Commentarius in Cantica canticorum

PL 15, cols 1851A–1962B, 40 chunks (0000–0039), translated 2026-08-14 by eight agents working five chunks apiece. This file is the merge of the eight per-batch files cruces-0000.md … cruces-0035.md, which it replaces; nothing from them was dropped, and every table below is reproduced verbatim from the batch that compiled it.

Read §1 before touching the English, §2 before "fixing" anything that looks inconsistent, and §3 before re-opening the one open editorial question.

1. Work-wide conventions (fixed at merge)

Migne prints an Old Latin / LXX-derived Song of Songs throughout, and Ambrose's exposition repeatedly argues from the divergent word. The controlling rule for this work is therefore: one English word per Latin word, held across all forty chunks, because the moment two Latin words share an English word the author's own argument — and our [var: …] apparatus, which exists to show Migne's Latin against the Vulgate's — stops meaning anything.

1.1 The equivalence table (binding for the whole work)

Second person follows Pattern 17 work-wide: thou/thy for a grammatically singular addressee, you for plurals — by number, never by dignity.

1.2 What the merge actually had to repair

Four divergences between agents were found and reconciled in the English chunks, not merely noted here. Each is recorded so a later sweep does not churn it back.

1. proxima mea had three Englishes. "my neighbour" (0005, 0006, 0007, 0019), "my neighbor" (0011, 0012, 0023, 0024) and "my nearest one" (0025, three occurrences plus the gloss Nearest by charity). All now read neighbor, in the American spelling used by the style anchor (9741) and by the corpus at large. 0025's gloss now reads Neighbor by charity, a dove by simplicity, perfect by virtue, which is what Charitate proxima says. At 1864C the adjectival prose quae proxima eras was "thou who wast nearest" and now reads "thou who wast a neighbor", keeping the aliena / proxima antithesis on the same word as the lemma.
2. formosa had collapsed into pulchra. 0011, 0012 and 0019 rendered formosa "beautiful" — the word already carrying pulchra — so 1870A's formosus … et quidem pulcher and 1905C/1907A's formosa against the Vulgate's pulchra both went silent. 0011/0012/0019 now read well-formed (nine occurrences), matching 0006's Behold, my cousin is well-formed, and indeed beautiful.
3. speciosa had collapsed into pulchra and decora. 0004 and 0034 rendered speciosi/speciosa "beautiful"; 0011 "fair"; 0032 "beautiful"; 0033 "beautiful"; 0036 "comely". The apparatus at 1946D marks Migne's Quam speciosi pedes against the Vulgate's Quam pulchri, and at 1942B–C Migne prints speciosa sicut Hierusalem and then, one paragraph later, Pulchra ut Hierusalem — with both Englishes reading "beautiful", the [var:] note contradicted itself and the plate's own three forms of Cant 6:3 became one. All epithet uses now read shapely (0004 ×2, 0011, 0031, 0032 ×3, 0033 ×3, 0034 ×10, 0036 ×3), which is also what 0005/0006/0007 already had (How shapely are thy cheeks become, shapely and comely, blossomed the more shapely). ⚠ Deliberate carve-out: speciosus in free prose, where it is not an epithet of the bride or of the preacher's feet, keeps ordinary English — 1918B's speciosum aliquem locum ("some beautiful place") and 1918D's speciosus est usu ("beautiful in use"). "Shapely place" is not English. This is recorded so the carve-out is not read as residual drift.
4. decora inter mulieres was rendered "beautiful among women" once in 0004 against "comely among women" four times in the same chunk; now comely throughout. Two [var:] glosses of the Vulgate were normalized to the same table at 1953B/1953C (Vulg. pulchra → "beautiful as the moon", Vulg. formosus → "this well-formed one"), so that a gloss never renders a Latin word differently from the running text.

Checked and found already conformant: chunks 0000–0004, which were translated before the equivalence table existed — frater meus is "my brother" at 1860B, and consobrinus, fraternus and acclinatio do not occur in them at all; charitas is "charity" and dilectus "beloved" throughout. Chunks 0013–0018, 0020–0024, 0026–0029, 0037–0039 use none of the reconciled words in a contested slot. node scripts/verify-english.mjs 6963 passes after the edits (40 chunks, columns, notes and sections aligned).

1.3 The equivalence tables as the agents recorded them

1. Fixed equivalences adopted for this stint (merge-time reconciliation note)

Migne's text uses three different words for the Vulgate's dilectus, sometimes within one sentence, and the exposition sometimes turns on the kinship sense (1879C: ut non aliae, sed consobrina sola sibi ejus vindicaret adventum). One English word is fixed per Latin word and never interchanged:

• frater meus → my brother
• consobrinus meus → my cousin
• fraternus meus (substantival, 1873C, 1873C–D) → my kinsman

Likewise: proxima mea → my neighbour (Vulg. amica mea, "my friend"); acclinatio / reclinatio → reclining-place (Vulg. lectulus, "little bed"); decora / speciosa / bona / pulchra / formosa are each kept distinct (comely / shapely / good / beautiful / well-formed), because §48 argues from the difference between bona es and decora es.

Agents taking later chunks of 6963 should follow this table; it is a within-work coherence requirement (translation-style.md, Song of Songs boundary note), not a cross-work harmonization.

Register decisions fixed for this batch

• salire = "to leap (upon)", transilire = "to leap over", exsilire = "to leap forth". The contrast is load-bearing at 1879D (salit super Adam, transilit super Synagogam) and 1880B (et si transilit, ut te transitus ejus custodiat), so the three are never interchanged.
• retia = "nets", laquei = "snares" throughout §§43–44, 57, where the pair is enumerated (Quot vitia, tot retia: quot peccata, tot laquei).
• portio = "portion" (never "share" or "part") through §§64–66; pars in the Deut 10:9 quotation is "part", as printed.
• proxima = "neighbor", the gloss word at 1883C; formosa = "beautiful"; columba = "dove". Fixed at 1882C and held.
• grossi = "green figs" (1884D, 1885D, 1887D), the unripe fruit the whole fig excursus turns on.
• Second person follows Pattern 17: Ambrose's tu to the individual soul, reader, or monk takes thou (1880B, 1882D, 1891C, 1892C–D); vos takes you (1885A Videte quemadmodum, and the Matt 24:20 and Matt 19:28 quotations, which are plural in the Latin).

Recorded by the 1893–1907 stint, and still binding:

⚠ Within-work note for the merge pass (runbook step 4). Migne prints both amica mea (1901C, ch. IV §1) and proxima mea (1905C, 1907A) for the same address. Because the printed Latin differs, the Englishes must differ (Song-rule case 1): this stint renders amica mea "my love" and proxima mea "my neighbour." That is a divergence in the plate, not drift between agents — do not reconcile the two into one English word. What does need checking at merge is that other agents' chunks follow the same mapping.

Also for the merge pass: taciturnitas is rendered "silence" (1901C, 1901D), propago "shoot" throughout (1896A, 1896C, 1898A–B), capillamentum "hair" (1903B, 1903D), tonsae "shorn ones" (1903B, 1903C, 1905B).

The closing stint (1949B–1962B) recorded that it followed the table without exception: frater meus → my brother, charitas → charity / dilectio → love, dealbata → made white (matching 0029), and that nothing in its five chunks required a new work-level equivalence.

2. ⛔ Plate divergences deliberately left UNHARMONIZED

These are not agent drift. Migne's text visibly sits on two Latin traditions at once, and that is the finding this edition exists to show. A tidy-minded pass that conforms any of them deletes it. Each is rendered as printed at every occurrence.

1. amica mea (1901C) vs proxima mea (1905C, 1907A) for the same address. Migne prints both; the English prints "my love" and "my neighbor" respectively. Do not merge them.
2. Cant 5:10 ex millibus (§§58–59, 1934A) vs ex decem millibus (§60, 1934C). Ambrose needs "ten thousands" for the Ephraim/David argument and quotes it that way two paragraphs after printing "thousands."
3. Rom 13:10: plenitudo legis est Christus (1873B) vs …est charitas (1878D), and again plenitudo enim Legis charitas est (0035 §12). The work carries both predicates.
4. Cant 5:1: manducavi panem meum (1909C, 1917D, 1919A, 1920B) vs Comedi favum (1918B); bibi vinum meum (1917D, 1920A) vs bibi potum meum (1919A). "my bread" / "the honeycomb", "my wine" / "my drink".
5. Cant 6:3 in three forms within two columns — the LXX-derived Beneplacita es… speciosa sicut Hierusalem, admiratio sicut ordinata (1942B §6), then the Vulgate form Pulchra es, amica mea, suavis et decora in the very next paragraph (1942B §7), then Pulchra ut Hierusalem (1942C §7).
6. Cant 7:4 Nares tuae (1948D §9) then Nasus tuus (1949A §10), one paragraph apart — plural/feminine against the Vulgate's singular; and Nares ejus again at 1949A §11.
7. Cant 7:2 both readings four lines apart inside one paragraph (1948C §8) — the Vulgate's numquam indigens poculis and the LXX's non deficiens mixto, with Ambrose glossing the LXX one.
8. Cant 8:13 masculine Qui sedes in hortis (1959C §28) vs the feminine Quae sedes that Ambrose reports from Symmachus and Aquila (§31). He is knowingly holding both, and §31 exists to say so.
9. Cant 8:5 in three forms in two columns (1951D §1, 1952B §2, 1953B §6) — ascendit candida / ascendit dealbata / ascendit with no epithet: "white" / "made white" / bare.
10. Cant 4:16 perfluunt (indicative) vs perfluant (subjunctive), 1916B, four lines apart — "and my spices flow forth" / "and let my spices flow forth".

The full within-work lists compiled by the stints follow, unaltered.

2a. 1907D–1921B (as recorded by that stint)

Within-work variation in the same lemma — NOT harmonized

Migne prints the same verse differently at different columns. Each occurrence is rendered as printed; none has been conformed to any other.

• Song 5:1: manducavi panem meum cum melle meo (1909C, 1917D, 1919A, 1920B) beside Comedi favum cum melle meo (1918B). The English carries "my bread" and "the honeycomb" respectively.
• Song 5:1: bibi vinum meum cum lacte meo (1917D, 1920A) beside bibi potum meum cum lacte meo (1919A). "my wine" / "my drink".
• Song 5:1: Descendat fraternus meus (1916B) beside descendat frater meus (1917C, 1918D, 1920A). Fraternus is an adjective standing where a noun is needed; it does occupy the slot and so takes no [sic:] under Pattern 12's in-place test. Rendered "my kinsman" against "my brother" elsewhere, so the divergence stays visible.
• Song 4:16: perfluunt aromata mea (indicative, 1916B, first citation) beside perfluant aromata mea (subjunctive, same section, second citation). Rendered "and my spices flow forth" / "and let my spices flow forth". The plate carries both moods four lines apart; neither has been assimilated to the other.
• Prov. 16:24: Favus mellis sermones boni (1909C) beside Favi mellis sermones boni (1918A). Singular/plural rendered as printed.
• Song 4:16 introduced twice, once as aromata mea (1916B) and once as unguenta tua (1917C), both under Ambrose's own citation.

2b. 1949B–1962B (as recorded by that stint)

D. Within-work inconsistencies left UNHARMONIZED

Each of these is Migne printing the same verse two ways within a few columns. Per the scripture policy, each is rendered as printed and none is harmonized to the other.

1. ⚠ Cant 8:13, the gender of the addressee (0038 §28 vs 0039 §31). §28 prints Qui sedes in hortis (masculine, the Bridegroom) and expounds it that way; §31 then reports that Symmachus tamen et Aquila interpretati sunt, quod Christus dicit ad Ecclesiam: Quae sedes in hortis" — the feminine, the Vulgate's reading, with a second exposition built on it. Ambrose is knowingly holding both, and the paragraph exists to say so. Conforming §28 to the Vulgate would delete the disagreement §31 is about. This is the clearest 7a″ trap in the batch.
2. Cant 7:8 (0035 §12): Ascendam in palmam in the lemma, ascendam ad palmam in the re-quotation nine words later, with Dixi absent from the first and present in the second.
3. Cant 7:12 (0035 §17 vs §18): diluculo surgamus in vineis / diluculo surgamus ad vineas.
4. Cant 8:5 (0036 §1, §2, §6): ascendit candida / ascendit dealbata / ascendit with no epithet at all — three printings in two columns. Rendered "white" / "made white" / bare, respectively.
5. Cant 8:6–7 (0036 §3 vs §5 vs §9, and 0037 §18): signaculum super cor tuum, super brachium tuum at §3, but in cor tuum, in brachium tuum at §9 and at 0037 §18. And Aqua multa excludere non poterit charitatem, et flumina non obruent eam at §3 against Aquae multae non poterunt excludere charitatem, et flumina non inundabunt ipsam at §5 — singular for plural, and a different verb, two columns apart.
6. Cant 8:5b (0036 §7): Sub arbore malo in the lemma, sub arbore mali twice in the gloss.
7. Cant 8:9 (0038 §21 vs §22): sculpamus super eam tabulas cedrinas in the lemma, Aedificemus super eam tabulas cedrinas when the same clause is re-quoted one paragraph later — and the verb aedificemus belongs to the murus half of the verse, not the janua half.
8. Cant 8:9/8:10, receptacula vs propugnacula (0038 §21): the lemma prints receptacula argentea; Ambrose's gloss six lines later argues from argentea ei propugnacula requiruntur, i.e. the Vulgate's word. Both stand.
9. Cant 8:10 (0038 §23 vs §25): ubera mea turres / ubera mea sicut turres.
10. Cant 8:14 (0039 §30 vs §32): similis esto tu capreolae aut hinnulo cervorum, super montes / similis efficere cervo aut hinnulo cervi, in montes.
11. II Cor 2:15 (0039 §30 vs §32): Bonus odor Christi sumus Deo / Christi enim bonus odor sumus Deo.
12. Rom 13:10 (0035 §12): plenitudo enim Legis charitas est — agreeing with 1878D, and differing from 1873B's plenitudo legis est Christus, which the peer recorded. Left as printed; the work carries both predicates.

