: on a work whose twin has been silently repaired, NO MARKER COUNT MEANS ANYTHING UNTIL THE PATCH PASS HAS RUN. Quoting one before is quoting the digitization's opinion of Migne, not Migne.

F3. THE PLATE — 90 columns, none of which had been opened when the band began

Every stint read its own leaves; plate-gate refuses a marker on a column nobody opened, and it passes work-wide. What the leaf did that no other instrument could:

• It withdrew markers. Six candidates in Band A's own reckoning, and in Band B at least sixteen more, every one of which read as a confident Migne defect from the Latin alone. One stint had six [sic:] candidates and the plate took five.
• It reversed the direction three times in one range — two candidates withdrawn as ours, intrepretatur confirmed as Migne's when it read like the classic transcriber artifact.
• ⭐ It stopped a false [var:] against Migne. solennitatis @0960A looked like a divergence from the Clementine; the plate shows Migne prints solemnitatis, agreeing with it. CC had de-diphthonged him and manufactured the appearance of a divergence.

### ⚠ A SECOND IMAGING HAZARD, MEASURED — italic e images as c Found by the 0044–0048 stint (Ostcndisti, omncs, Vallcs), briefed into wave 2, and it then withdrew eleven false accusations against Migne across the band (omncm, opcrantis, Zcb, Eccc, undc, verbornm, agom, nocent among them). Without a same-letter control on the same line, this one hazard would have manufactured a defect class that does not exist. It belongs beside the h/b warning — which also fired in reverse this band, our TEI carrying biatu where the plate reads hiatu.

⛔ The turned-sort pair Viuo and Clamabuut is deliberately NOT in that class: an n cannot become a clean bottom-arched u by failing to ink. The distinction is what makes the hazard rule usable rather than a licence to decline everything.

### The mod-4 clipping rule stays withdrawn, and is now refuted 22 more times Band A drew "column ≡ 1 (mod 4) is clipped" from four consecutive leaves. Band B opened ~90 columns and found exactly one clipped column in the whole band, 0981 — and the marker resting in its margin was WITHHELD and logged for Gallica, not fired. That is the rule's real failure mode (silent withholding) being made visible instead of being allowed to happen.

F4. OUR LATIN IS NOT MIGNE'S — 149 patches over 73 columns

~1.4 departures per collated column, agreeing with Band A and with 9003's independent collation of the sibling PL 114 Psalter. The class runs three ways, all three found here:

1. CC moves Migne ONTO the Clementine — invisible to any Vulgate collation by construction. ⭐⭐ conculcabit→conculcabis @1000B is the one to remember: the gloss's very next words are Serpentem calcat Ecclesia, the third person the whole exposition is built on. Audivit→Audivi @1007B erased a first/third-person divergence before any instrument looked.
2. CC REPAIRS Migne's defects, destroying markers we are entitled to fire — including a compositor eye-skip @0991C, where CC reconstructed the sense Migne lost, so nothing in our twin could ever have shown the plate was broken.
3. CC MANUFACTURES defects and readings. ⭐ @0942A it INSERTS Sina into a lemma whose whole point is that Sina is absent — Migne's next gloss reads Quidam libri non habent mons, sed a facie Dei Sina Israel — making Migne argue against a text he did not print. And @0999D/1000A it manufactured a dittography at the column turn where Migne breaks Ideo-|que; no [d:] is owed, because the doubling is ours.

F5. WHAT THE PATCH PASS ITSELF FOUND — scripts/patch-audit.mjs

chunk-work.mjs refuses a find that matches zero or many times. That is uniqueness, and uniqueness is not location. The new audit resolves each patch's EDIT offset to its enclosing <pb/> and compares with the declared column. On its first run it found:

• 24 of Band A's 43 applied patches had NO column recorded at all, so the check had never been possible on more than half the file.
• One Band A patch was filed at the wrong column (Gen. III, filed 0855C, lands 0853D) — the 9003 failure recurring. The repair was aimed right; the RECORD was wrong, which would have sent anyone re-reading that leaf to the wrong page.
• Four Band B sites were one BAND letter off, and a patch pair at 0928B would have deadlocked the queue — one edit fell inside the other's find string, so the second would have matched zero times and aborted chunking outright.

⭐ The ordering rule was then vindicated by measurement: re-chunking with all 192 patches left Band A's 39 Latin chunks and Band C's 29 byte-identical, all 38 of Band B's changed, and no chunk boundary moved — so Band A's filed English was untouched and Band C's lemma briefs did not need regenerating.

F6. CONVENTIONS — inherited, confirmed, and the exceptions taken

• §A2 Diapsalma ROMAN — confirmed by twelve more plate reads across the band.
• §A4 Psalmus David ≠ Psalmus ipsi David — ⭐ now evidenced from Migne's own plate rather than merely obeyed: at 1010A he prints Psalmus ipsi David exactly, four columns from the Psalmus David that took a [var:]. He sets both forms.
• In finem → "Unto the end." held throughout. ⚑ One recorded non-engagement: @0976D Migne does not print In finem at that titulus at all (Intellectus pro his qui commutabuntur), so the ruling is not engaged there.
• usque ad — three stints independently took exceptions, and Wilson's ruling of 2026-09-04 (§A7) settles the class: the etc. splice test replaces italic membership. Re-audited across the band afterwards: 274 formula, 18 non-formula, and every English rendering is consistent with the ruling. The two as far as outside a splice are genuine sense-renderings (spread as far as the ends of the earth).
• gens tenebrarum does not occur anywhere in Band B; ruled by Wilson on Band A's single site as "people of darkness" (§A5).
• Convention 3 exception, named: Deus quis similis? @0982C — expanding the stump would silently decide the est/erit question the gloss exists to argue.

