Commentaria in Cantica canticorum

Translated 2026-08-08 by twenty-one Opus agents over 126 chunks (134,020 Latin words), each range of six logging to its own cruces-NNNN.md. Those twenty-one files are the record and are unabridged; this file is the merge. It exists because a work split across twenty-one agents who cannot see one another will drift on its load-bearing vocabulary, and reconciling that is a required pipeline step (translation-runbook.md step 4), not a tidy-up. Concatenating is the easy half.

Work totals after the eleven-reader blind polarity read and the whole-apparatus audit (§7): 101 [var:] · 50 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:] · 719 [n:] (719 in the Latin twins — exact parity). verify-english.mjs clean at 126/126: columns, notes and sections aligned, no duplicate paragraphs. EN/LA word ratio 1.56 work-wide (209,013 / 134,020).

§1 · WORK-WIDE CONVENTIONS — RULED AND APPLIED. DO NOT RE-DECIDE.

These are settled and swept. They are set down here with their reasons so that a later sweep does not churn them, and so that a reader can tell a considered choice from an accident.

Divisions. tomus = "Tome" / liber = "book" is inherited from 11638 — but this work has no tomi. Its divisions are liber = book (six of them), caput = chapter, volumen = volume (135 instances, none of them a division of the work). The tomus/liber collapse that cost 11638 sixteen sweep sites cannot arise here in that shape; the 11638 rule stands only in case a tomus ever turns up. INCIPIT = "HERE BEGINS", EXPLICIT = "Here ends".

CAPUT PRIMUM = "CHAPTER THE FIRST." — ruled and swept 2026-08-08. Seven sites: chunks 0002, 0003 ((cont.)), 0026, 0047, 0069, 0091, 0108. The ranges had split 3–3 against "CHAPTER ONE."; the three were swept. "Chapter the First" is the right side of that split because PRIMUM is an ordinal spelled out, and it is the spelled-out-ness that carries Migne's own contrast with the 179 numeral chapters that follow. "Chapter One" is a cardinal and flattens exactly the distinction the plate makes. ✅ Verified: 7 ## CAPUT PRIMUM in the Latin, 7 ## CHAPTER THE FIRST in the English, 0 ## CHAPTER ONE.

⭐ amica mea (Cant. 1:8) = "my friend", NOT "my love." Ruled from Caput V, where Philip glosses the word and proves it from John 15:14's amici mei. The lemma recurs at Cant. 2:10, 2:13, 4:1, 4:7, 5:2, 6:3 — nearly every range met it. ✅ 20 Latin amica mea, 17 English "my friend"; the three shortfalls are re-quotations absorbed into Philip's own syntax, not drift. The one English "my love" in the whole work (0122 @0480D) renders mea dilectio, not amica mea, and is correct — checked, no edit.

⭐ muraenula stays italic Latin. Caput VIII's whole argument is that the word names a fish (Muraenula piscis est…) and the chapter runs tail-to-mouth → ring → net-mesh → chain → Moses' head-and-tail offering. "Chains of gold" makes seventeen paragraphs explain nothing. ✅ 22 Latin instances (0032 ×21, 0033 ×1); the five English instances not carrying their own asterisks are either the frontmatter incipit (verbatim Latin) or nested inside a wider italic run — nesting, not drift.

⭐ vermiculatas argento IS translated ("vermiculated") — the companion call, decided the other way on purpose, because English carries the whole gloss chain: vermiculate → vermilion → little worm → Ego autem sum vermis. Italic Latin is reserved for words with no English exponent; this one has one. ✅ 6/6.

⭐ The bed triple, and it is a triple on purpose. lectus = "bed" · lectulus = "little bed" (40 Latin / 41 English) · cubile = "couch." 0279B's whole humility argument turns on the diminutive (non tamen Virgo lectum, sed magis lectulum), and "bed" is not available for cubile because a third word collapsing into the pair would void that argument. See §4.5 for the two sites where this had not in fact been swept.

Other ruled equivalences, all held work-wide: cancelli = "lattice" (10/10; 0312D identifies the lemma with the architectural sense and "chancel" would sever the argument) · emissio = "emission", not Douay's "plants" · vicus = "lane" (9/10) and platea = "street" (18/19), kept distinct against Douay, load-bearing at 0318D · ferculum stays italic Latin, on the muraenula precedent · maceria = "dry-stone wall", distinct from murus · adolescentulae = "maidens" · charitas = charity (62 Latin / 65 English), never "love" · candidatio is translated, not carried.

⭐ The colour triple, kept strictly distinct: nigra = black · fusca = swarthy · formosa = comely. fusca is never "black" — and in this work that matters more than in 11638, because Cant. 1:5 prints nigra sum where the Vulgate has fusca sim (0230D): the colour word and the mood both changed, and Philip builds four columns of nigredo exposition and the whole nigra sed formosa architecture on it. Marked [var:], never conformed.

Pattern 3, "Holy Scripture" — both words capitalized, for divina, sacra, divinae Scriptura(e) and the bare form alike. 22 instances in the English; 0 instances of "divine Scripture" or "sacred Scripture", which is the defect 11638 shipped. The bare form is the one 11638's sweep missed and this work's brief warned about in advance: it fires on a bare in Scripturis (0260A), a bare genitive Scripturae (0307D), a bare nominative Scriptura (0331B), and sixteen times in chunks 0078–0083 alone, fourteen of them bare.

Pattern 17: thou = singular, you = plural. A number marker, not a reverence marker. Anchor case: 0454D, Luke 10:23 vos videtis → "you see".

⭐ Migne's mixed capitalization is MIRRORED PER INSTANCE, never regularized. Sponsus/sponsus, Sponsa/sponsa, Virgo/virgo, Dilecto/dilecto, Spiritus sanctus/sancti Spiritus — he runs both cases inside single sentences. This was named in advance as the likeliest drift point across twenty-one agents, and it did not drift: every range checked its own plate word by word and reported so. The worked example that closed the question is 0305B Spiritus sanctus against 0356C sancti Spiritus — two ranges appeared to disagree, both were right, and there was nothing to reconcile. Before reporting a capitalization disagreement in this work, check whether Migne simply prints both. He usually does.

Guillemet asymmetries are Migne's and are mirrored, never balanced. Five chunks are asymmetric — 0006 (8 «, 9 »), 0011 (11/10), 0044 (4/3), 0083 (5/4), 0089 (10/11) — and the English reproduces each 1:1 (422 « and 421 » on both sides). At 0307C Migne himself breaks one quotation into two guillemet pairs around inquiens, doing what Pattern 5 forbids us to do; reproduced.

§2 · THE TYPOGRAPHY — three findings, none of which one range could have made

2.1 · The marks divide BY FUNCTION, not by book

11638 changes how it marks scripture at a book boundary — italics in Books I–II, guillemets « » from Book III. 11632 is not built that way. It runs both marks in parallel from chunk 0000, divided by function: guillemets carry a quotation adduced as an authority (scripture is the commonest case, not the rule — the Delphic oracle, Virgil, and two bare pronouns displayed as words are all guilleted at 0251D–0254B), and italics carry the running Song lemma plus book titles, words discussed as words, printed locators, and Migne's marginal source-edition page numbers.

The book boundaries were checked and carry no seam — LIBER SECUNDUS. at 0246C (cruces-0024), LIBER TERTIUS. at 0303A (cruces-0042), LIBER QUARTUS. at 0354C (cruces-0066), LIBER QUINTUS. (cruces-0090). Both marks run identically on either side of each. The question is closed: later book heads are ordinary text.

The proof that the division is functional is that the Song ITSELF takes guillemets when quoted as a proof-text rather than as the lemma — Cant. 4:11 @0220A, Cant. 5:11 @0226A, Cant. 2:14 @0226C, Cant. 1:12 @0270B, Cant. 5:11 again at 0432C, each with the italic lemma a few lines away. Five instances, four ranges.

2.2 · ⭐⭐⭐ BUT THE GUILLEMETS STOP — and the seam is nowhere structural

Found independently by the 0096–0101 and 0102–0107 agents, then verified against all 126 Latin chunks.

From chunk 0097 to the end of the work there is not one guillemet — 29 chunks, ~73 columns, ~50 notes, none. The last « in the book stands at col. 0416D. And Philip does not stop quoting: adduced proof-texts continue at full density, each with its inquit/ait and a correct note, printed in italics.

⭐ Internal control, exact and on one verse: Isa. 53:7 is guilleted at 0399C and italic at 0440A–B — same verse, same siglum, forty-one columns apart.

⚑ The seam falls mid-Caput VII of Book V, 29 columns from either book head. So this does not overturn §2.1's no-seam-at-a-book-boundary result; it strengthens it. Migne's practice changes where nothing structural happens, which is why nobody predicted it and why it had to be measured across the whole work rather than reasoned from one range.

⛔ The practical consequence, which was pushed into the briefs of every range from 0090 on: "italic run = probably just a lemma" is true to 0089 and false after it. In the last 29 chunks italics carry the lemma and every quotation, and they are told apart by function, not by mark. A range applying the earlier rule would have mis-classed every quotation it met and collated none of them.

2.3 · ⭐⭐⭐ THE DISPLAY RULE — it dissolves the "counter-instances"

Waves 3–5 kept logging classical and patristic authorities printed bare — Ovid at 0259, Sallust at 0314A (adduced explicitly, verax sententia… ethnicorum), Ambrose by name at 0337B, Virgil at 0385A — and read them as counter-instances to §2.1, killing one explanation after another, including the tempting "Migne declines to guillemet pagans", which a Father printed unmarked disposed of. They were never counter-instances. The discriminator is typographic DISPLAY.

• A quotation Migne sets displayed, on its own line or lines, is italicised.
• A quotation Migne runs into the paragraph is bare.

⭐ The control is one line of Horace printed twice. At 0022 the Siculi tag is displayed on its own line and italicised (Invidia, inquit, Siculi non invenere tyranni / Tormentum majus.); at 0447D the same line is run into Philip's sentence — majus quo nec Siculi, juxta quemdam ethnicum, tormentum invenere — and printed bare. Same poet, same line, two treatments, one variable. ✅ Checked against all 126 chunks: every displayed block in the work is italicised, and every bare authority is run-in.

⚑ The rule is PERMISSIVE, not predictive, and that has not changed: reproduce what is printed, and never add a mark because the function seems to call for one. What it changes is what counts as a finding. A bare adduced authority is expected; only a displayed quotation printed bare, or a run-in one italicised, would be real.

⚑ Migne's [n:] is usually inside the closing » but not always — outside at 0235B and 0239B; a bare locator inside a run with no [n:] wrapper at 0232D and 0236D; a printed locator standing inside a guillemet run at 0203A and 0222B–C. Reproduced as printed in every case.

§3 · PHILIP'S BIBLE — the copy-text readings

A divergence attested in more than one range is evidence about Philip's copy of the Bible, not a local slip. These are the merge's own results; no single agent could have made any of them.

⭐⭐⭐ Cant. 2:17 — montes BETHEL for the Vulgate's Bether (0338D, again at 0341A). Philip spends four columns on the etymology of the divergent word: Beth, domus; el vero Deus interpretatur; et Bethel domus Dei… Domus autem Dei non incongrue interna Virginis conscientia judicatur. Conform the lemma and the back half of Caput XVIII has nothing to stand on. Printed twice, which is the copy-text shape — the shape that made Zech. 4:14 and Gen. 3:19 findings rather than slips on 11638.

⭐⭐⭐ Eccles. 9:1 — the pair REVERSED in six of its seven printings. utrum odio dignus sit an amore against the Vulgate's utrum amore an odio. ✅ Verified against all 126 Latin chunks: reversed at 0029, 0042, 0062, 0070, 0074 and 0125; only 0049 carries the Vulgate order. So the copy-text shape comes with its own internal control — Migne contradicts himself exactly once. And the seventh printing (0489A, the last in the work) is recast: different wording, the reversal surviving three re-castings. Statistically the strongest member of the class. Marked nowhere it is unmarkable: at 0489A the span carries neither guillemets nor italics nor a note, and under §8a's rule it is logged, not marked.

⭐⭐⭐ Cant. 5:9 — and it is the only member of the class with an AUTHORIAL witness. At 0429B Migne prints ex dilecto once where the Vulgate prints it twice. At 0430B–C Philip discusses the omission himself: he reports that most of the codices he could find carry the phrase twice, and declines to say whether the repetition is the author's or a scribe's. This is not our inference from his exposition — it is his own collation note, printed. [var:] fires; Migne stands.

⭐⭐⭐ Cant. 7:13 — A NEW SHAPE: Migne DIVIDES the verse differently. At 0473B he prints Mandragorae dederunt odorem in portis nostris, where the Vulgate gives the gates to the apples (In portis nostris omnia poma). Every word is present — only the break moved. All of Caput XXXI expounds the gates as the mandrakes' (necdum in aedibus, necdum in thalamo… sed in portis), and Caput XXXII opens its lemma at Omnia poma with the gates already spent. ⚠ This shape is not greppable the way a divergent word is. Nobody was looking for it for 113 chunks. It is found only by reading the lemma against the Vulgate's own division — which is now a standing instruction for lemma-and-gloss commentaries.

⭐⭐ Cant. 5:7 — vigiles for custodes (0427A), which is Cant. 3:3's word, conformed onto 5:7. The proof it is copy-text and not a compositor's reach is Migne's own cross-reference: (sup. l. III, c. 22) resolves exactly to the Cant. 3:3 lemma in chunk 0065. Philip then builds two columns on the resulting distinction (custodes murorum kept a vigilibus), which the Vulgate's single word makes unwritable.

⭐⭐ Cant. 8:2 at 0475C — the lemma carries a whole clause the Vulgate does not have (et in cubiculum genitricis meae), the LXX/Old Latin reading, and five chapters of exposition stand on it. Cant. 8:12 at 0486C is its opposite number: a deletion (his dropped from et ducenti his qui custodiunt) which changes who the two hundred are, and Philip's exposition follows the printed reading.

Others attested and marked: Sponsa mea for speciosa mea (Cant. 2:13, 0323C, with Philip's next sentence proving it is copy-text) · in cavernis plural (Cant. 2:14) · grex for greges (Cant. 4:1–2, and the chapter is an argument about the singular) · pellis Salomonis singular (Cant. 1:4, and Caput XVIII expounds one skin flayed from one animal) · Sunamitis for Sulamitis (seven printings; Sulamit- occurs nowhere in the work) · matri suae dative for matris suae genitive (Cant. 6:8a, printed four times, the whole chapter built on the dative — and the work knows the genitive, printing it correctly at 0020) · Sion added to Cant. 6:8b at 0454B, with Philip expounding the addition in his next clause · conturbata est for conturbavit me (Cant. 6:11, voice reversed, gloss carrying both).

⭐ Cant. 1:12 — Cum for the Vulgate's Dum, ruled here. Migne prints Cum four times: three in chunk 0033 and again at 0270B where the verse returns as a guilleted proof-text with Philip's own gloss picking the word up. No [var:] fires, and that is right — the sense is untouched, which is Pattern 14's threshold. But four independent printings is copy-text evidence, so it is recorded here as an attested reading rather than a slip. ⚠ The fifth occurrence, at 0397C, prints only the verse's second half and supplies no evidence either way; a range reporting "met, agrees" would have been counted as an attestation of something it never witnessed. Name the words, not the verse.

⭐ An expounded quotation Migne leaves unmarked is LOGGED, never marked — ruled once for all twenty-one ranges. The class: Ps. 90:5–6 fused at 0245A, with the next sentence glossing the fusion; Lam. 4:20 at 0291B printed with the Old Latin ante faciem nostram against the Vulgate's oris nostri, expounded three times in two sentences, and carrying no guillemets, no italics and no [n:]; 1 Cor. 2:9 at 0372D; Deut. 32:33 at 0469A; the Chalcedonian formula at 0444A. And one member is a DELETION, which is worse: John 19:39 at 0399D prints ferens misturam myrrhae et aloes libras centum for the Vulgate's quasi libras centum, unmarked in every channel, twelve lines before Philip argues in pondere et numero… certo pondere, perfecto numero consummatum. His whole four-sentence argument needs the hundred to be exact, and what has gone is the Vulgate's hedge. Every other member substitutes; a substitution at least leaves a strange word on the page. A deletion leaves nothing at all, and no mechanical channel can reach it. The ruling stands: log the candidate with its evidence, place no [var:]. A [var:] asserts a divergence inside a quotation, and where Migne marks the words as neither quotation nor lemma we cannot establish the quotation's boundary. Marking it would have us assert a boundary the plate does not print.

§4 · WHAT THE MERGE CHANGED — the reconciliations

Five live divergences between ranges. Each is ruled here, with the reason, and the exact edits are listed. The edits are proposed, not applied — they are handed to Wilson.

4.1 · ⭐ aromata = "spices". "Aromatical spices" comes out. — 12 edits

aromata is one Latin word; "aromatical spices" is two English words, and it is Douay's phrase at Cant. 3:6 and 5:1. Fourteen chunks render the root; nine of them say "spices" and five say "aromatical spices". This is the work's characteristic defect class — the Douay arriving over the plate — showing up in vocabulary rather than in a verse.

⭐ The ruling is forced from inside chunk 0105, which contains both words and contradicts itself in three lines. Migne prints areolae aromatum (noun) in the lemma and arboribus aromaticis (adjective) in the gloss. Our English gave the lemma "beds of aromatical spices" and then, two sentences later, glossed it "the Virgin compared him to beds of spices" — the same phrase, two ways, on one page. If aromata takes two words, the genuine adjective aromaticus has nothing left to be.

RULE: aromata (the noun, any case) = "spices". aromaticus (the adjective) = "aromatic". A bare aromata never takes "aromatical".

One judgement call, listed separately. 0105.md @0437D reads "herbs and spices are fitly brought forth for aromatical uses" for Migne's herbae et species aromatibus congrue proferuntur. "Uses" is supplied and "aromatical" is doing adjectival work for a noun in the dative. Proposed: "herbs and spices are fitly brought forth for spices" is unreadable, so the least-bad literal is "…are fitly brought forth for the making of spices", with the supply visible. ⚠ Flagged rather than ruled — it is the one site where the rule does not mechanically apply, and it wants a translator's hand.

Not touched, and correct: 0105.md @0438B "brought forth aromatic spices" renders aromaticas species — genuine adjective + genuine noun. Leave it.

4.2 · ⭐ suavis = "sweet". The 0107 mapping wins. — 6 edits

Two ranges mapped the pair in opposite directions:

• 0097–0099: dulcedo = "sweetness", suavitas = "pleasantness".
• 0107: suav- = "sweet", dulc- = "delightfulness".

⭐ 0107 is right, and the reason is that its mapping is fixed by the plate while 0097's is free. Chapter XXVI's lemma is Cant. 5:16 Guttur illius suavissimum — "His throat is most sweet" — and the whole chapter is an exposition of that superlative (non tam suavior quam suavissimus). The chapter then locks the two words into a chiasmus that forces distinct English: dulcedo suavior, suavitas dulcior non habetur. If suavitas is "pleasantness" there, the lemma the chapter expounds has changed word underneath it.

⚑ And 0097 is already inconsistent with itself: it renders suavi otio "sweet leisure" twice (@0419A, @0420D) while rendering suavitatem of the same page "pleasantness". Half that chunk is already on 0107's side.

⚑ The work-wide counts settle the direction rather than the sweep. Latin: 198 suav- against 88 dulc-. English: 279 "sweet", 73 "delight", 19 "pleasant". So suav- is the majority word and "sweet" is its standing exponent almost everywhere; the two collide only where Philip puts them in one period. RULE: suavis/suavitas = sweet/sweetness. Where dulcedo stands in the same period as a suav- word, dulcedo = delight/delightfulness. Elsewhere "sweetness" for dulcedo* is left standing — a full 88-site sweep buys nothing and risks damage out of proportion to the gain (the 11638 principle applied).

0099.md @0424A "the sweetness of it" already renders ejus suavitas under the ruling — no edit. 0107.md is ruled correct throughout — do not touch it.

4.3 · donec ipsa velit — TWO edits, not three. Migne prints three phrases.

The report was that chunk 0121 renders the formula three ways. Checked against the plate, two of the three are Migne's, not ours — this is the 11638 QUALITER lesson recurring, and a majority-rule "fix" here would have been a silent normalization of the Latin.

And the lemma is printed at three places in the work: 0046 @0299D "till she herself will", 0068 @0353A "till she please", 0121 @0477D "till she please". Only 0046 renders ipsa. Douay's "till she please" drops both the emphatic pronoun and the volition, and Philip's own exposition at 0478C renders it in full — so the lemma should show what the exposition expounds.

RULE: donec ipsa velit = "till she herself will", at all three lemma sites.

And doctores = "teacher", one edit. Work-wide: 10 Latin doctor, 10 English "teacher(s)", 1 English "doctors"* — a single outlier, in the same chunk and the same chapter as the above, eleven words from two correct "teachers".

4.4 · ⭐ The catchword at 0124 that does not match its lemma — fix the LEMMA

Migne prints De qua subjungit: «Vinea fuit,» etc., then the head CAPUT XLVII. and the lemma Vinea fuit pacifico, in ea quae habet populos (Cant. 8:11). Our English gave the catchword "There was a vineyard, etc." and the lemma "The peaceable had a vineyard, in that which hath peoples." A reader following the catchword cannot find the passage, because the two share no opening words.

⭐ The cause is not the catchword. The catchword is literal; the lemma is Douay ("The peaceable had a vineyard" is Douay Cant. 8:11, which turns Migne's dative of possession into an English transitive). Repairing the catchword would conceal the pull-through; repairing the lemma removes it and makes the two match automatically.

Leave There was a vineyard, etc. exactly as it stands — it is a faithful rendering of Migne's two-word re-quotation, and after this edit a reader following it lands on the lemma's first four words. ⚠ Second-best alternative, if the lemma is held sacrosanct: change the catchword to The peaceable had a vineyard, etc. This restores navigability but reproduces four English words where Migne prints two, and leaves the Douay in place. Recorded so the choice is visible.

4.5 · cubile = "couch" — two sites the earlier sweep missed

The ruling was made mid-run (cruces-0060 set "couch"; cruces-0066 rendered it "bed" eleven columns later) and the brief named the remaining sites as chunks 0082, 0083, 0115. All three are correct. Chunk 0070 was never on the list. ✅ Count: 14 Latin cubil- (0063 ×4, 0070 ×2, 0082 ×5, 0083 ×2, 0115 ×1) against 13 English "couch" — and one of the thirteen is not a cubile at all (below).

⭐ The control is internal and exact: 0063 @0342A renders Isa. 57:8's dilatasti cubile tuum "thou hast enlarged thy couch"; 0070 @0357C renders the same verse "thou hast enlarged thy bed." Same verse, same word, fifteen columns apart.

⚠ 0070 @0357B is NOT an edit. "The bed is straitened" there renders Isa. 28:20's stratum — a fourth Latin word (10 instances work-wide), which under the lectus/lectulus/cubile ruling has nowhere to go. It is left as "bed" because lectus does not occur in that sentence, so nothing collides. Recorded as a crowding, not repaired: a fourth English word for stratum would be invented rather than found, and the merge declines to invent one.

4.6 · The Gallemart colophon at 0490B — a proposed house ruling

Migne closes the work with Joannes GALLEMART S. Theol. Doctor, et catecheseos professor in Universitate Duacensi, hac 7 Febr. 1618. — the subscription of a Douai censor, printed italic. It is not Philip's text and not scripture; it is a documentary signature, and our English currently reads "John GALLEMART, Doctor of Sacred Theology, and professor of catechesis in the University of Douai, this 7 Feb. 1618." — half-Englished, and half in Migne's capitals.

Proposed ruling: in a documentary subscription, the NAME stands as printed and the OFFICE is Englished.

Reasoning. A signature identifies a historical person in a record, and Englishing half of one manufactures a person — "John Gallemart" signed nothing. The capitals are Migne's typography and fall under the same mirroring rule as Sponsus/sponsus (§1). The office, by contrast, is ordinary descriptive Latin with an ordinary English exponent, and leaving it in Latin would tell an English reader nothing. This is a house ruling, not a work ruling — it will govern every censor's approbatio, imprimatur and editor's subscription in the corpus, which is why it is worth deciding once here rather than per work.

§5 · WHAT THE MERGE RATIFIED — checked, correct, do not churn

• The CAPUT numeration is mirrored as printed and not tidied — CAPUT PRIMUM spelled out, CAPUT II–L in numerals, and CAPUT VIII / CAPUT IX / CAPUT XV / CAPUT XXXI printed without their full stops at 0052, 0053, 0075, 0114, 0119. Reproduced without them.
• Migne's own [ alias …] variant bracket is reproduced, and no [var:] fires under it — he is carrying the Vulgate reading in his own apparatus (0054 @0321B, in obscuro [ alias caliginoso] loco, 2 Pet. 1:19).
• ⭐ But it is NOT a singleton, and the brief's work-wide check was too narrow. Migne prints a second shape at 0122 @0481B: cor est incibile [ f. lucibile] — f. for forte, his own conjecture on his own non-word. The 2026-08-08 check that declared the class a singleton searched for alias and found what it searched for. A sweep is only as wide as its pattern list — the same lesson 11638's Pattern 3 sweep taught, recurring in a different register. Both instances are reproduced as printed; neither takes a [sic:] of ours stacked on top (brief §10.5).
• Philip's own alia translatio is on the same footing — at 0378C he introduces Ibo mihi ad montem myrrhae as another version. No [var:] fires; the divergence is his apparatus, not Migne's text.
• Migne's lower-case after a misplaced full stop stands, ruled on 11638 and explicitly not to be churned.
• The three mountain etymologies (Amana, Sanir, Hermon) are translated, not carried as Latin — the same call as vermiculatas and for the same reason.
• A quotation broken around inquit/ait keeps its breaks. Pattern 5: 0044 @0297A reads as split gibberish in English («Stay therefore me, she says, up with flowers») — but Migne prints three italic runs and our English mirrors them 1:1. A blind reader proposed the smoother reading and it was declined, because the repair moves sacred marker boundaries. Recorded so the proposal is not made a third time.
• The colour triple and the bed triple hold work-wide (§1), including 0103's se nigras et abjectas.
• Chunk-boundary (cont.) on a repeated head is ours and is merged at assembly — never a defect, never translated.

§6 · MIGNE'S CITATION SIGLA — counted across all twenty-one ranges

⭐ No siglum in this work is systematically wrong, and that is a weaker and far more useful fact than a running error count implies. It could only be established by counting the RIGHT ones, which every range was instructed to do.