Further unharmonized doublets recorded elsewhere in this file, listed here so the whole class can be seen at once: Fusca sum et decora / Nigra sum, et decora (1861A, 1862B); Osculetur me osculo / ab osculis / ab osculo oris sui (1851A, 1852A, 1854A); Unguentum exinanitum beside vel oleum effusum (1856B); cubiculum against cellaria, adjudicated by Ambrose himself at 1860D; Frater meus mihi in the lemma and Dilectus meus mihi three lines later (1888D, 1890B); in nocte / in noctibus (1891B, 1891C, 1892B); domus panis / domus Dei (1881B–C); Umbilicus ejus / Umbilicus tuus (1948C §7 vs §8); et totus desiderium (1938C) against the Vulgate form et totus desiderabilis nine lines later (1939A–B); receptacula argentea in the lemma against propugnacula in the gloss (1957C); Ps 36 quoted in two forms (1863B, 1864B); Prov 16:24 Favus / Favi mellis (1909C, 1918A); II Cor 2:15 in two word orders (1960B, 1960C).

3. SETTLED — 1893B–C, nex for nec, and the two sites that made it a class

⚠ The ruling below REVERSES the merge-pass ruling of earlier the same day. That first ruling — carry the type, let the English stand affirmative, log the consequence — was made on the printed page alone. The blind polarity read then brought corpus evidence the merge did not have, and it decides the question.

Ruling (2026-08-14, blind polarity read): carry the negation, keep the marker. The English now reads She who was speaking Christ, [sic: nex] was not stripped of her cloak.

The evidence that overturned it is inside the work and did not need an outside witness. §8 of this same passage prints et non exutus est pallio, "and he was not stripped of his cloak"; and chunk 0028 asks qui castae animae pallium tollunt? The crux Ambrose is worrying at, across thirty-seven columns, is that this soul kept her cloak. The affirmative English contradicted both its own paragraph and a passage far downstream. Note that this is exactly the alternative the translating agent proposed and did not take.

And it is not one case — it is a class, found the same afternoon by three independent readers in three chunks far apart:

• 1893B — [sic: nex] swallowed the negative; the English asserted the opposite.
• 1899A — [sic: car] swallowed cur; a reproach ("why does he not take away…?") became an assertion that he does, in a paragraph closing Non potest igitur haec dici justa misericordia. Now reads [sic: car] why does he not take away…
• 1917A — [sic: eccidit] swallowed the clause's only verb, leaving "which … the prophets" with no predicate. Now reads which slays [sic: eccidit] the prophets.

The principle, now written into translation-style.md as 7a⁗: a [sic: …] quarantines the defective TYPE, never the FORCE that type governs. The marker isolates what the plate got wrong; it is not a place to park a negation, an interrogative, or a clause's only verb and let the English go silently affirmative. The reader still meets the broken type at the point of loss — nothing is concealed — but the sentence still says what the author is arguing.

This does not reopen Pattern 7. Supplying the force the marked run governs is not a conjectural emendation of the plate: the negation is in the printed run, which is why the run is marked at all. What was conjectural was the earlier reading, which had the English assert a thing the surrounding Latin twice denies.

The stint's own entry, in full:

###### 1893B–C — nex for nec (§7) ⚠ READ THIS ONE Migne prints: Quae Christum loquebatur, nex exuta est pallio, et quem quaerebat, invenit.

nex ("slaughter, violent death") is a real Latin word with no grammatical slot here — a bare nominative noun stranded between a relative clause and a finite passive verb (Pattern 12, test 1). Carried untranslated in italics.

Conjecture, near-certain: nec. Ambrose's own preceding sentence sets the question — qua ratione hic ei non tulerint pallium, haereo reperire ("by what reasoning they did not take away her cloak here, I am at a loss to find out") — so the clause must read nec exuta est pallio, "she was not stripped of her cloak," and §8's closing sentence repeats the point affirmatively: et non exutus est pallio, postea quam coepit esse perfectior.

⚠ Consequence for the English, stated plainly per Pattern 7a′. Because nex is not a negative particle but a noun, carrying it leaves no negation in the English clause, and our page therefore says that she was stripped of her cloak — the contrary of the argument the paragraph is conducting. This is not a repair hidden in a marker: the marker sits exactly where the negation belongs, and the reader sees the broken type at the point of loss. It is logged here in full because it is the one place in this stint where the plate's defect inverts the sense. If a later reader judges that the printed letters do carry the negation (i.e. that nex is simply nec with a broken c), the remedy is to render "was not stripped of her cloak" and keep the [sic:] — not to leave the affirmative standing silently.

4. Vulgate / LXX divergence apparatus — the edition's headline deliverable

Roughly 230 [var: …] markers stand in the English of this work. The tables below are the complete record, reproduced verbatim from the eight stints, in plate order (which is also the order of Ambrose's exposition). Nothing here is summarized. Items marked ✅ / ✓ / ✚ / [var:] carry a marker in the running text; the rest are recorded only, because a lemma recurs and the marker fires once per divergence per chunk.

4.1 Finding a verse

Song of Songs, by chapter, in the tables at §4.2:

Verse-level rows are keyed by column inside each table; where a stint gave the verse number it is in the row.

Every other book, verse → column, derived from the tables at §4.3 (the column given is where the row's divergence stands; rows spanning several columns list them all):

• Genesis — 6:3 (1906D) · 48:14 (1934C) · 48:19 (1877D) · 48:20 (1877D) · 49:9 (1878D) · 49:12 (1904C) · 49:21 (1898B)
• Exodus — 1:5 (1906C) · 3:5 (1865B, 1947C) · 13:21 (1875D) · 15:1 (1868A) · 15:23 (1905A)
• Numbers — 1:7 (1945B) · 18:20 (1889C) · 20:17 (1877A)
• Deuteronomy — 10:9 (1889C) · 15:9 (1866A) · 32:13 (1888A) · 33:17 (1934C) · 33:23 (1898B)
• Kings — 11:12 (1946B)
• Job — 1:21 (1897A) · 2:10 (1897B) · 33:4 (1937B)
• Psalms — 6:7 (1892A) · 11:7 (1851A, 1955A) · 15:2 (1955D) · 17:12 (1892B) · 18:3 (1892B) · 18:6 (1859A, 1880A) · 18:7 (1859A) · 18:10 (1938C) · 22:2 (1890C) · 22:5 (1920C) · 23:1 (1862B) · 23:7 (1923A) · 23:9 (1938C) · 29:1 (1943A) · 33:9 (1876D, 1919B, 1938C) · 33:16 (1883A, 1958D) · 33:17 (1891D) · 36:5 (1863B) · 36:6 (1864B) · 41:3 (1935B) · 44:3 (1942C) · 44:4 (1948A) · 49:11 (1878C) · 49:19 (1878C) · 54:8 (1953A) · 62:2 (1935B) · 62:7 (1931A, 1956A) · 64:10 (1920C) · 64:11 (1920C) · 65:10 (1956B) · 65:12 (1956A) · 67:12 (1937A) · 67:27 (1935C) · 74:4 (1938A) · 75:1 (1857A) · 75:5 (1881A) · 75:7 (1867D) · 79:9 (1863C, 1920A) · 86:1 (1960B) · 91:13 (1870C) · 100:1 (1939A) · 103:4 (1953A) · 103:15 (1907B) · 112:5 (1872A) · 117:5 (1892D) · 118:12 (1857B) · 118:91 (1939C) · 118:103 (1857D, 1909D) · 118:131 (1855B) · 118:146 (1870A) · 120:6 (1870C) · 121:7 (1958B) · 123:5 (1944D, 1947A, 1953B) · 124:2 (1880B) · 126:3 (1874D) · 132:2 (1931A) · 133:2 (1931A) · 138:5 (1937B) · 139:6 (1883B) · 140:2 (1960B) · 140:3 (1910C) · 141:4 (1883B) · 213:5 (1947A)
• Proverbs — 1:28 (1853B) · 3:16 (1877A) · 3:17 (1877B) · 4:8 (1947B) · 5:15 (1914C) · 5:16 (1931B) · 5:18 (1914C) · 9:1 (1853D) · 9:5 (1916C) · 16:24 (1875A, 1909C) · 20:6 (1864D) · 22:20 (1904B) · 23:9 (1910A) · 27:6 (1876B, 1931C)
• Ecclesiastes — 2:6 (1914D) · 2:14 (1934D)
• Wisdom — 1:4 (1853B) · 9:15 (1945B)
• Ecclesiasticus — 3:33 (1916A) · 18:6 (1932C) · 23:18 (1944B) · 23:26 (1944B)
• Isaiah — 6:3 (1923D) · 11:1 (1885B) · 11:2 (1938A) · 24:18 (1881D) · 27:3 (1894B, 1957D) · 35:1 (1872C) · 43:25 (1870A, 1938D) · 47:2 (1946D) · 48:20 (1928B) · 49:4 (1939D) · 49:6 (1858C) · 49:16 (1958D) · 50:4 (1858D) · 52:7 (1946D) · 54:1 (1931C) · 63:1 (1952A)
• Jeremiah — 5:8 (1907A) · 9:21 (1882D) · 16:16 (1880D) · 20:9 (1954D)
• Lamentations — 4:20 (1919B, 1962A) · 4:21 (1926A)
• Ezekiel — 1:20 (1945D) · 6:11 (1854B) · 13:18 (1877B) · 18:4 (1944B) · 21:14 (1854B)
• Hosea — 10:12 (1914B)
• Habakkuk — 3:8 (1946C) · 3:15 (1867C, 1946C)
• Zechariah — 13:7 (1866D)
• Tobit — 12:7 (1942A)
• Matthew — 3:9 (1866B) · 3:11 (1947C, 1955B) · 3:15 (1860C) · 4:3 (1953A) · 4:5 (1953A) · 4:9 (1941A) · 5:8 (1896D, 1931D) · 7:21 (1941B) · 8:20 (1872D, 1888B) · 8:29 (1881C) · 9:2 (1869D) · 9:22 (1869D) · 10:9 (1890A) · 11:8 (1929C) · 11:29 (1867B, 1946A) · 18:10 (1933A) · 19:21 (1940C) · 19:27 (1940D) · 19:28 (1886C) · 22:30 (1906D) · 23:27 (1955C) · 23:37 (1917A, 1955C) · 24:20 (1885A) · 26:18 (1893A) · 26:31 (1866D) · 28:6 (1894D) · 28:20 (1943A)
• Mark — 8:33 (1866C) · 14:27 (1866D)
• Luke — 1:35 (1860A) · 1:42 (1871B) · 2:35 (1928B) · 7:40 (1936C) · 7:47 (1855B, 1954C) · 8:46 (1869C, 1895A) · 10:40 (1936C) · 13:32 (1888B) · 15:7 (1938D) · 17:21 (1865C) · 23:31 (1921C) · 23:34 (1869B) · 24:5 (1894D, 1901B) · 24:32 (1885B, 1955A)
• John — 1:14 (1939C, 1956D) · 1:16 (1935B) · 1:30 (1947C) · 2:19 (1883C) · 3:19 (1939A) · 4:8 (1878D) · 6:27 (1954B) · 7:23 (1878A) · 9:4 (1892C) · 10:9 (1958A) · 10:38 (1926B) · 12:15 (1914B) · 13:8 (1925B) · 13:14 (1925B) · 14:8 (1940C) · 14:30 (1890A, 1929A) · 16:32 (1923C) · 17:21 (1943C) · 17:23 (1943C) · 20:27 (1922B) · 21:16 (1866D) · 21:22 (1956A)
• Acts — 3:6 (1890A) · 4:32 (1943D) · 17:28 (1858B, 1961B) · 22:17 (1908C)
• Romans — 5:20 (1912C) · 7:24 (1899C) · 8:35 (1952D) · 8:38 (1952D) · 10:15 (1946D) · 13:10 (1878D) · 13:12 (1888A)
• Corinthians — 2:4 (1936D) · 2:15 (1871A, 1960C) · 4:10 (1926C) · 4:19 (1926C) · 4:20 (1936C) · 5:16 (1937D) · 7:5 (1938B) · 8:9 (1878A) · 10:4 (1905A) · 13:4 (1937C) · 13:5 (1954C) · 15:41 (1932D) · 15:48 (1946C)
• Galatians — 3:13 (1939D) · 3:28 (1862B) · 4:19 (1954A) · 4:27 (1931C) · 6:14 (1889D)
• Ephesians — 4:26 (1873B)
• Philippians — 2:7 (1890D, 1939B) · 2:16 (1939D)
• Colossians — 1:13 (1900C) · 2:20 (1858A, 1919D) · 3:3 (1962A) · 3:9 (1925A, 1937A) · 3:11 (1862B)
• 1 Timothy — 2:5 (1939B)
• Hebrews — 4:13 (1898C)
• Peter — 1:12 (1894C) · 2:2 (1910B) · 2:21 (1919B)
• Apocalypse — 3:21 (1949B) · 21:10 (1930B) · 21:11 (1930C) · 21:12 (1930C)

4.2 The Song of Songs lemmata

4.2a — cols 1851A–1866A (chunks 0000–0004)

5. Divergences from the Vulgate — the deliverable (Pattern 7a″ / 14)

Ambrose is expounding an Old Latin text of the Song, not the Vulgate, and the exposition repeatedly depends on the divergent word. Every item below was checked word by word against Migne's printed Latin before the English was written. Items marked ✅ carry a [var: …] in the running text; the rest are recorded here only, because the lemma recurs and the marker is fired once per divergence per chunk.