F7. ⛔ STILL OWED

1. The blind polarity read has NOT run on Band B. Band A's found six of our own defects, four invisible to every other check, and one stood in a column the cruces had already logged. A logged column is not a checked column.
2. Five reported sites remain unpatched, each because the report itself declined or could not rule: Nafrubimus @0962B (the letter would not settle as f or t at 1200 dpi, and a patch must not guess a letter any more than a marker may), Malivolenti : @0929A, 0926B's self-contradictory report, res alienas/res aliena @0971D, and (Rom. IX)→(Rom. IV) @0971A — where CC CORRECTED Migne, so restoring it needs a citation-corrections.json entry landed with it (⛔ column written WITHOUT its leading zero, or it matches nothing and still reports 0 unparsed).
3. Five Gallica sites, one IIIF request each: f443, f449, f453 (Band A), 0932B (Misericordia/Misericordiam, in col 0932's over-inked outer margin) and 0981C (filiis/filii Excelsi omnes, in the band's one clipped column).
4. Residual prose punctuation listed by three stints as "recorded, not argued" is not applied; the sense-bearing and lemma-internal marks are.
5. Band C — chunks 0077–0105, 29 chunks, 28,516 words. Its briefs are current and unaffected by the patch pass. It carries the gradual psalms, and it inherits §A6 (the plate where read), §A7 (the etc. splice test) and the two imaging hazards.

⛔ Nothing ships until all three bands are merged. Nothing here is pushed or deployed.

Four Opus readers, fenced from cruces.md, every stint file and every brief, split 0039–0047 / 0048–0057 / 0058–0066 / 0067–0076. Every one named its clean chunks and its rejected candidates. 28 sites, 24 ours. Recorded via polarity-record.mjs.

⛔⛔ IT AUDITED THE PATCH PASS THAT HAD JUST RUN, AND THE PATCH PASS FAILED

Seven of the twenty-eight are defects I introduced or left behind earlier the same day, every one found by a reader who could not see the patch report, the pending file, or the cruces:

1. A [sic:] fired on the WRONG WORD. [sic: Eccesiam] @0920C sat after "kingdom", telling the reader Migne's plate is defective at regnum. The defective type is Ecclesiam = Church. A misplaced marker is a false public accusation against Migne at a word he set correctly.
2. liberavit @0927A — I patched the Latin to Migne's doubled perfect and left the English reading "will deliver", so our English manufactured the perfect-for-future gloss the plate does not have, in the one place where Migne's own parallel (dedit, id est, dabit) proves the defect.
3. quid @0929D — patched the Latin, left the English reading qui, demoting an interrogative to a relative modifier with no marker.
4. sint erant @0944A — patched away CC's supplied relative, left the English rendering CC's fluent clause, silently repairing a broken column.
5. : et @0989C — I listed the English edit as owed in the patch-pass record and never made it.
6. biatu @0924B — a patch its stint reported and I never built.
7. The unmatched CC parenthesis @0936D — same.

⭐ The lesson is a mechanism, not an exhortation. The stint reports are PROSE and I extracted 149 patches from them by hand; nothing reconciled reported sites against applied ones, and nothing checked that a fired marker sits on the word it accuses. Two of the seven were pure omission, and I would not have found them by re-reading my own work, because my own work is where the omission is. The blind read is the only instrument in this pipeline that audits the pass that precedes it.

The systematic finding — THREE readers, three ranges, one habit

Fourteen lemmas that Migne prints SHORT and our English silently completed from the Vulgate: In finem pro his → "for those who shall be changed" · Nonne → "Shall not my soul be subject to God" · Nolite sperare in → "in iniquity" · Te decet → "A hymn becometh thee" · Anticipaverunt vigilias → "My eyes prevented the watches" · Cibabis → "Thou wilt feed us" · manus in cophino → "his hands" · In sancto → "In my holy one" · Sicut sol in conspectu → "in my sight" · Opprobri servorum → "of thy servants" · In sinu → "In my bosom" · In pennis → "Under his feathers" · Adorate Dominum in atrio → "in his court".

⚑ How much the convergence is worth, stated honestly: all four readers were told to watch for "a supplied possessive the Latin never printed", so their agreement on the CLASS is one brief speaking three times, not three independent findings ([[feedback_briefed-agents-are-not-independent]]). What none was told is the PATTERN — that these cluster specifically in truncated lemmas and run toward the Vulgate. And one site carries its own proof without any convergence: at 0965B Oculi mei is itself a separate lemma which the glossator then expounds twice, and Migne's third gloss offers inimici mei as the alternative subject — which only makes sense if the head lemma carried none. Supplying it pre-empts the gloss it exists to introduce.

⛔ This is 7a″ conformation inside the apparatus rather than inside a quotation, and it is the failure profile predicted for verse-by-verse commentary. All fourteen are now rendered as printed.