• (I Cor. II) was wrong at 0200D, 0240A and twice more — and correct at 0305C and 0318A. (II Cor. IX) correct at 0314C. (Luc. IX) names Matt. 8's words at 0240B and is correct at 0331A. (Psal. XCII) is wrong at 0275B and correct at 0281A, eleven columns later. (Isa. LI) is wrong at 0328C and right at 0332D–0333A. (Isa. LIII) is wrong once at 0421A and correct five times elsewhere. A tally of failures alone would have manufactured a pattern that is not there.

Two real shapes among the scripture sigla:

1. Off by exactly one chapter — five of seven wrong sigla in a single range; a mechanical profile. (Job. II) for Job XI (0372D) belongs to it.
2. The note cites one book, the words are another book's — four attested: Luke 9/Matt. 8 (0240B), Deut. 32/Ps. 91, Isa. 42/Matt. 12 (0318D), and Exod. 3/Acts 7:34 verbatim (0456B, the cleanest of the four).

⭐ The strongest instance is an INTERNAL control and needs no appeal outside the work. At 0283B Migne prints « Vere Filius Dei erat iste » under [n: (Marc. XV)] — verbatim Matt. 27:54, where Mark 15:39 reads Vere hic homo Filius Dei erat. Decisive because the same six words appear at 0267D under [n: (Matth. XXVII)]: Philip quotes Matthew twice and Migne labels the second one Mark.

(II Paral. III) at 0238A, off by thirty-two chapters, belongs to neither shape. A third shape is now named once and once only: at 0412C an (ibid.) carries Psalm 118 onto a quotation of Psalm 31 standing eleven words away — a positional error, not a naming one.

⭐ Migne's INTERNAL cross-references are a separate class with a real profile: 3 right, 2 wrong, and BOTH wrong ones are off by exactly one book. Wrong: (supra l. I, c. 2) at 0437A should be Book II; (l. IV, c. 10-12) at 0445B should be Book V. Right: 0477D and 0479B (both checked against their targets), and one that reaches back 179 columns and lands exactly. ⚑ The correct ones are the harder cases, which is why the class cannot be dismissed as unreliable.

Nothing is corrected anywhere. §10.2 governs: a following Ibid. resolves against the printed sequence at index time, and "fixing" an antecedent makes the Ibid. inherit the corrected key and resolve wrong.

§7 · THE BLIND POLARITY READ — eleven readers, and what it proves

Runbook step 4a, done 2026-08-08. One reading pass over all 126 chunks, split across eleven Opus readers at ~12 chunks each plus a whole-apparatus marker auditor, every one of them fenced off from these twenty-one cruces files, data/briefs/11632.md, next-session-resume.md and git history. Findings were formed blind and reconciled afterwards, never before. The full record is BLIND-READ.md; this is its result.

144 ours-side sites. Repairs applied there, not here.

⭐ The result that matters

One dropped negation in the entire work — 0122 @0481D, where Migne's appareat non deesse shipped as "may appear to be wanting", asserting the exact opposite of a paragraph that exists to show the lamp serves the neighbour. Repaired.

~110 necdum/nondum sites, one flattening — 0070 @0358C, nondum satis appropinquant as "yet do not draw near", in a chunk that renders nondum tamen plene sciunt correctly twelve words later. The particle was grepped first in every range, ahead of non/nec/nisi, precisely because it had been caught flattening during translation.

⚑ What that pair of numbers means. The prose held where it was being watched. Almost every other ours-side site was lexical or syntactic — a -urus participle read as passive, a dative read as a nominative and eight verbs pluralized to agree with it (0046 @0301A, Capreae), postpositive gratia read as the noun (twice, ranges apart), tota turpis as "wholly" instead of "all" inside a four-term scheme built on the one word. The failures clustered in mis-parsed grammar and in Douay wording arriving over the plate — not in polarity. Migne's own defects were rendered as printed and logged, never repaired, in every case the readers checked.

And one whole 46-word sentence was missing (0091 @0404D). Nothing could see it: column anchors matched, [n:] counts matched, and the word ratio was 1.61 — dead centre for that range. The reader's tell was that the Latin carried four question marks and the English three. ⭐ verify-english.mjs now checks question-mark parity as a WARN; it cost one regex and immediately found a second real item at 0051.

The marker audit: 50 [sic:] checked, 45 sound; 102 [var:] checked, 94 sound (101 stand now — one was deleted in repair). The failure that dominated 11638 is absent here: all 102 sat adjacent to their own quotation, none sat on Philip's prose, all four ellipses showed the divergent words. Reference format is uniform — every note carrying a reference uses arabic chapter:verse, against 11638's 71 roman / 59 arabic / 8 none.

⭐ How four unmarked defects were found, which no per-range reader could have done: the auditor built a frequency baseline over all 100 Latin works in src/latin/, isolated 11632 hapaxes whose single-edit neighbour is high-frequency, and read each in context. The rest were genuine rare Latin.

§8 · OPEN — recorded honestly, NOT resolved

These are Wilson's to rule. They are set down with the argument on both sides, because in each case the rulebook is genuinely ambiguous and a merge that decided them silently would be worse than one that does not decide them.

8.1 · ⛔ Pattern 12's boundary — four [sic:] on real Latin words

Pattern 12 says a real word that is merely wrong gets NO marker and is rendered as ordinary English. The marker audit accordingly wants all four removed (and Incipit @0345B reclassified with them).

They are being LEFT as [sic:], on this reading: Pattern 12's stated test is that such a word can be rendered — the rule exists because "is repeated" for repetitur is available and honest. None of these four construes in place. Ama is an imperative in an indicative clause; sternis is a second-person verb where the verse needs an adjective; Pecte and signa leave their sentences without a subject. Rendering them "as ordinary English" would produce English that is not merely odd but ungrammatical, and would hide a broken type behind a word the reader will read past.

⚠ But the translating agents themselves were not consistent about it, and said so: cruces-0060 records that Incipit at 0345B "is a real word, so Pattern 12 keeps it", and cruces-0030 explicitly asked the merge to rule the class. Answering it in five separate edits guarantees an inconsistent result. The inventory is complete; the boundary is Wilson's to draw, and whichever way it goes it should be drawn for all five sites in one commit.

8.2 · ⛔ Supplied em-dashes — ~562 sites, ONE policy decision

Migne prints ZERO em-dashes anywhere in this work. Our English prints 562, an average of 4.5 per chunk. This was measured rather than reported: the 0072–0083 reader counted the whole work and correctly declined to list its own ~40, because doing so would have buried its real findings.

Also measured: English 3,780 colons against Latin 3,667 — printed colons are routinely replaced and occasionally supplied; and English 224 question marks against Latin 206, with five supplied at 0121/0123/0125 and further ones at 0097 and 0107 (all now surfaced by the verifier's parity check).

This is one Pattern 8 policy decision, not 600 edits. Pattern 8 makes the deletion direction the live one — four deleted colons at 0118 are already logged, and one deleted full stop at 0311A. Whether a supplied dash for a printed comma is a Pattern 8 violation or an ordinary translator's licence is exactly the question, and it is not answerable one site at a time. Recorded, measured, undecided.

8.3 · ⛔ [sic: informen t] at 0100 @0428A may mark OUR OWN artifact

The marker reports broken type. But the Latin source file itself reads informen t — the "break" is italics markup in our own text, not Migne's. The plate almost certainly prints informent, an ordinary word. This is our digitization's damage attributed to the plate, which is precisely the error the corpus rules exist to prevent. Needs a plate read at PL 203 col. 428A; deliberately not guessed. If the plate prints informent, the [sic:] comes out and a data/tei-patches/11632.json entry goes in, in the same commit.

8.4 · The split/join asymmetry — six carried, three normalized

Six word-splits are carried as [sic:] — a beo, a vida, hila rescat, in cutiens, informen t (see 8.3), a dolescentulae. But the word-join aforis (0077) is silently normalized, as are persequenter (0046 @0300C) and the lemma-level myrrhoe (0069) — while the parallel lemma defect fovum (0093 @0407A) is carried as [sic:]. Same defect class, two treatments, inside one work. ⚠ verify-english cannot see any of this: it strips asterisks and compares substrings. Not repaired, because it belongs with 8.1 — both are one question about where our carrying threshold sits.

8.5 · Readings argued on both sides, not flipped

Five construals where a blind reader argued rather than asserted, and was right not to assert. Listed so a later editor knows the argument was had: ille mundus at 0240C ("that world" or "that Clean One" — both grammatical, and ad quem nihil immundum attinet makes the jingle audible) · non vere facies coloratur at 0314D (adverb, or the ablative of ver — and the paragraph is built on verna suavitas, so it may be Philip's pun) · factus a Virgine at 0338B (agent or separation) · printed cum for eum at 0351B and 0339A, handled two incompatible ways, with a bracketed [him] supplied at 0067 and a silent reading at the other — one of the two must move · laetitia cordis ejus (his or hers, split between lemma and exposition with no note, which is the part that is indefensible either way).

§9 · MEASURED — the numbers, so nobody re-derives them

The [n:]-only class, reported as a null where it was one. Ranges reported their hit rate on [n:]-only spans even when zero, which is what makes the class measurable: 0/6, 0/20, 0/38, 0/19, 1/12, 1/14, 1/7, 1/4, 1/10 across the ranges that tallied — roughly 3–4 real quotations in ~124 [n:]-only spans. The 0018–0023 agent found all nineteen of its spans to be Philip's own prose and said so. ⛔ That null is exactly why the separate prose pass is mandatory rather than advisory — the real finds (Lam. 4:20, John 19:39, Virgil at 0385A, Deut. 32:33) were in spans the extraction could not list at all.

⚠ The extraction's CLASS LABELS are unreliable in a chunk with a « / » asymmetry. An unclosed « breaks the bracket-matching, so a genuinely guilleted span is listed as [n:]-only — two were, in chunk 0083. The text is still listed and nothing is lost; only the label is wrong.

§10 · OWED

1. Plate read at PL 203 col. 428A — §8.3, informen t. Blocking nothing, but it is the one marker in the work that may be asserting our own damage as Migne's.
2. Wilson's ruling on §8.1 (Pattern 12's boundary) and §8.2 (em-dashes). Both are one decision applied to a class; neither should be settled piecemeal.
3. node scripts/index-work.mjs 11632 after any of the above edits land — rule 9, index from day one, no retroactive passes.
4. node scripts/scan-raw-markers.mjs before deploy. Markers reach the reader as raw brackets unless a builder transforms them, and that has happened twice on PG pages because nothing checks rendered output.
5. The badge. workStatus for 11632 gates the "First English translation" claim; absent a verified none, it renders "New English translation" and that is correct. Do not go hunting to earn the strong claim.

§11 · METHOD DATUM — why the quotation lists were worth their cost

Every agent was handed a verbatim list of every quotation printed in its own range, extracted mechanically from the Latin, with the instruction that the divergence list was the most valuable thing it would produce, and with the requirement that a span collated and found to AGREE be logged as such — so that the absence of a marker is recorded evidence rather than silence.

Result: 101 [var:], plus several hundred spans collated and explicitly logged as unmarked across the twenty-one range files, plus roughly 250 further weaves found on the mandatory prose pass that the extraction could not list at all (22 in one range, 24 in another, 31 in a third, and so on). The blind read then found the apparatus sound: 94 of 102 [var:], zero drift, zero markers on Philip's prose.

⭐ The finding that justifies the whole design is that the mechanical list was not enough and everyone knew which half was missing. Two of this work's three strongest §8a items — Lam. 4:20 and John 19:39 — carry no guillemets, no italics and no [n:], and reached a human only through the prose pass. A review that reads only the extractor's list misses that shape entirely.

⛔ And the defect can be an ADDITION you write while looking straight at it. The 0096–0101 agent drafted Eph. 5:29 at 0430D as "no man ever hated his own flesh", silently restoring the Vulgate's unquam, which Migne does not print — having already written "unquam dropped" into its own collation while the English still said "ever." Knowing the divergence did not prevent reproducing the familiar wording; only the separate word-by-word pass caught it. That is the single most useful sentence in this file for whoever translates the next work.

§12 · RULED 2026-08-09 — the four [sic:] on real Latin words STAY

The marker audit proposed stripping [sic:] from four sites because Pattern 12 reserves the marker for carried non-word type and these are all real Latin words. Ruled the other way: all four keep the marker, and Pattern 12 has been amended to record the test the ruling turns on.

The test is "does it render IN PLACE?", not "is it a Latin word?"

The first three have no grammatical slot at all. Pattern 12's own examples, fuges for sugas and munde for mundo, are not like them: those occupy their slot and merely mean the wrong thing.

⭐ 0048 is the instructive one and it goes the same way for a different reason. ut signa prius ardeat sets a plural noun against singular ardeat. "That signs may burn" is perfectly good English — and that is exactly the problem. English may burn is number-neutral, so rendering signa "signs" silently repairs the disagreement and leaves the reader nothing to see. Where English has no exponent for the defect, rendering it IS the repair, and the marker is the only treatment that does not quietly fix Migne.

⚑ This runs opposite to Pattern 9, which normalizes precisely where English has no exponent — and the difference is worth stating, because a later sweep will meet both. Pattern 9's cases (grammatical gender, section labels) are systemic features of Latin and carry no information about this plate. A number disagreement in one clause is a fact about this printing. Normalize the language; mark the page.

Consequence for the split/join asymmetry recorded above: it is now decidable on the same test rather than by counting precedents. aforis (0077) renders in place as "from without" and takes no marker; persequenter (0046 @0300C) and the lemma-level myrrhoe (0069 @0353D) are non-words and should be marked — both are currently normalized silently, and both remain owed.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range 1 of 21: the Prologue, the Proem, the explanation of the title, and Caput I–IV on Osculetur me osculo oris sui. This range opens the work, so several conventions below are proposals the merge should ratify or overturn, and they are flagged ⚑ where that is so.

Marker totals for the range: 6 [var:] · 3 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on 0000–0005 (no errors); ratio warnings on 0004 (1.64) and 0005 (1.68) — see §7.

§1 · THE QUOTATION COLLATION — all 43 listed spans, itemized

Every span in data/quotations/11632/0000-0005.md was collated word by word against the Vulgate. Spans recorded as AGREES were collated and found to agree; the absence of a marker on them is a finding, not a silence.

Chunk 0000 — 6 spans

1–4. 0181A, four italic spans (Sumite, / materiam vestris, qui scribitis, aequam / Viribus; et versate diu, quid ferre recusent, / Quid valeant humeri.) — not scripture. Horace, Ars poetica 38–40, quoted as verse and set by Migne on three lines; the line structure is preserved in the English. Migne prints humeri where Horace has umeri — an attested medieval orthography, Pattern 9, deliberately unmarked.

1. 0182A « Homo non percipit animalis [n: (I Cor. II)]. » — 1 Cor 2:14. Vulg. Animalis autem homo non percipit ea quae sunt Spiritus Dei. Migne's span is truncated to four words and inverted (homo … animalis), and the object (spiritalia) sits outside the guillemets in Philip's own clause. Every word printed is a Vulgate word and the sense is unchanged. Collated; no [var:] — a quotation cut short to fit a host sentence is citation practice, not a variant reading. Rendered « The animal man does not perceive », with spiritalia outside, exactly as printed.

1. 0182A « Adsum. » — one word, no locator, no [n:]. Nearest received text is Isa. 58:9 Ecce adsum. Not a lemma; rendered « I am here. » No marker.

Chunk 0001 — 1 span

1. 0184A videntes — not a quotation. A word discussed as a word (quos ampliore gratia videntes vel prophetas voluit appellari; cf. 1 Reg. 9:9). Rendered seers, italics kept by function, per the 11638 §1 rule that italics survive by function. It is not carried Latin and takes no [sic:].

Chunk 0002 — 11 spans

8–9. ⭐ 0185C Cantica, inquit, canticorum Salomonis. — [var:] FIRED, and it is the most load-bearing divergence in the range. Vulg. Cant. 1:1 title: Canticum canticorum Salomonis, singular. Migne prints the plural Cantica, and Philip's whole paragraph is an argument from the plural: Canticum vel Cantica in ipso titulo superscripsit … hic non solum Cantica, sed Cantica posuit canticorum. Conforming the lemma to the singular voids the paragraph that expounds it. Rendered "Songs, he says, of songs of Solomon." ⚑ And Migne is inconsistent with himself on the same page: the chunk's own ## head prints the singular genitive TITULI CANTICI CANTICORUM EXPLICATIO immediately before quoting the plural title. Both stand as printed — the head renders "THE SONG OF SONGS", the lemma "Songs of songs". Not harmonized in either direction.

1. 0185C pax — etymological gloss of Salomon (Salomonis nomine pax signatur), not scripture. Italic word-as-word, translated ("peace").

1. 0185D adversarius — etymological gloss of Satan, not scripture. Translated ("adversary").

1. 0185D « Habebat, inquit, pacem ex omni parte in circuitu, a Dan usque Bersabee [0186B] [n: (III Reg. IV)]. » — collated; AGREES, but it is a conflation of two verses. 3 Reg. 4:24 has habebatque pacem ex omni parte in circuitu; a Dan usque Bersabee is from 4:25. Each half agrees with the Vulgate word for word; the enclitic -que is dropped because the quotation opens there. No [var:] — a conflated citation is not a divergent reading. Note that the column anchor [0186B] falls inside the guillemets, between the last word and the [n:]; reproduced in place.

1. 0186B « Requiem, ait, dedit mihi Deus meus per circuitum, et non est Satan neque occursus malus [n: (III Reg. V)]. » — [var:] FIRED. Vulg. 3 Reg. 5:4: Nunc autem requiem dedit Dominus Deus meus mihi per circuitum, et non est satan, neque occursus malus. Migne drops Dominus and moves mihi before Deus meus. The second half agrees exactly, negations included (non est … neque), and both are carried in the English.

1. 0186C pacificus — etymological gloss of Salomon, not scripture. Translated ("the peaceable").

15–16. 0186C Parabolae, inquit, Salomonis filii David [0186D] regis Israel. — collated; AGREES with Prov. 1:1 word for word. Rendered per Pattern 2 as a title (The proverbs … of Solomon the son of David, king of Israel). The anchor [0186D] falls inside the italic run; reproduced in place.

17–18. 0187B Verba, inquit, Ecclesiastae filii David regis Hierusalem. — collated; AGREES with Eccl. 1:1. Hierusalem for Jerusalem is orthographic (Pattern 9), unmarked.

Chunk 0003 — 13 spans

19–20, and its recurrences. Cant. 1:1 Osculetur me osculo oris sui — collated; AGREES exactly, at every one of its SEVEN occurrences in this range (0003 ×3, 0004 ×3, 0005 ×1). Rendered uniformly "Let him kiss me with the kiss of his mouth." ⚑ This is the running lemma of the entire commentary and the merge should hold every agent to one English form of it.

• At 0190B Philip re-quotes it with his own connective inserted — Osculetur igitur me osculo oris sui — rendered "Let him therefore kiss me…". The igitur is his syntax, not a variant reading: no marker.
• At 0192B and 0194A the lemma is broken around inquam (Osculetur me, inquam, osculo oris sui). Pattern 5 applies to the italic run as it does to guillemets: the interruption is translated in place and the run is not re-punctuated.

21–23. 0190C–0191A, the three-stage « Ave, inquit, gratia plena » → « …Dominus tecum » → « …benedicta tu in mulieribus [n: (Luc. I)] » — collated; all three AGREE with Luke 1:28. Each stage is a true prefix of the Vulgate verse; the staged truncation is Philip's rhetorical device (the angel greets, then adds, then adds again) and is not a textual variant. No markers.

1. 0191A Luke 1:29 Quae cum audisset … qualis esset ista salutatio — collated; AGREES.
2. 0191B Luke 1:30 Ne timeas, Maria, invenisti enim gratiam apud Deum — collated; AGREES.
3. 0191B Luke 1:31 Ecce concipies in utero, et paries filium, et vocabis nomen ejus Jesum — collated; AGREES.
4. 0191C Luke 1:34 Quomodo fiet istud, quoniam virum non cognosco? — collated; AGREES. Note that here alone the [n: (ibid.)] sits outside the closing guillemet; position reproduced.
5. 0191D Luke 1:35 Spiritus sanctus superveniet in te, et virtus Altissimi obumbrabit tibi — collated; AGREES.
6. 0191D Luke 1:38 Ecce ancilla Domini, fiat mihi secundum verbum tuum — collated; AGREES.

Chunk 0004 — 9 spans

1. Cant. 1:1 — agrees (see 19–20).

1. ⭐ 0193B « Prophetam, inquit, suscitabit Deus de fratribus vestris similem mei; ipsum audietis [n: (Deut. XVIII)]. » — [var:] FIRED, four divergences at once. Vulg. Deut. 18:15: Prophetam de gente tua et de fratribus tuis sicut me suscitabit tibi Dominus Deus tuus: ipsum audies. Migne prints the plural de fratribus vestris … ipsum audietis, and similem mei for sicut me — i.e. the form in which the verse is quoted at Act. 3:22 (and 7:37), not the form of Deuteronomy, though Migne's own [n:] cites Deuteronomy. The plural is visible in the English ("from among your brethren … him shall you hear") under Pattern 17, which is exactly why the marker matters: a Pattern-17 reading of the singular Vulgate would have made Moses address one man.

1. 0193B « Ipse, inquit, tanquam Sponsus procedens de thalamo suo, exsultavit ut gigas [n: (Psal. XVIII)]. » — collated; AGREES with Ps. 18:6. Vulg. et ipse tamquam sponsus procedens de thalamo suo; exsultavit ut gigas ad currendam viam. The leading et is dropped at the quotation opening and the verse is cut after gigas; tanquam/tamquam is orthographic. No [var:]. Migne capitalizes Sponsus where the Vulgate has lower case; capitalization mirrored, see §4.

1. 0193C « Os, inquit, Domini [0193C] locutum est haec [n: (Isa. I)]; » — [var:] FIRED. Vulg. Isa. 1:20 (and 40:5, 58:14) read quia os Domini locutum est with no object. Migne adds haec. The addition is small but it is a word that is not in the received text, and Philip's sentence turns on Isaiah speaking these announced things. The column anchor falls inside the quotation and the semicolon inside the closing guillemet; both reproduced (Pattern 8).

1. ⭐ 0193C « Ecce, inquit, virgo concipiet et pariet filium, et vocabitis nomen ejus Emmanuel [n: (Isa. VII)]. » — [var:] FIRED. Vulg. Isa. 7:14: et vocabitur nomen ejus Emmanuel — passive, "his name shall be called". Migne prints vocabitis, second person plural active, "you shall call". Rendered as printed. This is the range's cleanest 7a″ trap: the familiar English of Isa. 7:14 is passive, arrives before the Latin is read, and would have erased both the person and the number.

1. 0193C « Faciet, inquit, Dominus novum super terram, et femina circumdabit virum [n: (Jer. XXXI)]. » — [var:] FIRED. Vulg. Jer. 31:22: quia creavit Dominus novum super terram: femina circumdabit virum. Migne prints Faciet — future for perfect, "shall make" for "hath created" — and supplies et before femina. The tense change is material in a chapter arguing that the thing is still to come.

1. 0193D « Fiat mihi, ait, secundum verbum tuum [n: (Luc. I)]. » — collated; AGREES with Luke 1:38 (the second half of the verse quoted at 0191D).

37–38. Cant. 1:1 at 0194A — agrees.

Chunk 0005 — 3 spans

1. Cant. 1:1 at the head of CAPUT IV — agrees.

1. 0195C « Inventa est, inquit, habens in utero de Spiritu sancto. » — collated; word order diverges, and [var:] was DELIBERATELY NOT FIRED. Vulg. Matt. 1:18: inventa est in utero habens de Spiritu Sancto. Migne prints habens in utero. Every word is the Vulgate's, in the same clause, and the sense is identical; Pattern 14 fires on divergence "in a way that changes the sense," and this does not. Recorded here so the absence of the marker is a decision on the record rather than an oversight. A later reviewer who fires one here should say so and not treat it as a fresh discovery.

1. 0195C « Quod enim in ea natum est, de Spiritu sancto est [n: (ibid.)]. » — collated; AGREES with Matt. 1:20 word for word.

Unguilleted, unitalicised citations also collated (not on the list)

The extraction lists only marked spans; these are woven into Philip's prose with only an [n:], and were checked because the argument leans on them.

• 0187A, quia vir obediens debet magis loqui victorias quam tacere [n: (Prov. XXI)] — Prov. 21:28 reads vir obediens loquetur victoriam (singular, and without debet magis … quam tacere). This is Philip's paraphrase inside his own sentence, not a marked quotation; rendered as printed, no marker (a [var:] on unmarked prose would misrepresent it as a lemma).
• 0186B, [n: (Ephes. V)] on angularem lapidem se objecit. The corner-stone is Eph. 2:20 and utraque unum fecit just before it is Eph. 2:14; Migne's note says chapter V. Reproduced verbatim, no correction proposed — brief §10.2. Logged only because a later agent will meet it again.
• 0188A, [n: (Gen. XXVIII)] / ipsum, ut ait, Deum videt facie ad faciem et salvatur — the face-to-face saying is Gen. 32:30, not the Bethel ladder of Gen. 28; Philip runs them together in his own prose. No marker.
• 0194D, Gen. 2:7 vitae spiraculum … in ejus faciem inspiravit — Vulg. inspiravit in faciem ejus spiraculum vitae. Philip's own word order in his own sentence, unmarked as a quotation. No marker.

§2 · THE TYPOGRAPHY QUESTION — 11632 does NOT open in an italics-only regime

⭐ Brief §7 asks where 11632's seams fall. In this range there is no seam, because both marks are in use from the first columns, and the division between them is FUNCTIONAL, not sequential. Counted across 0000–0005: 21 guillemeted spans, 22 italicised — and they never compete for the same job.

• « » marks scripture quoted inside the running prose. All 21 guillemeted spans in this range are scripture; not one is anything else.
• Italics mark four other things: (a) the running Song lemma being expounded (Osculetur me osculo oris sui, 7×); (b) book titles quoted as titles (Cantica canticorum Salomonis, Parabolae Salomonis…, Verba Ecclesiastae…); (c) words discussed as words (videntes, pax, adversarius, pacificus); (d) classical verse (the Horace) and Migne's marginal source-edition numbers.