The Song of Songs lemmata

The five that matter most, and why:

1. 1861A, non est intuitus me sol. The Vulgate says the sun has discolored her; Migne prints that the sun has not looked upon her (LXX παρέβλεψέν με ὁ ἥλιος). The negative is the hinge of the whole of §§21–24 — "the sun of justice has forsaken me… I did not open the windows… when I shall have opened them, he will enlighten my eyes." On the Vulgate text the exposition is unintelligible. The non is carried everywhere it is printed.
2. 1864A, quem dilexit anima mea. Migne's perfect against the Vulgate's present — and Ambrose's next sentence is Cur quem dilexit anima mea dicis, et non quem diligit? The commentary is an argument about the tense. A conformed lemma would have destroyed the paragraph.
3. 1861B, Fusca sum et decora. §24 turns on the conjunction itself ("Ambrose Bassus" vs "Ambrose and Bassus"), and the Vulgate prints sed. The et is load-bearing type.
4. 1859C / 1860D, cubiculum against cellaria. Ambrose adjudicates the divergence himself at 1860D: Graecus in promptuarium suum et in cellarium suum habet — the plate records the author noticing the very variant this apparatus tracks.
5. 1865B, in calceamentis gregum. "In the sandals of the flocks", against the Vulgate's "after the footsteps of the flocks" — the Greek πτέρναις ("heels") read as footwear. §34's entire gloss is about unshod feet and Moses's loosed sandal, so the divergent word is the peg the exposition hangs on.

4.2b — cols 1866A–1879C (chunks 0005–0009)

3. Vulgate divergences — the Song of Songs lemmata (Pattern 14 [var:])

Every lemma in these five chunks was checked word by word against the Vulgate. Marked ones carry a [var:]; the rest are recorded here.

Note on the adjuration lemma (1875B vs 1878C). Migne prints the positive protasis — si suscitaveritis et resuscitaveritis — where the Vulgate has the prohibition ne suscitetis. This is not a dropped negation on our side: neither printing of the lemma in these chunks carries one, and §28 (1878D–1879A) builds an entire exposition on rousing charity in the Old Testament and rousing it up again in the New, which only works from the positive text. Rendered as printed.

4.2c — cols 1879D–1893B (chunks 0010–0014)

⚠ This stint tabled its Song lemmata and its other citations together; the table below therefore also carries Jer, Isa, Ps, Matt, John, Deut, Num, Gal and Acts rows, which are not repeated at §4.3.

Scripture divergences from the Vulgate

Migne prints an Old Latin / pre-Vulgate Song of Songs throughout, and his supporting citations diverge repeatedly. Those that are material — that is, where Ambrose's exposition rests on the printed word — carry a [var: …] in the English. Those listed here without a marker are recorded but not marked.

Marked with [var: …]

Recorded, not marked (immaterial to the argument)

• 1880A (0010 §32) Ps 18:6-7 agrees with the Vulgate as printed; no divergence.
• 1881C (0010 §38) Matt 8:29 printed Quid venisti ante tempus torquere nos — the Vulgate's opening (Quid nobis et tibi, Jesu Fili Dei?) is simply not quoted. Truncation, not divergence.
• 1882C (0011 §41) Stat post parietem tuum for the lemma's nostrum — Ambrose adapting the verse to his second-person address, not a text-form divergence.
• 1885A (0012 §50) flores visi in terra (no sunt) against the same chunk's flores visi sunt in terra at 1884B and 1887C. Migne's own inconsistency; both rendered as printed.
• 1885B (0012 §50) Luke 24:32 cum aperuisset in nobis Scripturas for the Vulgate's aperiret nobis Scripturas; also in via moved. Minor.
• 1888A (0013 §59) Suxerunt mel de petra et oleum de firma petra — Deut 32:13 has de saxo durissimo. Woven, unmarked by Migne, and the exposition does not turn on the adjective.
• 1888B (0013 §60) Luke 13:32 Ite, dicite vulpi illi omits the Vulgate's et.
• 1891D (0014 §2) Ps 33:17 omits the Vulgate's autem.
• 1892D (0014 §5) Ps 117:5 In tribulatione for the Vulgate's De tribulatione.
• 1893A (0014 §6) Matt 26:18 agrees as far as it is quoted.
• 1890C (0013 §67) Ps 22:2, 1888A (0013 §59) Rom 13:12, 1883A (0011 §41) Ps 33:16, 1881A (0010 §36) Ps 75:5, 1880B (0010 §33) Ps 124:2, 1885B (0012 §50) Cant 2:1, 1886D (0012 §53) Cant 2:13b, 1890B (0013 §67) Cant 2:16b, 1890D (0014 §68) Phil 2:7, 1890C (0014 §68) Cant 2:17, 1892A (0014 §3) Ps 6:7, 1892B (0014 §4) Ps 17:12 and Ps 18:3, 1890A (0013 §66) Matt 10:9 — all agree with the Vulgate as far as printed. Checked word by word.

4.2d — cols 1893C–1907C (chunks 0015–0019)

Vulgate divergences — the deliverable list

Migne prints an Old Latin / pre-Vulgate Song of Songs throughout, and quotes the Psalms, Gospels and Pentateuch in forms that repeatedly leave the Vulgate. Every entry below is rendered as printed; those marked ✚ carry a [var: …] in the English because the sense turns on the divergence.

Song of Songs (the lemmata Ambrose expounds)

4.2e — cols 1907D–1921B (chunks 0020–0024)

Divergences from the Vulgate — the deliverable list

Ambrose's Song of Songs text throughout these columns is Old Latin, not Vulgate, and it diverges word by word. Material divergences carry a [var: …] in the English at the point of divergence; those marked (logged only) were judged immaterial to the sense and carry no marker. The list is exhaustive for these five chunks.

Song of Songs (the lemmata Ambrose expounds)

4.2f — cols 1921C–1935B (chunks 0025–0029)

A. Vulgate divergences in the Song of Songs lemmata (Cant. 5:2–12, 8:5)

Migne prints a pre-Vulgate (Old Latin / LXX-derived) Song throughout this stretch. Every lemma below was checked word by word against the Vulgate before translating. Marked [var: …] at first occurrence per chunk; repetitions of a refrain (the Exui tunicam meam refrain recurs eight times in 0026) are left unmarked to avoid marker spam, and are covered by this entry.

4.2g — cols 1935C–1949A (chunks 0030–0034)

A. Vulgate divergences in the Song of Songs lemmata (Cant. 5:13–7:4)

Every lemma checked word by word against the Vulgate before translating. Migne continues to print an Old Latin / LXX-derived Song, and it diverges at nearly every verse in this stretch.

4.2h — cols 1949B–1962B (chunks 0035–0039)

A. Vulgate divergences in the Song of Songs lemmata

Every lemma was checked word by word against the Vulgate before translating. Marked [var: …] at first occurrence in each chunk; re-quotations within the same chunk are covered by the entry and left unmarked to avoid marker spam.

4.3 Psalms, prophets, Gospels, Epistles, Pentateuch

4.3a — cols 1851A–1866A

Non-Song citations

Note on Ps. 36 (1863B and 1864B): the plate quotes the same psalm twice, in two different forms — educet … justitiam tuam … judicium tuum the first time, Deduces eorum justitiam … judicium eorum the second — and the second form, with "their", is what §31's argument needs ("for them thou hast been made noonday, who are fed in thy riches"). Both are rendered as printed; neither is harmonized to the other or to the Vulgate.

Note on Deut. 15:9 (1866A): this is the closing argument of the chapter — that Nosce teipsum is Mosaic and not Apollo's. The whole point rests on Attende tibi being Scripture's own words. The Vulgate's cave would have made the argument unrecognizable, and the divergence is therefore marked rather than merely logged.

4.3b — cols 1866A–1879C

4. Vulgate divergences — Psalm, prophetic and Gospel citations

Marked with [var:] where the sense turns on the wording; otherwise logged here.

• 1866B, Matt. III, 9 — Potens enim Deus est… (Vulg. potens est Deus…): word order only, unmarked.
• 1866C, Mark VIII, 33 — Veni retro me, Satana for Vulg. Vade retro me, Satana. ✓ marked: "Come" for "Go" is a real change of the command.
• 1866D, John XXI, 16 — Pasce agniculos meos for agnos: the diminutive is Ambrose's, and the following sentence turns on Peter's place at the right hand. ✓ marked.
• 1866D, Zach. XIII, 7 — Migne prints Percutiam pastorem, et dispergentur oves gregis, i.e. the Matt. XXVI, 31 / Mark XIV, 27 form (first person future) under a Zechariah reference; the Vulgate Zechariah has the imperative Percute pastorem. ✓ marked. Per the standing rule, no correction is proposed to the printed reference.
• 1867B, Matt. XI, 29–30 — Tollite jugum meum super vos, quia leve est; et onus meum, quia suave est. The Vulgate's epithets are exchanged: jugum enim meum suave est, et onus meum leve. ✓ marked. Note that §43 (1868C) then argues from leve est enim jugum Christi — the exposition follows Migne's printing, not the Vulgate's, which is exactly the 7a″ trap and the reason the lemma was not conformed.
• 1867C, Habac. III — Immisisti in mare equos tuos, turbantes aquas multas (Vulg. Hab. 3:15 Viam fecisti in mari equis tuis, in luto aquarum multarum) and Ascendisti in equos tuos, equitatus tuus sanitas (Vulg. 3:8 quadrigae tuae salvatio). Both LXX-shaped. ✓ both marked. sanitas is rendered "health," not "salvation" — the whole horses-and-chariot exposition depends on the cavalry word.
• 1867D, Psalm LXXV, 7 — agrees with the Vulgate.
• 1868A, Exod. XV, 1 — projecit for Vulg. dejecit. ✓ marked (thin, but the verb is the only word quoted).
• 1869B, Luke XXIII, 34 — quia nesciunt quid faciunt for non enim sciunt. ✓ marked.
• 1869C, Luke VIII, 46 — Sentio virtutem de me exiisse for nam ego novi virtutem de me exiisse. ✓ marked; Ambrose's balsam simile depends on sentio.
• 1869D, Matt. IX, 22 — Constans esto, filia, remissa sunt tibi peccata tua conflates Matt. 9:22 (Confide, filia, fides tua te salvam fecit) with Matt. 9:2 (remittuntur tibi peccata tua), and the argument is drawn from the forgiveness clause. ✓ marked. Reference left as printed.
• 1870A, Psal. CXVIII, 146 — the quoted words are v. 17 (Retribue servo tuo…) in a form diverging from the Vulgate (ut vivam for vivifica me; mandata for sermones). ✓ marked on the wording; reference untouched.
• 1870A, Isai. XLIII, 25–26 — substantially LXX-shaped throughout. ✓ marked.
• 1870C, Psalm CXX, 6 — cited inline, in italic, not as an [n:] note, and Migne's column anchor [1870C] falls inside the parenthesis. Reproduced exactly, anchor included. Not tagged [f:] (Pattern 4 does not apply to a commentary's inline scripture).
• 1870C/D, Psalm XCI, 13 — sicut cedrus quae in Libano est for sicut cedrus Libani. ✓ marked.
• 1871A, II Cor. II, 15 — word order only (Bonus odor Christi sumus Deo); unmarked.
• 1871B, Luke I, 42 — agrees.
• 1872A, Psalm CXII, 5 — agrees.
• 1872C, Isai. XXXV, 1 — imperatives (Laetare… exsulta… floreat) where the Vulgate has futures, plus sterilis for deserta et invia. ✓ marked.
• 1872D, Matt. VIII, 20 — non habet ubi caput suum reclinet (Vulg. ubi caput reclinet); at 1878A quoted again in the imperfect, non habebat. Logged, unmarked.
• 1873B, Ephes. IV, 26 — agrees.
• 1874D, Psal. CXXVI, 3 — agrees.
• 1875A, Prov. XVI, 24 (unreferenced, "sicut scriptum est in Proverbiis") — Favi mellis sermones sunt boni for Favus mellis composita verba. ✓ marked.
• 1875D, Exod. XIII, 21–22 — Deus praeibat illos for Dominus autem praecedebat eos, and the whole clause re-shaped. ✓ marked.
• 1876B, Prov. XXVII, 6 — Utiliora vulnera amici, quam voluntaria oscula inimici for Meliora sunt vulnera diligentis quam fraudulenta oscula odientis. ✓ marked; the voluntaria/fraudulenta difference reverses what is wrong with the enemy's kiss.
• 1876D, Psalm XXXIII, 9 — Gustate et videte quoniam suavis est Dominus: agrees, unreferenced.
• 1877A, Num. XX, 17 — Via regali ibimus, non divertemus… for per viam regiam gradiemur, nec declinabimus…: synonym-level, logged, unmarked.
• 1877A/C, Prov. III, 16 — Longitudo vitae for Vulg. Longitudo dierum. ✓ marked, and this is the most consequential citation divergence in the batch: §25 (1878A–B) argues at length that wisdom's right hand gives longitudo vitae aeternae, an exposition that has no purchase on longitudo dierum.
• 1877B, Ezech. XIII, 18 — Vae his qui assuunt cervicalia for Vae quae consuunt pulvillos sub omni cubito manus. ✓ marked.
• 1877B, Prov. III, 17 — viae bonae … in pace for viae pulchrae … pacificae. ✓ marked.
• 1877D, Gen. XLVIII, 19 — divergent throughout, and note frater junior where the Vulgate has frater ejus minor. ✓ marked.
• 1877D, Gen. XLVIII, 20 — Faciat vos Deus (plural) for Faciat tibi Deus (singular). ✓ marked; a number change, and Ambrose's dicentibus posteris takes the plural.
• 1878A, II Cor. VIII, 9 — quia cum dives esset, pauper factus est for quoniam propter vos egenus factus est, cum esset dives. ✓ marked.
• 1878A, John VII, 23 — agrees.
• 1878C, Psalm XLIX, 19 — Et species agri mecum est for Vulg. Ps. 49:11 et pulchritudo agri mecum est: synonym; logged, unmarked. Reference as printed.
• 1878D, Rom. XIII, 10 — plenitudo legis est charitas for plenitudo ergo legis est dilectio. Logged, unmarked (the substitution is universal in Ambrose). ⚠ Compare 1873B, where the same maxim is printed plenitudo legis est Christus — a different predicate entirely, rendered as printed.
• 1878D, I John IV, 8 — Charitas Deus est for Deus caritas est. ✓ marked: here the reversal is the argument (charity as the subject being roused).
• 1878D, Gen. XLIX, 9 — tamquam leo, et catulus leonis; ut Judae ascendat ex germine for Catulus leonis Juda: ad praedam, fili mi, ascendisti. ✓ marked. The following Quis suscitabit illum? (Gen. 49:9 quis suscitabit eum?) is printed without a note and left as printed.