The sense-level defects — the class with no grep signature

Ten, every one of which read perfectly in English:

• Judicandi igitur ante precentur absolvi @0962C — a nominative plural gerundive read as a genitive, so the English lost who is praying.
• quibus imperat et ipse dilectus @0942B — a nominative subject turned into a predicate, so this gloss and Migne's own next gloss three lines below said opposite things about who the Beloved is, and only in English.
• sub spe religionis @0984C rendered "under the show of religion" — specie for spe. It converts an unfulfilled hope into hypocrisy, a different charge, in a passage whose proof-text is about a hope carried through the fall.
• idem corpus sanat, et lavat caput @0984D — idem pulled off the word it stands next to; the clause's whole force is that the body healed piecemeal now is the SAME body healed at the end.
• Superiora Dominus @0927C — nominative read as genitive, turning a statement about who SPEAKS into one about whose things they are, in a sentence entirely about who is speaking.
• jacta in Dominum, non alium qui portus est @0925A — the harbour attached to "another", and the same sentence then ended "not a harbour", contradicting itself two clauses later.
• quod adhuc hodie facit @0949B — an unanchored facit whose nearest English referent was Christ, so the page read as Christ casting the gall.
• cui et dicet @0937A — the supplied "he" resolved to God, so the page had God praying "Hear us, thou who art the Saviour."
• congregata ab aliis rapiuntur @0933C — the agent attached to the wrong verb, so others did the gathering and nobody did the snatching.
• et quod Trinitas @0993A — a supplied "he was" carried the subject over from Verbum and stated that the incarnate Word is the Trinity.

A marker WITHDRAWN, and it was mine

[d: videantur, et] @0991C is withdrawn. [d:] marks type Migne prints twice adjacently; the earlier videantur is eight words back in a different clause and et … et is a correlative pair. Firing it deleted a finite verb and one limb of the correlative, and lost the point — that they still seem to be friends. Now rendered as printed. The reader was right and I was wrong, on a call I had made in the merge two hours earlier.

⚑ Where a reader was RIGHT ABOUT THE FACT AND WRONG ABOUT THE RULE — twice

The standing pattern for a fenced reader, and both instances are instructive:

• It asked for [sic: PSLMUS] on the Psalm C head. Declined: Pattern 9 normalizes a section label in English precisely because the facing Latin preserves the misprint. The observation (the defect is invisible in English) is correct; the remedy is not.
• It filed the @0936D parenthesis as Migne's, deleted by us. It is CC's, inserted — the reader could see our TEI but not the leaf, so it read an addition as a deletion. It is now patched out, which is what its stint had asked for.

Two sites left OPEN by the read

• quidam mundi defecerunt @0995B — mundi is genitive of mundus or nominative plural of mundus "clean". In a passage about prelates slain and the just plundered, "certain cleansed ones failed" fits better and would reverse who is failing. The reader raised it rather than filing it: it needs the plate.
• Non per agentes promissa @0980D — the sense wants "[they lied] by not performing what was promised", but Migne prints the non on the prepositional phrase. Moving it inside would be a Pattern 7 repair. Left as printed, with the strain recorded.

Result: 28 sites, 24 ours, all applied except the two left open above. verify-english, plate-gate and patch-audit clean.

After the blind polarity read, seven open sites were closed by re-opening the leaf on the on-disk scan (raw/scans/pl113/patrologiaecurs04migngoog.pdf), rendered at 150–1200 dpi. Every one had been correctly refused earlier — a patch must not guess a letter, and a report that contradicts itself must not be acted on — and every one fell to a single render.

⚑ The lesson for Band C: a site a stint declines is not necessarily hard. Several of these were simply never re-opened.

Closed at the plate

• ⭐⭐ Nafrubimus @0962C. The 0054–0058 stint declined to transcribe the marker because the letter between Na and rubimus would not resolve as f or t even at 1200 dpi. Re-read at 600 and 1200 dpi: the glyph carries a tall ascender with the hooked top of an italic f, standing to the height of the b of bimus, against the shorter straight-stemmed t of tua on the same line. ⭐ And the leaf supplies the control the stint lacked: Migne sets Narrabimus. Invocabimus nomen tuum. CORRECTLY a few lines above in the B band, so both forms stand in one column. [sic: Nafrubimus] now fired.
• (Rom. iv) @0971A. Migne prints iv; Rom 9:27 is the true locus, so CC had corrected him. Restored, with a citation-corrections.json entry so the index still resolves to Rom 9:27 — the policy of 2026-07-18 exactly. ⚑ The Ibid. hazard was checked, not assumed: the next note in the chunk is an explicit (Prov. II), so nothing inherits a corrected key. ⛔ The column is written 971a, without its leading zero — the normalisation that otherwise matches nothing while still reporting 0 unparsed.
• res aliena @0971D. Confirmed as Migne's. No marker: the concord mismatch has no English exponent (Pattern 12(2)), and with the reading confirmed his, Pattern 9 governs. English unchanged.
• et attribuam @0926B. Its stint's report contradicted itself about which form the plate carried; at 500 dpi Migne prints id est, et attribuam. CC's ei is the reading that construes, which is why it was substituted. [cj: et; read ei] fired.
• Malivolenti : @0929A/B. Filed as "a departure needing a ruling, not a defect", because Malivolenti is a real form. At 700 dpi Migne prints Malivolenti :, ending -i, with his usual space before the colon. The ruling: the parallel member is peccatum, operis, a GENITIVE, so the construction wants malivolentiae and the printed form has no slot. [cj: Malivolenti; read malivolentiae] fired.