This is a materially different opening from 11638, whose Books I–II italicise scripture and switch to guillemets only at INCIPIT TERTIUS TOMUS. Anyone reconciling the two works should not expect 11632 to repeat that history, and an italic run in a later 11632 chunk is not by itself evidence of a pre-seam regime — it may simply be a lemma. Under Pattern 6 nothing was tidied in either direction; guillemets are 1:1 with the Latin twin in all six chunks (verified mechanically: 0/0, 2/2, 2/2, 9/9, 6/6, 2/2 pairs).

No guillemet asymmetry occurs in this range — every « has its », and none is stray or unopened.

Migne's marginal source-edition numbers 101–110 run through the range (101 in the 0000 head, then 102 0182B, 103 0183D, 104 0185B, 105 0187A, 106 0188D, 107 0190C, 108 0192B, 109 0194A, 110 0195D). Kept italic, verbatim, at the exact word they interrupt — 108 in particular falls between excellentiam and non contaminans, and the English clause was built around it rather than moving it. Same class as 11638's 303–313.

§3 · PATTERN 7 — printed readings rendered as printed, conjectures kept here

Three [sic:] fired. Each is a non-word, and none of the three is also translated in the running text (brief §8's named failure — a [sic:] on a word that was ALSO Englished passes every check and reports nothing).

1. 0185B [sic: molodia] — non vocis molodia, sed morum consonantia. Read melodia, near-certain: the phrase is a melody/concord antithesis and molodia is not a word. Carried untranslated in italics.
2. 0186C [sic: tilulo] — insuper et in tilulo regem regis filium se appellat. Read titulo, certain: the same sentence-run has in ipso titulo at 0185C and in hujus quoque libri titulo at 0187A. Carried.
3. 0191C [sic: eumdum] — ipse tuus uterus eumdum Sponsum hominem generabit. Read eumdem, certain. ⚠ The English deliberately reads "thy very womb shall beget [sic: eumdum] Bridegroom as man" and does not also say "that same": supplying the English demonstrative alongside the marker would have translated the defect and quarantined it at the same time.

Rendered as printed, no marker (real words, merely wrong or merely odd):

• 0181B narratoria — Quod si virium consideratio narratoria est in carminibus aethnicorum. Rendered "is a narratorial thing". The word is formable from narrator/narratio and construes as a predicate adjective, so Pattern 12's boundary puts it on the English side; but the sense wanted is plainly "necessary/required" and necessaria is the obvious conjecture. The English is ours and reads strangely on purpose. ⚑ Second opinion welcome at merge; do not silently smooth it to "necessary".
• 0182A, the misplaced semicolon. Migne prints serius quisque [0182A] videatur; non tam nova cudere, quam furari potius praeinventa — a semicolon cutting videatur from its own infinitives. Pattern 8: preserved, so the English reads "…would seem; not so much to strike out new things…". It is meant to look wrong.
• 0189B longe suavis obdormisse — the construction requires the comparative adverb suavius (longe suavius, "far more sweetly"), and suavis is the nominative adjective, which cannot govern obdormisse. Under 7a the printed word may not be read as its comparative to rescue the clause, so the English reads "slept far sweet". ⚠ This is the sentence in the range most likely to be "fixed" by a later reader. The conjecture is suavius and it lives here, not in the text.
• 0195A laborem fingere valeam in praeceptum — rendered literally, "may be able to shape labour into a precept". Construable but strained; no emendation offered and none made.

Negation audit (brief §9). Every printed non/nec/neque/nisi/nullus/ nihil/nunquam in the six chunks was carried, including the ones that make the English clumsy: nec debeam, nec praevaleam (0189A), nullus penitus adnotatur, qui non … praeloquatur (0189D), nisi suo soli Domino non concessum (0190C, rendered "not granted save to her Lord alone" — the Latin's redundant negative is preserved in the "not … save"), non defectum, profectum vero (0194D), and the double non praestabit … non negabit at 0182B. No negation was supplied that the Latin does not print.

§4 · CONVENTIONS SET BY THIS RANGE — ⚑ for the merge to ratify

Brief §5's ruled vocabulary was applied as given. These are the new decisions the opening forced, which §5 does not cover.

1. ⚑ liber = "book", caput = "chapter", volumen = "volume" — and 11632 has no tomi. The work divides into libri (## LIBER PRIMUS. → BOOK ONE.) and capita (## CAPUT PRIMUM. → CHAPTER ONE.). Note that 11638's tomus/liber problem does not arise here in the same shape: at 0186C–0188B Philip uses liber and volumen interchangeably for Solomon's three books (In tertio volumine … liber ille primus), and both are rendered by their own English word, never merged.
2. ⚑ Migne's own numbering inconsistency in the heads stands. He prints CAPUT PRIMUM in words, then CAPUT II, CAPUT III, CAPUT IV in numerals. The English mirrors it: CHAPTER ONE. then CHAPTER II. etc. Pattern 9 normalizes a section label only where English has no exponent for the defect; word-vs-numeral has one, so nothing was normalized. Do not tidy this to a uniform series.
3. INCIPIT = "HERE BEGINS" (brief §5), applied at the 0000 head.
4. ⚑ PROOEMIUM IN CANTICA. → "PROEM ON THE SONGS." — plural, because Cantica is the plural Philip argues for at 0185B–C. The number of Cantica / Canticum / Cantici is tracked per occurrence throughout, never harmonized; see §1 items 8–9.
5. (cont.) on the repeated CAPUT PRIMUM head is untranslated and carried verbatim (brief §10.7).
6. [CANT. I.] at the head of 0003 is Migne's own bracketed marginal locator. Reproduced verbatim, untranslated, in place.
7. ⚑ Migne's mixed capitalization of Sponsus/sponsus, Sponsa/sponsa, Virgo/virgo is MIRRORED, not regularized. He runs both within single sentences (0193D advenientis sponsi… beside 0193C adventu Sponsi; 0194A sponsam matrem Virginem). The English capitalizes exactly where he does. This is the most likely drift point across 21 agents and is the one convention here that a mechanical sweep could check.
8. Pattern 17 applied throughout, and it does real work in this range. The Annunciation dialogue is singular on both sides — thou/thee/thy for Gabriel to Mary and Mary to Gabriel, including Osculare me osculo oris tui at 0194A — while Philip's own frame is plural: postulatis/deposcatis/ vestris precibus (0182A–B), vos veteres nostis (0182B), Vos igitur in hoc auditorio … qui astatis (0195A–B), sicut nostis and vos, fratres, nostis (0195D, 0196B) all take you. The payoff case is item 31 above: Deut. 18:15 printed in the plural, rendered "him shall you hear", visibly different from the singular the Vulgate would have given.
9. Pattern 3. Scripturis meaning the canon occurs twice in the range (0184A vera legerent in Scripturis; 0190D scripturis inhaerentem) and is rendered the Scriptures both times. ⚑ Neither instance uses divina or sacra — which is precisely the form 11638's sweep pattern missed — so a later sweep of 11632 must match the bare Scriptura(e) too.
10. Register vocabulary held to brief §5 where it appears: clemens = "clement" (0192A) kept distinct from misericors = "merciful" in the same sentence; adolescentulae = maidens (0188B, 0188C).

§5 · THINGS DELIBERATELY NOT REPORTED (brief §10 classes, checked once)

• 0002's column anchors run [0188A] [0188B] [0188C] and then [0187D]. This is brief §10.1's band-arithmetic false-positive class exactly. Checked: no text is broken mid-word or mid-construction across any anchor in this range, and the 0187D anchor closes the chunk after two empty heads. No [ed:], no patch, no plate check. Recorded only so the next agent does not spend a session on it.
• noteCount matches the body [n:] count in all six chunks (1/1/7/18/8/3), verified mechanically.
• Attested medieval spellings left alone, unmarked, per Pattern 9: aethnicorum, tanquam, Hierusalem, humeri, spiritalis/spiritales beside spiritale, coelestia.
• No [f:]. 11632 is lemma-and-gloss, not a florilegium (brief §10.4).

§6 · OPEN — for the merge or a later plate read

1. ⚑ The Cantica / Canticum number problem is work-wide, and this range only sees its first page. The plural title at 0185C is argued for; the head at the same column is singular; the running work-title in 0181B is the singular ablative Cantico canticorum. Every later agent will meet the same pair. The merge should collect the numbers rather than impose one, and should check whether any later agent silently conformed the 0185C lemma.
2. The [n: (Ephes. V)] at 0186B and the Gen. 28 / Gen. 32 run-together at 0188A are Migne's references, reproduced. Neither has a following Ibid. depending on it, so both are safe to correct upstream if a citation-corrections entry is ever wanted — but that is not this file's business and no correction was proposed.
3. ⛔ The blind polarity read has not been done and cannot be done by me. Findings for it must be formed blind and reconciled against this file afterwards (runbook 4a). The three passages most likely to be wrongly repaired are named in §3: narratoria (0181B), the semicolon (0182A), and especially longe suavis (0189B).

§7 · RATIO NOTE

EN/LA by chunk (verifier's own figures): 0000 1.55 · 0001 1.55 · 0002 1.52 · 0003 1.56 · 0004 1.64 · 0005 1.68. The verifier warns on the last two. The cause is genre, not padding: Philip writes rhymed, heavily parallel prosa rimata, and the last two chunks are its densest stretch (purificet, magnificet, laetificet; me deliniat, me reficiat, me serenet; utilis, fertilis, et jucunda), where a Latin participle or a syncopated ablative absolute needs a finite English clause each time. 11638 ran 1.35–1.63 on the same author. No clause was expanded that the Latin does not carry, and nothing was summarized.

Philip of Harvengt, Commentaria in Cantica canticorum, Tome I, Capp. V–XI: the exposition of Quia meliora sunt ubera tua vino (Capp. V–VIII, with VIII running over the 0009 chunk boundary) and of Oleum effusum nomen tuum (Capp. IX–XI).

Marker totals for this range: 15 [var:] · 1 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six chunks (one ratio warning, 0009 at 1.61, inside 11638's shipped 1.35–1.63 band). Guillemet, column-anchor and [n:] parity confirmed 1:1 against the Latin twin for every chunk, including the two asymmetries recorded in §3.

§1 · THE QUOTATION COLLATION — all 64 listed spans

Every span in data/quotations/11632/0006-0011.md was collated word by word against the Vulgate. The spans recorded as AGREEING are the point of this section: an unmarked quotation in these six chunks has been checked, not skipped.

1a. Divergent — marked [var:] (15)

1b. Collated and AGREEING with the Vulgate — no marker (24)

Recorded so that the absence of a [var:] is evidence.

• 0006 — Cant. 1:1 Quia meliora sunt ubera tua vino (the second, corrected printing of the lemma) · Deut. 32:33 Fel draconum vinum eorum, et venenum aspidum insanabile · 3 Reg. 18:21 Accedens… Elias ad omnem populum, ait: Usquequo claudicatis in duas partes? Si Dominus est Deus, sequimini eum; si autem Baal, sequimini eum (agrees but for the second eum where Vulg. has illum — no English exponent, so logged here and not marked) · Gal. 2:11 Cum venisset Cephas Antiochiam, in faciem ei restiti, quia reprehensibilis erat (Vulg. autem displaced by the interrupting inquit) · Gal. 2:14 (agrees; gentiliter et non judaice vivis transposes Vulg. gentiliter vivis, et non iudaice, same sense, and Migne prints no terminal ? — the English follows him, Pattern 8).
• 0007 — Cant. 1:1 (×3, the running lemma) · John 1:14 Verbum caro factum est · Ps. 84:11 Misericordia et veritas obviaverunt sibi · Ps. 83:12 Quoniam misericordiam et veritatem diligit Deus, gratiam et gloriam dabit Dominus (Vulg. quia for Quoniam; no English exponent).
• 0008 — Cant. 1:1–2 …fragrantia unguentis optimis · Ps. 84:11 Justitia et pax osculatae sunt · Gen. 27:27 Ecce odor filii mei sicut odor agri pleni, cui benedixit Dominus (verbatim; but see §4 on the printed locator) · Heb. 4:12 (verbatim from vivus et efficax on — the opening Vivus est enim sermo Dei is displaced by Philip's own frame Est enim odor ille, so nothing is divergent, only truncated).
• 0009 — Ps. 26:8 Exquisivit te… facies mea, faciem tuam, Domine, requiram · 2 Tim. 4:7–8 Cursum consummavi, reposita est mihi corona justitiae (abridged across fidem servavi. In reliquo; the words printed agree) · Rom. 8:24 Spe salvi facti sumus (Vulg. Spe enim).
• 0010 — Cant. 1:2 Oleum effusum nomen tuum · Luke 1:31 Ecce concipies in utero et paries filium, et vocabis nomen ejus Jesum · Matt. 1:21 ipse enim salvum faciet populum suum (truncated before a peccatis eorum) · Titus 3:5 Non ex operibus justitiae quae fecimus nos, sed secundum suam misericordiam salvos nos fecit · Ps. 142:11 Propter nomen tuum vivificabis me (Vulg. Domine displaced by the interrupting inquit) · Ps. 142:12 In misericordia tua disperdes inimicos meos · Ps. 24:11 Propter nomen tuum, Domine, propitiaberis peccato meo · Ps. 5:12 Gloriabuntur… in te omnes qui diligunt nomen tuum (Migne adds the vocative Domine — an addition that asserts nothing new, logged not marked) · Isa. 48:9 Propter nomen meum longe faciam furorem meum · Ps. 66:3 Ut cognoscamus in terra viam tuam, in omnibus gentibus salutare tuum · Ps. 143:5 Domine, inclina coelos tuos et descende; tange montes et fumigabunt.
• 0011 — Cant. 1:2 (×2) · Matt. 16:16–17 (agrees; abridged — Tu es Christus stops before Filius Dei vivi, which is what Philip's argument needs, and Beatus… Simon Bar-Jona drops the copula es) · Ps. 96:7–8 Adorate eum, omnes angeli ejus; audivit et laetata est Sion · Ps. 71:17 Ante solem permanet nomen ejus and Et benedicentur in ipso omnes tribus terrae · Luke 1:68 Benedictus Dominus Deus Israel · Rom. 9:5 Qui est super omnia Deus benedictus in saecula · Ps. 21:15 Sicut aqua effusus sum; et dispersa sunt omnia ossa mea · Isa. 45:15 Vere tu es Deus absconditus. Deus Israel Salvator (agrees; Migne's full stop and note fall inside the verse — see §3).

1c. Listed spans that are NOT scripture (and were not collated)

• 0197B (Job II) — an italicised printed locator, not a quotation.
• 0198A occurrit digna relatu / Pompa; senem potum pota trahebat anus — an elegiac couplet from quidam poeta, Philip's own citation. Rendered as verse; the italic run structure is mirrored 1:1.
• 0201A Qui properant, / nova musta bibant; mihi fundat avitum / Testa merum. — likewise a classical hexameter fragment, cited as E quibus unus. Three italic runs in, three out.
• 0203A (Psal. [0203B] LXXXIV) — a locator, italic, printed inside the guillemets and containing a column anchor. Reproduced exactly.
• 0207D Salutaris · Salvator · eleos · misericordia, 0209C unctus, 0212A Oleum effusum nomen tuum, / Ideo, — words displayed as words and the closing lemma. See §5.
• The Job III speech at 0197B–C (Quare, inquit Job, non in vulva miser ego mortuus exspiravi…) is Philip's paraphrase, not a quotation; the extractor did not list it because Migne prints no opening «. See §3.

1d. Scripture woven into the prose unquoted (no marker; recorded for the index)

Not in the extracted list, since Migne marks none of it, but each is an adapted verse and each was checked so that no conformation crept in: Ps. 77:47 (percussit… vineas iniquorum), Deut. 32:32, Jer. 35, Exod. 21:24 / Matt. 5:39, Jas. 2:13, Job 28, 1 Cor. 8:1 (cited I Cor. II), Isa. 35:3, Phil. 3:13, 1 Cor. 9:27, Heb. 12:6 (quem Pater diligit Filium — Vulg. quem enim diligit Dominus, castigat), Ps. 83:3, 1 Cor. 5:7–8, Luke 10:34, John 9:6, Ps. 6:7–8, Rom. 10:12, Isa. 10:27, Ps. 44:8 (unxit eum Deus… prae consortibus suis for Vulg. unxit te Deus Deus tuus… prae consortibus tuis, third person for second), John 1:9 (word order), Ps. 18:5, Ps. 75:2, Ps. 59:10, Ps. 111:9 + Ps. 77:25, 1 Pet. 1:12, Rom. 11:29 (see §4).

§2 · THE THREE FINDINGS THE MERGE SHOULD CARRY

⭐⭐ Ps. 82:19 — nomen tibi DEUS, not Dominus (0011 @0211B). This is the proof-text of Caput XI, whose whole thesis is announced in its first line: Nomen Sponsi, Deus est. The Vulgate reads Et cognoscant quia nomen tibi Dominus. Conforming the lemma would leave the chapter arguing that the Bridegroom's name is God from a verse that says it is Lord — the exact 7a″ failure. If another range meets Ps. 82:19, expect Deus and do not conform it.

⭐ Isa. 53:4 — infirmitates nostras, not dolores nostros (0008 @0204B). Migne's reading is the Matt. 8:17 form. It is adduced to prove why the Bride says ointments, plural — languores multos, multas infirmitates nostra sibi miseria comparavit — and the next 200 words run on quae infirmitas, quo unguento indigeat. The Vulgate's dolores would leave the plural argument without its keyword.

⭐ The head lemma of Caput V prints ubera SUA (0006, col. 0196C). The Vulgate, the frontmatter incipit, and every other printing of the lemma in this range (0007 ×3, 0008, and the re-quotation inside Caput V itself) read ubera tua. The commentary that follows the defective printing expounds the second person outright — longe esse meliora ubera tua vino — so the divergence is confined to the lemma line. Rendered as printed ("his breasts") with a [var:]. Worth watching across the other twenty ranges: if the running lemma is printed sua again elsewhere it is Philip's text; if this is the only instance it is the compositor's, and this range holds the datum.

§3 · TYPOGRAPHY — the seam question, answered for this range

11632 does not divide its two quotation marks sequentially the way 11638 does. It divides them by FUNCTION, and does so from Tome I onward. Across cols 0196D–0212A both marks are in continuous simultaneous use:

• Guillemets « » carry every quotation Philip adduces as a witness — scripture, and only scripture, in all 45 runs in this range.
• Italics carry (a) the running Song lemma at the head of every Caput and at its recapitulations, (b) classical verse (0198A, 0201A), (c) words displayed as words (0207D, 0209C), (d) printed locators — (Job II) at 0197B and (Psal. LXXXIV) at 0203A, the latter inside a guillemet run, and (e) Migne's marginal source-edition page numbers 111–119.

So an italic run in this work is not evidence of a pre-seam book, and a range that meets both marks on the same column has not crossed a seam. I crossed no Explicit/Incipit — chunks 0006–0011 are continuous inside Tome I, Capp. V–XI — and I report no seam. Whether a seam exists later is still open; this range shows only that the starting practice already uses both marks.

Guillemet asymmetries, mirrored and not balanced (Pattern 5 corollary):

1. Unopened » at 0197C. Philip's paraphrase of Job 3 (Quare, inquit Job, non in vulva miser ego mortuus exspiravi…) is closed by a » that was never opened: …uberibus sum lactatus [n: (Job III)]? ». Reproduced as printed. The chunk's counts are therefore 8 « against 9 » and this is correct.
2. Unclosed « at 0211D–0212A. « Vere, inquit, tu es Deus absconditus [n: (Isa. XLV)]. Deus Israel Salvator. opens and never closes; the run swallows the whole of Philip's closing exposition to the end of the chunk. Reproduced as printed — 11 « against 10 ». This is also why the extractor listed only 14 spans for 0011: the Isa. 45 quotation is invisible to it.

Two guillemet runs swallow Philip's own commentary (the 11638 @0569A phenomenon, twice in this range). At 0207D–0208A a single « … » encloses Joel 2:32, Ps. 117, the eleos/misericordia etymology, the whole argument that our salvation is of mercy and not of justice, and Titus 3:5 — 110 words of which perhaps 25 are quoted. At 0211C a run opening at Luke 1:68 encloses et benedictum nomen majestatis ejus (which is Ps. 71:19, under a (Luc. I) note) and then Philip's own gloss Nomen humilitatis ejus, Homo est; nomen vero majestatis, Deus est. Both reproduced 1:1. Likewise 0210B, where At ille sits inside the Matt. 16 quotation.

§4 · MIGNE'S REFERENCES — printed as printed, not corrected

Recorded, not acted on (refDisplay keeps what he prints; §10.2 of the brief forbids touching an antecedent a following Ibid. depends on — and both of these have one).

• 0204A [n: (Gen. XVII)] on Ecce odor filii mei — the verse is Gen. 27:27. Reproduced verbatim.
• 0211C [n: (Luc. I)] covers Benedictus Dominus Deus Israel (Luke 1:68) and, under the following [n: (ibid.)], et benedictum nomen majestatis ejus, which is Ps. 71:19. The ibid. chain is therefore anchored to a citation that is right for the first half of its run and wrong for the second. Untouched.
• 0200D [n: (I Cor. II)] on scientia non sanare, sed inflare — the sentence is 1 Cor. 8:1. Untouched.

§5 · RANGE CONVENTIONS (offered to the merge, not settled unilaterally)

Words displayed as words: the Latin is translated, the italics are kept. Four fall in this range, all in the same etymological construction that §5 of the brief ruled on for candidatio:

• 0207D Jesus, Latina lingua Salutaris dicitur, vel Salvator → "Saving, or Saviour". The argument runs Jesus → Saviour → ipse enim salvum faciet populum suum, and Latin left standing would break it.
• 0209C Christus autem unctus dicitur → "anointed". The whole of Caput X runs anointed → oil → ipse unctus et ungens.
• 0207D eleos autem Graece, Latine misericordia est → the Greek* word is kept (eleos; it is named as Greek and is the foreign term), the Latin gloss is Englished: "eleos in Greek is, in Latin, mercy*". Two italic runs in, two out.

This is the same call 11638 §2.3 made and I have made it the same way; it is worth a line in the merge because 11638's own §1 lists Eleos and misericordia among its surviving italics without saying whether they were Englished.

Pattern 17. The range is almost entirely third person. Every second-person form is inside a quotation and every one is singular in the Latin (ubera tua, tu erudieris, faciem tuam, nomen tuum, inclina coelos tuos, tu es Deus absconditus) — so thou throughout. The two plurals are Philip addressing his readers (Nunquid non vobis… videbatur 0198B; De quo legistis 0197C; ut legistis 0199D) and the Matt. 23 Vae vobis, all rendered you. No mixed-number sentence occurs in this range.

Ruled vocabulary held throughout (brief §5): misericordia = mercy (~40 instances, never "pity" where the noun is used) kept distinct from clementia (does not occur here); charitas = charity (0206D a charitate Christi; 0211D ejus charitas se diffudit) — never "love", which is reserved for amor/dilectio, both of which also occur (0205C inducit… amorem, 0210B illius dilectio); Sponsus = Bridegroom, Sponsa = Bride; unguentum = ointment (never "perfume"), species = spices, confectio = confection/compounding.

§6 · PATTERN 7 — printed readings rendered as printed, conjectures kept here

1. 0202C, Pecte igitur virgo ubera sponsi vino dicit meliora. [sic: Pecte] — carried untranslated in italics. Pecte is a real Latin word ("comb!", imperative of pecto) but cannot fill the adverb slot the sentence gives it, and the conjecture is certain: Recte, which is exactly how the parallel sentence opens the same argument at 0196C (Recte, inquit, adventum tuum… praeoptavi) and at 0199C (Recte ergo Jesus dicitur, 0207D). I judged this over the line Pattern 12 draws between "a real word merely wrong" (render it) and carried type: rendering "Comb" would put a considered-looking reading on the page that misrepresents the plate more than the carried run does. This is the boundary Pattern 12 itself flags as worth watching (the 11081 @0030B case); if the merge prefers the other side of it, the fix is one marker.

2. 0204D–0205A, coepit tuam statim turpitudinem abhorrere. Rendered as printed — "began at once to abhor thy foulness" — inside a paragraph that is otherwise wholly third person and reflexive: the next clause is suae merito foeditatis ipsa sibi vehementius displicere, and the one after suo ipsius odio. The conjecture is suam. Rendered literally, not repaired; a real word wrongly inflected takes no marker under Pattern 12.

3. 0203D, quae Deus, honorum semper amans, legitur odoratus. Rendered "ever loving honours". The conjecture is bonorum ("ever loving good things"), which the context — God smelling the odour of sacrifices — plainly wants. A real word, so rendered, not carried.

4. 0206C, unitus patientia, certam velut fiducialiter et humiliter jam praesumit. certam is a feminine accusative with no noun anywhere in the sentence to govern it. Rendered as closely as the words allow ("presumes upon a sure thing, as it were"); the likely loss is certam victoriam or certam salutem (the latter is what the parallel at 0207A supplies — quasi datam teneat jam salutem). Not supplied in the text.

5. 0206B, ejus labores et agones fingit gratia misericorditer in praecepto. fingit rendered "fashions"; figit is possible and would give "fixes… in a precept". Both construe; no repair made, no marker.

6. 0207B, non apud se, ad modum invidorum; clauso spiraculo retentare. Migne's semicolon falls inside the phrase, between invidorum and its participial clause. Preserved (Pattern 8), which leaves the English punctuated oddly at the same point.

7. 0202A, Gal. 2:14 ends with a full stop where the sense is a question (quomodo gentes cogis judaizare.) — 0006 @0199A. The Vulgate prints ?. Preserved as a statement per Pattern 8; not supplied.

Negation audit. Every non / nec / neque / nisi / nunquam / nullus / nihil in the six Latin chunks was matched against the English on a second pass, with particular attention to the dense clusters at 0196D (non sponsorum… non lactari), 0198D–0199A (Nolebat… non solum… quod fas non est), 0208D (noluit misereri… non avertit… non convertit), 0209D (non principio, non fine, non loco, non tempore) and 0210A–D (non prius… non reddiderit se immundum… non Judaeo tantum, non Graeco… nec est ab aliquibus… interclusum). No negation was dropped and none was supplied. The one place where carrying a negation makes the English read strangely is 0210C — remotiores quoque ab eadem exclusi gloria non habentur — and it is carried.