4.3c — cols 1879D–1893B → tabled with the Song lemmata at §4.2c

4.3d — cols 1893C–1907C

Psalms, Gospels, Epistles, Pentateuch

4.3e — cols 1907D–1921B

Other scripture

4.3f — cols 1921C–1935B

B. Vulgate divergences outside the Song

4.3g — cols 1935C–1949A

B. Vulgate divergences outside the Song

4.3h — cols 1949B–1962B

B. Vulgate divergences outside the Song

5. Carried type — [sic: …] (Patterns 7 / 10 / 12)

Every non-word the plate prints, with its conjecture. The conjectures live here only; not one is carried into the text.

5a — cols 1851A–1866A

1. Carried type — [sic: …] (Pattern 7 / 10 / 12)

Four runs are not words and cannot be rendered; each is carried untranslated in italics inside a [sic: …], and each is verbatim in its Latin twin.

###### 1.1 — 0000, col 1854A/B: deret > …quae omnia fecisset ut sponsum [1854B] deret

deret is no Latin verb. The clause needs a transitive verb governing sponsum; the likeliest intended reading is videret ("that she might see the bridegroom"), the loss of vi- falling exactly at the column break — which is where this plate's type is worst. adiret is a weaker alternative. Rendered: "…so that she might [sic: deret] the bridegroom". Conjecture only; not carried into the text.

###### 1.2 — 0001, col 1856A: cupieos > coelestibus cupieos inhaerere mandatis

cupieos is a non-word. Sense and syntax require the participle cupiens ("desiring to cleave to the heavenly commandments"), agreeing with the subject of ait. Rendered: "…and as though despising all his own delights and pleasures, [sic: cupieos] to cleave to the heavenly commandments, he says…". The English therefore has no participle in that slot; that is the honest consequence, since supplying "desiring" would silently repair the plate.

###### 1.3 — 0002, col 1859B: attrabatur > Prudenter ergo rogat ut attrabatur

attrabatur for attrahatur (passive subjunctive of attraho) — the verb the whole section is expounding (Attrahe nos, Cant. I, 3). A compositor's b for h. Rendered: "Prudently, therefore, she asks that she may be [sic: attrabatur]…".

###### 1.4 — 0004, col 1864B: damnanationem > ego crucem non suscepi ut mundi redemptionem, sed ut noxii damnanationem

damnanationem for damnationem — an intra-word syllable doubling (-na-na-). Not treated as Pattern 11 dittography, because the doubled run is a syllable inside a word rather than a repeated run of text, so there is nothing to print twice; it is broken type, and takes [sic: …]. Rendered: "…but as the [sic: damnanationem] of a guilty man."

5b — cols 1866A–1879C

2. Carried type — [sic: …] entries

###### 1867C — duo decajugum Printed: O bonorum equorum duo decajugum mirabile. As printed the run does not construe: duo is nom./acc. plural, decajugum a singular neuter noun/adjective, and mirabile is singular. Conjecture: duodecajugum, "twelve-yoke team" (twelve apostles drawing the four-wheeled Gospel chariot of the following clause, which is the sense the whole paragraph builds). Carried untranslated in italics under Patterns 7 and 10 (split type), marked [sic:].

###### 1868A — per sequebantur Printed: qui non sequebantur legem, sed per sequebantur. Split type for persequebantur, "but were persecuting [it]" — the antithesis sequebantur / persequebantur is the point. The whole broken run is carried per Pattern 10, so that the division itself, not a repaired word, is what the reader meets.

###### 1875B — gutluri Printed: Et fructus ejus dulcis gutluri meo. Non-word for gutturi ("to my throat"), and note that this is the Vulgate's own word — Ambrose has been quoting in faucibus meis ("in my jaws") for three sections and here quotes the other form. The defect therefore sits exactly on a lemma divergence (see §3 below); carried untranslated, [sic:].

###### 1879B — collonquentem Printed: cum aliquibus pariter viantibus collonquentem. Non-word for colloquentem. Carried untranslated per Pattern 7's non-word class.

5c — cols 1879D–1893B

Carried type

• 1891C (0014 §2) — David per singulas nocctes lectum suum lacrymis diluebat. nocctes is not a word; the same phrase is printed correctly as per singulas noctes twice in the next section (1892A, 1892B), and again in the Ps 6:7 quotation. Carried untranslated in italics under Patterns 7 and 12: [sic: nocctes]. Conjecture: noctes, certain.

5d — cols 1893C–1907C (the nex entry is at §3)

###### 1895D — floregratior for flore gratior (§18) Nonne omni floregratior resurrectio mortuorum? Run-together word division (Pattern 10): flore gratior, "more pleasing than every flower," parallel to the preceding omni melle dulcior. Carried whole, per Pattern 10's rule that the unit is the broken run, not either fragment; the English reads "Is not the resurrection of the dead [sic: floregratior] than every one?", where "than every one" renders the surviving ablative omni.

###### 1899A — car for cur (§28) …laxat exsilium; car latrocinandi qua potest clementiore via non eripit facultatem… Non-word type; conjecture cur, "why." Carried untranslated.

###### 1900A — generali for generati (§32) …quod superiores videantur qui ante generali sunt, infirmiores qui postea. generali is a real word (dat./abl. sg. of generalis) but occupies no grammatical slot: it cannot be a predicate of sunt under ante, and the clause will not construe. Conjecture generati sunt, "who were begotten earlier" — which is exactly what the following sentence glosses (qui in virtute juventutis nati sunt). Carried under Pattern 12, test 1.

###### 1903D — animasum for animarum (§10) …eminentia quaedam justarum animasum… Non-word; conjecture animarum, confirmed by justarum immediately before it and by §11's animae justorum. Carried untranslated.

###### 1906C — Aeyyptum for Aegyptum (§20) Inside the Gen. XLVI lemma: Omnes autem animae quae intraverunt cum Jacob in Aeyyptum. Non-word; carried untranslated. Because it sits inside an already italic scripture lemma it takes no second italic marker (Pattern 10), so the English prints [sic: Aeyyptum] unasterisked.

5e — cols 1907D–1921B

Carried type ([sic: …])

1907C/D (chunk 0020, §25) — hoc est, egredere de corpore, et totam te exne. exne is not a Latin word. The construction (totam te + imperative, parallel to egredere) requires a second-person imperative governing totam te; the near-certain intention is exue, "strip thyself wholly / put thyself wholly off," which suits both the sense (going forth out of the body) and the letter-shapes (u/n confusion in Migne's fount is routine). Carried untranslated in italics per Pattern 7 and marked [sic:]; the conjecture stays here.

1916D–1917A (chunk 0023, §54) — sed haec Hierusalem terrena est, quae eccidit prophetas. eccidit is not a word. Two readings compete: occidit ("slew the prophets" — the Matthean context, quae occidis prophetas, which the very next sentence quotes) and excidit. Occidit is near-certain on sense. Carried untranslated in italics and marked [sic:]; not emended, because the plate's own type is the evidence.

5f — cols 1949B–1962B

E. Markers fired

[sic: …] — carried type (Pattern 7 non-word class / Pattern 12)

• 0037 1955A §12 — presecutor. Printed Denique apostolus factus est, qui presecutor advenerat. Not a word; for persecutor, which the antithesis (apostle / persecutor) requires and which Migne prints correctly elsewhere in the work. Carried untranslated in italics, marked.
• 0037 1956B §16 — lcareos. Printed aestu solis cera resoluta lcareos volatus penna relabente destituat. A lowercase l set for capital I: Icareos, "Icarian" — the sentence is the Icarus fable in full (wax loosed by the sun's heat, the feather slipping). Carried untranslated in italics, marked.

[d: …], [ed: …] — none fired

No dittography in these columns (the 0036 §1 candidate is discussed at §C.1 and was rejected). No hole in the digitization: every ## head the Latin twin carries has a body, no empty elements, no truncation at a column boundary.

(No [sic: …] was fired in 1921C–1949A; that stretch's difficult readings are all real words in real slots and stand at §6.)

6. Printed readings rendered literally — no marker, no emendation (Patterns 7 / 7a / 8 / 12)

The Pattern 12 boundary: a real word occupying a possible slot is translated as ordinary English however wrong it looks, and the conjecture is recorded here; only a non-word, or a real word with no slot at all, is carried in a [sic:].

6a — cols 1851A–1866A: real words that are probably wrong

2. Real words that are probably wrong — rendered literally, NOT marked (Pattern 7)

These occupy their grammatical slot and can be rendered, so under Pattern 12's boundary the word on the page is our English, not Migne's type. No marker; the conjecture lives here only.

###### 2.1 — 0001, col 1856B: agentibus > Hoc unguentum exinanitum est supra Judaeos, et collectum est agentibus: > exinanitum in Judaea, et redoluit in omnibus terris.

Rendered literally: "and was gathered up for those acting". The antithesis of the sentence (Judaea ↔ omnibus terris) wants gentibus, "and was gathered among the nations", and the initial a- is very likely picked up from the preceding est. Rendered as printed; conjecture recorded, no emendation.

###### 2.2 — 0000, col 1853D: note reads (Pov. IX, 1) Migne's note prints Pov. for Prov. Reproduced verbatim inside the [n:] marker per the verbatim rule. No [sic: …] stacked on it — the defect is inside a note whose contents are untouched by policy, and the facing Latin column shows it.

###### 2.3 — 0002, col 1859B: note reads (Joan. XIII, 36; Matth. XIV, 30) The second locator does not answer to Non potes me sequi modo. Recorded, not corrected — Migne's printed reference stands, per refDisplay policy.

###### 2.4 — 0004, col 1865B: Esai. LII, 7 cited for wording that is Romans' Migne prints Quam speciosi pedes evangelizantium pacem under a note citing Isaiah 52:7. That wording is Rom. 10:15's; the Vulgate Isaiah reads quam pulchri super montes pedes adnuntiantis et praedicantis pacem. Note reproduced verbatim, text rendered as printed. Recorded, not corrected.