⭐ The site whose QUESTION was misdiagnosed — quidam mundi @0995B

The blind reader raised it and said it needs the plate. The plate settles nothing: it prints quidam mundi defecerunt exactly as our TEI has it. Both readings use the same letters, so no transcription check could ever have decided it — it is a parse, not a reading.

What decides it is internal. mundi is either the genitive of mundus (world) or the nominative plural of mundus "clean". Two things favour "clean": (a) the lemma it glosses is ab emendatione, where the Vulgate reads ab emundatione — our own [var:] two words away records it — so the gloss's mundi picks up the cleansing root its lemma has lost; and (b) mundi … in terram redacti sunt is a real antithesis (clean → earth), while "certain of the world failed and were brought down to the earth" is close to tautologous. Rendered "certain clean ones have failed." The alternative is recorded here, not buried.

Confirmed in passing, from leaves opened for other reasons

Super et fundatam @0979A · bene siccata @0980A · the lemma manus in cophino with no ejus · perpercit @0971A · Cynomia · Spiritus vadens et non revertens · De manu tribulationis · and ⭐ Quoniam qui audivit? @0929C with Quis audivit? quoted twice below it — the Class-1 patch verified independently from the leaf rather than from the report.

⛔ What the plate could NOT close

• Col 0981 is genuinely clipped, verified by rendering from x=0: characters are missing from the line starts and the loss is ragged, which no rectangular crop produces. filiis/filii Excelsi omnes @0981C stays owed to Gallica, and the 0064–0068 stint's report is vindicated rather than taken on trust.
• Non per agentes promissa @0980D — the plate confirms the words verbatim, so this too is a parse and not a reading. Migne prints the non on the prepositional phrase; moving it inside would be a Pattern 7 repair. English unchanged, strain recorded, as its reader recommended.

The Gallica list is unchanged at five — f443, f449, f453 (Band A), 0932B and 0981C. None of the seven sites closed above was a Gallica site: all seven were answerable from the copy already on disk, which is precisely why they should not have stayed open.

Nothing below is abridged. Each stint's own reasoning, declined list and checks stand as written — including the candidates it raised and rejected, which are the measured instrument.

§G.0039 — stint file cruces-0039.md, verbatim

Range: src/latin/8967/0039.md–0043.md, cols 0918C–0930C. Psalms L–LVIII. Lemma brief: data/briefs/8967-lemmata-0039.txt.

⚠ This file is one stint's range. It must be merged into src/english/8967/cruces.md.

§0 · SPAN COUNT RECONCILIATION (the required check)

My brief's header states 384 spans for chunks 0039–0043. Counting … runs in my own Latin:

The brief's total and my own count agree exactly. The split is sound for this range. The English carries 384 spans in the same order and the same places (checked mechanically with the [var: …] markers stripped, since their contents are italic and inflate a naive count).

Anchors: 12 · 4 · 11 · 11 · 10 = 48, all reproduced verbatim, in order, and each checked against the word it stands beside in the Latin, not merely for presence (test 2b). Question marks: 1 · 0 · 3 · 1 · 4, matching the Latin band for band. noteCount matches the body in all five once the frontmatter incipit's own [n:] is discounted — the documented off-by-one false positive, not a finding.

§1 · THE PLATE — seven leaves, and what they changed

Read: PDF pages 464–470 at 200 dpi full page plus 400 dpi crops, raw/scans/pl113/ patrologiaecurs04migngoog.pdf. Corner numbers read FIRST on p. 464 (917 / 918), confirming PDF page = (column + 11) / 2, then on every subsequent leaf. Recorded in data/plate-reads.json as six full reads covering cols 0917A–0928D; 0929–0930 were already on record.

⭐ Not one column in cols 0917–0930 is clipped at the inner margin. 0921 and 0925 are both 1 (mod 4) and both perfectly clean, which is one more counterexample to the withdrawn rule. No foot-of-page conjecture apparatus of either layer on any of the seven leaves.

18 word-level CC departures from the plate over 13 columns ≈ 1.4 per column — Band A's measured rate. Every one is reported in data/briefs/8967-PENDING-TEI-PATCHES.md with its column and leaf. None of them took a marker, because a [sic:] on any of them would publicly accuse Migne of the digitization's error, and because [sic:] content must appear verbatim in the Latin twin, which by construction it does not.