§7 · WHAT A LATER BLIND READ SHOULD LOOK AT FIRST

1. 0207D's 110-word guillemet run. If any repair has hidden anywhere in this range it is there: Migne's quotation marks enclose Philip's own prose, which means the reader cannot tell quotation from commentary and neither could I. Read it against the Latin as one unit.
2. The [var:] at 0199C–D (the Ps. 18/Ps. 44 fusion) is the marker in this range making the largest claim, and it asserts a conflation rather than a variant. If the merge decides a fusion under one note is not a Pattern 14 case, that marker is the one to withdraw.
3. [sic: Pecte] at 0202C — §6.1, the one place I chose the less conventional side of a Pattern 12 boundary.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering CAPUT XII–XVII: the exposition of Cant. 1:2–4 (Ideo adolescentulae dilexerunt te → Nigra sum, sed formosa). Six chunks, 6,782 Latin words.

Marker totals for this range: 15 [var:] · 1 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six (one ratio warning, 0013 at 1.61 — inside the 1.35–1.63 band 11638 established for this author's rhymed synthetic prose). Column anchors, [n:] contents and guillemet counts all verified 1:1 against the Latin twins by diff, not by eye.

§1 · TYPOGRAPHY — this range crosses NO seam, and the reason matters

⭐ 11632 does not switch marks the way 11638 does. It prints both, side by side, throughout, and the division is by FUNCTION. Every chunk in my range carries both, and a count across the whole work confirms the pattern holds from chunk 0000 onward (guillemets and italics appear together in 0000, 0002, 0003 … 0029); there is no run of italic-only chunks followed by a run of guillemet-only ones.

The functional division, consistent in all six of my chunks:

• Italics — the Song lemma under exposition, both when quoted as the chapter's head-text (Ideo adolescentulae dilexerunt te., Trahe me; post te curremus…, Introduxit me rex in cellaria sua., Nigra sum, sed formosa…) and when re-quoted piecemeal in the gloss (Curremus, … Trahe me. … Memores, … in te;). Also Migne's marginal source-edition page numbers (120–128 run through this range) and one printed locator (§4 below).
• « » — every other scripture quotation, without exception in this range: 53 guillemeted spans, all of them non-Song scripture (or, at 0220A, the Song quoted from elsewhere in the book — Mel et lac sub lingua tua, Cant 4:11, which is not the lemma under exposition and therefore takes guillemets).

That last case is the proof. At 0220A Philip quotes the Song itself and Migne guillemets it, because it is a proof-text, not the lemma. The mark tracks the quotation's role, not its book. Do not "tidy" 11632 toward 11638's book-boundary model — the two works are marked on different principles, and a sweep that assumes a seam in 11632 will manufacture one.

Guillemet asymmetries: none in this range. All six chunks balance (7/7, 6/6, 4/4, 13/13, 13/13, 10/10) and every pair is reproduced 1:1 in the English. Pattern 5 inquit-interruptions are frequent here — Philip's habit is to drop inquit / ait / inquit Petrus / ait Isaias inside the quotation — and all of them stay inside the guillemets, untouched.

§2 · THE COLLATION — all 80 spans, every one accounted for

The lemma repeats (Trahe me, Curremus, Recti, Exsultabimus, etc.) are short re-quotations of the head-text already collated at its full occurrence and are not re-listed individually below; each was checked against the head-text and none re-quotes it with different words.

0012 (cols 0212B–0214B) — 10 spans

Unguilleted weaves in 0012, checked and left as Philip's own prose (he does not quote them, he writes with them): Wisd 9:15 (corpus quod corrumpitur…), 1 Cor 9:27 (castigat corpus suum). Both carry [n:] and both are recast into his third-person syntax; neither is presented as a quotation.

0013 (cols 0214C–0217A) — 11 spans

Unguilleted weaves checked: Ps 147:15 (cujus velociter sermo currit), Cant 2:8 + Ps 18:6 fused (qui saliens in montibus, colles transiliens ut gigas caeteris antecurrit, with [n: (Cant. II)]). Philip's prose, not quotation.

0014 (cols 0217B–0218D) — 6 spans

Weaves checked: Job 36:5, Wisd 12:18 (subest ei posse cum voluerit — recast to third person from Vulg. subest tibi cum volueris posse; Philip's prose, [n:]-noted, not guilleted), Col 2:3, 1 Cor 2:12, 2 Cor 5:16. All in his own syntax.

0015 (cols 0219A–0221C) — 23 spans

⭐ The two Psalm markers are one finding, not two. Ps 15:9 and Ps 65:17 sit sixteen lines apart at 0219D–0220A, and Migne's Ps 15:9 has acquired exactly the two features of Ps 65:17 — the sub lingua and the first-person verb — while Ps 65:17 has in exchange acquired Ps 15:9's verb exsultavi for the Vulgate's exaltavi. The two verses have contaminated each other, and Philip builds directly on the result: « Ad ipsum… exsultavi sub lingua mea. » Sub lingua, dicit in corde. His whole distinction between laetari (of the heart) and exsultare (of the work) — set out over a full column, then qualified by the observation that exultation is sometimes inward too — rests on a phrase that in the Vulgate belongs to the other psalm and to a different verb. Both markers are placed at their own quotation and neither elides the clause it reports.

Weaves checked: 1 Tim 4:8 (vitae quae nunc est et futurae, [n:]-noted, recast).

0016 (cols 0221D–0224C) — 17 spans

Weaves checked: Ps 113:8 (idols), Deut 32:4 as cited but Ps 91:16 as worded — Rectus Dominus Deus noster, et in eo iniquitas non habetur stands unguilleted in Philip's own prose with [n: (Deut. XXXII)]; the wording is Ps 91:16's (rectus Dominus Deus noster, et non est iniquitas in eo), recast. Reproduced, not corrected. Acts 3:2–8, Num 17:8, Isa 11:1 cited as [n: (Isai. XXI)] (reproduced verbatim), Cant 3:6, Acts 14:21.

0017 (cols 0224D–0227D) — 13 spans

§3 · TYPE DEFECTS

⭐ 0015 @0220A — Desalutari (Luke 1:47). The one [sic:] in the range. Migne prints « Exsultavit Spiritus meus in Desalutari meo ». Deo salutari has been run together and lost its o: Desalutari is not a Latin word at all, so Pattern 10 governs (run-together division is non-word type) and it is carried into the English untranslated in italics inside [sic: …], with the surrounding in … meo rendered around it. Intended reading, near-certain: Deo salutari, "in God my saviour", the Magnificat's own words. The English therefore reads "« My spirit hath exulted in my [sic: Desalutari] »", which is ugly and is meant to be: the alternative is silently repairing a compositor's fusion in the one verse in this chapter that every reader knows by heart. ⚠ Note that the verse is otherwise in full agreement with the Vulgate — this is a plate defect, not a divergent reading, and it takes no [var:].

⭐ 0016 @0222D — percincti against Migne's own praecincti at 0223B. Luke 12:35 is quoted twice in one chunk, eight lines apart, and Migne spells the participle two different ways: percincti at 0222D, praecincti at 0223B. The Vulgate has praecincti. No [sic:]: percingo is an attested (if rare) verb, so this is Pattern 12's "real word that is merely wrong" class — rendered as ordinary English ("girt about" against "girded" at the second occurrence) and marked [var:] against the Vulgate, not [sic:] against the plate. The self-correction inside a single chunk is what makes it evidential: whatever Philip wrote, Migne's compositor knew the right form and set it eight lines later.

0012 @0213D — a misplaced full stop, preserved (Pattern 8, and 11638 §3). Migne prints cum in me totum Deum, Verbum Patris. Dei Filium introspexi — a stop dropped into the middle of an appositional series, breaking Verbum Patris from Dei Filium. The English carries the stop where the plate sets it, and the resulting fragment stands. Intended reading, obvious: …Verbum Patris, Dei Filium introspexi, "the Word of the Father, the Son of God, I beheld within". Not normalized — English has an exponent for a stop in the wrong place, so Pattern 9's normalization licence does not reach it, and 11638 ruled explicitly that this construction stands.

§4 · A PRINTED LOCATOR INSIDE A QUOTATION — 0016 @0222B–C

Migne prints « Bonus Israel Deus his qui recto sunt corde (Psal. [0222C] LXIX). » — the citation set in italics, inside the guillemets, with a column anchor falling in the middle of it, and no [n:] wrapper. It is therefore not one of the chunk's twenty notes (the count is 20 without it, verified against the body), and it is reproduced exactly as printed, anchor and all, italics and all. This is 11638 §1's "italics survive by function" case: an italic run inside a guillemet quotation that is a locator, not scripture.

⚠ The locator is also wrong — Bonus Israel Deus his qui recto sunt corde is Ps 72:1, not Ps 69. Reproduced verbatim and not corrected; the citation is refDisplay matter and correcting it upstream would break any following anaphora (brief §10.2).

Migne's citation errors in this range, all reproduced verbatim, none patched: (Ose. II) for Hos 11:4 · (II Cor. IX) twice for 1 Cor 9:26 and 9:24 · (Psal. LXIX) for Ps 72:1 · (Isai. XXI) for Isa 11:1 · (Deut. XXXII) on wording that is Ps 91:16's. Five distinct works misnamed in six chunks is a rate worth carrying to the merge — see §5.

§5 · THREE THINGS THE MERGE SHOULD KNOW

1. The mechanical quotation list does not see bare [n:]-noted quotations, and one of my markers is on a span it never listed. Rom 8:10 at 0226B (spiritus autem vivit propter justitiam for the Vulgate's propter iustificationem) is set with neither guillemets nor italics — it sits naked in Philip's sentence with only its note — so it does not appear among the 80 spans I was handed. It is nevertheless a genuine, expounded divergence: Philip repeats vivit propter justitiam eight lines later as the phrase he is glossing, so the Vulgate's iustificationem would take the keyword out of the argument (the 7a″ mechanism exactly). Other ranges should read their chunks for [n:]-noted spans the list could not catch. I found this one only by reading the notes against the prose after finishing the listed collation; there may be more of the class in this work, and no script will surface them.

2. Migne's citation sigla are unusually unreliable in this stretch — five misnamed references in six chunks — and two of them are the SAME error twice. (II Cor. IX) is printed for 1 Cor 9:26 at 0216C and again for 1 Cor 9:24 seven lines later. A repeated wrong siglum on one column is evidence about the copy-text or the compositor's exemplar, not a slip, and it is worth the merge counting the class across all 21 ranges before anyone proposes a citation-corrections.json entry. Nothing was corrected here.

3. 11632's marking practice is functional, not sequential — the brief's open question, answered for this range. Italics = the Song lemma under exposition (plus marginal page numbers and printed locators); guillemets = every other quotation, including the Song itself when quoted as a proof-text from elsewhere in the book (Cant 4:11 at 0220A, Cant 5:11 at 0226A, Cant 2:14 at 0226C — all guilleted, while the lemma five lines away is italic). If the whole work holds to this, 11632 has no seam to find, and the 419/635 corpus counts in the brief reflect a two-channel system running in parallel from column one, not a switch. The merge should test this against the ranges at the Explicit/Incipit boundaries before concluding either way.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range 4 of 21, covering CAPUT XVIII–XXIV: the exposition of Sicut tabernacula Cedar, sicut pellis Salomonis (Cant. 1:4) through Nolite me considerare (1:5), Filii matris meae pugnaverunt contra me and Posuerunt me custodem in vineis (1:5), to Indica mihi… ubi pascas, ubi cubes (1:6). Six chunks, 5,293 Latin words.

Marker totals for this range: 6 [var:] · 1 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six — no errors, and no ratio warnings (the six sit inside the 1.35–1.63 band). Column anchors (49), [n:] markers (55) and guillemet pairs (37 « / 37 ») were verified 1:1 against the Latin twins mechanically, not by eye.

§1 · TYPOGRAPHY — no seam here either; §7 of the brief holds throughout

Both marks run side by side in every chunk of this range, divided by function exactly as wave 1 described. The proof case recurs and is worth naming, because it is the cleanest instance yet:

⭐ 0239C — Cant. 2:1 is GUILLEMETED while the running Song lemma two columns earlier is italic. Quorum primus et praecipuus ait: « Ego flos campi [n: (Cant. II)]. » The chapter's own lemma, Indica mihi ubi pascas, ubi cubes, is italic at 0237C, and the Song quoted as a proof-text from elsewhere in the book takes « ». Same chunk, same column band, both marks. 11632 is not marked by book, it is marked by role, and no seam should be hunted here.

Two printed locators sit inside guillemet runs with no [n:] wrapper, the 0016 @0222B phenomenon, and both are reproduced exactly:

• 0232D « Non sunt, ait, recordati foederis fratrum (Amos I). » — italic locator, inside the quotation, not counted among the chunk's four notes (verified: III Reg. III, Ephes. III, Joan. X, Psal. LIV). Here the locator is correct (Amos 1:9), unlike the 0222C case.
• 0236D (Osee XII) — an italic locator inside Philip's own prose (Hos 12:1, Ephraim pascit ventum), again with no [n:].

Guillemet asymmetries: none. All six chunks balance (6/6, 2/2, 3/3, 6/6, 3/3, 17/17). Pattern 5 inquit/ait/inquiunt interruptions are frequent and every one stays inside the marks.

The [n:] does not always sit inside the closing ». At 0235B (…filiis sceleratis! » [n: (Isa. I.)]) and 0239B (…manduces pascha? » [n: (Marc. XIV.)]) Migne prints the note after the closing guillemet, where elsewhere he prints it before. Mirrored in both places; not tidied.

Marginal source-edition page numbers 129–135 run through the range, one or two to a chunk, kept italic and in place.

Capitalization mirrored, per brief §7a. Virgo/virgo alternates within single paragraphs (0229D–0230A has virginem, virgo, virginis lower case in the middle of a chapter that is otherwise Virgo; 0231B virgo prudens; 0235D conversa virgo) and the English follows word by word — "the Virgin" / "the virgin", "the Bridegroom" / "the bridegroom" (0229D sponsi thalamus). ⭐ The sharpest case is 0238D, where John 3:29 is guilleted with lower-case sponsam sponsus inside a sentence whose surrounding prose has capital Sponsus three times. Rendered "he that hath the bride is the bridegroom" inside the marks, "the Bridegroom" outside.

§2 · THE COLLATION — all 80 listed spans, plus the [n:]-only sweep

Every span in data/quotations/11632/0018-0023.md was collated word by word against the Vulgate. Spans recorded as AGREES were checked and found to agree; the absence of a marker on them is a finding, not a silence. Short re-quotations of a head lemma already collated at its full occurrence are noted once and not re-listed.

0018 (cols 0227D–0230A) — 12 spans

Weaves checked and left as Philip's own syntax (no conformation admitted): Ps 103:2 (extendit coelum sicut pellem), Ps 113:16 (coelum coeli), Ps 44:3 (forma prae filiis hominum speciosus), Ps 102:5 (instar aquilae juvenescit), Rom 9:20–21 (plasma plasmatori, testa figulo), Col 3:1–2 (in his quae supra sunt), 2 Cor 4:17 (pressuram momentaneam).

0019 (cols 0230A–0232C) — 8 spans

⚠ A Pattern 8 item, and it is Migne's own editorial parenthesis, not scripture. 0232A prints quod dictum est: Decoloravit me sol, pro eo potest accipi, quod est (decolorabit) praeteritum scilicet pro futuro. The parenthesized (decolorabit) is a word displayed as a word and is translated, not carried ("(he shall discolour)"), per the brief's candidatio ruling: the argument is the tense, and Latin left standing makes the inference unavailable. It is not italic on the plate and is not italicized here.

Weaves checked: Ps 68:2 (usque ad animam intrant aquae), Luke 2:35 (pertransit gladius passionis), Eph 5:27 (rugam et maculam), Apoc 21:4 (omnem lacrymam), Ps 111:7 (mentem… securam).

0020 (cols 0232C–0234C) — 9 spans

Weaves checked: Matt 15:14 (caeci caecorum duces), Wisd 9:4 (suarum sedium assistricem), Matt 26:67 / 27:30 (sputis illinunt, ingerunt alapas).

0021 (cols 0234C–0235D) — 8 spans

Weaves checked: Isa 5:2 and Luke 13:8 (ponit vites in ordine, sepit maceria, putat, fodit), Luke 2:35 again (eam eorum gladius pertransivit), the Exsultet's Felix culpa (liturgical, not scripture).

0022 (cols 0235D–0237C) — 12 spans

Weaves checked: Isa 51:8 / Job 13:28 (ut vestimentum tinea), 1 John 3:18 (non verbo, non lingua).

0023 (cols 0237C–0240A) — 31 spans, 29 notes

This chunk carries 29 [n:] markers in 1,111 words — the densest citation stretch in the work so far — and nine of the range's nineteen [n:]-only entries. Every one of the nine turned out to be Philip's own prose.

Weaves checked: Ps 44:3 (ille pulcher prae filiis hominum), Joel 3:2 (valle Josaphat), and the classical sella breviore et avaro tripode.

§3 · PATTERN 7 / PATTERN 8 — printed readings rendered as printed

1. 0233D — a comma inside its own phrase, preserved. Migne prints non de bono ut aiunt, opere, sed de blasphemia persequuntur: the parenthetical ut aiunt has been dropped between the adjective and its noun, and the commas set around the wrong words. The English carries the interruption at exactly that point — "not for a good, as they say, work, but for blasphemy" — and reads as oddly as the Latin does. Pattern 8; not normalized, because English has an exponent for a comma in the wrong place.

2. 0238A — divisisque ab invicem turris, non consummatur magnitudinis infinitae. The ablative absolute divisisque ab invicem has no noun anywhere in the sentence to agree with (the sense requires the builders, or the tongues, neither of which is in the nominative or ablative here), and turris stands against a genitive magnitudinis infinitae. Rendered as closely as the printed words allow — "and, they being divided one from another, a tower of infinite greatness is not completed" — with "they" supplied only as English grammar's minimum. Conjecture: an ablative such as linguis or artificibus lost from the phrase. Not repaired, not marked — a real construction with a missing referent is not carried type.

3. 0238B — in qua filii Israel Deo non tibi, sed Moloch superstitione misera deferebant. As printed, the children of Israel paid honour to God, and then "not to thee, but to Moloch" — which contradicts itself inside six words, since the tu addressed throughout the chapter is God. Rendered literally, contradiction included: "wherein the children of Israel paid honour to God, not to thee, but to Moloch." Conjecture, near-certain: non Deo tibi, sed Moloch — "not to thee, their God, but to Moloch" — i.e. non has migrated one word to the right. Not repaired in the text (brief §9: never translate a word as something other than what it means to rescue a sentence), and no marker, because every word printed is a real Latin word correctly inflected.

4. 0235D — [sic: a beo], the one [sic:] in the range. Migne prints conversa virgo ad Sponsum vult a beo qui sint ejus plenius edoceri. A beo is not Latin: the space has migrated one letter to the right in ab eo, which is Pattern 10's split-type class (the defect is the division, so the whole run is carried, a included). Carried untranslated in italics inside [sic: …] with the sentence built around it — "wills to be more fully taught [sic: a beo] who are his." Intended reading, certain: ab eo, "by him", the Sponsum named four words earlier. Rendering "by him" and saying so in a crux was considered and rejected: it puts a considered-looking reading on the page where the plate has none.

Negation audit. Every non / nec / neque / nisi / ne / nunquam / nullus / nihil / nondum / necdum in the six Latin chunks was matched against the English on a second pass, with particular attention to the dense clusters at 0232D–0233A (non verentur… in nullo miserentur… noluerunt… non sorori exhibent, non denique homini), 0233B–C (nec dilexit… nec sensu mollitur… nec Scripturae testimoniis, nec naturae imperiis… nec Deum timent, nec hominem reverentur… nec virginem in conceptu, nec post partum virginem), 0236A–B (non verbo, non lingua, non solo strepitu… non in omnibus locis, non in cunctis pascuis… non habere… non senserit abhorrere) and the whole of 0237D–0238B, which is a chain of five successive non pascis denials. No negation was dropped and none supplied. The two places where carrying one makes the English read strangely are §3.3 above and 0236A quae vitalem succum noverit non habere / quae a vita non senserit abhorrere — a double negative Philip means positively; both are carried.

§4 · MIGNE'S REFERENCES — reproduced verbatim, none corrected

Per brief §10.2 and the refDisplay principle; two of these have a following ibid. chain hanging on them (0239A Ezech. XXXIV → (ibid.), 0239B Marc. XIV → (ibid.)), which makes correcting them actively harmful.

§5 · THE THREE THINGS THE MERGE SHOULD CARRY

1. ⭐⭐ Cant. 1:5 prints nigra sum for the Vulgate's fusca sim (0230D), and this is the range's most important datum. Brief §5 rules nigra = black and fusca = swarthy, strictly distinct, fusca never "black" — which means the Vulgate's Cant. 1:5 is not a blackness verse and Migne's is. Philip's CAPUT XIX–XX are four columns of nigredo built on it, and the whole nigra sed formosa architecture of Capp. XVII–XXI depends on the colour word being the same one as in v. 4. If another range meets Cant. 1:5, expect nigra sum and do not conform it — and if any range meets fusca, that is a fact worth reporting, because it would mean Migne is inconsistent with himself across the work.

2. All nineteen [n:]-only spans in this range were Philip's own prose, not quotations — and the class is still worth every minute it costs. Wave 1 found Rom 8:10 in this class and it was a genuine expounded divergence; I found none. That is the useful result to report, not a null one: it tells the merge that the class's hit rate is low but that the reading pass is cheap, and it means the range's six [var:] markers all sit on spans Migne himself marked. Two of the nineteen are nevertheless worth the index's attention because they are citation-practice data rather than variants: the Ps 40:10 + Ps 54:14 fusion at 0234B under a single (Psal. LIV) note, and the Ps 110:2 gloss at 0239B where Philip writes voluntates meas for the psalm's eius while glossing a different psalm.

3. 11632's two-channel typography holds through CAPUT XXIV, and 0239C is the cleanest proof in the work so far. Cant. 2:1 (Ego flos campi) is guilleted as a proof-text in the same chunk whose own Song lemma stands in italics — same chapter, same speaker, same book, different marks, because the roles differ. Three wave-1 agents converged on this and my six chunks add nothing against it. Two subsidiary facts for the merge's typographic sweep, both new here: Migne prints the [n:] outside the closing » at 0235B and 0239B (elsewhere always inside), and he prints two locators inside guillemet runs with no [n:] wrapper at all ((Amos I) 0232D, and (Osee XII) 0236D in bare prose). Neither is an asymmetry and neither was tidied, but a script counting notes by their position relative to guillemets will mis-bin them.

§6 · WHAT A LATER BLIND READ SHOULD LOOK AT FIRST

1. 0238B Deo non tibi, sed Moloch (§3.3) — the one sentence in the range that says the opposite of what the argument needs, rendered literally and unrepaired. If any silent repair has hidden anywhere here, this is where the pressure to make one was greatest.
2. The Mark 14:14–15 compression at 0239B — the span I judged a compression rather than a variant, and the closest call among the eighty.
3. [var:] on Isa 5:7 Sabaoth (0235A) — the marker in this range making the smallest claim; if the merge decides a Hebraism substituted for its Latin equivalent is not a Pattern 14 case, that is the one to withdraw.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering the end of Book I (Caput XXV–XXVI, on Indica mihi ubi pascas, ubi cubes in meridie and Ne vagari incipiam post greges sodalium tuorum), the Explicit of Book I and the head LIBER SECUNDUS., and then Book II Caput I–V, on Si ignoras te, o pulcherrima inter mulieres, egredere, et abi post vestigia gregum → Et pasce haedos tuos juxta tabernacula pastorum → Equitatui meo in curribus Pharaonis assimilavi te, amica mea (Cant. 1:6–8). Six chunks, 6,853 Latin words.

Marker totals for this range: 9 [var:] · 2 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six (no errors); two ratio warnings, 0026 at 1.66 and 0028 at 1.63, inside the band wave 1 shipped (0004 1.64, 0005 1.68). Column anchors, [n:] contents and guillemet counts were diffed 1:1 against the Latin twins mechanically, not by eye: anchors identical in order and place in all six; [n:] strings byte-identical and in the same order (11 / 2 / 2 / 2 / 14 / 13, matching each frontmatter noteCount); guillemets 3/3, 2/2, 1/1, 1/1, 16/16, 12/12. No guillemet asymmetry anywhere in this range — every « has its », none stray, none unopened.

§1 · ⭐ THE BOOK BOUNDARY — I CROSS ONE, AND THERE IS NO TYPOGRAPHIC SEAM AT IT

This is the range brief §7 asks for. Chunk 0025 ends Book I and prints ## LIBER SECUNDUS. as its last head (col. 0246C), with an empty body; chunk 0026 opens ## CAPUT PRIMUM. under it. Migne even prints the transition in Philip's own voice — primo huic tandem libro terminum imponamus … curabo quiescentem calamum retorquere.

Nothing about the marking changes across it. Guillemets and italics are both in use on both sides, in the same functional division wave 1 established:

• Book I side (0024–0025): italic lemma (Indica mihi…, Ne vagari incipiam…), italic marginal numbers 136–139, italic printed locator; guillemets on Gen. 49:9, Wisd. 9, Matt. 17:4, Cant. 6, Luke 12:32.
• Book II side (0026–0029): italic lemma and its piecemeal re-quotations, italic marginal numbers 140–145; guillemets on twenty-nine quotations.

So the brief's ruling holds at the one place in the work where 11638's book-boundary model predicted a switch. 11632 has no seam. A later range that crosses the Book II/III boundary should test the same thing and say so; on this evidence the merge can stop looking.

1a. ⚑ But the FUNCTIONAL rule needs widening: guillemets are not scripture-only

Wave 1 stated the division as "« » carries scripture, italics carry everything else," and recorded that every guilleted span in cols 0181–0227 was scripture. That is no longer true from col. 0251D onward, and the counter-examples are not marginal — they are the argument of a whole chapter. In 0028 Migne guillemets:

• « Scito, inquit, te ipsum, homo. » (0251D) and « Scito, ait, te ipsum, homo. » (0252B) — the Delphic oracle, twice, attributed by Philip to unum e septem Graeciae sapientibus or to Apollo. Not scripture, not italic.
• « Dii, si qua est pietas coelo. » (0253A) — a line of classical verse (Aen. 2.536, si qua est caelo pietas), introduced sicut ille poeta.

The prologue's Horace and the two poets at 0198A / 0201A were italic. So classical verse takes italics in Book I and guillemets in Book II. The stable description is therefore not "scripture vs. everything else" but « » = a quotation adduced as an external AUTHORITY (whatever its source), italics = the lemma under exposition, printed locators, marginal numbers, and words displayed as words — with 0029's exception below.