6b — cols 1851A–1866A: syntax that will not construe

4. Syntax that will not construe — rendered as closely as the words allow

###### 4.1 — 0000, col 1854D: the anacoluthon of §4 > …jamdudum infusionem sibi divinae praesentiae et gratiam Verbi salutaris exoptet, > commacerari eam quod sero veniat, et affligitur; ideo … conversa ad Patrem, rogare > ut mittat … et causam … declarare dicentem

The sentence begins as an Igitur vel … vel … construction, shifts to accusative + infinitive (commacerari eam, rogare, declarare), and once lapses into a finite indicative (et affligitur) inside the infinitive chain. Rendered literally with the shift preserved ("…that she is worn away because he comes late, and is afflicted; and therefore … that she asks God…"). No emendation.

###### 4.2 — 0001, col 1857A: sentence without a main verb > Hoc itaque nomen quod vasa Judaeorum angustiis suis coercitum continebant.

A relative clause with no governing predicate; Migne punctuates it as a full sentence. Rendered as the fragment it is ("This name, therefore, which the vessels of the Jews, confined within their own narrows, contained.") and the full stop kept, per Pattern 8. Note also that coercitum (neuter, agreeing with nomen) sits inside the clause whose subject is vasa; the concord is the plate's.

###### 4.3 — 0000, col 1853A: amore pio cudere cudere ("to hammer, to forge") is a real word in a startling place — the Virgin is to be hammered by holy love. Rendered literally; the metaphor is continuous with the heifer and the yoke immediately before it, so it is very likely what was meant.

###### 4.4 — 0004, col 1864D: double negative ne … non agnoscat > ne qui mentem tuam interrogat, non agnoscat auctorem

Both negatives carried: "lest he who questions thy mind should not recognize the author." The doubled negation stands as printed rather than being resolved into a single one.

###### 4.5 — 0000, col 1853A: verbo … verbere The heifer is accustomed verbo lasciviae before verbere disciplinae — a sound-play on verbum/verber that English cannot carry. Rendered by sense ("the word of playfulness … the lash of discipline"); the play is lost, and is recorded here rather than glossed in the text.

6c — cols 1866A–1879C

5. Readings rendered literally under Pattern 7 (no marker, conjecture here only)

###### 1868C — ira for ita Printed: Itaque sicut aurum bonum, ira Ecclesia cum utitur, detrimenta non sentit, sed magis fulgor ejus augetur. ira is a real word occupying a real slot (ablative with utitur), so it is rendered in place — "when the Church makes use of wrath" — per Pattern 12's boundary for a real word that is merely wrong. Conjecture: ita ("so the Church, when it uses [gold], feels no loss"), which the sicut … ita correlation and the following donec Christus veniat both want. Not emended.

###### 1878C — in virtutibus e fortitudinibus agri The same lemma at 1875B prints et. Here e is a preposition governing the ablative fortitudinibus and does construe, so it is rendered "from" rather than carried as broken type. Conjecture: et, by the parallel three columns earlier.

###### 1872A — necubi eum in pectoris tui convalle reperias Rendered with its negative intact ("lest anywhere thou find him in the valley of thy breast"), which reads against the encouraging drift of the sentence (humilibus enim frequenter inhalatur, "he is often breathed upon lowly things"). necubi may be doing duty for an indirect question ("whether somewhere thou mayest find him"). The negation is printed and is carried; the difficulty is recorded rather than smoothed.

###### 1867A — sed absoluto pro equitatu Elliptical: an ablative absolute with no expressed noun, the point being that the singular equa stands for the collective equitatus, "as we have often said." Rendered "but a word put absolutely for the cavalry"; the two supplied English words carry no sense the Latin does not, but the ellipsis is noted here.

###### 1869C–D (§47) — the antithesis series terram movebat, et haerebat in ligno … vulnus erat et fluebat unguentum: scarabaeus audiebatur, et Deus agnoscebatur. Rendered flat and paratactic as printed; scarabaeus ("beetle," the old Latin rendering at Hab. 2:11) is kept as the plate has it and not glossed in the text.

6d — cols 1879D–1893B: flags — printed readings that contradict their own context

Flags — printed readings that will not construe, or contradict their own context

Each is rendered as printed. None is repaired in the English.

1. 1885A (0012 §50) — Videte where Ambrose's own gloss requires Orate. Migne prints Matt 24:20 as Videte ne fiat fuga vestra hieme vel sabbato. At 1891A (0014 §69) Ambrose paraphrases the same verse: et ideo suadet orandum, ne fiat fuga nostra hieme aut sabbato — "and therefore he urges that prayer must be made." The paraphrase presupposes Orate (Vulgate), which the quotation five columns earlier does not print. Rendered "See that your flight be not…" as printed, with [var:]; the contradiction with §69 is left standing.

1. 1884C (0011 §48) — a negation missing from the third member. quia non sequuntur Domini passionem, non tollunt crucem suam, et Christum sequuntur. The first two members are negative and the third is affirmative, so the sentence asserts that those who do not follow the passion nevertheless follow Christ — the opposite of the argument (they are the ones whose flight is in winter). nec for et is near-certain. Rendered as printed: "they do not follow the Lord's passion, they do not take up their cross, and they follow Christ." This is a Pattern 7a case and the English carries the contradiction deliberately.

1. 1881B–C (0010 §§37–38) — domus panis / domus Dei, and a doubled clause. The last sentence of §37 and the second sentence of §38 are the same clause printed twice: salit enim super Ecclesiam quae est domus panis, eo quod fidelium corda confirmet / salit enim super Ecclesiam, quae est domus Dei, eo quod fidelium corda confirmet. The lemma is super montes Bethel, and Bethel is domus Dei; domus panis is Bethlehem. Both are rendered as printed. No [d: …] — the two runs are not identical (the variant word is the whole interest of the doubling), and Pattern 11's marker wraps a repeated run, not a variant one. Whether §37's sentence is a compositor's anticipation of §38 or Ambrose's own doublet cannot be settled from the plate.

1. 1888D (0013 §63) — Frater meus in the lemma, Dilectus meus in the gloss. The verse 16 lemma is printed Frater meus mihi, et ego illi; three lines later the same verse is quoted Dilectus meus mihi et ego illi, and again at 1890B Dilectus meus mihi, et ego illi. Migne's plate carries both text forms of Cant 2:16 within one section. Both rendered as printed ("My brother", "My beloved"); no harmonization.

1. 1891B / 1891C (0014 §1) — in nocte in the lemma, in noctibus in the re-quotation. The block lemma reads In cubili meo, in nocte quaesivi; the re-quotation eleven lines later, and again at 1892B (§3), reads In cubili meo in noctibus quaesivi. Singular against plural in the same section. Both as printed.

1. 1879C (0010 §31) — ego me vulneratam charitatis puto. A bare genitive with vulneratam; the construction wants charitate (as at 1882A §40, vulneratur charitatis decore, where the genitive depends on decore). Rendered "I deem myself wounded of charity", which keeps the genitive visible. Not emended.

1. 1879D (0010 §32) — no stop between two sentences. Videamus salientem Salit de coelo in Virginem. The plate prints no punctuation and a capital S. Under Pattern 8 no stop is supplied: "Let us see him leaping He leaps from heaven into the Virgin."

1. 1880B (0010 §33) — mixed persons. si non potest esse mons et non praevales, esto vel collis — third-person potest and second-person praevales/esto in one conditional. Rendered literally ("if one cannot be a mountain and thou dost not prevail, be at least a hill"); both negations carried.

1. 1881B (0010 §37) — maturus exspectes. Si in tempore messis quo diriguntur isti metere, maturus exspectes… The nominative maturus is stranded (the ears of corn, not the watcher, are what ripens in the surrounding figure). Rendered "thou shouldst wait ripe", which keeps the case relation as printed.

1. 1889A (0013 §64) — Quomodo … est. punctuated as a statement. Quomodo idem et mihi sacerdos et Deo est. An interrogative adverb closed by a full stop. Under Pattern 8 the mark stands: "How the same is a priest both for me and for God." No question mark supplied.

1. 1886D (0013 §53) — dum ligatur, exuitur. The three-member figure runs dum ligatur, exuitur: dum humiliatur, attollitur: dum reciditur, coronatur, and the two following members invert their verbs where the first does not (bound → stripped is not an inversion). erigitur (as in the immediately preceding vine clause, ut vinea dum ligatur, erigitur) would give the expected sense. Rendered as printed: "while it is bound, it is stripped."

1. 1886A (0012 §51) — the fig-tree excursus will not fully construe. Namque succedentibus aliis quae quasi degenerantia despuuntur, et arenti infirma radice, renovatis quibus sit succus utilior, exuuntur. Two ablative absolutes, a relative clause without a clear antecedent, and a plural exuuntur with no expressed subject. Rendered as closely as the printed words allow ("For when others succeed, which are spat out as though degenerate, and the weak root being dry, they are stripped away, those being renewed whose sap is more serviceable"). No word supplied, none dropped.

1. 1891D (0014 §2) — first person inside a second-person address. Tunc maxime quando tenebrae in circuitu meo, et parietes me operiunt — in a paragraph otherwise addressed to tu (et tu totam noctem…, Ne dixeris ergo…). Rendered as printed ("when darkness is round about me and walls cover me"); the shift is Migne's.

1. 1890C (0013 §67) — montes convallium. "Mountains of the valleys," against the lemma's lilium convallium and the work's steady mountain/valley opposition (1880A: Deus montium est, non vallium). Rendered literally.

1. 1888B (0013 §61) — Allophylorum. Migne prints the LXX transliteration (ἀλλόφυλοι) where the Vulgate of Judges 15 has Philisthinorum. Kept as Allophyli in italic Latin under the register's technical-term rule rather than resolved to "Philistines", which would silently supply the Vulgate's word.

1. 1888A (0013 §59) — note reference (Rom. XXIII, 12). Rom 13:12 is meant (Nox praecessit, dies autem appropinquavit); Romans has sixteen chapters. Migne's citation is reproduced verbatim per the [n: …] rule; not corrected in the text, and no citation-corrections entry is proposed here.

1. 1885D / 1884D (0012) — the section number 50 is printed twice. Two consecutive sections are both headed 50. in the plate (the second beginning Ficus protulit grossos suos), and the numbering resumes at 51. Reproduced as printed; the heads frontmatter records the duplication.

1. 1891A (0014) — CAPUT TERTIUM. has no text. The chapter head is followed immediately by section 1 with nothing between. That is the plate, not a digitization hole; no [ed: …].

6e — cols 1907D–1921B: other readings

Other readings

1909A–B (§29/§30) — Favum itaque mellis tua opera componant is printed twice, once at the close of §29 (where no honeycomb lemma has yet been quoted) and again at the head of §30, immediately after Favus distillans, labia tua, sponsa, where it belongs. The §29 occurrence is anticipatory and reads as a compositor's doubling. Both are rendered. No [d: …] was fired: Pattern 11's marker exists so a repetition inside one run of prose is not read as our build defect, and these two sit a section-head and a lemma apart, where nobody will take the second for a rendering artifact. Recorded here so the doubling is not lost.

1909B (§30) — nec soli tibi, sed etiam pluribus congreges quid scis enim quando a te tua anima reposcatur — the plate prints no punctuation between congreges and quid scis enim, and the two clauses do not join. Rendered without supplied punctuation per Pattern 8; the English is correspondingly rough at that seam.

1916C (§53) — sciens quod ortus suus Christo displicere non possit. The parallel passage at 1918D (chunk 0024, §6) prints the same sentence with hortus, and the whole section is about the garden. H-dropping is an attested orthographic convention, so this falls under Pattern 9, not Pattern 7: rendered "her garden," no italics, no marker.

1918C–D (Cap. V §5, §6) — §6 opens with the column anchor set hard against the first word, [1918D]Bona debet. Reproduced with a space in the English; nothing is lost.

1919A (§6) — radicesque suas quas tinxerat… et in bonos fructus… pullulaverint is an anacoluthon: the relative clause has no governing verb and pullulaverint switches subject. Rendered as printed, with the break visible.

1921A (§13) — the sentence haereditas enim ejus est somnus sanctorum, feriatus ab omnibus corporis voluptatibus. closes with a full stop, and the next line resumes lowercase across the column break: ab omni animi perturbatione, tranquillitatem menti invehens…, which continues the participial string of the previous sentence. The printed stop is preserved and the lowercase resumption is preserved: the English begins a new sentence with "from every disturbance of the mind." This is the one place in the batch where a construction is broken across an anchor, and the break is in the punctuation, not in the word division.

1920D (§13) — Jacob sanctus divina dormiens videbat mysteria… de coelo ad terras aerem praevium sanctis, respicientem Dominum, et pollicentem terrae ejus possessionem. The accusatives depend on videbat with no verb of their own and the ladder of Gen. 28 is nowhere named; aerem praevium sanctis is obscure. Rendered literally rather than supplied from Genesis.

6f — cols 1921C–1935B

C. Printed readings rendered literally under Pattern 7 (no emendation)

1. 1925A §30 — nescit quemadmodum inquinari NON possit. The argument wants "she does not know how she can be defiled" (that is the sense of the whole paragraph, and of §32's parallel quemadmodum eos inquinare non possis, where the negative belongs). Migne prints the negative here as well, producing "she knows not how she may not be defiled." The negation is carried (Pattern 7a): the English reads she knows not how she may not be defiled. Conjecture: nescit quemadmodum inquinari possit. Not emended.