What the plate settled that the Latin alone could not

• hiatu / biatu @0924B — and it cut the OPPOSITE way from the standing warning. The brief warns that lowercase h images as b in this scan family. Here our Latin carries biatu and the plate reads hiatu. I settled the letter against the h of habitaculis standing on the same printed line, per the same-line control rule, not against the type in isolation. The English renders "the gulf of earthly desire".
• deservit / deserit @0921A — checked, clean. Non deserit malum, qui diligit super benignitatem reads deserit on the leaf, and nothing in this range is dative-governed. The standing check was run and returns a negative; recording the negative is the point.
• Laetctur @0923B — a candidate I DECLINED after reading it. The e of Laetetur images as c. That is a broken sort on this copy, the possit→rossit class Band A refused: it is not reported for patching and it takes no marker. Our Latin's Laetetur is right and stays.
• The heads. CC has the terminal periods backwards on three of four in this range — PSALMUS L. / PSALMUS LI on the plate against PSALMUS L / PSALMUS LI. in our Latin, and PSALMUS LIV. against PSALMUS LIV. My English follows the Latin twin, per §A3 (PSALM L, PSALM LI., PSALM LIV), so three English heads move when the Latin is patched. Flagged so the patch pass does not leave the twins disagreeing.
• Quoniam qui audivit? @0929C — the plate reads qui, our Latin quis, the Clementine (Ps 58:8) quis. Class 1: the divergence was erased before any Vulgate collation could see it. Migne's own gloss quotes the verse back as Quis audivit? twice nine lines below, so the plate carries both readings. No [var:] fired, because the word our page prints is CC's, not Migne's; the marker becomes available the moment the patch lands, and that is said in the patch queue.

§2 · THE DIVERGENCE LIST — every lemma word, against the Clementine

⭐ This section names what I checked, not that I looked. All 384 spans were walked. Below are (a) the divergences that took a marker, (b) the divergences I found and DECLINED with reasons, and (c) the lemmata I checked and found to agree exactly. Sources: sources/vulgate/clementine-flat.txt.

2a. [var:] fired — eleven

All eleven stand on columns recorded in data/plate-reads.json; plate-gate.mjs passes on all of them.

2b. Divergences found and DECLINED — with the reason, because a blind reader will look here

Pattern 14's bar is a divergence that changes the sense. These do not clear it, and firing them would bury the eleven above in noise:

• In ore stultorum est cor eorum @0920D vs Sir 21:29 In ore fatuorum cor illorum. Three differences, and the sense is unchanged. ⛔ But the decisive reason is other: the plate reads stulorsorum, a non-word, so the word our page prints is CC's repair, not Migne's. A [var:] here would compare the Clementine against a reading Migne never set. Declined, and reported for patching instead.
• Non potest abscondi civitas supra montem posita @0923C vs Mt 5:14 Non potest civitas abscondi supra montem posita. Word order only. No sense turns on it.
• Nuntiabo @0924B vs Ps 54:18 annuntiabo. The prefix only, and Migne himself writes annuntia at 0924D. Not a sense change.
• a tempestate @0923B vs Ps 54:9 et tempestate. The Vulgate's noun is governed by the same a; the preposition repeated changes nothing.
• Pro nihil @0925D vs Ps 55:8 pro nihilo. A real word, and the gloss reads it adverbially either way (quasi nihil inde tibi sit).
• Deus meus es, tu @0926B vs Ps 55:10 Deus meus es. An appended emphatic pronoun.
• In tituli inscriptione ipsi David cum tenerent eum @0925B vs Ps 55:1 David in tituli inscriptionem, cum tenuerunt eum. Case and mood, no sense change. ipsi David is deliberately kept distinct per §A1.4 and is not the divergence.
• Et coepit Jesus taedere et pavere @0927B vs Mc 14:33 coepit pavere et taedere. Migne reverses the two verbs and supplies Jesus. No sense turns on the order — but note that this is the cinis et pulvis shape from 7a″, so: my English keeps Migne's order and does not restore the Vulgate's.
• In hoc natus sum, ut testimonium perhibeam veritati @0924C vs Jo 18:37, which reads Ego in hoc natus sum, et ad hoc veni in mundum, ut …. A compression, not a divergence.
• Nonne duodecim horae sunt diei? @0925D vs Jo 11:9 Nonne duodecim sunt horae diei?. Word order.
• sperantes in divitiis praevaluerunt in vanitate sua @0929B vs Ps 51:9 speravit in multitudine divitiarum suarum, et praevaluit in vanitate sua. Deliberately pluralized by the gloss to apply the verse to the fortes of Ps 58:4. An adaptation the gloss makes on purpose is not the class Pattern 14 marks.
• Recte @0928A vs Ps 57:2 recta judicate. Adverb for neuter plural; the gloss argues from neither.
• Circumibunt @0929C vs Ps 58:7 circuibunt. Orthography.
• Fortitudinem meam … susceptor meus es, Deus. @0929D — the trailing Deus is the head of Ps 58:11 pulled into the lemma. Not a variant reading.

2c. Checked and found to agree with the Clementine exactly

Every remaining lemma in the range. Named so that this list cannot be mistaken for the work of an agent who only reported findings:

• Ps 50 (0918C–0920C): Miserere mei, Deus · Secundum magnam misericordiam tuam · Magnam misericordiam · Miserationum · Amplius lava me · Munda me · Quoniam iniquitatem (meam) · Contra me · Tibi soli peccavi · Coram te · Ut justificeris · Vincas cum judicaris · In iniquitatibus · Ecce enim · Veritatem · Incerta et occulta · Occulta · Asperges me · Hyssopo · Super nivem · Auditui meo · Averte · In visceribus meis · Ne projicias me a facie tua · Spiritu principali · spiritum rectum · Spiritum sanctum · Docebo · De sanguinibus · Justitiam tuam · Laudem tuam · Sacrificium Deo · Contribulatus · Cor contritum · Benigne fac · Muri · Tunc acceptabis sacrificium · oblationes · holocausta · Tunc imponent · vitulos · Altare. ⚑ spiritum principalem @0920A is not a divergence: it is the gloss inflecting its own lemma into the accusative after per. The lemma proper at VERS. 14 is Spiritu principali, which is Ps 50:14 verbatim.
• Ps 51 (0920C–0921C): Doeg · Quid gloriaris in malitia · Cogitavit · Sicut novacula · Dilexisti malitiam · Lingua dolosa · Propterea Deus · Et radicem tuam de terra viventium · Videbunt justi (both) · Ego autem · Oliva fructifera · In aeternum et in saeculum saeculi · Fecisti, et exspectabo nomen.
• Ps 52 (0921C–0922A): Dixit insipiens · Insipiens · Deus de · Si est · Omnes declinaverunt · Nonne scient · Deum non invocaverunt · Ossa eorum qui hominibus placent — Ps 52:6 verbatim (the plate's qni is a turned sort, reported for patching) · Confusi sunt · Quis dabit ex.
• Ps 53 (0922A–0922C): Ziphaei · Deus, in nomine · Deus, exaudi · Percipe · Alieni · Ecce enim · Adjuvat · Averte mala inimicis · In veritate · Voluntarie sacrificabo. ⚑ In virtute is not a divergence: Migne labels it himself, secundum aliam litteram. ⚠ Mihi autem, @0922C I could not place in Ps 53. It reads as a lemma pickup and the nearest Clementine home is Ps 138:17. No marker; logged as unplaced rather than guessed.
• Ps 54 (0922C–0925B): In finem in carminibus · Intellectus · Exaudi, Deus, orationem meam · Ne despexeris deprecationem · In exercitatione · Conturbatus sum a voce inimici (spans the 54:3/54:4 verse break, which is Migne's clipping, not a variant) · Cor meum conturbatum · Timor et tremor · Contexerunt me tenebrae · Sicut columbae · Ecce elongavi · Et mansi in solitudine · Exspectabam eum · Praecipita · Divide · In civitate · Et labor in medio ejus · Usura · Quoniam si inimicus · Abscondissem me forsitan ab eo · Tu vero, homo · Unanimis · Notus · Dux · Notus meus. Qui simul · Cibos · Veniat · Veniat mors · Descendant in infernum viventes · Viventes · In habitaculis · in medio · Ego autem · Vespere et mane · mane · meridie · Narrabo · Appropinquant mihi · Quoniam inter multos erant mecum · Exaudiet Deus · Humiliabit illos qui est ante saecula · Non enim est illis commutatio … · Contaminaverunt · Ira · Cor · Molliti sunt sermones · jacula · Ipsi sermones · Curam tuam · Fluctuationem · Contristatus sum in exercitatione · Tu vero · Dies suos, ego autem · Laetare, sterilis, quae non paris (Gal 4:27 verbatim).
• Ps 55 (0925B–0926D): Allophyli · Geth · Miserere mei, Deus · Impugnans tribulavit me · Ab altitudine · Altitudo diei · sperabo · Tota die · Inhabitabunt · In ira populos confringes (both) · Deus, vitam meam annuntiavi tibi · Posuisti lacrymas · In quacunque die · In Deo laudabo verbum · In Domino · Laudabo sermonem · In me sunt · Quoniam eripuisti · Ut placeam.
• Ps 56 (0926D–0928A): In finem · Miserere mei · Et in umbra · Misit de coelo · Liberavit · dedit · in opprobrium conculcantes me · Misit Deus · Misericordiam · Veritatem · Misericordiam suam · Dormivi · Conturbatus · Filii hominum · Lingua · Tristis est anima mea (Mt 26:38) · Exaltare · Laqueum paraverunt · Pedibus · Incurvaverunt · Ante faciem meam foveam · Paratum · cantabo · Exsurge, gloria mea · exsurgam diluculo · Exsurge, psalterium et cithara · Ad coelos.
• Ps 57 (0928A–0929A): In finem ne disperdas David in tituli · Si vere utique justitiam loquimini · Manus vestrae · Concinnant · Alienati sunt peccatores a vulva · Furor illis · Surdae et obturantis · Terra · cauda · Venefici incantantis · Deus conteret · Molae fractae · Dentes · In ore · Molas leonum confringet Dominus. Ad nihilum devenient · Confringet · Ad nihilum · Tanquam aqua · Intendit arcum · Priusquam spinae · Rhamnum (both) · Sicut viventes, sic · Sic in ira absorbet eos · Laetabitur justus cum viderit vindictam · Et dicet homo · Deus judicans eos. ⚑ Medicamenti medicati a sapiente is Migne's own labelled alia littera, not a divergence.
• Ps 58 (0929A–0930C): In finem ne · Ne disperdas · Quod scripsi, scripsi (Jo 19:22) · Eripe · De inimicis · Eripe me · Fortes · Iniquitas · peccatum · Exsurge · In occursum · Vide · Convertentur · Ad vesperam · Circumibunt · Et gladius · Gladius · Deridebis eos ad nihilum · Quis audivit · Ostendit · Ne occidas · insertis · Ne quando obliviscantur · In virtute tua · Depone · Delictum oris eorum · delictum · Et comprehendantur · In superbia · Et de exsecratione · Annuntiabuntur in consummatione · In ira consummationis · consummationis · consumptionis · Et scient · Convertentur ad vesperam · Murmurabunt · Ego autem · Et exaltabo · Misericordiam · Ad vesperum demorabitur fletus, et ad matutinum laetitia, Ps 29:6 verbatim.