1b. ⚑ And guillemets can carry a single word displayed AS a word

0254B prints « Meos » dicit … and dicit « tuos » — two bare pronouns, guilleted, each being discussed as the word of the lemma (pasce haedos tuos, against John 21's agnos meos). Rendered « my » and « thy ». Under wave 1's rule these would have been italic. Reproduced as printed; do not tidy in either direction, and note that 0029 therefore has 12 guillemet pairs where a reader counting quotations would expect 10.

1c. A printed locator inside the guillemets, again

0244C prints « Nolite timere, pusillus grex (Luc. [0244C] XII). » — italic locator inside the quotation, containing the column anchor, with no [n:] wrapper. Third instance of the class (cf. 0203A, 0222B–C). This is why 0025's noteCount is 2 and not 3. Reproduced exactly, anchor and italics and all.

1d. Marginal source-edition numbers

136 (0240D) · 137 (0242D) · 138 (0244B) · 139 (0246B) · 140 (0247D) · 141 (0249C) · 142 (0251A) · 143 (0253A) · 144 (0254D) · 145 (0256B). Continuous with wave 1's 101–128. Each kept italic and verbatim at the exact word it interrupts — 137 falls between vobiscum and et laborat, 143 between tamen and virum, and the English clause was built round it rather than moving it.

§2 · THE COLLATION — all 70 listed spans, plus the unlisted [n:] weaves

Every span in data/quotations/11632/0024-0029.md was collated word by word against the Vulgate. Spans recorded as AGREES were collated and found to agree; the absence of a marker on them is a finding, not a silence. Lemma re-quotations (Egredere, et abi, Si ignoras, Equitatui meo, etc.) are short repetitions of a head-text collated at its full occurrence; each was checked against its head-text and none re-quotes it with different words.

2a. Divergent — [var:] fired (9)

2b. Collated and AGREEING — no marker (recorded so the absence is evidence)

0024 — Cant. 1:6 Indica mihi ubi pascas, ubi cubes in meridie (the lemma; truncated at the opening, quem diligit anima mea having been expounded in the preceding range at 0022 — truncation, not variant) · Matt. 17:4 « Bonum est nos hic esse » (opening Domine falls outside) · Wisd. 9 « Da mihi… sapientiam ut mecum sit, et mecum laboret » — a conflation, not a variant: Da mihi… sapientiam is 9:4 (dropping sedium tuarum assistricem), ut mecum sit et mecum laboret is 9:10, and every word printed is the Vulgate's. Per the cruces-0000 item-12 precedent, a conflated citation takes no [var:].

0025 — Cant. 1:6 Ne vagari incipiam post greges sodalium tuorum — agrees word for word. · Cant. 6:5 « dentes tui sicut grex ovium » — agrees, truncated before quae ascenderunt de lavacro (the first half of the same span is the [var:] above) · Luke 12:32 « Nolite timere, pusillus grex » — agrees.

0026 — Cant. 1:7 Si ignoras te, o pulcherrima inter mulieres, egredere, et abi post vestigia gregum — agrees exactly (and see §3.2: the same head at 0250B does not) · Luke 1:28 « Benedicta tu, inquit, in mulieribus » — agrees.

0027 — Cant. 1:7 Et pasce haedos tuos juxta tabernacula pastorum — agrees (also at 0254A).

0028 — Gen. 18:27 « Loquar, inquit, ad Dominum meum, cum sim pulvis et cinis » — agrees, opening at loquar · Ps. 70:16 « Quoniam non cognovi, ait, litteraturam, introibo in potentias Domini; Domine, memorabor justitiae tuae solius » — agrees word for word · Dan. 4:27 « Nonne haec est… Babylon magna, quam aedificavi in robore fortitudinis meae, et in gloria decoris mei? » — drops ego and in domum regni; abridgment, no marker · Job 29:13–16, 29:25, 30:1 — the long « Cor viduae… » run agrees clause by clause, abridged across 29:14 (et diademate iudicio meo) and 29:17–24, and cumque sederem for cum sederem; a stitched citation, not a divergent reading · Ecclus. 11:27 « In die… malorum ne immemor sis bonorum » — agrees · 2 Cor. 12:2–4 « Scio, inquit, hominem in Christo raptum usque ad tertium coelum… » — every word Pauline; compressed (drops ante annos quattuordecim and one of the two nescio, folds raptum… usque ad tertium caelum into v. 2). ⚠ Checked specifically for negation loss under brief §9: the surviving nescio and non licet are both carried in the English. No marker · 2 Cor. 12:5 « Pro hujusmodi gloriabor » — agrees · John 19:26 « Mulier… ecce filius tuus » and 19:27 « Ecce mater tua » — both agree · Ezek. 34:31 « Vos autem greges mei, dicit Dominus » — agrees for the words printed; truncated before greges pascuae meae homines estis, with dicit Dominus pulled up from the verse's end. No marker.

0029 — 1 John 1:8 « Si dixerimus… quia peccatum non habemus, nos ipsos seducimus, et veritas in nobis non est » — quia for quoniam, nos ipsos for ipsi nos; sense untouched, logged, no marker · Jas. 3:2 « In multis… offendimus omnes » — agrees (Vulg. In multis enim) · John 21:15 « Simon Joannis, diligis me? pasce agnos meos » — agrees, dropping plus his; note that this is the correct interrogative form of the verse Migne printed as a conditional at 0249B · Exod. 32:7 « Descende, peccavit populus tuus » — agrees, dropping the preceding Vade · Cant. 1:8 Equitatui meo in curribus Pharaonis assimilavi te, amica mea — agrees word for word · John 15:14 « Vos amici mei estis, si feceritis quae praecipio vobis » — agrees, dropping ego · Exod. 15:4 « Currus Pharaonis et exercitum ejus projecit in mare » — agrees exactly · Exod. 15:13 « Dux fuisti… in misericordia tua populo quem redemisti » — agrees · Hab. 3:8 « Qui ascendis super equos tuos, et quadrigae tuae salvatio » — Vulg. ascendes, future. A one-letter tense difference on which nothing in the exposition turns, and both forms are attested in Vulgate transmission; logged here, deliberately NOT marked. Recorded so a later reviewer who fires one says so rather than treating it as a fresh discovery.

2c. The [n:]-only class — read against the prose after the listed collation

All fourteen ⭐ spans on the list were traced back to their real left edge. Thirteen of the fourteen turn out to be Philip's own prose, written with scripture rather than quoting it (accusative-and-infinitive frames, third-person recastings, passive recastings). None takes a marker; a [var:] on unmarked prose would misrepresent it as a lemma. But two are worth the merge's attention:

1. ⭐ 0240B, Luke 9:58 cited, Matt. 8:20 worded. Quia ergo vulpes foveas, nidos coeli volucres habere dignoscuntur [n: (Luc. IX)]. Luke 9:58 reads volucres caeli tabernacula ; nidos is Matthew's word. Philip's own syntax, so no marker — but the note names the gospel whose wording is not printed. Same class as cruces-0012's Deut. 32:4-cited / Ps. 91:16-worded case.
2. ⭐⭐ 0245A, Ps. 90:5–6 fused, and Philip then expounds the fusion. He writes vel sagitta in tenebris, vel incursu meridiano [n: (Psal. XC)]. The Vulgate has a sagitta volante in die, a negotio perambulante in tenebris, ab incursu et daemonio meridiano — the arrow flies by day; what walks in darkness is the negotium. Migne's arrow flies in the dark, and the very next sentence glosses the fused form: Sagitta volans in tenebris, quae ferit graviter dum nescitur, occulta est suggestio. The whole hidden-suggestion / open-assault antithesis of Caput XXVI rests on it. No marker, because Migne marks it as neither quotation nor lemma** — it is Philip's prose with a bare note, exactly the class the extractor cannot see — but this is the single most consequential textual observation in my range and the merge should carry it. If a later range meets Ps. 90:6, expect the arrow in darkness.

The remaining twelve, listed so nobody re-opens them: 0240D 1 John 5:19 + Luke 1:35 woven; 0240D John 19:26 recast; 0241C John 13:23 ([n: (Cap. XIII)] — a bare chapter with no book named, reproduced verbatim); 0242A John 20:22; 0242A Acts 2:3; 0244A Ps. 93:20 recast into the passive (labor fingitur in praecepto for qui fingis laborem in praecepto; Philip used the active at 0195A, so the recasting is his); 0244A Matt. 13:43 recast to the singular (justus… sicut sol lucidus resplendescit for iusti fulgebunt sicut sol); 0247A Hag. 1:6 recast; 0249B Matt. 25:33; 0254A Luke 17:10; 0254B Matt. 25:33 again; 0256A 1 Cor. 1:23 recast.

Further unmarked weaves checked for silent conformation and left as Philip's own prose: Matt. 23:15 (duplo filia sit gehennae, 0244D), Luke 10:18, 2 Cor. 11:14 (0245C), Ps. 68:29 / Apoc. 3:5 (deleri de libro viventium, 0248B), 2 Cor. 1:22 (pignus Spiritus et arrhas, 0248C), Luke 1:30 (0248C), John 6:27 (0242C), Ps. 127:2 (0242C), Heb. 12:9 (Pater spirituum, 0241B), Col. 2:9 (0250A), Ps. 117:15 (0249C), Ezek. 28:13 (0250D), Ovid Met. 1.85 (os homini sublime, 0252A), Ps. 31:9 (equum et mulum, 0252B), 1 Cor. 13:10 (evacuato quod ex parte est, 0254D), Ps. 83:8 (de virtute in virtutem, 0254D), Matt. 11:30 (levi jugo et suavi, 0256A). None was conformed.

2d. Listed spans that are NOT scripture

0251D and 0252B « Scito te ipsum, homo » (the Delphic maxim, §1a) · 0253A « Dii, si qua est pietas coelo » (Virgil, §1a) · 0254B « Meos » and « tuos » (words displayed as words, §1b) · 0244C (Luc. XII) (a printed locator, §1c) · all the italic lemma fragments.

§3 · THE FOUR FINDINGS THE MERGE SHOULD CARRY

⭐⭐ 3.1 · Cant. 6:4 at 0244C — the dropped grex is the word the sentence exists to prove. Philip's point is that the faithful are called a grex singular and the false a greges plural: sciens quia convenientius nomen justis attribuitur singulare. He then adduces two proof-texts for the singular. Migne prints the first as « Capilli tui sicut caprarum » — with no noun at all where the Vulgate has sicut grex caprarum. The keyword has fallen out of the proof-text adduced for it, leaving a bare genitive and only the second half (dentes tui sicut grex ovium) carrying the argument. Rendered as printed, "Thy hair as of goats", with the [var:].

⭐ 3.2 · The head lemma of Book II Caput III prints pulchra where Caput I prints the Vulgate's superlative — and Philip's exposition follows the short form. At 0245D the lemma is o pulcherrima inter mulieres, the Vulgate's superlative. At 0250B, four columns later and heading the same verse's exposition, it is o pulchra inter mulieres, positive. And the gloss between them is unambiguously on the positive: O pulchra, inquit, inter mulieres (0247A), Quamvis quod pulchra inter mulieres dicitur (0247B), then o pulchra four more times at 0253A. So the divergence is not a one-off compositor's slip on a head line (the shape cruces-0006 found for ubera sua at 0196C): six of the seven printings of the phrase in this range read pulchra, and the Vulgate's superlative appears only once, at the head that opens the book. [var:] fired at 0250B only, since 0245D is the one that agrees. ⚑ Worth a count across the later ranges: if pulchra keeps winning, it is Philip's text and the 0245D superlative is the outlier, which reverses the usual expectation.

⭐ 3.3 · Luke 17:10 at 0252D — omnia bona, and the next sentence expounds it. Migne prints « Cum omnia bona feceritis, dicite, quia servi inutiles sumus » where the Vulgate has cum feceritis omnia quae praecepta sunt vobis. Philip's immediately following gloss is built on the substituted phrase: non omnia bona quivis facere comprobatur, si non istud dicere bonis caeteris apponatur; et bonorum congeriem… non colorat. Conforming the lemma to "all things commanded you" would leave four sentences arguing about good things from a verse that no longer mentions them — the 7a″ mechanism exactly. Rendered as printed.

⭐ 3.4 · Exod. 15:1 at 0255B — contamination from the verse quoted two lines later. Migne prints « Equum et ascensorem projecit in mare » (Vulg. deiecit), and then, in the same sentence, « Currus Pharaonis et exercitum ejus projecit in mare », which is Exod. 15:4 and where proiecit is correct. The first verse has taken the second's verb. The sense is not changed — both verbs mean "cast" — and I record explicitly that I fired the [var:] on the word, not on a sense-shift, because this is the same evidential class as cruces-0012's Ps. 15:9 / Ps. 65:17 mutual contamination: two adjacent quotations of the same song, one having borrowed from the other. ⚑ If the merge rules that a sense-neutral lexical swap does not earn a Pattern 14 marker, this is the one to withdraw — the finding survives in this file either way.

§4 · PATTERN 7 — printed readings rendered as printed, conjectures kept here

Two [sic:] fired. Neither carried run is ALSO translated in the running text (brief §8's named failure).

1. 0241C [sic: lehari] — ne a gloria deprimamur, dum ad formam lehari stulte invadimus altiora. Read Icari, near-certain: the sentence's own frame is the Daedalus myth (Inter utrumque ergo non Dedalus, sed Pater spirituum doceat nos volare), and the antithesis is exactly Daedalus's counsel — too low into mud and dregs on one side, too high after Icarus's fashion on the other. lehari is not a Latin word, so Pattern 7 governs: carried untranslated in italics, and the English deliberately does not also say "Icarus".
2. 0254D–0255A [sic: vidilicet] — ut vidilicet sequatur exemplo. Read videlicet, certain. ⚠ This one sits on a boundary and I record the choice rather than hide it: an e/i interchange in an unstressed syllable is the sort of thing Pattern 9 exempts as period orthography, and a case could be made for rendering it "namely" and marking nothing. I marked it because wave 1 marked exactly this shape — [sic: eumdum] for eumdem at 0191C — and a work that carries one and renders the other is incoherent on its own page. If the merge prefers Pattern 9 here, it should also revisit eumdum.

Rendered as printed, no marker (real words, merely wrong or merely odd):

• ⛔ 0250B Nolens Sponsus gradatim ad profectus Virginem promovere, docet eam quomodo mereatur proficiendi gratiam obtinere. The printed Nolens contradicts its own sentence: the Bridegroom is said not to will to promote her, and in the same breath to teach her how to obtain the grace of promotion, which the whole of Caput III then does. The conjecture is Volens, and it is as near certain as the parallel constructions allow. Under brief §9 the negation is CARRIED — the English reads "The Bridegroom, not willing to promote the Virgin by degrees to advances, teaches her how she may deserve to obtain the grace of advancing." It is meant to look wrong. ⚠ This is the sentence in my range most likely to be silently repaired by a later reader, and the repair would be invisible: dropping one participle's negative gives a perfectly smooth chapter opening.
• 0253A quod quidem difficile est, sed tamen 143 virum est. virum is accusative and fills the slot of a neuter predicate; the clause does not construe. The conjecture is verum ("but yet it is true"), which is what the argument requires — Philip has just said Si here is affirmative, not dubitative. virum is a real word, so Pattern 12 keeps it on the English side and it is rendered, not carried: "which indeed is difficult, but yet is a man". ⚠ Second place most likely to be tidied. Do not.
• 0242D cum invitat justus secum patientiam ad laborem. The proof-text three lines above is Wisd. 9's ut mecum sit, et mecum laboret, of sapientia — so patientiam is almost certainly sapientiam. A real word that construes in its slot; rendered "invites patience along with himself to labour", no marker.
• 0250B si, quae tam obtinuit, apud se… sic reponat. tam where the sense wants jam. Rendered "which she has so obtained".
• 0247B de qua nasci voluit, sepe carnis novo genere non abrupta. sepe = saepe, ablative of saepes, "hedge" — attested medieval orthography, Pattern 9, deliberately unmarked. Rendered "the hedge of her flesh not being broken".
• 0252C Et Adjunxit: « Pro hujusmodi gloriabor ». Migne capitalizes the verb mid-sentence. The English supplies the pronoun English requires ("And he added"), which relocates the capital and so cannot mirror it; logged here rather than forced. Not a Pattern 8 punctuation case and no mark was deleted.

Negation audit (brief §9), second pass, against the Latin. Every printed non / nec / neque / nisi / nunquam / nullus / nihil was matched, with particular attention to the dense clusters at 0242B (non nisi meridie cubare), 0243D (non nisi dies una non proficiens non deficiens dilucescit, cujus non fuscatur splendor, calor splendidus non tepescit), 0250C (vel non quati ariete, vel flatu quantumvis modico non moveri… nunquam animo consentire), 0250D (si non de his quae acceperat superbiret… Deo paria non sentiret), 0252D (quia non omnia bona quivis facere comprobatur, si non istud dicere… apponatur… qui eam vanitatis vitandae gratia non ignorat), 0255C (minasque Pharaonis non evadunt tali praesidio… equo aut rota volubili non transitur… sicut non navigium, sic nec vehiculum), and 0256A–B (non culpam tibi contrahis, non peccatum, non frangeris opprobriis, non iter deseris… Nec solum caput tuum non premit… sed nec rivus praeteriens). No negation was dropped and none was supplied.* The one place where carrying it makes the English contradict itself is 0250B's Nolens*, above, and that is the point.

§5 · CONVENTIONS SET OR HELD IN THIS RANGE

1. ⚑ amica mea (Cant. 1:8) = "my FRIEND", not "my love", and the gloss forces it. Caput V opens by glossing the word — Amicam dicit, cujus animus auditui penitus concordat — and proves it from John 15:14, « Vos amici mei estis, si feceritis quae praecipio vobis ». Douay's "my love" would leave the chapter arguing from friends to a lemma that says love, and would also collide with amor and dilectio, both of which are in use in the same columns. This lemma recurs many times later in the Song (2:10, 2:13, 4:1, 4:7, 5:2, 6:3), and the merge should hold every range to one English form of it. Flagged as the range's one work-wide vocabulary decision.
2. LIBER SECUNDUS. → "BOOK TWO.", after wave 1's LIBER PRIMUS. → "BOOK ONE." The head sits at the end of 0025 with an empty body and its Caput I opens the next chunk; nothing was moved.
3. Migne's CAPUT PRIMUM in words, then CAPUT II–V in numerals, mirrored and not tidied — and note that Book II restarts the numbering while Book I ended at CAPUT XXVI. Preserved as printed.
4. ⚑ Sponsus/sponsus, Virgo/virgo capitalization mirrored word by word (brief §7a). The live case in this range is 0243B–C, where Migne prints lower-case sponsum — sponsum tenere sinu, affectu, velut brachio circumplecti — inside a paragraph that has Sponsus capitalized four times before and after it. The English reads "bridegroom" there and "Bridegroom" everywhere else. Likewise 0252D quae nondum dat filius (lower case, rendered "the son") beside Sponsus capitalized in the same sentence. Mechanically checkable against the Latin; do not regularize.
5. Pattern 17 does real work here. Caput I–V of Book II is the Bridegroom speaking to the Virgin, singular throughout — Tu, ait, quae majorem cunctis mulieribus gratiam jam adepta (0247D), attende diligentius quid tibi jam meo munere sit collatum, si… memoriter consideraveris te habere, assimilavi te, te addixi, te benedixi, tuas… maculas, te perduco, tibi contrahis, caput tuum, vera mater facta es — so thou without exception. Philip's own frame is plural and takes you: sicut nostis (0241C), cum vobis, fratres and the whole 0242C–0243A exhortation (castigatis, alligatis, relegatis, laboratis, manducatis, dicitis, aestuatis), Videtis Paulum (0252C), ut… noveritis (0252D). The one mixed-number sentence is 0255B's « Vos amici mei estis, si feceritis quae praecipio vobis » — plural, inside a chapter otherwise addressed to one woman, and the English shows it.
6. Brief §5 vocabulary held: charitas = charity (0241D sua erga eos charitas) kept distinct from amor / dilectio = love (0241C amoris gratia, dilectionis floribus, where Philip's own doublet forces the repetition and it was not varied for elegance); misericordia = mercy (0255D in misericordia tua), clementia does not occur; fusca = swarthy — live at 0247B, matrem elegerit fuscam, luscam, and at 0243D, cujus non fuscatur splendor, both rendered "swarthy"/"made swarthy", never "black"; Sponsus = Bridegroom, Sponsa = Bride; equitatus = cavalry, quadriga = four-horse car, kept distinct from currus = chariot (0255C–D runs all three in two sentences).
7. Migne's Si discussed as a word at 0253A is NOT italicized on the plate (Quod enim dicit Si, non dubitativa, sed affirmativa conjunctio est), unlike every other word-as-word in this work. The Latin word is kept in the English and no italics were supplied — adding a run the plate lacks would break italic parity to make a point the sentence already makes.

§6 · THINGS DELIBERATELY NOT REPORTED (brief §10, checked once each)

• 0026's anchors open [0245D] and then jump to [0246D], and its colContext is 0246C while its colFirst is 0245D — i.e. the chunk begins at a column before its own context. This is brief §10.1's band-arithmetic class exactly (two works interleaving by band). Checked: no text is broken mid-word or mid-construction across any anchor in any of the six chunks. No [ed:], no patch, no plate check. Recorded only so the next reader does not spend a session on it.
• noteCount matches the body [n:] count in all six chunks (11/2/2/2/14/13), verified mechanically, not by eye.
• Migne's references reproduced verbatim, none corrected (refDisplay, brief §10.2): [n: (Luc. IX)] on Matthew's wording at 0240B; [n: (Cap. XIII)] at 0241C, a chapter number with no book named; [n: (Job. II.)] and [n: (Dan. IV.)] printed with a full stop inside the parenthesis where the other notes have none. The two [n: (ibid.)] chains (0252C on 2 Cor. 12, 0253B on John 19, 0255C on Exod. 15) all resolve correctly against their printed antecedents and none was touched.
• [n:] position varies and is reproduced, not normalized: it falls inside the closing guillemet in most cases, but outside it at 0251A (Dan. 4) and 0251C (Job 2) — …decoris mei? » [n: (Dan. IV.)]. Mirrored.
• Attested medieval spellings left alone and unmarked (Pattern 9): tanquam, coelum/coelestia, aethnicum/aetheris, moerentium, sepe, spiritaliter, Aggaeus, Nabuchodonosor, Hierusalem (not in this range), charitas.
• No [f:]. 11632 is lemma-and-gloss, not a florilegium (brief §10.4).
• No [d:], no [nt:]: no dittography and no prose note in the range — all 34 notes are bare citations.

§7 · OPEN — for the merge or a blind read

1. ⛔ The blind polarity read has not been done and cannot be done by me. Findings must be formed blind and reconciled against this file afterwards (runbook 4a). The three passages most likely to be wrongly "repaired" are, in order: Nolens at 0250B (§4), virum est at 0253A (§4), and the carried lehari at 0241C. An entry defending an unusual rendering deserves more scrutiny than an unlogged passage, not less.
2. ⚑ §1a and §1b widen wave 1's typography rule and the merge should adopt or reject the widening explicitly, because "guillemets = scripture" is stated as settled in three range files and is false from col. 0251D. The safe form is "guillemets = a quotation adduced as an authority; italics = the lemma, locators, marginal numbers and words-as-words" — with 0254B's « Meos » / « tuos » standing as a genuine exception to the last clause.
3. ⚑ The pulchra / pulcherrima count (§3.2) is work-wide and this range only sees its first columns. Cant. 1:7 heads several later capita. The merge should collect the readings rather than impose one, and should check whether any later range silently conformed the 0250B head to the superlative.
4. Ps. 90:6's arrow in darkness (§2c.2) is unmarked because Migne marks it as nothing at all. If the merge decides that an expounded [n:]-only weave earns a [var:] — the question cruces-0012 opened with Rom. 8:10 and answered the other way — this is the range's candidate, and the marker would go at 0245A.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering Capp. VI–XII: the exposition of Cant. 1:9–13 (Pulchrae sunt genae tuae sicut turturis → Botrus Cypri dilectus meus mihi, in vineis Engaddi). Six chunks, 7,684 Latin words.

Marker totals for this range: 5 [var:] · 4 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six (no errors; one ratio warning, 0035 at 1.65, inside the 1.35–1.68 band this work has already shipped — 0004 ran 1.64, 0005 1.68, 0026 1.66). Column anchors (10/12/14/14/10/13), [n:] contents (5/5/19/17/10/7) and guillemet counts (5/4/11/10/9/3 pairs) were diffed 1:1 against the Latin twins by script, not by eye. Every italic run in the English is accounted for against the Latin's own runs plus the four [sic:] and the five [var:].

§1 · THE TERM DECISION THE MERGE MUST RATIFY — muraenula stays Latin

⭐⭐ This is the most consequential thing in the range and it is work-wide, not range-local. Caput VIII (chunk 0032) expounds Cant. 1:10 Muraenulas aureas faciemus tibi, vermiculatas argento, and its entire argument is that the word names a fish:

Muraenula piscis est, qui nutritus in aquis… Cujusmodi piscem forte Sponsus promittere Virgini videtur… quod intellectum ad litteram nobis penitus interdicit, cum eumdem piscem aureum vel argenteum esse dicit.

Then the fish's shape carries the whole chapter — tail to mouth, a ring, meshes of a net, a chain, Moses' head-and-tail offering, ut integretur macula, fiat annulus, muraenula retundatur. Rendering the lemma "chains of gold" (the 10379 solution) makes seventeen paragraphs explain nothing: a chain that "is a fish" is not an argument, it is a non-sequitur.

So this range follows 10804's precedent, which translation-style.md names as correct for this very word: muraenula / muraenulae stays in italic Latin, in the lemma and in the prose alike (17 instances in 0032). This is the Song ruling's case 2 — the gloss IS the etymology, the stylus/epistylium and murenulae/ferculum class — and not a Pattern 12 carry-through, so no [sic:] and no crux-per-instance.

⚑ The lemma recurs later in the work and every later agent must render it the same way. Within a work: one voice (the style file's own boundary clause). If the merge prefers "little lampreys" or "chains", the change is mechanical, but it must be made everywhere or Caput VIII becomes incoherent on its own page.

The companion decision, and it went the other way: vermiculatas argento is translated, as "vermiculated with silver". The gloss chain here runs vermiculat → vermiculatio → vermilio → vermiculus → vermis (Ps. 21:7 Ego autem sum vermis), and English happens to carry the whole chain visibly — vermiculate / vermiculation / vermilion / little worm / worm. Nothing is lost, so nothing is left in Latin. Brief §5's candidatio rule governs: italic Latin is only for words with no clean English equivalent, and this one has one.