1. 1923D §25 — a full stop inside a quotation frame. The plate prints …Cherubim et Seraphim, qui dicunt. Sanctus, sanctus, sanctus… — a terminal period where a colon belongs. Preserved (Pattern 8: never delete a printed mark). The English reads "…are present, who say. Holy, holy…".

1. 1922A §18 and 1923B–C §24 — Nonne … sentences closed with a full stop. Both interrogatives are printed with terminal periods, not question marks. The English keeps the interrogative construction and the printed period.

1. 1926C §35 — sed cito redit, EI iterum pulsat. ei is almost certainly a misprint for et ("and knocks again"), which the asyndeton and the parallel clauses require. But ei is a real word occupying a possible slot (dative, "to him"), so under Pattern 12's boundary it is rendered as ordinary English, not carried: "to him he knocks again." No [sic:]. Conjecture: et.

1. 1927D §40 — praesentia enim Verbi HAUSIT animae virtutem. haurire is "drain, draw off," and the comparison that follows (Mary's presence erudivit John in the womb) requires the Word's presence to impart strength, not remove it. Rendered literally, "drank in the strength of the soul." Not emended; no plausible single-word conjecture (hausit may be used loosely for "drew in, absorbed").

1. 1933C §56 — quo diutius abest QUI DESIDERAT. The sense demands qui desideratur, "he who is desired" — it is the beloved's absence that inflames the waiting soul, and the clause's own apodosis says so (exspectantis desideria). Printed reading carried: "the longer he who desires is absent." Conjecture: qui desideratur.

1. 1935A §62 — Hos ALIOS ac profundos… pascit Dominus. alios pairs absurdly with profundos; altos ("high and deep," the standard doublet, and the paragraph is about profunda prudentia and alta aperire) is near-certain. alios is a real word in a possible slot, so it is rendered, not carried: "These others, and deep ones, the Lord feeds…" Conjecture: altos.

1. 1933B §56 — spes PROPRIOR. For propior. Attested as a scribal/print variant and with no English exponent for the doubled r; translated normally under Pattern 9 ("a nearer hope"). No marker, no emendation.

1. 1932B §54 — Visu COEPIT fulgorem. coepit for cepit is a standard late-Latin orthographic confusion (Pattern 9, attested spelling); rendered "took in." No marker.

1. 1921C §15, 1921D §16 — ferrum non ascendit. Carried literally ("no iron comes up"), echoing Judg 13:5 novacula non ascendet, rather than smoothed to "no razor touches." Not a defect; noted because the phrase recurs.

1. 1927B §38 — stibium. Antimony eye-paint. Rendered "the antimony" rather than a modern cosmetic word; the noun is concrete, not technical-institutional, so it is Englished and not left in italic Latin.

1. 1934C §60 — the David/Saul period. Denique etiam David juvenculae Mariae auctorem ex cujus successione Christus est natus per Virginis partum, in decem millibus praedicabant has no expressed subject (the women of 1 Sam 18:7) and an appositive that outruns its case. Rendered as closely as the printed words allow, with the appositive kept as an aside. Not emended, no marker.

6g — cols 1935C–1949A

C. Printed readings rendered literally under Pattern 7 (no emendation)

1. 1939A §77 — Hoc autem judicium est, QUA lux venit in hunc mundum. Migne's own note points at John 3:19, where the Vulgate and every witness read quia. qua is a real word occupying a possible slot (ablative feminine of qui, "whereby"), so under Pattern 12's boundary it is rendered as ordinary English, not carried: "whereby the light came into this world." Conjecture: quia. No [sic:], no [var:] — the divergence is a printing matter, not a witness to a different scriptural text.

1. 1944D §15 — Torrentem PERTRANAVIT anima nostra. For Ps 123:5's pertransivit. Unlike case 1, pertranare ("swim across") is a normal Latin verb with a clean English exponent, and the surrounding exposition is about temptations that are tolerabiles — the soul gets across them. Rendered literally, "hath swum through the torrent," and marked [var:]. Not emended.

1. 1937C §73 — Onir autem propter castitatem…. Onir is the onyx (LXX λίθου / Gk ὄνυξ, latinized), and it is not in the lemma Migne has just printed — Venter ejus eburneus, distinctus saphiris names only sapphire. Ambrose glosses a stone his own quoted verse does not contain. Rendered "the onyx"; the lemma is not supplemented and the gloss is not cut. Not emended, no marker.

1. 1941D §4 — ut quid mihi eum de quo veniat diabolus et dicat. Meus est. The period after dicat falls where a colon belongs, and ut quid mihi eum has no governing verb. Punctuation preserved (Pattern 8) and the clause rendered as closely as the printed words allow: "to what end is he mine, concerning whom the devil may come and say, He is mine." Not emended.

1. 1946B §21 — idem pater Naason, id est, serpentum. Ambrose derives Naason from Hebrew naḥash, "serpent," where §17 (1945B) had given the same name its Numbers referent, the prince of Judah. The two etymologies stand side by side unreconciled in the printed text; both rendered as printed.

1. 1948C §7 — Umbilicus EJUS tamquam crater tornatilis, glossed with Et venter EJUS, against §8's Umbilicus TUUS. The plate shifts the lemma from second to third person and back within two paragraphs. Rendered as printed each time ("her navel" / "thy navel"); not harmonized.

1. 1947D §4 — divites cum saeculi istius terreno rege senserunt. sentire cum = "to side with"; rendered "sided with," not "perceived." The clause is printed with no negative and it carries none in the English.

1. 1948B §6 — verae autem virtutis insignis spiritualibus cervices fidelium adornavit. insignis stands where the parallel specie… tamquam manu artificis wants an ablative plural (insignibus), and the subject of adornavit is left to be supplied from the preceding Sive… sive pair. Rendered as closely as the printed words allow, keeping the sive… sive period intact. Not emended, no marker; the mismatch has no English exponent (Pattern 9).

1. 1942A §5 — Erat homo abscondens sermones suos. An unmarked allusion (cf. Sir 20:5–7 / Prov 11:13) printed as narrative, not as a quotation, and with no note. Rendered as narrative; not treated as scripture, not marked.

6h — cols 1949B–1962B

C. Printed readings rendered literally under Pattern 7 (no emendation, no marker)

1. 0036 1951D §1 — Quae est quae haec ascendit candida. The plate transposes haec: the same verse is printed Quae est haec quae ascendit at 1952B (§2) and again at 1953B (§6). Rendered as closely as the printed words allow — "Who is she who this one comes up white" — and not repaired to the reading Migne himself prints twice within two columns. Considered for [d: …] and rejected: the two quae are not an adjacent repeated run, and dittography is not what the plate did here.
2. 0036 1953D §7 — flagrantiam spiritus non habebat. flagrantia is "blazing," fragrantia "fragrance," and the whole sentence is a contrast of odors (confessionis suae odorem diffundebat longius), so fragrantiam is near-certain. But flagrantia is a real word occupying a real slot, so under Pattern 12's boundary it is rendered, not carried: "the burning of the spirit." No [sic:]. Conjecture: fragrantiam.
3. 0036 1952A §1 — the anacoluthon at Quomodo ea quae posuit caput suum…. A sentence of six clauses whose subject ea… anima never reaches a finite verb; it closes on the infinitive phrase hoc est enim caput suum… constituere. Rendered as printed ("In what manner she who has set…"), with no supplied apodosis.
4. 0037 1956C §17 — Haud facilis volatus omnibus. The haud is printed and is carried ("Flight is by no means easy for all"), against the encouraging drift of §16's quia volandi nobis data est copia. Not smoothed.
5. 0037 1956B §15 — Ipsi Petro… dicit: Transi, recumbe. Sed ille dicit: Igne nos examinavit…. The exchange does not construe as a dialogue: Peter answers an invitation with a third-person psalm verse, and Sed marks a contrast that is not there. Rendered flat and as printed. No conjecture is defensible on one clause.
6. 0038 1959A §26 — spiritalia sub jugum Verbi corde mansueto corda subjecit. spiritalia… corda is split around two prepositional phrases, and the subject is the congregatio of the preceding clause. Rendered as printed, with the word order normalized only as far as English requires.
7. 0039 1960C §32 — effugitans canum. effugito governs the accusative; canum is genitive. Rendered "and fleeing of dogs," which keeps the genitive visible, rather than "fleeing the dogs," which would silently supply the case. Conjecture: canes.
8. 0039 1960D §32 — Ubi ergo Christus quaeritur, in pectore prudentis scilicet sacerdotis? A question that contains its own scilicet answer and closes the paragraph with no connection to what follows. Kept as printed, question mark included (Pattern 8).
9. 0038 1958A §22 — the John 10:9 citation is INLINE, not an [n:] note, and Migne's column anchor falls inside the parenthesis: (Joan. [1958A] X, 9). Reproduced exactly, anchor included, the locator untranslated. Not tagged [f: …] — Pattern 4 does not apply to a commentary's inline scripture (the same judgment the peer recorded for Psalm CXX, 6 at 1870C).
10. 0035 1949D §13 — Quod sequitur secundum profectum: tertium secundum perfectionem? A question mark closing what is grammatically a statement. Preserved (Pattern 8: never delete a printed mark).
11. 0037 1957A §18 — the lemma splices two verses. Pone me ut signaculum… quae nova et vetera servasti mihi joins Cant 8:6 to Cant 7:13. Ambrose's own conflation, not a defect; rendered as printed, unmarked.

7. Punctuation carried as printed (Pattern 8)

Never delete a printed mark. Additional cases are embedded in §6.

Punctuation carried as printed (Pattern 8)

###### 1893B–C — full stop breaking a sentence (§8) Qui vero angelos transivit. Verbum invenit; — a full stop where the sense wants a comma or nothing at all (Qui vero angelos transivit, Verbum invenit). The stop is preserved: "But he who passed beyond the angels. Found the Word;". The English is deliberately left without a supplied subject in the second sentence, because supplying one would be an emendation dressed as syntax.

###### 1896B — full stop followed by a lower-case relative (§20) …de orationibus sanctorum [n: (Apoc. VIII, 4)]. quae incensa referuntur… Preserved: full stop, then lower-case "which are offered as incense…". Not merged into one sentence.

###### 1899C — exclamation mark inside a quotation (§30) Infelix ego homo! quis me liberabit… — Migne's ! where the Vulgate has a comma. Preserved, and the following clause left lower-case as printed.

8. Migne's own references — reproduced, never corrected (refDisplay rule)

No printed reference in this work has been emended, and no citation-corrections.json entry is proposed for any of them. No Ibid. chain has been touched.

8a — cols 1893C–1907C

Citations — printed as Migne gives them, not corrected (refDisplay rule)

None of the following is emended in the text or proposed for citation-corrections.json; they are recorded so the index has the reasoning.

• 1894A §10 — [n: (Psal. XVII, 12)] covers two verses run together: Posuit tenebras latibulum suum is Ps. XVII, 12, but nox nocti indicat scientiam is Ps. XVIII, 3. Migne's single note stands.
• 1894D §14 — [n: (Matth. XXVIII, 5 et seq.)] is attached to a quotation whose closing clause (quid igitur quaeritis viventem cum mortuis) belongs to Luc. XXIV, 5. Marked with a [var:] at that point; the note is untouched.
• 1900C §34 — [n: (Joan. XVIII, 17)] for Ego in hoc natus sum, which stands at Joan. XVIII, 37. Printed as given.
• 1905A §15 — [n: (Num. XX, 11)] attached to Ideo et manna manducaverunt; Num. XX, 11 is the striking of the rock, and the manna is Exod. XVI. Printed as given. Note also that this section carries an inline italic citation (Exodi XV, 25) in the running text rather than as an [n:] note; reproduced verbatim in place, untranslated, and not tagged [f:] — a commentary quoting and expounding a lemma does not meet the Pattern 4 trigger.
• 1905A §15 — the two [n: (I Cor. X, 3 et 4)] notes in this section are both printed with the same locator although the second attaches to Petra erat enim Christus (I Cor. X, 4). Printed as given.

8b — cols 1907D–1921B

Migne's own references — left as printed

Per the standing rule, no printed reference has been corrected, and no Ibid. chain has been touched.

• 1913B (§41) — the lemma of Song 4:13–14 is cited (Vers. 3, 4.). The verses are 13 and 14.
• 1918D (§6) — Exsurge, aquilo… (Song 4:16) is cited (Cant. IV, 10).
• 1920A (§10) — the note (Psalm. LXXIX, 9) is attached to Vindemiavi myrrham cum unguentis meis (Song 5:1) but belongs to Vineam ex Aegypto transtulisti, which follows two sentences later. Reproduced at the position Migne prints it.
• 1914C (§44) — (Ibid., 18) resolves to Prov. V, and is left as printed.