§3 · NEGATION — every printed negative, and the word it is on

I named the host word for each non/nec/neque/nisi/nunquam/nullus/nihil/ne in the range and checked the negative sits on that host in the English, not merely that it is present (test 2, and test 2a's hole). The sites where the check bit:

• @0918C non cadendi, sed, etsi cecideris, resurgendi ne desperes — two negatives, on cadendi and on desperes, and English apposition would happily drop the second. Rendered "not of falling, but of rising again if thou hast fallen, that thou despair not."
• @0920A robur solum petit non auferri — ⚑ the exact 2a trigger: a negative on an infinitive with the quantifier solum beside it. solum modifies robur, not the asking: he asks that the strength alone not be taken. Rendered "he asks that its strength alone be not taken away" — the negative on auferri, the quantifier on robur.
• @0925C non ait: Non timebo, sed, sperabo in te — two negatives, one of them inside a quoted refusal. Both carried; the sed correlative is what fixes the scope.
• @0928B Non solum non faciunt, sed nec audire volunt — three negatives, and the first belongs to solum, not to faciunt. Rendered "Not only do they not do, but they will not even hear."
• @0930A Non eos, sed delictum occide — ⚑ the sharpest one in the range, and the §4 class exactly. The non is on eos, not on occide: the command is to kill, and to kill the sin instead of the men. "Slay not them, but the sin" reads the negative onto the verb and inverts Augustine. Rendered "It is not them, but the sin, that thou art to slay", which puts the scope where the Latin puts it.
• @0925B Vae! qui peccant et non poenitent — the apposition trap from §4. Both the ! (kept where Migne sets it, after Vae, not moved to the end of the sentence) and the non carried.
• @0922B qui non nisi temporalia noverunt — non and nisi both printed, both rendered: "who know nothing except things temporal."
• @0921A Non deserit malum, qui diligit super benignitatem · @0919A Cujus verba non transeunt · @0919B Christo nullum habenti peccatum · @0919C ut nec peccata sint impunita · @0921B quae non fructum parit / nihil habebunt cum sanctis · @0926A nullus inde tibi labor / quasi nihil inde tibi sit / pro nullis eorum meritis · @0926C Vota … quae a nobis nunquam recedunt · @0928D non vera, sed falsa … videntur vivere, nec vivunt / non est ira comparatione futurae · @0930A non ipsi / non surgeret Paulus, nisi prius caderet Saulus · @0930A nec fieri potest ut superbus non invideat — all checked host by host, all carried, none supplied.

No negative in the range was translated as its opposite, and none was relocated to a neighbouring verb to make a clause run. Nothing in these five chunks was smoothed.

§4 · UNEXPRESSED SUBJECTS — the pronouns I supplied, and who they are

Per the brief's dominant-defect rule, each supplied "he/she/it/they" named:

• @0925B–0926D, Psalm LV throughout: the subject is the CHURCH, feminine. Orat Ecclesia … dicit se non timere, dicit se a superbis conculcari, laudem Dei cantaturam se dicit. The gloss names her once and then runs on subjectless for a column. Rendered "she" throughout, with Ecclesia as the antecedent — not the psalmist, and not Christ, who enters only at Ps LVI.
• @0924B Alienae sunt, non naturales — feminine plural, and the antecedent is nequitiae from Ps 54:16 (nequitiae in habitaculis eorum), which is in the verse and not in the sentence. Rendered "They are alien" of the wickednesses, not of the dwellings (habitacula, neuter) and not of the men.
• @0927D dum eis compatiens sterilis reddebatur ab eis — the subject is Christ, carried from Domini two clauses up; the eis twice are the persecutors.
• @0928C horum Molae fractae — horum are the asps and lions of the preceding clause, i.e. the Jews, named there.
• @0921B qui cum eo nihil habebunt cum sanctis — qui are the ministri, the immediately preceding word, not Antichrist.

§5 · CRUCES — passages where the Latin will not simply construe

Rendered as closely as the printed words allow, per Pattern 7. None was repaired inside the translation.