§2 · THE COLLATION — all 97 listed spans, every one accounted for

Every span in data/quotations/11632/0030-0035.md was collated word by word against the Vulgate. Spans recorded as AGREES were collated and found to agree; the absence of a marker on them is a finding, not a silence.

0030 (cols 0256C–0258D) — 7 spans

Unguilleted weaves in 0030, checked and left as Philip's own prose (no marker, recorded so a later reader knows they were read): Matt. 11:11 (major quo inter natos mulierum non surrexit), Gen. 2:24 / Eph. 5:31 (secum factus est caro una), Eph. 2:3 (irae filius ex natura).

0031 (cols 0259A–0261D) — 8 spans

Also checked and left unmarked: the maxim juxta ethnicum, non honor est, sed onus (Ovid, Her. 9.31) — Migne sets it in neither guillemets nor italics, so there is nothing to reproduce; Luke 1:38 (ancillam humilem se fatetur), Eph. 2:14 + 4:5 (duo parietes una fide, uno baptismate), John 2:7–9 (aqua implentur hydriae). All in Philip's syntax.

⭐ Pattern 3 fires here on a BARE Scripturis. 0260A: unde tamen vel mandatum in Scripturis, vel formam in praecessoribus non haberet → "in the Scriptures". Neither divina nor sacra is present — which is exactly the form 11638's sweep pattern missed, and cruces-0000 §4.9 warned would recur. It has. A later sweep of 11632 must match bare Scriptura(e) too.

0032 (cols 0262A–0265B) — 26 spans

⚠ Typography worth naming: at 0265B Migne breaks his own italics to isolate a verb. He prints Muraenulas aureas faciam tibi, sed faciemus. — faciam set in roman inside the lemma, because the whole point of the sentence is that the Bridegroom did not say it. The English mirrors the run structure exactly (Muraenulae of gold I will make for thee, but we will make.).

0033 (cols 0265C–0268D) — 25 spans

0034 (cols 0269A–0271B) — 18 spans

Weaves checked: Matt. 27:32 (non tanquam Simon in angaria), Matt. 26:41 (Spiritus promptus). Philip's prose.

0035 (cols 0271C–0274C) — 13 spans

Weaves checked: 1 John 2:27 (docet de omnibus unctionem), Ps. 10:8 (qui nonnisi justitias diligit), Prov. 4:18 (diei deducit ad perfectum), Rom. 13:13 + John 12:35 (sicut in die honeste ambulet, et non eum tenebrae comprehendant), Matt. 25:21 (frui jam illo sui Domini gaudio).

§3 · THE FIVE [var:], IN FULL

1. ⭐⭐ Ps. 11:7 at 0264C — eloquium SINGULAR, and it is [n:]-only. Migne prints Habetis quia eloquium Domini eloquium castum, argentum igne examinatum, probatum terrae, purgatum septuplum [n: (Psal. XI)]. The Vulgate reads the plural Eloquia Domini, eloquia casta. The singular is the word Philip goes on to expound, twice, within eight lines: Petri eloquium ignis non sonorum… Joannis vero eloquium non satis lucidum invenitur. His whole point is that one man's speech and another man's speech are each assayed like silver, and the plural would dissolve the comparison into a general predication about scripture.

⛔ This span is set in NEITHER guillemets NOR italics — it is the class the brief warns about at §8 and that cruces-0012 §5.1 found Rom. 8:10 in. The extractor listed it only because the note is attached; it could not know where it began, and the opening is eloquium, not Habetis.

2. Isa. 43:2 at 0264C — two words, and Philip's next clause uses both. Migne: « Si transieris, ait, per ignem, non combureris; et flamma non nocebit in te ». Vulg.: cum ambulaveris in igne, non combureris, et flamma non ardebit in te. (transieris has migrated from the preceding water clause, cum transieris per aquas.) The marker matters because the sentence Philip builds on it reads illaesum transiens, "passing through unharmed" — which picks up transieris and nocebit precisely, and neither of which is in the Vulgate's wording of the verse.

3. Lam. 3:65 at 0268A — tense, plus an added vocative. Migne: « Domine, ait, dedisti eis scutum cordis, laborem tuum ». Vulg.: Dabis eis scutum cordis, laborem tuum — future, no vocative. The tense is material in the place Philip puts it: the verse is the third of three proof-texts (utquid dereliquisti me · Laboravi sustinens · this) adduced to show that the incarnate life was non quietis haec habitatio, sed laboris — an accomplished labour, not a promised one. Same class as the 0193C Jer. 31:22 marker in wave 1.

4. Prov. 8:17 at 0273B — Ergo for Ego. Migne: « Ergo, inquit, diligentes me diligo ». Vulg.: Ego diligentes me diligo. Under Pattern 5 the interrupting inquit sits inside the quotation, so Ergo is the quotation's first word, not Philip's connective outside it. Rendered "Therefore" as printed; the marker records the emphatic pronoun the received text has and this printing does not. (This is the one marker in the range where the divergence is most likely a compositor's transposition rather than a copy-text reading — but Pattern 14 forbids speculating about why, so the note states both readings and nothing more.)

5. Luke 8:46 at 0273B — sensi for novi. Migne: « Tetigit me, inquit, aliquis: ego enim sensi virtutem de me exisse ». Vulg.: nam ego novi virtutem de me exiisse — "I know", not "I felt". Material here because the verse is adduced for a touch, and the paragraph around it is built on sensation: cum tangitur affectu placito diligentis… qui si tactu et amore digne credentium compungatur. The Vulgate's novi would put knowledge where Philip's argument needs feeling.

§4 · PATTERN 7 AND PATTERN 10 — four [sic:], and what was NOT marked

The four carried runs. None of the four is ALSO translated in the running text (brief §8's named failure).

1. 0030 @0256C [sic: a vida] — et alter alterum non fastidit, sed fauce semper a vida satiatur. Pattern 10, split type: a vida is not a Latin phrase (vida is not a word), and the run is avida, agreeing with fauce — "with an ever-greedy throat". The whole broken run is carried, spacing included; the English reads "with a throat ever [sic: a vida] is satisfied" and does not also supply "greedy".
2. 0031 @0260B [sic: munos] — Ipsa prior, ipsa felicior accipit grande munos. Read munus, certain (grande munus). Carried untranslated; the English does not also say "gift".
3. 0032 @0263A [sic: cam] — nulla adversitas a proposito cam praevalet removere. Read eam, certain. The English reads "no adversity prevails to remove [sic: cam] from her purpose" and supplies no pronoun.
4. 0035 @0273A [sic: idco] — idco forte divinus sermo vineas appellavit. Read ideo, certain. The English does not also say "therefore".

Rendered as printed, no marker (real words, merely wrong or merely odd) — conjectures live here and NOT in the text:

• 0263C, fornax exstitit et incudo. The slot wants the nominative incus; incudo is a real word (1 sg. of incudo, "I forge on an anvil") and cannot fill it. Migne himself sets the correct form two columns later (tanquam in incude macula tunditur, 0265A). Rendered "anvil" as an inflectional error under Pattern 12's "real word merely wrong" clause; ⚑ if the merge judges incudo a non-word in this slot, the fix is one [sic:] and I flag it rather than defend it.
• ⚠ 0265A, non viandi completio, non gaudium ei acceptio, sed promissio dat coronae. This clause does not construe as printed and it is the passage in the range most likely to be quietly repaired. gaudium is accusative in a list of nominatives, and coronae cannot be the object of dat. Rendered as closely as the printed words allow — "not the completing of the wayfaring, not joy, the receiving of it for her, but the promise gives of the crown" — which is meant to read wrong. Conjecture: non gaudii ei acceptio… sed promissio dat coronam. No emendation made, none smuggled into the English.
• 0262C, Inquieta undarum profunditas… non sufficiunt admirari, quod festinant adversus Virginem… congregari. Rendered actively ("cannot wonder enough that they hasten…"), the waves personified — which is what obstupescunt attendentes eleven words later requires. A passive reading ("are not enough to be wondered at") is possible; logged, not adjudicated in the text.
• 0266B, nec de utero altero concepisse. Rendered "nor had conceived of another womb". Odd as printed (de viro altero is the obvious sense); a real word, so rendered, not carried.
• 0273B, temptationis oculis — see §2, 0035 table.
• 0274A, in luctu fortior habeatur. Rendered "the stronger in the wrestling": luctus for lucta is attested medieval usage (Pattern 9, no marker), and the paragraph is unambiguous — athletam strenuum, cum recens accingitur in agone, with the wrestler's oil three lines above; the same chunk-neighbour uses lucta correctly (0270A, sopita lucta carnis).
• 0273B, volverit aut valuerit perscrutari. Rendered "shall have turned it over or been able to search it out" — volvere "to turn over in the mind" is good Latin and construes with ipsum. The rhyme partner makes voluerit the likely intention; not read that way, logged here.
• 0272B, Migne's misplaced full stop, PRESERVED with its lower case. He prints potius quam secunda, ut aiunt. qui viderunt, vini peroptimi tam fecunda. The English reads "…rather first than second, as they say. who have seen it, so fruitful of the very best wine…". Brief §5's ruling is explicit — lowercase after a misplaced stop STANDS, because English has an exponent for it — and 11638 §3 and cruces-0012 §3 both settled the construction. ⚠ Do not churn this.
• 0272B, the second stop, likewise preserved: ut ejus gaudium excellentius caeteris liquido cognoscatur. Botrus, de quo sermo est… — a stop cutting a purpose clause from the main clause it depends on. Two sentences in the English, as printed.

Negation audit (brief §9). Every non / nec / neque / nisi / nullus / nihil / nunquam / haud in the six Latin chunks was matched against the English on a separate second pass. The dense clusters are 0256D (non laedit macula, non corruptio… non sordescit… non pallescit), 0257A–B (nullius meriti, nullius… alterum ultra nescit… nec sic alterum vult habere, nec in ligno viridi vult sedere), 0257D–0258A (non se tamen Christum, non esse Sponsum… eligit non mentiri… non se Sponsum asserit, sed amicum), 0258B (non potest non gaudere — a double negative carried as a double negative), 0258C (non germinat, non justo germine fecundatur, nisi Sponso largiente), 0262C (Impellunt, non expellunt; convellunt, non evellunt; concutiunt, non excutiunt), 0262D (Nullum quippe scrupulum, nullum offendiculum), 0264B (nec vermiculi nec argenti), 0265B (haud dubium quin et Spiritus sanctus — carried as "there is no doubt but that"), 0269C–0270A (non solum non eligant consentire… Neque enim possumus… non penitus relegare), 0272C (aliis varie diffidentibus non diffisa), 0272D (non consuetudo, non necessitas), 0273A (Engaddus vineas dicitur non habere), 0274B (nulli par esse creditur non secunda). No negation was dropped and none was supplied.

§5 · TYPOGRAPHY AND CAPITALIZATION — this range crosses no seam either

Brief §7 holds without exception across cols 0256C–0274C, and this range adds two more instances of the decisive proof. Guillemets carry every non-lemma scripture quotation (42 runs); italics carry the running Song lemma, printed locators and Migne's marginal page numbers 146–155. The Song quoted as a proof-text takes guillemets — « Cum esset, inquiens, Rex in accubitu suo [n: (Cant. I)] » at 0270B and « Tenui eum, nec dimittam [n: (Cant. III)] » at 0271A, both guilleted while the italic lemma sits a few lines away in the same column. That is now Cant. 4:11 @0220A, 5:11 @0226A, 2:14 @0226C, 1:12 @0270B and 3:4 @0271A — five instances, three ranges, one rule. 11632 has no seam and no one should go looking for one.

Guillemet asymmetries: none in this range. All six chunks balance and every pair is 1:1 with the Latin twin. Pattern 5 inquit / ait / inquiens / ait sacerdos interruptions are very dense in this stretch — Philip drops one into roughly every second quotation — and every one stays inside the marks, in the italic lemma runs (Cum esset, inquit, Rex in accubitu suo) exactly as in the guillemeted ones.

Two printed locators inside guillemet runs, neither wrapped in [n:]: (ibid) at 0260D and (Matth. [0267C] XVI) at 0267B — the latter with a column anchor inside it. Both reproduced exactly, italics and all; neither counts toward noteCount (verified against the body).

⚑ §7a mirroring — the drift point — held word by word. The range is full of mixed case inside single sentences and none of it was regularized: Virgini… virgo diligens hilarescit (0258A) · cum ea sponsus est (0258B, lower case in a chunk that capitalizes Sponsus eleven times) · in utero virginis Deus homo (0259C) · Patrem cum filio et sanctum aeque Spiritum (0265B) · Sponsae virgini loquens (0265B) · nardus virginis (0268A) beside Virginis throughout 0266 · Dilectum vero suum virgo (0270A) · amore sponsi (0269D) · Bonus iste Fasciculus (0270B, capital) · nostra virgo (0268D) and hujus virginis (0273D, 0274A) in a chapter that capitalizes Virgo nowhere after 0272A. The Cant. 1:13 lemma prints Dilectus capitalized once and lower case three times. All mirrored. This is mechanically checkable against the Latin twin and I would check it first in any review.

Pattern 17. Philip's own frame is plural throughout and takes you (accepistis 0259C, vobis videtur 0259D and 0263C, nostis 0261A and 0264A, Videtis 0258A, 0263C, 0269C, Habetis 0264C, Nostis 0269A, legistis 0270C, accepistis 0273A). Every second-person singular is inside a quotation and every one is singular in the Latin — vincula colli tui, captiva filia Sion; si transieris… non nocebit in te; dereliquisti me; dedisti eis… laborem tuum; consentiens adversario tuo, dum es in via; faciemus tibi — so thou throughout. The one plural inside a quotation is Matt. 11:29 Tollite jugum meum super vos, et discite, rendered you, sitting four lines from the singular Isa. 52:2 — the number distinction is visible on the page exactly as Pattern 17 intends.

Ruled vocabulary held (brief §5): charitas = charity throughout 0032, including the doublet dilectio sive charitas (0263B) which is split "love, or charity"; misericordia = mercy (0266D, 0273C); torcular = winepress (0263D, 0272B); species = spice (0269A, of myrrh); adolescentulae = maidens (0265C, 0274B); recti, cellaria, clementia, nigra/fusca/formosa do not occur. Sponsus = Bridegroom, Sponsa = Bride, unguentum = ointment, nardus = nard (and nardus pistica = spikenard at 0268C, where Migne's own Latin distinguishes them).

One new fixed equivalence this range had to set, offered to the merge: accubitus = "repose", with accubuit / accumbens / accubitui rendered "reposed" / "reposing" / "his repose" throughout Caput IX. The chapter turns on the single root — in accubitu suo → in meo utero Rex ille gratanter accubuit → cujusmodi accubitus accumbenti complacuit → Sponso accumbente in dextera divinae majestatis — so a Douay-style "at his repose" for the lemma and "reclined" for the verbs would have severed it. recumbenti (John 12, of Christ at table, 0268C) is kept distinct as "reclining", because Migne's Latin distinguishes it.

§6 · THINGS DELIBERATELY NOT REPORTED (brief §10, checked once)

• Column-band arithmetic: no anchor in this range is followed by text broken mid-word or mid-construction. The bands run cleanly A–D. No [ed:], no patch, no plate check. §10.1.
• noteCount matches the body [n:] count in all six chunks (5/5/19/17/10/7), verified by script — and note that the two printed locators (§5) are correctly outside that count.
• Migne's sigla, all reproduced verbatim, none corrected (§10.2): (Rom. XIII) for Rom. 11:16 @0258C · (Matth. XXI) for Matt. 26:10 @0268D. Two in six chunks — notably lower than the five-in-six rate cruces-0012 §5.2 recorded for cols 0212–0227, which is itself a datum for the merge's count of that class.
• Attested medieval spellings left alone, unmarked (Pattern 9): mirrha, muraenula (for murenula), tanquam, quodcunque, utquid, praetiosae beside pretiosi, Quodpropter, siquis, coelo, spiritalis, luctus for lucta, aethnicum/ethnicum.
• No [f:]. 11632 is lemma-and-gloss, not a florilegium (§10.4).

§7 · WHAT A LATER BLIND READ SHOULD LOOK AT FIRST

1. 0265A, non viandi completio, non gaudium ei acceptio, sed promissio dat coronae (§4). The English is deliberately unconstruable because the Latin is. This is the single most likely place in the range for a well-meaning repair.
2. 0264C, the Ps. 11:7 [var:] (§3.1). It is on an [n:]-only span, which means nothing on the page marked it as a quotation and nothing mechanical would ever have found it. If it is wrong, it is wrong invisibly — check the left edge of the quotation, which I judged to begin at eloquium and not at Habetis.
3. The Cum / Dum lemma of Cant. 1:12 (§2, 0033). Four printings, no marker fired. Either the merge rules it Philip's Bible (brief §6 shape) or it is a compositor's habit; the decision needs the other twenty ranges' evidence, not mine.
4. §1's muraenula decision. It is the only term choice in this range that would be expensive to reverse late.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering Book II, Capp. XIII–XXI: the exposition of Cant. 1:14–2:3 (Ecce tu pulchra es, amica mea → Ecce tu pulcher es, dilecte mi, et decorus → Lectulus noster floridus; tigna domorum nostrarum cedrina → Ego flos campi, et lilium convallium → Sicut lilium inter spinas → Sicut malus inter ligna silvarum). Six chunks, 6,284 Latin words. The Song's own chapter boundary is crossed: Migne prints his bracketed [CAP. II.] at col. 0281B, inside chunk 0038.

Marker totals for this range: 4 [var:] · 1 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs clean on all six (no errors); one ratio warning, 0037 at 1.61, inside the band this work has already shipped (0004 1.64, 0005 1.68, 0026 1.66, 0035 1.65). Column anchors (11/10/10/4/9/16), [n:] contents (8/6/3/7/4/13, each matching its frontmatter noteCount) and guillemet counts (3/2/0/4/3/12 pairs) were diffed 1:1 against the Latin twins by script, not by eye; all six frontmatter blocks are byte-identical to the Latin. Every italic run in the English is accounted for against the Latin's own runs plus the one [sic:] and the four [var:].

§1 · THE COLLATION — all 65 listed spans, every one accounted for

Every span in data/quotations/11632/0036-0041.md was collated word by word against the Vulgate. Spans recorded as AGREES were collated and found to agree; the absence of a marker on them is a finding, not a silence.

0036 (cols 0274D–0277B) — 14 spans

Weaves checked and left unmarked (Philip's own prose, no notes attached, recorded so a later reader knows they were read): Ps. 44:3 (speciosus forma prae filiis hominum, 0274C, and again formam prae filiis hominum pulchriorem, 0276B) · Isa. 53:2 (non habentem speciem vel decorem, 0276B) · Horace, A.P. 365 (juxta poetam decies repetita, 0275B — Migne sets it in neither guillemets nor italics, so there is nothing to reproduce).

0037 (cols 0277C–0279D) — 10 spans

0038 (cols 0280A–0282B) — 7 spans

Also in this chunk: Migne's own bracketed [CAP. II.] at 0281B, the Song's chapter division, printed immediately after the column anchor and outside the italic lemma. Reproduced verbatim and untranslated, on the footing of [ al., …] — it carries no [n:] wrapper and does not count toward noteCount (verified: 3 notes in body, 3 in frontmatter). The same form recurs at 0063, 0074, 0091, 0108, 0114, 0120 for Capp. III–VIII, so later ranges will meet it.

Weaves checked, unmarked: 1 Tim. 6:16 (lux lucem habitans inaccessam — Migne's accusative lucem… inaccessam for Vulg. lucem inhabitat inaccessibilem; Philip's own apposition) · Isa. 11:1 (Radix Jesse germinavit et flos prodiit de radice) · Ps. 71:6 (descendente, ut in vellus, pluvia spiritali) · Matt. 5:15 (in conclavi sub modio).

0039 (cols 0282C–0283B) — 9 spans

0040 (cols 0283C–0285C) — 6 spans

Weaves checked, unmarked: John 1:5 (lux lucere in tenebris), Rom. 13:12 / 1 Thess. 5:5 (filia lucis lucet, non habens opera tenebrarum), Ps. 136:8 (filiae fuerant Babylonis).

0041 (cols 0285D–0289C) — 19 spans

Weaves checked, unmarked: John 1:12 (eos qui eum receperunt… filii), Rom. 8:29 (in multis fratribus ipse primogenitus), Heb. 2:16 (non angelum… sed semen Abrahae… apprehendit), Eph. 3:9 / Gal. 4:4 (absconditum a saeculis, nostri plenitudine temporis), Acts 4:12 (nec erat nomen aliud in quo datum fuerit nos salvari), Gen. 1:26 (ad imaginem et similitudinem Dei), Gen. 2:9 (lignum vitae… in medio paradisi).

§2 · THE [n:]-ONLY CLASS — hit rate, and two logged candidates

⚑ Hit rate reported as brief §8a requires. The range carries 21 [n:]-only spans. Nineteen are Philip's own prose, itemized above. Two are genuine weaves whose wording diverges from the Vulgate in a way that matters — and per brief §8a both are logged with full evidence and NOT marked, because the plate marks neither with guillemets nor italics and we cannot establish the quotation's left edge without asserting a boundary Migne does not print.

That is 2 of 21 ≈ 10%, against wave 1's Rom. 8:10 (1 hit) and the 0018–0023 agent's null result (0 of 19). The separate second pass over the [n:] notes against the prose, which the brief mandates, is what produced both — neither was visible from the extractor's clause alone, because in both cases the printed clause the script hands you begins in the wrong place.

2.1 ⭐⭐ 0277B — Wisd. 7:26 + Heb. 1:3 fused, with imago veritatis for imago bonitatis

Migne prints, inside Philip's own quia-clause:

Pulcher, inquam, es, quia candor lucis aeternae, speculum sine macula divinae majestatis, paternae splendor gloriae, figura substantiae, imago veritatis [n: (Hebr. I)]

Three things at once:

1. The note names Hebrews only, but the first two members are Wisdom 7:26 (candor est enim lucis aeternae, et speculum sine macula Dei maiestatis, et imago bonitatis illius), and only splendor gloriae, figura substantiae is Heb. 1:3. A two-book fusion under a one-book siglum.
2. Wisdom's imago bonitatis is printed imago veritatis. The change is real and is not a truncation: bonitas and veritas are different predications of the Son.
3. Wisdom's Dei maiestatis is printed divinae majestatis.

Not marked, and the reason is exactly brief §8a's: the run is set in neither guillemets nor italics, it is grammatically a string of appositives inside Philip's own sentence (the finite verb is his es), and the extractor's left edge — Pulcher, inquam, es — is his prose, not the quotation's. If the merge rules the class markable, this is the site in this range that takes the marker, and imago veritatis is the word to put in it. Nothing in the surrounding exposition picks the word up, which is why I do not think it rises to §3's level even if the class were opened.

2.2 ⭐ 0278D — Prov. 31:19, manum ad collum where the sense requires colum

Migne prints:

illa mulier cujus fortitudo praedicabilis commendatur, manum ad collum mittere, et fusum digitis apprehendere memoratur [n: (Prov. XXXI)]

The Vulgate of Prov. 31:19 reads Manum suam misit ad fortia, et digiti eius apprehenderunt fusum. Two separate observations:

• ad collum, "to the neck", cannot be right beside fusum, "the spindle". The word wanted is colum, the distaff — a real Latin word one l away, and the reading behind the Hebrew and the Old Latin of this verse. Migne's doubled l makes it the neck.
• Even colum would be a non-Vulgate reading, since the Vulgate has ad fortia here.

Rendered as printed — "to put her hand to the neck, and to take hold of the spindle with her fingers" — under Pattern 7. No [sic:]: collum is a real word merely wrong in its slot (Pattern 12's explicit exclusion, the fuges / munde class), and no [var:] because the span is [n:]-only. The conjecture colum lives here and not in the text. The tail of the same span, panem suum non comederit otiosa, is Prov. 31:27 (panem otiosa non comedit) recast into Philip's eo quod clause, with the negation carried.

§3 · THE FOUR [var:], IN FULL

1. ⭐ Luke 23:34 at 0284D — ignosce for dimitte, and quia for non enim. Migne: « Pater, inquit, ignosce illis, quia nesciunt quid faciunt ». Vulg.: Pater, dimitte illis: non enim sciunt quid faciunt. The main verb is a different word — ignoscere, "to pardon", against dimittere, "to remit" — and it is the Old Latin form of the most famous sentence in the Passion, which is precisely the 7a″ conformation hazard: the remembered English arrives before the printed Latin is read. The marker matters here because the sentence Philip builds round it reads cum manu misericordiae spinarum turpitudinem abradebat — a hand of pardon scraping the thorns, not a debt remitted. Rendered "pardon them" so that the divergence is visible in the English, per Pattern 7a's rule that a [var:] may not elide what it exists to report. The negation is intact on both sides (non enim sciunt / nesciunt) and is carried.

2. ⭐⭐ Isa. 35:1 at 0285A — germinabit for florebit, and Philip expounds the divergent verb twice. Migne: « Laetabitur, inquit, deserta et invia, et exsultabit solitudo; et germinabit quasi lilium ». Vulg.: et florebit quasi lilium. This is the range's clearest marker. Within nine lines Philip writes quia juncta lilio quasi lilium germinabit, and then spe fructus exsultabit habens primitias germinantes — the whole close of Caput XX runs on budding-into-fruit, not on blossoming, and florebit would put the argument's keyword back in a verse that no longer contains it. Everything else in the verse agrees word for word (deserta et invia, exsultabit solitudo), which is what makes the one substitution legible as a reading rather than a paraphrase.

3. Acts 10:26 at 0286D — an added clause, sicut et tu. Migne: « …Petrus autem levavit eum dicens: Surge, et ego ipse homo sum, sicut et tu ». The Vulgate ends the verse at et ego ipse homo sum; the comparative clause is in no edition. It is material in its place: Philip's whole point is that Peter refused adoration by asserting parity with Cornelius, and sicut et tu states the parity the argument needs. (Also in the same span: Migne supplies the subject Cornelius, which the Vulgate carries in the preceding verse, and prints levavit / autem for elevavit / vero — neither marked.) ⚑ Note the marker's content deliberately quotes the Vulgate's wording, not Migne's, so that it cannot collide with Pattern 14's mirror test against the Latin twin.