8c — recorded in the divergence tables above

Reference-vs-text mismatches also stand at 1854B (Ezech. XXI, 14 for 6:11), 1853D (Pov. for Prov.), 1857B (Ps. 118:12 for wording answering to 118:68), 1859B (Matth. XIV, 30), 1865B (Esai. LII, 7 for wording that is Rom 10:15's), 1866D (Zach. XIII, 7 printed in the Matt/Mark form), 1870A (Psal. CXVIII, 146 for v. 17), 1888A (Rom. XXIII, 12), 1926C (II Cor. IV, 19), 1931A (Ps. CXXXII, 2), 1935B (Psalm. XLI, 3 for Ps 62:2's wording), 1936C (Luc. VII, 40-42 for Luke 10), 1936D (II Cor. II, 4 for 1 Cor 2:4), 1938C (Psal. XXIII, 9 for Ps 33:9), 1944B (Eccli. XXIII, 26), 1946B (IV Reg. XI, 12 for 2:12), 1947A (Psal. CCXIII, 5 — there is no Psalm 213), 1935B (Cant. VIII, 7 for 8:5, and a note with an empty verse number, (Cant. I, )), 1913B (Vers. 3, 4 for 13–14), 1918D (Cant. IV, 10 for 4:16), 1955C (Matth. XXIII, 27 for 23:37), 1953A (Matth. IV, 3 for the pinnacle at 4:5), and the verse numbers running one ahead of the Vulgate's through 0038.

9. Pattern 9 — attested spelling, gender and orthographic variance: no action

3. Orthographic and structural variance — no action (Pattern 9)

• 0003, col 1861B: the same lemma is printed filiae Hierusalem and, eleven lines later, filiae Jerusalem. English has no exponent for the variance; both render "daughters of Jerusalem". No marker, no [var:].
• 0003, cols 1861A / 1862B: the lemma appears as Fusca sum et decora and then as Nigra sum, et decora. Not an error: §24's whole grammatical argument depends on the printed et, and the two adjectives are the two Old Latin renderings the author is working across. Both rendered as printed ("dusky" / "black"); the divergence from the Vulgate is marked at 1861B (see §5.14).
• 0000, col 1853B: ## INCIPIT TEXTUS. carries no body before ## CAPUT PRIMUM. The empty section is reproduced as printed.
• 0000, col 1853B: the anchor [1853B] recurs after [1854A]/[1854B], and the inner edition number 1547 likewise recurs, at the transition from the Prologue to the beginning of the text proper. Both bands are transcribed where the plate prints them; the text does not break mid-word or mid-construction across either. Normal, per the column-band rule. No action.

Pattern 9 — logged, no edit, do not churn

• 1893C–D §8, et non exutus est pallio: masculine participle where the subject through the paragraph is Ecclesia / anima (feminine). Latin grammatical gender has no English exponent; the English reads "he was not stripped of his cloak" following the printed masculine. No emendation to make or undo.
• 1904B §12, Istae sunt vere qui geminos creant: feminine demonstrative with a masculine relative. Same class.

Not defects — recorded so a later pass does not re-open them

• The band letters in 1893C→1893D→1894A etc. run in order through this stint; no anchor breaks a word or a construction. Nothing to report.
• 1575–1583: the Benedictine-edition column numbers Migne sets in italics inside the running text. Reproduced verbatim and in place, at the same word boundary as the Latin where English word order allows.
• 1898D §28 Non misereberis illius against Vulg. Deut. XIX, 13 eius — a pronoun variant with no effect on sense; rendered normally, no [var:].
• 1897C §23 nec (ut dicitur) ave dicat — Migne's own parenthetical hedge on the idiom, translated in place as his comment, not marked.
• No guillemets are printed anywhere in 1893C–1907C; none are supplied. All quotation in these chunks is marked by italics alone, and the English follows suit rather than adding " ".

10. Candidates investigated and REJECTED — do not churn these

10a — cols 1879D–1893B

Considered and NOT marked

• 1885A (0012 §50) discesit sibi — single s against discessit sibi three lines later in the same section and at 1884B. A one-letter consonant degemination is standard medieval orthographic variation (Pattern 9), not type unique to this plate; translated normally, no marker.
• 1887A (0012 §56) and 1888A (0013 §59) Integimento / integimento. This looked at first like Pattern 10 run-together type for in tegimento (which the plate prints separately at 1887A §55 and 1888A §59). It is not: integimentum is a real word (= integumentum, "covering"), and the ablative construes without a preposition in both places — Integimento enim corporis Christi … munimentum … invenit, Hoc integimento animae piorum indutae. Attested spelling, Pattern 9. Rendered "covering", no marker. Recorded so a later sweep does not churn it.
• 1884B (0011 §48) proeteriit and its recurrence at 1884D (0012 §50) — oe for ae is a printing convention of the volume, not a defect.

10b — cols 1949B–1962B

F. Candidates investigated and REJECTED — do not churn these

Recorded so a later sweep does not re-open settled ground.

1. 0035 — [1951D] sits at the end of §21, with §22 and the ## CAPUT OCTAVUM. head following it unanchored. Column-band positional guide, not a defect. No text is broken across it.
2. 0039 — [1961B] is followed by [1962A]; columns 1961C–D never appear. Checked: a ## 34. head intervenes and no word or construction is broken across the gap. Band arithmetic; not reported.
3. ## CAPUT OCTAVUM. (0035) is followed by an empty body in the Latin chunk — the chapter's text opens the next chunk. The English keeps the head and the same empty body, exactly as the peer handled ## CAPUT SECUNDUM. at 0006. Heads count 13 : 13.
4. Note punctuation irregularities in 0038 — [n: (Vers 12.)] with no period after Vers, [n: (Vers. 10)] with no period after the numeral. Reproduced verbatim; no normalization.
5. The plate's verse numbers run one ahead of the Vulgate's through 0038 — (Vers. 11.) at §24 and (Vers 12.) at §25 both stand over Cant 8:10. References left untouched.
6. (Matth. XXIII, 27) for Matt 23:37 and (Matth. IV, 3) for the temple pinnacle at Matt 4:5 (both 0036–0037). Reference-vs-text mismatch is on the do-not-correct list; logged at §B, untouched.
7. 0036 §4 — ministros suos ignem urentem, sed non exurentem. The non looked like a candidate defect against Ps 103:4. It is Ambrose's own antithesis, not a quotation error; carried, no marker.
8. 0037 §14 — ipsum autem ex nullo. Ellipsis, not a hole: pendet carries over from the preceding clause. Rendered "while itself hangs upon none." No marker.
9. Pattern 4 [f: …] was considered for the inline (Joan. X, 9) at 0038 §22 and rejected — a commentary quoting scripture inline does not qualify, however the citation is typeset. See §C.9.
10. Migne prints no ( sic ) of his own anywhere in 1949B–1962B, so no question of stacking ours on his arose.

10c — rejected elsewhere in this file, gathered

• [d: …] was considered and rejected three times: at 1881B–C, where §37 and §38 print the same clause with domus panis / domus Dei (Pattern 11 wraps a repeated run, not a variant one); at 1909A–B, where Favum itaque mellis tua opera componant is printed twice, a section-head and a lemma apart; and at 1951D §1, where Quae est quae haec ascendit is not an adjacent repeated run. No [d: …] is fired anywhere in the work.
• [ed: …]: none fired. Every empty element in the work — ## INCIPIT TEXTUS. (1853B), ## CAPUT SECUNDUM., ## CAPUT TERTIUM. (1891A), ## CAPUT SEXTUM., ## CAPUT SEPTIMUM., ## CAPUT OCTAVUM. — is the plate's, not a digitization hole.
• Missing section heads at 0027 (## 43.) and 0034 (## 6.) were NOT restored — supplying them would break marker parity; the heads: frontmatter records the gap. The section number 50 is printed twice at 1884D/1885D and reproduced as printed.
• Band-gap arithmetic is not a defect: [1853B] recurring after [1854A]/[1854B] at the Prologue/text transition; [1951D] sitting at the end of §21 with §22 and ## CAPUT OCTAVUM. following unanchored; [1961B] followed by [1962A] with 1961C–D never appearing. Each was checked for a word or construction broken across the gap; none is. Not reported.
• discesit (1885A) — single s against discessit three lines later: consonant degemination, Pattern 9, no marker. Integimento (1887A, 1888A) — a real word (= integumentum), construing without a preposition; not run-together type for in tegimento, which the plate prints separately in the same sections. ortus for hortus (1916C) — h-dropping, Pattern 9, rendered "her garden". proeteriit (1884B, 1884D) — oe for ae is this volume's printing convention. proprior (1933B), coepit for cepit (1932B) — attested variants, no marker.
• Pattern 4 [f: …] was considered and rejected for every inline scripture citation in this work (Psalm CXX, 6 at 1870C; Exodi XV, 25 at 1905A; Exodi III, 5 at 1947A and 1947C; Joan. X, 9 at 1958A). A commentary quoting and expounding scripture inline does not meet the Pattern 4 trigger, however the citation is typeset — including where Migne's column anchor falls inside the parenthesis.
• Migne prints no (sic) of his own anywhere in the work, so no question of stacking ours on his ever arose.
• No guillemets are printed anywhere in 1851A–1962B. Migne marks every quotation with italics alone and the English does the same; none are supplied, in any chunk.

11. Structural notes, register and marker parity

11a — cols 1851A–1866A

6. Second person (Pattern 17)

Archaic thou is used throughout for singular addressees, modern you for plurals, per the number rule and not by dignity:

• Thou — the bridegroom/Word (ubera tua, Attrahe nos post te), the reader or the soul (Vis ad Christum aptare?, Cognosce te, anima, Disce, homo), the virgin (Accipe vas tuum, virgo), the Synagogue in §§31–32 (Cur dimisisti quem tenebas?), and Moses at 1865B.
• You — filiae Hierusalem (Nolite aspicere me, 1861A/B), Omnes tribus Israel, convenite (1861D), and the plural imperatives inside the Colossians quotation at 1858A (Ne tetigeritis…) and the Matthew quotation at 1859B (Venite ad me, omnes qui laboratis).

The transition is visible at 1861A–B, where the soul speaks of the bridegroom as thou and to the daughters of Jerusalem as you within the same paragraph — which is the distinction the rule exists to recover.

11b — cols 1866A–1879C

6. Structural notes

• ## CAPUT SECUNDUM. (0006) is followed by an empty body in the Latin chunk; the English keeps the head ("CHAPTER THE SECOND.") and the same empty body, so the heads count is 11 : 11.
• Bare column numbers 1557–1565 (Migne's inner numbering of PL 15) are reproduced verbatim, in italics, at the same point in the sentence. Two of them fall mid-phrase — 1561 between claritate and divina (1873C), 1564 between caput and suum (1878A) — and were placed at the nearest equivalent point in the English clause.
• Second person follows Pattern 17: thou/thy wherever Migne's addressee is grammatically singular (the soul, the virgin, the bride, the reader), you for the plurals (Tollite jugum meum super vos, Adjuravi vos, filiae Hierusalem, Introducite me, dedit vobis Dominus manducare).
• Marker parity verified against the Latin twins: column anchors 11/11/11/10/11 in order; [n: …] 12/12/11/6/16, identical strings in identical order; ## heads 8/11/11/8/9.

11c — cols 1921C–1935B

D. Structural observations (not defects, recorded so a later pass does not churn them)

• 0027 has no ## 43. head. The plate's §43 (the Invenerunt me custodes paragraph at 1928C–1929A) is printed without its numeral in our source; the heads: frontmatter records the gap (…"42.","44."…). The English mirrors the Latin exactly: eight ## heads, and §43's text runs as an unheaded paragraph inside §42. Not restored — the absence is in the chunk we were given, and supplying the head would break marker parity.
• 0027 prints its verse notes out of order — (Vers. 6.) at §39 precedes (Vers. 5.) at §40. Reproduced as printed.
• §35 and §37 (0026) expound the same verse twice with different Latin, as do §§58–60 (0029) for Cant 5:10. Both are Ambrose re-quoting, not a chunking artifact; neither text was harmonized to the other.
• No guillemets anywhere in 1921C–1935B. Migne marks every quotation in this stretch with italics alone, so the English carries none either (Pattern 6).
• Edition page markers 1593–1601 fall mid-sentence in the Latin and are reproduced at the nearest equivalent point in the English clause.

11d — cols 1935C–1949A

D. Structural observations (not defects, recorded so a later pass does not churn them)

• 0034 has no ## 6. head. The plate's §6 (1948A, Moduli femorum tuorum…) is printed inline as [1948A] 1610 6. Moduli…, and the heads: frontmatter records the gap (…"5.","7."…). The English mirrors the Latin exactly: twelve ## heads, and §6 runs as an unheaded paragraph after §5. Not restored — supplying the head would break marker parity. Same class as the missing ## 43. in 0027.
• ## CAPUT SEXTUM. (0031) and ## CAPUT SEPTIMUM. (0034) are both followed by two blank lines and no text before the next numbered head. Mirrored exactly.
• Two scripture citations in 0034 (§2 and §3) are printed as bare parentheses inside the italic lemma, (Exodi III, 5), rather than as [n: …] notes. Reproduced verbatim in place, untranslated — they are Migne's citations on the same footing as [n:] contents, and they are not counted in noteCount.
• [n: (Vers. 5-7.)] opens 0033 outside the italics, before Laudatur etiam Sponsa, because §10 is exposition rather than a lemma. Position preserved.
• The verse notes run behind the text in 0033: (Vers. 5-7.) heads §10, which expounds Cant 6:8; (Vers. 9.) at §13 for Cant 6:9. Reproduced as printed.
• No guillemets anywhere in 1935C–1949A. Migne marks every quotation in this stretch with italics alone, so the English carries none either (Pattern 6). Straight quotes are used nowhere; the Meus es / Tuus sum dialogue of §§3–4 (1940D–1941D) is printed unquoted and unitalicized and is rendered the same way.
• Edition page markers 1602–1610 fall mid-sentence in the Latin and are reproduced at the nearest equivalent point in the English clause.
• Pattern 17 (thou = singular). Ambrose addresses a singular tu throughout §§67, 71, 3–4, 9, and 21–22, and switches to plural vos only at 1943A §9 (orphanos non relinquam vos, quoting Matt 28:20 vobiscum). The English carries the switch: "thou/thine" up to Even if I have risen, then "you."