1. @0919C Hyssopus herba humilis, etc., usque ad in quibus est inflatio, sic humilitas superbiam. The abridgment tail has no verb: humilitas superbiam [purgat]. Rendered with the ellipsis visible — "so humility, pride" — rather than supplying purgat. The formula removes words; it does not license me to put one back.
2. @0920D Quae dum promittit innovationem, incidit ita ille dolosus, Antichristus scilicet, beatitudinem, et occidit. Two readings construe: incidit shared by both subjects (the razor cuts; so Antichrist cuts blessedness), or a second verb elided after dolosus (Antichrist promises blessedness and slays). I took the first, because incidit is the only finite verb before beatitudinem and et occidit pairs with it; the rival is named here because the apparatus does not record it. Rendered "so does that deceitful one, namely Antichrist, cut blessedness, and slay."
3. @0924A Corporales, qui solis iniquo poterant esse dulces. solis (nom./gen. of sol, or dat./abl. pl. of solus) against a dative singular iniquo does not agree. Confirmed on the leaf: Migne sets it thus. It is a real word, so per the addendum's rule it stays Migne's and takes no marker. Rendered by the evident sense, "sweet to the wicked alone", and logged.
4. @0925A sed jacta in Dominum, non alium qui portus est, vel saxum navi, non portus fluctuati in saeculo. Two readings: qui portus est describes the Lord (and the rest describes the alius), or it describes the alius throughout. Both construe; the second gives a sharper contrast. I rendered the printed order without choosing, keeping both negatives on their own nouns, and name the ambiguity here rather than resolving it silently in the text.
5. @0926A Deus, vitam. Confringes, et ego ad institutionem eorum vitam meam ante, quae semper mala. Verbless: the lemma's own annuntiavi has to carry over. I supplied "declare" and I am saying so, because an English clause cannot stand without a verb; the word supplied is the lemma's own, printed four words earlier, and nothing else is added.
6. @0926C Vel ita, ut Trinitas naturae personaliter per tria verba, quae ponit. Same shape, no verb. Rendered "stands personally in the three words which he sets down"; the copula is supplied and named here.
7. @0922D Quis dabit mihi pennas sicut? Migne prints the lemma clipped after sicut, with his own question mark. Rendered with the dangling comparative kept — "Who will give me wings as?" — rather than completed to sicut columbae, which he supplies himself two sentences later as a separate lemma. Convention 3 expands a fragment; it does not license completing a clipped comparative that the plate deliberately leaves hanging with a mark on it.
8. @0925C Diei Dies, Christus est — the lemma word diei and its gloss Dies stand inside one italic run. Rendered "Of the day. The Day, is Christ", keeping the run as one span.
9. @0929A Malivolentia: peccatum, operis. Two elliptical genitives, no verb: "Iniquity [is a matter] of ill-will; sin, of the deed." Rendered without supplying the copula. The plate reads Malivolenti, a dative, which is a further departure reported for patching.
10. @0929C et oraturo Judaeorum salus — a future participle in the dative with no obvious governor. Rendered "and, for him who is yet to pray, the salvation of the Jews", and confirmed on the leaf as printed.
11. @0930B ostenditur eis, quod non sine Christo, qui dominatur Jacob — verbless again. Rendered "that this is not so without Christ", keeping non on the predicate and supplying only the copula.

§6 · CONVENTIONS — inherited, applied, and one axis decided per site

• In finem → "Unto the end." 8 body occurrences in this range (0918C In finem psalmus, 0920C In finem pro, 0921C In finem pro, 0922C In finem in carminibus, 0925B In finem, 0926D In finem, 0928A In finem ne disperdas David in tituli, 0929A the clipped In finem ne), and 8 "Unto the end" in the English. Zero rivals. §A1.1 holds. ⚑ A raw grep returns 13 in the Latin: five of those are the frontmatter incipit of each chunk, which is copied verbatim and untranslated. The instrument was checked before the number was reported.
• usque ad → "as far as". 57 occurrences in this range (17 · 5 · 10 · 11 · 14), against 57 "as far as" in the English, all inside italic spans, all the abridgment formula, zero rivals. ⚑ I checked for the unitalicized look-alike class the addendum names and found none in these five chunks — a checked negative, worth recording.
• Diapsalmus — ROMAN, unitalicized, unanglicized. 6 occurrences (0922B, 0923B, 0927A, 0927D, 0929B, 0930B). Our Latin sets all six roman, the plate sets all six roman, and my English follows. §A1.2 holds and needed no override.
• sedes → "seat" @0928A (ubi sedes angelorum → "where is the seat of the angels"), per §A5.
• Heads. PSALMUS L → PSALM L, PSALMUS LI. → PSALM LI., PSALMUS LIV → PSALM LIV, and the rest with their periods. Roman numerals stay Roman and the twin's punctuation is followed, including where the plate disagrees with the twin (see §1).
• Psalmus ipsi David @0925B kept distinct from Psalmus David, per §A1.4.
• Archaism. thou/thee/thy used for the grammatically singular addressee only (God, the soul, the reader), you for the plural (et dicitis, quis audivit? @0929C → "and you say"). No -eth/saith was used anywhere in the range, in either voice. There are no « guillemets » in these five chunks, so the two-membership test reduces to the italic one here.
• gentes — decided per site, and the sites named: | col | Latin | rendering | why | |---|---|---|---| | 0921B | insertus est oleaster de gentibus | Gentiles | Rom 11, against Judaei in the same clause | | 0923A | avolavit ad gentes | Gentiles | against corda Judaeorum in the same sentence | | 0923B | mansi in solitudine, gentium | Gentiles | the solitude opposed to the synagogue | | 0924D | ut tollat locum et gentem | nation | singular ethnic gens, Jo 11:48 | | 0929A | famam habere in omnibus gentibus | nations | plainly generic, the in omnes gentes shape | | 0929A | orat etiam pro salute gentium | Gentiles | paired with Judaeorum salus at 0929B | | 0929B | Orata salus gentium … et Judaeorum salus | Gentiles | the opposition is explicit | | 0929C | convertere omnes gentes … omnes gentes Christianae futurae sunt | Gentiles | the Gentile mission set against the unbelieving Jews of the same gloss | | 0930B | sciunt se pares gentibus | Gentiles | explicit Judaei subject |
• ⛔ gens tenebrarum does not occur in this range. I checked: the only gens -initial match in the five chunks is inside tangens @0925B. Nothing to flag, and nothing decided.
• Em-dashes: zero in all five English chunks.

§7 · CANDIDATES I RAISED AND REJECTED