4. ⭐ Eph. 2:3 at 0288D — Sumus for eramus, present for imperfect. Migne: « Sumus, inquit, natura filii irae ». Vulg.: et eramus natura filii irae, sicut et ceteri — "we were". The tense is the point in the place Philip puts it: the sentence he is proving reads Nos autem miseros multo minores angelis culpa fecit… eo quod versum est vitium in naturam, an argument about our standing condition, not about a state Paul has left behind. The Vulgate's past tense would make the proof-text say the opposite of the sentence it is adduced for. ⚑ Cross-reference for the merge: cruces-0024 recorded Eph. 2:3 as an unmarked weave at 0030 (irae filius ex natura). This is the same verse returning as a guilleted proof-text with the tense changed — a second attestation, and worth counting.

§4 · MIGNE'S SIGLA — four misnamed references in six chunks (brief §8b)

All reproduced verbatim, none corrected. The count is the point.

1. ⚠ (Psal. XCII) at 0275B for « Deus, Deus meus ». Ps. 92 is Dominus regnavit, decorem indutus est and contains no such words; the doubled vocative is Ps. 21:2 or Ps. 62:2. And the same siglum is CORRECT eleven columns later at 0281A, where it points at Ps. 92:5 domum tuam decet sanctitudo. One siglum, two uses, one right and one wrong, inside one range.
2. ⚠ (Matth. XX) at 0274D for Matt. 22:40 — off by two chapters.
3. ⭐ (Marc. XV) at 0283B for wording that is verbatim MATT. 27:54. Migne prints « Vere Filius Dei erat iste »; Mark 15:39 reads Vere hic homo Filius Dei erat, Matthew Vere Filius Dei erat iste. The decisive evidence is in the previous range: cruces-0030 records the same six words at 0267D under [n: (Matth. XXVII)]. So Philip quotes Matthew twice and Migne labels the second occurrence Mark. This is the cleanest instance of the class so far in this work — the control is internal, not editorial.
4. ⚠ (Matth. XXV) at 0286C for fratres eos non confunditur vocare + cohaeredes, which are Heb. 2:11 and Rom. 8:17.

Also recorded as a form rather than an error: (cap. I) at 0281C, a bare chapter number with no book name, the book carried by Habacuc in the running prose. And the (ibid.) at 0286D resolves correctly to Acts X — do not propose a correction upstream of it (§10.2).

§5 · PATTERN 7 / 10 / 12 — one [sic:], and what was NOT marked

1. 0283A — [sic: ooopertus]. Migne prints Deus est in homine mundae carnis trabea ooopertus. Three o's: not a word in any Latin, and therefore Pattern 7's non-word class. Read coopertus, "covered", certain from the slot (trabea ablative of means) and from opertorium / occultavit five lines away. Carried untranslated in italics inside the marker, and the English does NOT also supply "covered" — the failure brief §8 names ("a [sic:] on a word that was ALSO translated passes"). The clause reads "is God in man, [sic: ooopertus] with the robe of clean flesh."

Rendered as printed, no marker (real words, merely wrong or merely odd) — conjectures live here and NOT in the text:

• ⚠ 0287C, « Conservus tuus suum, et fratrum tuorum… ». The Apoc. 19:10 re-quotation has lost its only finite verb: suum stands where sum belongs, and Migne himself prints the correct conservus enim tuus sum nine lines above at 0287A, in the full quotation of the same verse. So the conjecture is certain and the plate refutes itself on the same page. Even so, suum is a real word, which puts it in Pattern 12's explicit exclusion (the fuges / munde class: "rendered as ordinary English under Pattern 7, because they can be rendered"), so no [sic:]. The English reads « Thy fellow-servant his own, and of thy brethren that have the testimony of Jesus, adore God. » — verbless and wrong, deliberately, with Migne's commas preserved. ⛔ This is the single most likely place in the range for a well-meaning repair. ⚑ If the merge judges suum a non-word in this slot (the incudo question cruces-0030 §4 left open), the fix is one [sic:] and I flag it rather than defend it — the two cases should be decided together.
• 0283C, nonnullas videtur sentire functiones. functio has no sense here that construes cleanly against its rhyme-partner laesiones; medieval usage (onus, munus) would give "burdens". Rendered neutrally and literally as "workings", from fungi; conjectures concussiones / fluctuationes are possible and not supplied. Logged rather than adjudicated in the text.
• 0275A, ut humilem spiritu reddat virginitatis flosculus clariorem, flore claram virgineo, spiritu humilis pulchriorem. A chiasmus with only one finite verb: virginitatis flosculus is the first subject, and the second is spiritu humilis — a nominative adjective, where spiritus humilis would be expected. Rendered as printed ("the little flower of virginity renders her that is humble in spirit the brighter, and one humble in spirit renders her that is bright with the virginal flower the fairer"), the second verb supplied by English zeugma only. Conjecture spiritus humilis, not supplied.
• 0275B, nec videtur diligenti lectio haec justo amplius requisita. dilectio would be the obvious sense in a sentence whose subject two clauses earlier is dilectio; but lectio is a real word and construes with the Horatian tag that follows (juxta poetam decies repetita — A.P. 365, of a poem that will please though ten times re-read). Rendered "this reading." Logged, no emendation.
• 0279B, quem tu manu gratiae, ego laboris et industriae serenavi.* The verb is first-person singular and governs a second-person subject as well (likewise adornavi* in the next clause). English "have made serene" / "have adorned" carries both persons; the Latin zeugma cannot be reproduced exactly. Logged so the smoothing is on the record.
• 0288D, ut sancta confirmatio aeque vestigio permaneret. vestigio has no governing preposition. Rendered "should equally remain in its footing"; conjecture in vestigio, not supplied.
• 0277D, auctorem mortis Zabulum triumphares. Zabulus for diabolus is attested patristic and medieval Latin (Pattern 9): translated normally as "the devil", no italics, no [sic:].
• 0286B, copulantur nectura naturali. A rare verbal noun from necto; rendered "by a natural binding". Not marked.
• 0287A, nostri plenitudine temporis deopertum.* Likewise rare, and transparent (de- + operio*); rendered "uncovered". Not marked.

Negation audit (brief §9). Every non / nec / neque / nisi / ne / haud / nunquam / nullus / nihil / nil / nescit / necdum / nondum in the six Latin chunks was matched against the English on a separate second pass. The dense clusters are 0274C (non prior, non posterior ulla), 0275D (quibus haec naturae dona non omnibus omnia conferuntur — the double restriction carried as "not all of them bestowed upon all"), 0276A (non capitis lumen occiduum), 0276C (te non nisi deformem, non nisi hominem intuetur), 0276D (unum tamen, et non alterum), 0277A (nullius conceptio, nullius caruit vitio), 0277B (non possum non amare — a double negative carried as a double negative, "I cannot not love thee"), 0278B–D (nescit amor… non tam laborare… panem suum non comederit otiosa), 0280D (enodes… nunquam recalcitrat; scrupulus non habetur), 0281B (quid habet quod ab alio veraciter asserat non habere? — the negation is inside the indirect statement and stays there), 0281C (ne vel nesciri possem, vel a sciente non facilius apprehendi), 0281D (non nisi operi manuum respondet), 0282C (nec nisi super humilem), 0283C (nullas sustinet laesiones), 0284C (Nunquid non lilium… diffundebat? — both particles carried in the English interrogative), 0284D (Quis nesciat), 0286A (non est salus, non est vita, non indeficiens), 0287B (Non vult Christo angelus… non ligno fructuoso), 0287D (non tamen multum minor), 0289A (vident nec invident), 0289B (non fruges, non aries, non vitulus, non denique homo vel angelus; nec erat nomen aliud). No negation was dropped and none was supplied.

§6 · TYPOGRAPHY AND CAPITALIZATION

Brief §7 holds without exception across cols 0274D–0289C. Guillemets carry every non-lemma quotation adduced as authority (24 runs, all scripture in this range — no classical or oracular material occurs here); italics carry the running Song lemma with its piecemeal re-quotations, and Migne's marginal source-edition numbers 156–164. No seam, no exception, nothing tidied.

Guillemet asymmetries: none. All six chunks balance and every pair is 1:1 with the Latin twin (3/2/0/4/3/12 pairs). ⚑ Chunk 0038 carries zero guillemets across 1,004 words — an entire chapter and a half of exposition in Christ's own voice with not one adduced proof-text, which is worth noting only because a parity check that assumes at least one pair per chunk would misfire.

Pattern 5 interruptions (ait, inquit, inquam, inquiens, ait David) are as dense here as in the preceding range and every one stays inside the marks, in the italic lemma runs (Ecce, inquam, tu es pulchra,) exactly as in the guilleted ones (« Minuisti eum paulo minus ab angelis, ait David »).

[n:] placement varies and is reproduced as printed, per brief §7: usually inside the closing », but at 0283A it falls six words OUTSIDE, after Philip's sicut certum est Scripturam perhibere. At 0286C the column anchor precedes the note (amare cohaeredes [0286C] [n: (Matth. XXV)]). At 0283B and 0287D column anchors fall inside guillemet runs, splitting rex … Israel and servum… jura.

⚑ §7a mirroring — the drift point — held word by word. Mixed case inside single sentences, none of it regularized: pulchram virginem bis dixisse beside Virgo mirabiliter perseverat (0274D–0275A) · mater Sponsi, Sponsa filii (0277A — Sponsi and Sponsa capital, filii lower, in one four-word phrase) · hominem de homine genitum sine Patre… Dei Filium sine matre (0284B — capital Patre, lowercase matre, one sentence) · natus unicus fuit Patris… de virgine, unicus item Matris (0286C) · Cui flori et lilio virgo studet (0284A) and inter istos… virgo non frangitur (0284B) in a chapter whose lemma-word Sponsa is capitalized four lines away · Dilectus Sponsae praeeminet (0287C, capital) beside dilectus meus in the lemma. All mirrored.

⚑ Two mirrorings I want the merge to see, because both look like errors and are not:

• Migne capitalizes Cedri / Cedrus / Cypressus as common nouns at 0280B–0280D and then writes ad cedri et cypressi similitudinem lowercase at 0281A. Mirrored — "beams of Cedar… because Cedars, as they say" — which reads oddly in English and is meant to.
• 0288A, sed ille Spiritus immortalis, of the ANGELIC nature, not the Holy Spirit. Mirrored as "that immortal Spirit". I flag it because a reader may take it for the third Person and a sweep may "fix" it; the Latin capitalizes it and the referent is settled by angeli natura six words earlier.

Pattern 3. ⭐ Fires twice on bare forms, exactly as the brief warns. 0282C, inclinantque aurem Scripturarum monitis → "the admonitions of the Scriptures"; 0287B, angeli nonnunquam in Scripturis filii nominantur → "in the Scriptures" — the identical bare form cruces-0030 found at 0260A and that 11638's sweep missed. A third, singular, at 0287D: si quod Scriptura perhibet → "what Scripture declares". A later sweep of 11632 must match bare Scriptura and Scripturae as well as divina and sacra. Also at 0283A, sicut certum est Scripturam perhibere → "Scripture".

Pattern 17. Philip's frame is plural throughout and takes you (Videtis 0275C and 0286D, audistis 0275C and 0283C, noveritis 0283C, legistis 0288A, potestis — none in this range). Every second-person singular is inside the Bride's or the Bridegroom's address to the other, or inside a quotation, and every one is singular in the Latin (tu pulchra es · accendisti · descendisti · es constitutus · processisti · quem tu manu gratiae · Vide ne feceris · conservus tuus · Tu ergo non angelum, sed Deum adora · Minuisti eum) — so thou throughout. No plural second person occurs inside any quotation in this range, so the Dorotheus effect does not show here.

§7 · RULED VOCABULARY (brief §5) — held, plus three new fixed equivalences

Held: amica mea = "my friend" (0274C, 0283C, 0285C — the Cant. 2:2 occurrence the brief predicts) · misericordia = mercy (0279B, 0282B ×2, 0284C) kept distinct from clementia (does not occur) · charitas (does not occur) · Pattern 3 as above · Sponsus = Bridegroom, Sponsa = Bride, Virgo = Virgin, Deitas / deitas = Godhead. muraenula, vermiculatas, nigra / fusca / formosa, cellaria, recti, adolescentulae, torcular, candidatio, tomus do not occur in this range. Cant. 1:12 does not recur here, so this range contributes no evidence to brief §8c's Cum / Dum count.

Three new fixed equivalences this range had to set, offered to the merge:

1. ⭐ lectulus = "little bed", lectus = "bed", never interchanged. Forced by 0279B: non tamen Virgo lectum, sed magis lectulum nominavit; quo nomine legem illam… adimplevit, qua tam verbo quam opere humilia tenere consuevit. The diminutive is the humility argument. Fourteen instances across 0037–0038. The lemma is Cant. 1:15 and will recur.
2. laquearia = "panelled ceilings" (not "rafters", Douay's word), because Caput XVII glosses them as laid beneath the beams and stretched wall to wall as a binding — a rafter is the tignum, which is the other half of the same lemma and must stay "beam".
3. malus = "the apple tree" throughout Caput XXI, including where Migne writes neuter malum (0287B) — the chapter's three-term proof (umbra, fructus, odor) is about the tree, not the fruit, and poma is separately "apples" at 0286A.

Also fixed within the range for consistency: accubitum = "repose" (0278C), following cruces-0030's ruling · solitudo = "solitude" throughout 0285A–B, because Isa. 35:1's solitudo and Philip's nostra… solitudo are the same word and the gloss depends on it · scrupulus / scrupulosa = "scruple" / "scrupulous" (0280D), keeping the pun between the cypress's knotless wood and scrupulous obedience · convalles = "deep valleys" against valles = "valleys", which Migne's own Valles, imo convalles distinguishes.

§8 · THINGS DELIBERATELY NOT REPORTED (brief §10, checked once)

• Column-band arithmetic: no anchor in this range is followed by text broken mid-word or mid-construction; the bands run cleanly A–D and every chunk's colFirst / colLast matches its body. No [ed:], no patch, no plate check. §10.1.
• noteCount matches the body [n:] count in all six chunks (8/6/3/7/4/13), verified by script — and note that [CAP. II.] at 0281B is correctly outside that count.
• Attested medieval spellings left alone, unmarked (Pattern 9): tanquam, quodcunque, quidquid, coelum / coelestis, spiritalis, nunquam, Zabulus, Hierusalem, mirrha-class orthography, utcunque, nonnunquam, Absalon, Maecenas as a common noun, haereditario, praelatio.
• No [f:]. 11632 is lemma-and-gloss, not a florilegium (§10.4).
• No [nt:]. All 41 notes in the range are bare citation sigla.
• Migne's own ( sic ) does not occur in this range.

§9 · WHAT A LATER BLIND READ SHOULD LOOK AT FIRST

1. 0287C, « Conservus tuus suum… » (§5). The English is verbless because the Latin is, and Migne prints the correct sum nine lines above — which makes this both the most obviously repairable spot in the range and the one where a repair would be least visible. Decide it together with cruces-0030's incudo.
2. 0284D and 0285A, the two Luke/Isaiah [var:] (§3.1, §3.2). Isa. 35:1 is the confident one — Philip uses germinabit twice — and Luke 23:34 is the judgment call, because ignosce / dimitte are near-synonyms and the marker rests on the argument round it rather than on the lexicon.
3. 0277B, the Wisd. 7:26 / Heb. 1:3 fusion with imago veritatis (§2.1). [n:]-only, so unmarked under brief §8a — but if the merge opens that class this is where the marker goes in this range. Check the left edge, which I judged to begin at candor and not at Pulcher.
4. 0283A, « Deus autem ignis est » (§1, 0039 table). Guilleted, abridged, and deliberately unmarked. The closest call in the range.
5. The four sigla of §4, and particularly (Marc. XV) at 0283B, which the previous range's (Matth. XXVII) at 0267D turns from a suspicion into a demonstration. That pair is the strongest single piece of §8b evidence the work has produced so far and the merge should count it.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering Book II Caput XXII–XXVI (Cant. 2:3–7: Sub umbra illius quem desideraveram, sedi → Introduxit me rex in cellariam vinariam → Fulcite me floribus → Laeva ejus sub capite meo → Adjuvo vos, filiae Hierusalem), then the explicit-less transition to LIBER TERTIUS. and Book III Caput I–II on Cant. 2:8 Vox dilecti mei. Six chunks, 6,343 Latin words.

Marker totals for this range: 7 [var:] · 1 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs reports no errors on any of the six; four ratio warnings (0044 1.62, 0045 1.65, 0046 1.68, 0047 1.61), all inside the 1.35–1.68 band this work has already shipped (0004 1.64, 0005 1.68, 0026 1.66, 0035 1.65). Column anchors (13/10/8/9/13/4), [n:] strings (16/7/5/5/4/4, each matching its frontmatter noteCount) and guillemet counts were diffed 1:1 against the Latin twins by script, not by eye: anchors byte-identical in order and place, [n:] contents byte-identical in order. Every italic run in the English is accounted for against the Latin's own runs plus the seven [var:] and the one [sic:].

§1 · ⭐⭐ THE THREE FINDINGS THE MERGE SHOULD CARRY

1.1 · Lam. 4:20 at 0291B — an expounded quotation Migne marks with NOTHING AT ALL, and the extractor could not see it

This is the range's principal find and it came from the §8a second pass, not from the list. After the guilleted half-verse « In umbra, inquiunt, ejus vivemus in gentibus » [n: (Thren. IV)], Philip turns back to the first half of the same verse and writes:

Ideo praemittunt: Spiritus ante faciem nostram Christus Dominus.

The Vulgate reads spiritus oris nostri christus dominus — "the breath of our mouth". Migne prints ante faciem nostram, which is the Old Latin / LXX reading (πνεῦμα προσώπου ἡμῶν, spiritus faciei nostrae). And the divergent phrase is the one Philip expounds — it is the subject of his next two sentences, three times over:

Christus Dominus ante faciem nostram spiritus tunc habetur, cum eum nostra fides jam non secundum carnem, sed secundum spiritum intuetur… Ideo Paulus… Deum et ea quae sunt ejus, ante faciem proponenda.

Conform the lemma to oris nostri and the whole ante faciem / posteriora Christi architecture of the chapter — Peter putting Christ's hinder parts first, Paul setting God before the face — collapses.

⛔ Migne marks this span with NEITHER guillemets, NOR italics, NOR a note. It is not in data/quotations/11632/0042-0047.md in any class, because the extractor keys the [n:]-only class on the note and there is no note. No [var:] fired, per brief §8a: where the plate marks the words as neither quotation nor lemma we cannot establish the quotation's left edge, so marking it would assert a boundary the plate does not print. Logged in full so the merge can rule the class with this candidate in front of it. It is the exact shape of Ps. 90:5–6 at 0245A (cruces-0024 §2c.2) and of Rom. 8:10 at 0226B, and it is the third instance of the class in three ranges — which is itself the argument that the class is real.

⚑ A methodological consequence worth stating. Two of this work's three [n:]-only-class finds were attached to a note; this one was not. A range that reads only the extractor's ⭐ list will miss this shape entirely. The pass that found it was: read every scripture-sounding clause in the prose, note or no note.

1.2 · ⭐⭐ Cant. 2:7 at 0299D — the head lemma prints Adjuvo for Adjuro, and the correct form stands seven words earlier

The head lemma of Caput XXVI is printed:

**Adjuvo vos, filiae Hierusalem, ne suscitetis, neque evigilare faciatis dilectam, quoadusque ipsa velit.

Adjuvo means "I help". The verse is Adiuro vos — "I adjure you" — and the entire chapter is about the adjuration: humano more obtestatur durius et adjurat · astringit eas durius et interserit juramentum · Sponsi monita, quibus adjurat filias Hierusalem · filiae Hierusalem adjurantur (twice) · Mandatum vero sic adjurantis · si, per quod adjuratur… and the verse returns in full at 0302A as « Adjuro vos, inquit, filiae Jerusalem, per capreas, cervosque camporum. »

⚑ The decisive detail: Migne prints the right form immediately before the wrong one. The last words of chunk 0045 (col. 0299C) are the lemma tag Adjuro, etc. — so within seven words of printed text the plate has Adjuro and then Adjuvo.

Rendered as printed ("I help you"), which is meant to look wrong, with a [var:] naming the received reading. Not a [sic:]: adjuvo is a real word and Pattern 12's "real word merely wrong" clause keeps it on the English side. ⚠ This is the sentence in the range most likely to be silently repaired by a later reader, and the repair would be invisible — one letter restores a perfectly smooth chapter opening.

The head lemma also omits per capreas cervosque camporum (truncation, not a variant; the words return at 0302A) — which is itself notable, because the whole back half of the chapter is an exposition of the roes and harts.

1.3 · ⭐ Cant. 2:4 at 0292D — rex added, and cellariam for cellam

**Introduxit me rex in cellariam vinariam: ordinavit in me charitatem.

The Vulgate has introduxit me in cellam vinariam, with no rex.

• rex is Philip's text, not a slip. His first sentence glosses it: Quem et quare nominet regem, supra recolo me dixisse — he takes the lemma to name the king and refers back to his own exposition of Cant. 1:3. Conforming would strand the gloss.
• cellariam is the harder call, and it matters because brief §5 rules cellaria = storerooms against cella vinaria = wine cellar, and this lemma fuses them. Philip's own prose immediately separates them again: cum in regis cellaria introductam Virgo meminit se fuisse; quae… non tamen esse vinaria memorantur. Nunc vero in cellam vinariam Virgo se perhibet introduci — and thereafter tota cella vinaria (0293A), sed cella redundans; domus vinaria (0293A), Hoc in cella vinaria Christus beneficium operatur (0295B). Six printings of the prose form against one of the lemma form.

Rendered so the distinction survives on the page: the lemma reads "the wine storeroom", the retrospect reads "the king's storerooms", and every prose occurrence reads "the wine cellar" — three English forms for Migne's three Latin forms. [var:] fired on the lemma for both divergences.

§2 · THE COLLATION — all 72 listed spans, every one accounted for

Every span in data/quotations/11632/0042-0047.md was collated word by word against the Vulgate. Spans recorded as AGREES were collated and found to agree; the absence of a marker on them is a finding, not a silence. Lemma re-quotations (Sub umbra, Et fructus, Ordinavit, Fulcite me, Dextera, Adjuro vos, etc.) are short repetitions of a head-text collated at its full occurrence; each was checked against its head-text and none re-quotes it with different words — with the single, decisive exception of Adjuvo / Adjuro (§1.2), which runs the other way: the re-quotation is right and the head is wrong.

2a · Divergent — [var:] fired (7)

2b · Collated and AGREEING or abridged only — no marker (recorded so the absence is evidence)

0042 — Cant. 2:3 lemma Sub umbra illius quem desideraveram, sedi; et fructus ejus dulcis gutturi meo — agrees word for word (and at all four re-quotations) · Heb. 11:6 « Oportet, inquit, accedentem ad Deum credere quia est; et inquirentibus se remunerator sit » — every word the Vulgate's; enim dropped and credere moved behind accedentem ad Deum. Transposition, no marker · Heb. 3:14 « Si tamen, inquit, initium substantiae ejus usque ad finem firmum retineamus » — agrees exactly · Matt. 7:21 « Domine, Domine » — agrees · 1 Thess. 2:12 « Ut ambuletis, inquit, digne Deo » — Vulg. ambularetis, imperfect subjunctive in a past-sequence indirect command; Migne's present is the same construction in primary sequence, and nothing is asserted that the Vulgate does not assert. Logged, no marker · 2 Cor. 5:16 « Etsi novimus Christum secundum carnem, sed nunc jam non novimus » — Vulg. et si cognovimus secundum carnem Christum; verb of the same family, word order. Checked specifically for negation loss under brief §9: the single non is carried. No marker.

0043 — Luke 10:27 « Ex toto, ait, corde tuo, et ex tota anima tua, et ex omnibus viribus tuis » — agrees word for word for what is printed; truncated at both ends (diliges Dominum Deum tuum before, et ex omni mente tua after). No marker.

0044 — Cant. 2:5 lemma Fulcite me floribus, stipate me malis, quia amore langueo — agrees word for word (and at both re-quotations) · 2 Sam. 13:1–4 « Factum est, inquit, ut sororem Absalon Thamar adamaret Ammon filius David… » — a heavily compressed but faithful stitch: drops post haec, speciosissimam vocabulo, cur non indicas mihi, and fratris mei; Absalon / Ammon for Absalom / Amnon is transmission orthography. Every word printed is the Vulgate's. No marker; see §3 for the guillemet asymmetry inside it · Acts 5:29 « Oportet obedire Deo, magis quam hominibus » — agrees (Vulg. oboedire oportet, transposition) · Acts 4:20 « Non possumus quae audivimus et vidimus non loqui » — Vulg. non enim possumus quae vidimus et audivimus non loqui. ⚑ The two verbs are transposed and enim dropped; both negations are carried in the English. Sense untouched, so no marker — but recorded, because a transposition inside a double negative is exactly where a silent repair would hide.

0045 — Cant. 2:6 lemma Laeva ejus sub capite meo; et dextera illius amplexabitur me — agrees word for word (and at all three re-quotations) · Matt. 8:6 « Domine, inquit, puer meus jacet in domo » — agrees, Philip splitting off paralyticus et male torquetur into his own question-and-answer outside the marks · Matt. 8:8 « Non sum, inquit, dignus ut intres sub tectum meum; sed tantum dic verbo, et sanabitur puer meus » — agrees, dropping the opening Domine · John 5:3 — agrees for every word printed; Migne breaks the verse into four separate guillemet runs with Philip's own questions between them (see §3), and abridges caecorum, claudorum, aridorum. No marker · Ps. 15:11 « Delectationes, inquit, in dextera tua usque in finem » — agrees word for word. ⚠ The siglum reads (Psal. LI); see §2d.

0046 — Ps. 100:5 « Detrahentem, inquit, proximo suo, hunc persequebar » — Vulg. detrahentem secreto proximo suo hunc persequebar. One adverb dropped, and the clause asserts nothing the Vulgate does not. Logged, no marker — same treatment as the dropped cito at Matt. 5:25 (cruces-0030) · Cant. 2:7 re-quoted Adjuro vos… filiae Jerusalem, per capreas, cervosque camporum at 0302A — agrees word for word, and is the control that convicts the head lemma (§1.2). Note Jerusalem here against Hierusalem five times elsewhere in the same chunk: mirrored, not regularized.

0047 — Cant. 2:8 lemma Vox dilecti mei — agrees (both printings) · Isa. 52:6 « Ego ipse, ait, qui loquebar, ecce adsum » — agrees word for word; opens mid-verse after quia.