11e — cols 1949B–1962B

G. Structural notes and parity

• Marker parity verified against the Latin twins, all five chunks: column anchors 11 / 11 / 11 / 11 / 7, identical strings in identical order; [n: …] 11 / 15 / 11 / 15 / 6, identical strings in identical order; ## heads 13 / 10 / 9 / 10 / 6, 1:1. verify-english.mjs 6963 reports the whole work aligned.
• Chunks read and found clean (no defect, no divergence beyond what is tabled above, nothing requiring a marker): the Latin twins of 0035, 0036, 0037, 0038, 0039 were read in full, plus the style anchor pair 9741/0000 and the peer cruces files cruces-0005.md and cruces-0025.md. 0039 is clean of carried type entirely — its only findings are the three [var:] of §B and the two literal renderings at §C.7–8; its tail was re-checked against the Latin twin after a transport fault and is complete to servare se debet et custodire at [1962B], the work's colLast.
• Edition page markers 1611–1618 (Migne's inner numbering of PL 15) are reproduced verbatim, in italics, at the equivalent point in the English clause. Three fall mid-phrase: 1613 inside the Cant 8:6 lemma at 1952C, between durus and sicut inferi (the italic run is closed and reopened around it, as the Latin does); 1614 between unde and et Paulus ait at 1954A; 1617 at the end of §24, after the quotation.
• No guillemets anywhere in 1949B–1962B. Migne marks every quotation in this stretch with italics alone, so the English carries none either (Pattern 6). All quotation marks in these five chunks are the italics Migne prints.
• Second person follows Pattern 17 throughout: thou / thy for the grammatically singular addressee (the soul, o homo at §11, the sealed bride of 0037 §18), you for the plurals — Adjuravi vos, filiae Hierusalem, Quomodo dicitis, Ipse vos baptizabit, donec formetur Christus in vobis.

12. Provenance of this file

Merged 2026-08-14 from the eight per-batch cruces files, which were deleted after the merge was verified complete. Each stint's own preamble is preserved here.

Chunks 0000–0004 — from cruces-0000.md:

# Cruces — 6963, Ambrose, Commentarius in Cantica canticorum, chunks 0000–0004 PL 15, cols 1851–1866A. Batch: chunks 0000, 0001, 0002, 0003, 0004. Marker parity verified against the Latin twins: column anchors, [n:] contents, ## heads, and Migne's inner edition numbers (1545–1556) all 1:1 and in order. No guillemets are printed anywhere in these five chunks, so none are supplied. ---

Chunks 0005–0009 — from cruces-0005.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0005–0009 Columns 1866B–1879C. Ambrose expounds an Old Latin / pre-Vulgate Song of Songs throughout; the divergence list below is the principal deliverable of this batch. ---

Chunks 0010–0014 — from cruces-0010.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0010–0014 Columns 1879D–1893B. Song of Songs 2:8 – 3:2. ---

Chunks 0015–0019 — from cruces-0015.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0015–0019 Columns 1893C–1907C. Compiled with the translation; each entry gives column, printed reading, and the conjecture where one is defensible. Conjectures live here only — the running text renders what Migne prints (Pattern 7). ---

Chunks 0020–0024 — from cruces-0020.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0020–0024 Columns 1907D–1921B. Every entry gives the column, the printed reading, and the judgment. Nothing recorded here was repaired inside the translation.

Chunks 0025–0029 — from cruces-0025.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0025–0029 Columns 1921C–1935B. Batch translated 2026-08-14. Sections keyed by Migne column and the plate's own paragraph number. ---

Chunks 0030–0034 — from cruces-0030.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0030–0034 Columns 1935C–1949A. Batch translated 2026-08-14. Sections keyed by Migne column and the plate's own paragraph number. Continues cruces-0025.md; conventions set there (Old Latin / LXX-derived Song throughout, no guillemets, [var:] at first occurrence per chunk) are followed here without restating them. ---

Chunks 0035–0039 — from cruces-0035.md:

# Cruces — 6963 (Ambrose, Commentarius in Cantica canticorum), chunks 0035–0039 Columns 1949B–1962B — the closing stretch of the work (Cant. 7:4–8:14, the end of Book VII and the whole of Caput octavum). Batch translated 2026-08-14. Sections keyed by Migne column and the plate's own paragraph number. Migne prints an Old Latin / LXX-derived Song throughout, as in every earlier stint. The divergence tables in §§A–B below are the principal deliverable. Fixed equivalences of cruces-0005.md §1 were followed without exception: frater meus → my brother, charitas → charity / dilectio → love, dealbata → made white (matching 0029). Nothing in these five chunks required a new work-level equivalence. ---

Marker parity was verified against the Latin twins by every stint (column anchors, [n: …] contents, ## heads, and Migne's inner edition numbers 1545–1618), and node scripts/verify-english.mjs 6963 passes for the whole work after the merge edits of §1.2.

Nine Opus readers, blind to this file, read all 40 chunks against their Latin twins. 29 sites on our side, plus four secondary repairs (§5.3). Straight dropped negations came back zero in the entire work — every printed non / nec / nisi / haud / numquam / nullus / nihil / ne is carried, including the dense stacks at 0026 §29, 0032 §4, 0034 §3 and 0039 §§32–34. The damage here is speaker and subject mis-binding, mis-scoped particles and mis-bound cases, and markers that swallowed the force they governed (§3 above). Old Latin fidelity was hunted specifically in every range and held everywhere: 0 drift sites in ~60 checked lemmata. The thing most at risk in this edition is the thing the translators got right.

⚠ The read was interrupted by a session crash with chunks 0025–0039 unread and three reader reports destroyed. Those ranges were re-read from scratch, and 0000–0004 was re-read to recover a lost finding — which it did (the cudere site below), along with four more the first pass had not reported. Chunk 0003 is the only chunk in the work with a clean read and nothing logged against it.

5.1 Open — needs a plate, not a judgment

###### 1853B — cudere where the sense wants cudi ⚠ PLATE CHECK OWED Sic etiam Virgo nostra debuit prius amore pio cudere. Rendered "ought first to hammer with holy love" — the printed active. The English previously read "to be hammered", a silent passive: cudere is active, and it is the only one of four infinitives under debuit (cudere, mirari, cernere, haurire) that had been so treated, which made the Virgin the patient of the first and the agent of the other three inside one series.

The sense does want the passive — the whole surrounding image is passive on the bride's side (nubenti prius plauditur, quam imperatur; Bucula … prius assuescitur verbo lasciviae) — so cudi is the near-certain reading. But cudere is what our source prints, and sources/pl/tei/6963.xml prints it too, so this is not our pipeline's error and cannot be resolved from our files. cud- occurs nowhere else in the 40 chunks, so the corpus cannot settle it either. Owed: a read of the PL 15 plate at 1853B. If it prints cudi, correct the twin and restore the passive.

###### 1854B — [sic: deret] leaves a purpose clause verbless …quae omnia fecisset ut sponsum deret… The underlying word is not recoverable from the corpus (videret? adiret? teneret?), so unlike the four sites in §3 the marker cannot be given its force from internal evidence. Marker kept, deliberately. Same plate read settles it.

5.2 Ruled and applied

5.3 Secondary repairs — not polarity, but the same read found them

###### custodes was splitting into two English words, inside one verse The Song lemma at 1928C prints custodes twice — Invenerunt me custodes qui circumeunt civitatem … tulerunt pallium meum a me custodes murorum — and the English had "guards who go about the city" against "keepers of the walls". Across the work: 0015 all "guards" (5), 0027 mixed 5/2, 0028 all "keepers" (8). This invents a distinction the plate does not print, and 0028 §47 is built on the identity of the word — Fortasse sunt et alii custodes, "perhaps there are also other custodes", an argument that only runs if it is the same word. Unified to guards throughout (the dominant form, and the one 0015 already used for the same scene). This is the class the merge pass exists to catch and did not: the two chunks were written by different agents, neither could see the other, and both were internally consistent.

###### 0012's ratio warning — adjudicated, NOT padding The verifier flags 0012 as the work's highest EN/LA ratio. Its batch's report was lost in the crash, so this was closed here instead. Expansion is distributed, not localized: per-paragraph ratios run 1.48–2.19 across all 21 paragraphs, and the two highest are explained without residue. 1886D–1887A (Hanc vineam diligens agricola fodere, rigare, contendere consuevit…) is one dense synthetic period — five infinitives, three ne-clauses, two ablative absolutes — unpacked clause for clause with nothing added. 1887B's inflation is apparatus, not prose: the [var:] note carrying the Vulgate's in foraminibus petrae, in caverna maceriae against Migne's in tegimento petrae, juxta praemunitionem is 24 of its 44 extra words. The work's whole range is 1.50–1.87; 0012 is not an outlier in it.

###### 1877D (0009) — John 7:23 conformed to the familiar interrogative Migne prints dicens: Mihi indignamini quia totum hominem sanum feci in sabbato, indicative, no question mark — the Vulgate points it as a question, and the English had followed the Vulgate: "Are ye angry with me because…?" Now Ye are angry with me because I made a whole man sound on the sabbath. This one was never adjudicated by the first pass — its reader's report was lost in the crash — and it is exactly the 7a″ pressure (familiar wording arriving over the plate), reaching the page through punctuation rather than vocabulary.

###### 1872B (0007) — the one question mark KEPT against the plate Migne prints nonne inter asperitates laborum contritionesque animorum boni flos odoris assurgit; with a semicolon, and the English closes it with "?". Kept, and this is not the same case as 1927C: nonne is an unambiguous interrogative particle, so the clause is a question and English has no way to carry an interrogative without its mark, whereas 1927C's Quomodo transivit, id est… is a gloss on a lemma word and was never interrogative at all. The remaining 0002/0004 warnings are the same punctuation class, resolved in §5.4.

###### 1927C — a printed statement turned into a rhetorical question Migne prints Quomodo transivit, id est, mentis interna penetravit with a comma, as a gloss on the lemma word. The English had "How did he pass by? That is, he penetrated…". Migne punctuates questions freely in this very chunk (§38, §42), so the comma is a real signal, not plate noise. Now: How he passed by: that is, he penetrated… — which also resolves 0027's question-mark warning.

5.4 Ruled and deliberately NOT changed

###### 1959B — quam illi utique tribuit ipse quem diligit A reader proposed binding quem diligit to illi ("assigns to that one whom he loves"), on the strength of the parallel Ambrose immediately draws — sicut dilexit et Paulum, cui dedit — which makes the loved one the recipient. Rejected as an edit, logged as a crux. The Latin is genuinely muddled, the present English renders the printed word order literally, and the proposed reading requires a case mismatch (illi dative, quem accusative). Pattern 7 says render the printed reading and log the doubt; the reader's own confidence was low. Recorded so a later sweep does not churn it.

###### 1948B — verae autem virtutis insignis spiritualibus Rendered "with the spiritual badges of true virtue", which reads the printed insignis as insignibus. A quiet normalization with no crux — now logged. Not reverted: the clause is unconstruable as printed and no rendering of it is innocent, so the honest record is this entry rather than a churn.

###### The two question-mark warnings, 0002 and 0004 — resolved, nothing missing Both extra English ? are punctuation normalizations of interrogative clauses Migne closes otherwise: 1860B «Ego a te debeo baptizari, et tu venis ad me;» (semicolon) and 1865D–1866A Quid est se noscere, nisi ut sciat unusquisque… (full stop). Every Latin interrogative is present in the English in both chunks. This check exists because it caught a dropped 46-word sentence in 11632; here it caught punctuation.

###### 1960B–C — Migne's Quae sedes against his own Qui sedes At 1959C–D the plate prints the masculine Qui sedes in hortis (speaker: the soul, addressee: Christ). At 1960B, reporting Symmachus and Aquila, it prints the feminine Quae sedes — and Ambrose's whole appeal to them is that reversal (speaker: Christ, addressee: the Church). English "Thou that sittest" cannot inflect, so the distinction is invisible on the page; the frame sentence ("Christ says to the Church") carries the reversal, but not its textual basis. No inline marker: [var:] names an outside witness, and this is Migne disagreeing with himself — the verifier correctly refused it. Recorded here instead.