2c · The [n:]-only class — hit rate 0 of 20 on the listed spans

All twenty ⭐ spans on the list were traced back to their real left edge. Every one of the twenty is Philip's own prose, written with scripture rather than quoting it: accusative-and-infinitive frames, third-person and first-person recastings, his own rhymed clause-endings. None takes a marker; a [var:] on unmarked prose would misrepresent it as a lemma.

Reported per brief §8a even though it is a null result, and it is a real datum about the class: 0/20 here, against 1/14 for the 0024–0029 range and 0/19 for 0018–0023.

The twenty, listed so nobody re-opens them: 0042 — 0290C 1 Cor. 14:38 (ignorans… legitur ignorari, recast as a passive maxim) · 0291A Heb. 11:1 (fides optatae praesentia non est rei, sed sperandae substantia, his own antithesis; note the siglum, §2d) · 0291B Heb. 11:1 again (id maxime habebunt prae oculis quod habere promerentur nonnisi spiritales) · 0291C John 6:64 (eum qui vivificat, Deum spiritum novit esse) · 0291D Eccl. 9:1 (nescit enim miser homo utrum odio dignus sit an amore — his own adjective, and the Vulgate's amore an odio reversed, inside his own clause). 0043 — 0293A Ps. 22:5 (quam praeclarus ille calix inebriat) · 0293A Acts 2:13 (tanquam madentes musto linguosos efficiat) · 0293B Isa. 12:3 (ex ipso vel in ipso fonte haurit aquas gaudio pleniori) · 0293C 1 Cor. 13:4–5 (malum non cogitat, non agit perperam… ordinat charitatem — Paul's predicates strung into Philip's own list, with diligit aequitatem, iram non suscitat, amorem nutrit added; all six negations carried) · 0293D Matt. 22:39 (amat, sicut scriptum est, omnem proximum ut seipsam) · 0294B Eph. 5:29 (cum nemo carnem suam odio habeat, subjunctive inside his cum-clause). 0044 — 0297A Ps. 110:10 (et intellectus bonus est omnibus, non habentibus, sed facientibus intellectum — the antithesis is Philip's, and so is intellectum for the Vulgate's eum) · 0297B Acts 5:41 (Gaudeam et cum sensero vos ire gaudentes a conspectu concilii, second person, in his own cum-clause). 0045 — 0299C Ps. 16:15 (mane astabo vigilans… sub ejus dextera, apparente ipsius gloria satiabor — an ablative absolute for the Vulgate's cum apparuerit gloria tua, and sub ejus dextera imported from the lemma). 0046 — 0302A Gal. 6:2 (Sic sua invicem onera portantes legem Christi satagunt adimplere) · 0302C Num. 23:10 (ne compellar… resilire; see §2d for the note's position) · 0302C Ps. 83:3 (si anima mea morte justorum moriens et deficiens in atria Domini concupiscat). 0047 — 0303C Ps. 118:66 (docet disciplinam, docet et scientiam, docet etiam bonitatem — the Vulgate's bonitatem et disciplinam et scientiam doce me recast into three third-person clauses; note the siglum, §2d) · 0303C John 3:29 (Ego igitur amica Sponsi sto et gaudeo propter vocem Sponsi — the Vulgate's masculine amicus… sponsi qui stat et audit eum turned feminine and first person; see §5.1 on amica) · 0303C Isa. 64:1 (Utinam disrumperet coelos et veniret for utinam disrumperes caelos et descenderes — third person and a different verb, inside Philip's own utinam).

Further unmarked weaves checked for silent conformation and left as Philip's own prose (none marked, none conformed): Isa. 9:2 / Luke 1:79 (in tenebris et in umbra mortis, 0289D) · Acts 2:2 + Luke 1:78 (spiritus vehemens… ex alto visitans oriens… lumine matutino, 0290A) · 1 Cor. 13:12 (videt per speculum, 0290B) · Ps. 30:15 / Ps. 117:28 (Tu es ipse Deus meus, 0290C — see §4) · Mark 14:15 (in coenaculo grandi strato, 0292A) · Ps. 45:5 (fluminis impetus civitatem, 0293A) · Luke 1:39 (ad montana provehitur, 0295A) · Phil. 1:23 (vult dissolvi, 0299A) · Eccles. 1:3 (sub sole laborantem, 0299A) · Eph. 5:16 (dum dies mali sunt, 0299B) · Ps. 102:5 (sicut aquila, juvenescam, 0299B) · Ps. 62:2 (de mane vigilabo, 0299B) · 1 Pet. 5:3 (non dominari clero, sed formam subditis se praebere, 0300C) · Ps. 13:4 (sicut escam panis, plebem devorant perditorum, 0301C) · Luke 17:10 (servos tamen se inutiles dicunt esse, 0301D) · Ps. 118:32 (viam currunt mandatorum, cum corde et animo dilatantur, 0301D) · 2 Tim. 3:5 (sine virtute non est pietatis species fructuosa, 0297A) · Ps. 26:5 + Isa. 4:6 (ab aestu et turbine et pluvia in abscondito tabernaculi, 0302C) · 1 Cor. 13:12 (facie ad faciem, 0296D) · Phil. 3:8 (pro Christo judicans detrimenta, 0298B) · 1 John 2:27 (cui dat intellectum quem docet scientiam unctio spiritalis, 0298C).

2d · Migne's sigla — four to add to the merge's accumulating count (brief §8b)

Reproduced verbatim, none corrected. Thirty-seven of the range's forty-one notes name their reference correctly, which is a markedly better rate than cols 0212–0227. The four:

1. 0291A [n: (Hebr. II)] for Heb. 11:1 — off by nine chapters, and the chunk's own correct Hebrews sigla ((Hebr. XI) at 0290B, (Hebr. III) at 0290D) stand a few lines away.
2. 0298D [n: (Psal. LI)] for Ps. 15:11 (Delectationes in dextera tua usque in finem) — off by thirty-six psalms.
3. 0303C [n: (Psal. CXII)] for Ps. 118:66 (bonitatem et disciplinam et scientiam) — off by six.
4. 0302C [n: (Num. XXIII)] is correctly named but wrongly placed — it sits after resilire, while the Numbers matter (morte justorum, Num. 23:10) arrives in the next sentence, which carries its own [n: (Psal. LXXXIII)]. Reproduced in place; §10.2 forbids moving it.

⚑ Also for the count, and of a different kind: 0290D [n: (Habac. II)] names the right chapter of the wrong book for a lemma whose wording is Heb. 10:38 (§2a), and 0291C [n: (ibid.)] on (Matth. XVI) does the same for wording that is Mark 8:33. Both are cited-X-worded-Y, the class cruces-0024 §2c.1 and cruces-0012 opened. This range has two, and both are inside guillemets — which is why both took a [var:] where the earlier instances, being on unmarked prose, could not.

§3 · TYPOGRAPHY — the Book II/III boundary is crossed, and there is NO SEAM

⭐ This range answers the question cruces-0024 §1 left for a later agent. Chunk 0047 opens ## LIBER TERTIUS. at col. 0303A (empty body, exactly as LIBER SECUNDUS. sat at 0246C), and ## CAPUT PRIMUM. follows under it. Migne prints the transition in Philip's own voice at 0302D — stylus meus ad prosequendum caetera non festinet, sed dormiente Sponsa, iste paululum se reclinet — and names the division himself: cujus ab osculo primus liber sumpsit exordium, in somno finem accipiat hic secundus.

Nothing about the marking changes across it. Both marks run in the same functional division on both sides: Book II side (0046) has the italic lemma, the italic marginal numbers 171–172, and a guillemeted Ps. 100:5; Book III side (0047) has the italic lemma twice, the italic marginal 173, and a guillemeted Isa. 52:6. So 11632 crosses two book boundaries with no seam at either, and 11638's book-boundary model is now falsified twice over on this work. The merge can stop looking.

LIBER TERTIUS. → "BOOK THREE.", after LIBER PRIMUS. → "BOOK ONE." and LIBER SECUNDUS. → "BOOK TWO." CAPUT PRIMUM spelled in words then CAPUT II in numerals, in the same chunk, preserved as printed (brief §7a) — and note that Book III restarts the numbering while Book II ended at CAPUT XXVI.

3a · ⚠ The guillemet asymmetry at 0295C — a NESTED unclosed «

0044 prints 4 « against 3 », and the extra opening is inside an existing quotation run:

« Factum est, inquit, ut sororem Absalon Thamar adamaret Ammon filius David… Dixitque Ammon: « Thamar sororem Absalon amo [n: (II Reg. XIII)]. »

Migne opens a run at Factum est, and then, at the point where the narrative gives Ammon his own direct speech, opens a second « without ever closing the first. The single » at the end closes the inner one. Reproduced exactly as printed, unbalanced (Pattern 5 corollary), and logged here rather than repaired. It is the first guillemet asymmetry recorded in this work since wave 1.

3b · Four guillemet runs carved out of ONE verse — John 5:3 at 0298A

Migne dismantles the verse and reassembles it around Philip's own questions, guillemeting each fragment separately:

« In his, » ait, porticibus « jacebat multitudo magna. » Quare jacebat, exponit: « Languentium. » Quare in porticibus, adjungit: « Exspectantium aquae motum [n: (Joan. V)]. »

porticibus — Philip's word, pulled up from John 5:2 — sits outside the marks, between them. This is why 0045 has seven guillemet pairs where a reader counting quotations would expect four. Reproduced 1:1; the English keeps the same four fragments and the same interleaved prose, at the cost of some awkward word order ("« In these, » he says, porches, « there lay a great multitude. »"). Same class as 0254B's « Meos » / « tuos » (cruces-0024 §1b): the mark is doing grammatical work, not quotation work.

3c · A printed locator outside [n:], again — (Jacob II) at 0297A

sic fides sine opere legitur et intelligitur otiosa (Jacob II). Italic, inside the running prose, with no [n:] wrapper, and correctly not counted in 0044's noteCount of 5 (verified against the body). Fourth or fifth instance of the class in this work (cf. 0203A, 0222B–C, 0244C, 0260D, 0267B). Reproduced exactly, italics and all.

3d · Marginal source-edition numbers

165 (0290A) · 166 (0291D) · 167 (0293C) · 168 (0295B) · 169 (0297A) · 170 (0298C) · 171 (0300B) · 172 (0302A) · 173 (0303C). Continuous with 101–155 from the earlier ranges. Each kept italic and verbatim at the exact word it interrupts — 165 falls between torpentem and invenire, 170 between desiderantibus and spondet, 173 between praestolantibus and non tacet — and the English clause was built round each rather than moving it.

3e · ⚑ §7a mirroring — the drift point — held word by word

Mixed case inside single sentences, none of it regularized:

• 0291D virgo Mater — lower-case virgo, capital Mater, in a work that capitalizes Virgo almost everywhere. Rendered "the virgin Mother".
• 0294A, 0294B, 0295C Virgo mater — the same pair the other way up. Rendered "the Virgin mother" all three times.
• 0292D non modicam virgo fiduciam sibi sumit and 0293A cujus meriti virginem aestimamus — lower case in chapters that capitalize Virgo eleven times. Rendered "the virgin".
• 0290A nonnullos miserans spiritus, vehemens terret — lower case, the Acts 2:2 spiritus vehemens. Rendered "the vehement spirit", against capital Spiritus at 0291B, 0292A, 0293A, 0293D, 0294D, 0295A, 0298C.
• 0293D vera Mater and 0295A in Matre diligente — capital, rendered "Mother".
• Sponsus / Sponsa / Sponsi / Sponsae are capitalized everywhere in this range (thirty-odd instances, no lower-case exception), which is itself worth recording against 0243B–C where Migne dropped the capital.

This is mechanically checkable against the Latin twin and I would check it first in any review.

3f · Pattern 17

Both frames in this range are plural and take you: the Virgin addressing the adolescentulae (vos interim… convenite, fulcite, removete, refovete, floreatis, proferatis, vos ire gaudentes, digni habiti estis, 0296B–0297B) and the Bridegroom addressing the filiae Hierusalem (ne suscitetis, neque evigilare faciatis, revocetis, vobis, volo ergo vos, adjuro vos, 0299D–0302C). Philip's own frame to his brethren is plural too (ut audistis, ne putetis, vobis profero, ut legistis, sicut recolitis vos legisse, de quo legistis).

Every second-person singular in the range is inside a quotation and every one is singular in the Latin — Ne tradas regnum tuum · Absit a te · non sapis · Ex toto corde tuo… ex omnibus viribus tuis · ut intres sub tectum meum · in dextera tua · Quare sic attenuaris macie · Tu es ipse Deus meus — so thou throughout. The one plural inside a quotation is 1 Thess. 2:12 Ut ambuletis digne Deo, rendered you, sitting four lines from the singular Matt. 16 pair. The number distinction is visible on the page exactly as Pattern 17 intends.

§4 · PATTERN 7 — one [sic:], and the readings rendered as printed

One carried run, and it is NOT also translated in the running text (brief §8's named failure):

1. 0294C [sic: reete] — Hunc itaque proximum reete Virgo diligit ut seipsam. Read recte, certain: Philip uses recte correctly at 0290A, 0290C, 0292B, 0297A, 0299D and 0303A. reete is not a Latin word, so Pattern 7 governs: carried untranslated in italics, and the English deliberately does not also say "rightly" — it reads "This neighbour therefore the Virgin [sic: reete] loves as herself." The plate sets the broken word between proximum and Virgo; English word order puts it one word later, immediately before the verb it modifies.

Rendered as printed, no marker (real words, merely wrong or merely odd) — conjectures live here and NOT in the text:

• ⛔ 0290D, quia fides mortua, etsi non diligens, operatur. As printed this says that a dead faith, though it does not love, does work — which is the reverse of the Jas. 2:20 maxim Philip quotes correctly at 0297A (fides sine opere… otiosa) and sits awkwardly against his own next sentence, Dilectio quippe vita est fidei. Rendered exactly as printed ("since a dead faith, even though it does not love, does work"). Conjectures: quia fides mortua est, si non diligens, operatur, or quia fides mortua, etsi non diligens, non operatur. Neither supplied. ⚠ This is the single most likely place in the range for a well-meaning repair, and under brief §9 the printed negation count is exactly what stands: one non, on diligens, and no other.
• 0290C, Tu es ipse Deus meus. A four-word sentence standing outside the Ps. 115:1 guillemets, with no mark of any kind, in a paragraph otherwise in the third person. It reads as the continuation of the psalm-verse in the psalmist's voice (cf. Ps. 30:15 Deus meus es tu, Ps. 117:28), but Migne marks it as nothing. Rendered as printed ("Thou art thyself my God"), no marker, no quotation supplied.
• 0290A, insolubilis vi necturae. nectura is not classical, but it is Philip's own word — he uses it again at 0294B, natura, naturalique nectura proximum suum esse. Pattern 9: attested usage of this author, unmarked. Rendered "the unloosable force of her binding".
• 0298B, the doubled comma. Migne prints Quae ideo laeva satis congrue nominantur,, quia in hac praesenti vita…. A doubled comma has no English exponent (Pattern 9's test) and reproducing it would read as our own build defect; one comma is set and the doubling is logged here. ⚑ If the merge prefers Pattern 8's "never delete a mark", the fix is one character at 0298B.
• 0292C, nondum ad ventrem pervenit, quo perfecta satietas, invenitur. Migne sets a comma between subject and verb. Not reproduced (English cannot carry it without reading as a typo of ours); logged.
• 0291A, fides optatae praesentia non est rei. The genitive rei is split from praesentia by non est, which is Philip's rhyme-driven word order, not a defect. Rendered "faith is not the presence of the thing wished for".
• 0299D, timorem incutiat sacramentum. sacramentum here means the sworn oath (juramentum two words earlier), not the sacrament. Rendered "sacrament", the word Migne prints, because English carries the same ambiguity and glossing it would be an emendation. Logged.
• 0301A, Verum in hoc dramate. drama for the Song's dialogue. Rendered "in this play"; not italicized, because English has the word.
• 0295A, carne mundum et Spiritu Deo cunctis inaccessum genere counitum. The word order is very hard; cunctis inaccessum is taken with Deo, and genere with counitum, giving "clean in flesh and by the Spirit united in kind to God who is inaccessible to all". An alternative construal takes cunctis inaccessum with the child. Logged, not adjudicated in the text.
• 0296C, malagma sive oleum, sive ligatura reputabitur aegrotanti. Rendered "a poultice, or oil, or a binding"; malagma is a real English-available word (emollient plaster) and is not carried in Latin.

Negation audit (brief §9), second pass, against the Latin. Every printed non / nec / neque / nisi / nunquam / nullus / nihil / nil / nondum / necdum / ne was matched against the English. The dense clusters are 0290C (Sine hac fide Deo quispiam non placebit, nec est digna conscientia, cujus Deum scientia non tenebit… qui nescitur, a nesciente non potest acceptari… qui non crediderit, nec dici debet esse; qui autem non est, consequenter nec in umbra est… his qui non sunt… Non esse recte dicitur, qui Deum esse nescit), 0290D (Etiam non justus erit in fide, sed nonnisi justus vivet in fide — carried as "Even one that is not just will be in faith, but none save the just shall live in faith"), 0291C (non fiet istud, non morieris… non sapis quae Dei sunt), *0293B–C (non perturbat… non confundit… non obscurat, non aufert scientiam, sed infundit; malum non cogitat, non agit perperam… iram non suscitat, non infert odium — six in one sentence), 0294A (non jubetur modus aliquis, non mensura… non scriptor aliquis docuit, non Scriptura… videtur non quantum, sed ex quanto), 0294C (non amplexum pertulit, non juncturam… nullum sensit dolorem corporis, non laesuram… nec omnino potuit… nec erat quis tam felix), 0295A (ut seipsam illi non solum non praeferat, sed nec etiam conferat in amore — a triple, carried as a triple), 0297A–B (non habentibus, sed facientibus… Non possumus quae audivimus et vidimus non loqui… non capitis injuria, non membrorum paralysi… non febris angustia, non corporis quavis molestia… nihil ardentius, nihil vehementius… videtur nec tenetur… nec potest inveniri), 0300C–D (non audimus, vel magis audientes… nullatenus obedimus… non dominari clero… non curis exterioribus, non negotiis saecularibus… non causis forensium, non saecularium contentionibus… necessitas eos referat, non voluptas), and 0302C (ut non solum non suscitem nec evigilare faciam quoslibet dormientes, nec* mandati reus inveniar — carried whole). No negation was dropped and none was supplied.

§5 · CONVENTIONS SET OR HELD IN THIS RANGE

1. ⭐ amica Sponsi = "the friend of the Bridegroom" (0303C). Brief §5's ruling on amica mea is honoured in its other grammatical form: Philip recasts John 3:29's amicus sponsi into the feminine and the first person, Ego igitur amica Sponsi sto et gaudeo propter vocem Sponsi. "Friend", never "beloved" or "love" — the whole point of the sentence is that the Bride, who is about to hear the Bridegroom's voice, is standing in the Baptist's place. This is the same lemma-word the merge must hold across the work.
2. New fixed equivalences this range had to set, offered to the merge: - langueo / languor / languens = "languish / languor" throughout (Caput XXIV–XXV turn on the single root: quia amore langueo → amoris incendio Virgo dixerat se languere → ne amor durius langueat, languor moram sufferat duriorem → nihil undique languidum relinquetur). Kept strictly distinct from aegrotare = "to sicken" and deficere = "to fail", because the Ammon paragraph at 0295C runs all three of them in two sentences and the argument is the contrast between the sick lover and the languishing one. - fulcire = "to stay up" · stipare = "to compass about", fixed at the lemma and held through fulta vel stipata (0296D) and se stipari (0296C). - otium = "leisure" against negotium = "business" — Caput XXVI's whole argument is the pair (otio vel negotio spiritali, spiritali otio quamdam velut injuriam, vestra debere negotia… sua vero otia), and any variation would dissolve it. - laeva = "left hand" · dextera = "right hand", nouns not adjectives, after the lemma; sinistra (0299B, 0299C) likewise "the left".
3. Pattern 3 fires on a bare Scriptura at 0294A — quantum diligatur Deus, non scriptor aliquis docuit, non Scriptura. Neither divina nor sacra is present. Rendered "Holy Scripture", per brief §5's ⚑ and the 0260A precedent (cruces-0030 §2). This is the second attested bare-form instance in this work; a later sweep must match bare Scriptura(e) too.
4. Brief §5 vocabulary held: charitas = charity throughout (0293C ordinat charitatem, ordo charitatis; 0295A–B charitatem praecipuam ordinavit, excellentia charitatis; 0296D, 0302A affectus charitatis), kept distinct from dilectio / amor = love (0290D Dilectio quippe vita est fidei; 0293C in caeteris pluribus dilectio sit diffusa; 0301D geminae dilectionis mandatum) — and note 0294A, where dilectione and diligatur and diligendi run in three consecutive clauses beside charitas, which is exactly the doublet pressure the rule exists to resist · candidatur (0293B) is translated ("is made white"), per the candidatio rule · cellaria = storerooms, cella vinaria = wine cellar (§1.3) · adolescentulae = maidens (0296A, 0296C, 0296D, 0299D) · Sponsus = Bridegroom, Sponsa = Bride. nigra / fusca / formosa, clementia, misericordia, recti, torcular, muraenula, moralitas do not occur in this range.
5. ## CAPUT XXII.–XXVI. → "CHAPTER XXII."–"XXVI."; CAPUT PRIMUM. → "CHAPTER THE FIRST." — the numeral/word distinction preserved as printed inside chunk 0047, which carries both forms.
6. 0303A's frontmatter incipit begins ## CAPUT PRIMUM. — i.e. the chunker captured a head into the incipit field. Copied verbatim into the English frontmatter, unaltered, per the brief's frontmatter rule. Not a defect of ours and not to be tidied.

§6 · THINGS DELIBERATELY NOT REPORTED (brief §10, checked once each)

• Column-band arithmetic: 0042's colContext is 0289C while its colFirst is 0289D, and 0046's body opens with a bare [0299D] on its own line before the lemma. Both are brief §10.1's class. Checked: no anchor in any of the six chunks is followed by text broken mid-word or mid-construction. The bands run cleanly A–D throughout 0289–0303. No [ed:], no patch, no plate check.
• noteCount matches the body [n:] count in all six chunks (16/7/5/5/4/4), verified by script — and note that the printed locator (Jacob II) (§3c) is correctly outside that count.
• Attested medieval spellings left alone, unmarked (Pattern 9): tanquam, coelum / coelestium / coelis, quandiu, inpraesentiarum, nonnisi, quodpropter, Hierusalem (beside Jerusalem, §2b), charitas, mirrha-class orthography, spiritalis / spiritaliter, ammonere (0303A, for admonere), Absalon / Ammon, Isaiae, aethnicum-class forms.
• No [f:]. 11632 is lemma-and-gloss, not a florilegium (§10.4).
• No [d:], no [nt:], no [ed:]: no dittography in the range, and all forty-one notes are bare citations, not editorial prose.
• Migne's own ( sic ) does not occur here.

§7 · OPEN — for the merge or a blind read

1. ⛔ The blind polarity read has not been done and cannot be done by me. The three passages most likely to be wrongly "repaired" are, in order: (a) Adjuvo at 0299D (§1.2) — one letter makes a smooth chapter opening, and the correct form stands seven words earlier on the plate, which makes the repair feel authorized; (b) quia fides mortua, etsi non diligens, operatur at 0290D (§4) — the English is meant to read against the Jas. 2:20 maxim, because the Latin does; (c) the carried reete at 0294C.
2. ⚑ §1.1's Lam. 4:20 is this range's candidate for the class brief §8a left open, and if the merge rules that an expounded unmarked weave earns a [var:], the marker goes at 0291B on Spiritus ante faciem nostram Christus Dominus. ⚠ Note that it is a harder case than Ps. 90:5–6, because it carries no [n:] at all — so a rule framed as "an expounded [n:]-only weave" would not reach it. The class the merge needs to name is "an expounded weave Migne leaves unmarked", note or no note.
3. ⚑ The cited-X-worded-Y count is now at least five across four ranges (0240B Luke/Matthew · 0290D Habakkuk/Hebrews · 0291C Matthew/Mark · plus cruces-0012's Deuteronomy/Psalms). Two of this range's three are inside guillemets, which is new. Worth a work-wide tally before anyone proposes anything, per §8b.
4. ⚑ §3's no-seam result at the Book II/III boundary should be recorded in the merge as closing the question, together with cruces-0024 §1's result at Book I/II. Two boundaries, two negatives; brief §7's ruling is now confirmed rather than assumed.
5. The doubled comma at 0298B (§4) is a one-character judgment I made against Pattern 8's letter and in favour of Pattern 9's test. If the merge disagrees, it is a one-character fix and I flag it rather than defend it.

Philip of Harvengt, Commentaria in Cantica canticorum, PL 203. Range covering Book III(?) Caput III–IX — the exposition of Cant. 2:8b–13: Ecce iste venit saliens in montibus, transiliens colles → Similis est dilectus meus capreae hinnuloque cervorum → En ipse stat post parietem nostrum → Et dilectus meus loquitur mihi → Surge, propera, amica mea → Jam enim hiems transiit → Vox turturis audita est in terra nostra. Six chunks, 6,351 Latin words.

⚑ No book head falls in this range; the chunks carry only CAPUT heads, and the numeration runs III → IX continuously across the six, so no division boundary is crossed here.

Marker totals for this range: 2 [var:] · 3 [sic:] · 0 [ed:] · 0 [d:] · 0 [nt:] · 0 [f:]. verify-english.mjs reports no errors on any of the six; one ratio warning, 0052 at 1.61, inside the band this work has already shipped (0004 1.64, 0005 1.68, 0035 1.65, 0046 1.68). Ratios: 1.52 / 1.53 / 1.51 / 1.49 / 1.61 / 1.56.

Column anchors (13/13/11/6/11/6), [n:] strings (16/16/10/8/9/10 — each matching its frontmatter noteCount) and guillemet counts (13/2/7/2/2/11 pairs) were diffed 1:1 against the Latin twins by script, not by eye: identical in content, order and position in all six chunks. No guillemet asymmetry anywhere in this range — every « has its », none stray, none unopened. Every italic run in the English is accounted for against the Latin's own runs plus the three [sic:] and the two [var:] (asterisk counts: 8→12, 4→6, 10→10, 16→16, 12→12, 6→12, the deltas being exactly the marker contents).