Liber Sapientiae

Translated 2026-08-23, one translator, 7 chunks (6,456 Latin words), on Opus. Blind-read and repaired 2026-08-23 (§§4–5 below); the work was re-chunked after one TEI patch, 6,457 → 6,456 words, and chunk 0's words: frontmatter was re-synced to 1047. verify-english.mjs clean: columns, notes and sections aligned, no duplicate paragraphs.

Apparatus (after repair): 30 [var:] · 9 [sic:] · 1 [cj:] · 0 [d:] · 0 [ed:] · 0 [nt:] · 150 [n:] (exact parity with the Latin twins). Structure: PROLOGUS + CAPUT PRIMUM → CAPUT XIX (19 heads). Guillemets: zero in both. Question marks: 3 in the Latin (1170A, 1172D, 1173A), 3 in the English. One ! (1173A), carried. Em-dashes: zero. Punctuation compared band by band across all 61 column anchors — colons, semicolons, question marks, exclamations — zero mismatched bands (19 mismatches were found by that scan and all 19 repaired; a total-only check would have cancelled several of them).

Ruth (8968) §1 governs unchanged: VERS. n.-- passes through verbatim, the lemma is Englished, a fragmentary lemma is rendered as the phrase it stands for, the sigla are ordinary [n: …]. Tobit (8970) §1(b) governs the abridgment formula: usque ad → as far as, no verb ever supplied to a resumption tail. Neither was re-opened.

⚠ usque ad occurs 62 times in this Latin and only 55 of them are the formula. Seven are ordinary prose and are translated as prose, not as the abridgment mark: 1168D usque ad passionem Christi, 1172A usque ad adventum Christi and usque ad finem mundi, 1172A the lemma A fine ergo usque ad finem fortiter pertingit (Sap. 8:1 itself, not an abridgment — the single most dangerous look-alike in the work), 1181B usque ad exteriores tenebras, 1183A vindicta usque ad consummationem supervenit. The English "as far as" count is 55, matching the formula count exactly.

Plate status: coverage spot, checked zero. There are no [cn:] markers in this work and none were expected. That is not a claim that the plate carries no notes.

§1 · THE LEMMA DIVERGENCE TABLE

Every italic span in data/briefs/8969-lemmata.txt was walked against sources/vulgate/clementine-flat.txt, including the ones that agree. Verdicts below are mine, not the pre-scan's: a pre-scan ✓ was re-checked at the verse being glossed, and a pre-scan ⚠ was cleared or convicted on the verse, never on the flag.

Class note first, because it covers 51 of the 316 spans: the bare usque ad spans are Migne's abridgment mark, not scripture. They are not collated and take no marker. The four spans that carry a resumption tail inside the italics (1167B, 1170D, 1177A, 1181A) are collated below with the tail's own source.

Second class note — the Haec lemmata of CAPUT X (1173B–1174A). Sap. 10 opens five glosses with a bare Haec + accusative whose verb Migne has cut. English cannot carry a subject-plus-object with no verb, so convention 3 supplies the verb of the verse the fragment opens, and every supplied word is named here: 10:1 custodivit → "kept" (and formatus est → "formed"); 10:6 liberavit → "delivered"; 10:15 liberavit → "delivered". 10:13 is the hard one: its verse verb is negative (non dereliquit), so the expansion reads "She forsook not the just man that was sold." ⚠ The rival was to leave it verb-less and let the comma mark the truncation, and it lost because "She the just man that was sold," is not English at all, which is precisely the case convention 3 exists for. The tension with 7a — "no negation appears that the Latin does not print" — is real and is disclosed here rather than hidden: 7a governs the rendering of Migne's prose; convention 3 expands a lemma from its verse, and the negation is part of the phrase the fragment stands for. Dropping it would have made the lemma assert the opposite of Sap. 10:13. Two spans in this family need no verb at all and got none: 10:5 Haec et in consensu superbiae, and 10:8 Haec autem,.

CAPUT PRIMUM — CAPUT IV (1167B–1170A)

CAPUT V — CAPUT VIII (1170A–1172D)

CAPUT IX — CAPUT XI (1173A–1175B)

CAPUT XII — CAPUT XIII (1175B–1177A)

CAPUT XIV (1177B–1179B)

CAPUT XV — CAPUT XIX (1179B–1184B)

⭐ Summary of the collation: 29 [var:] fired, 3 [sic:] fired on lemma type (1175A, 1176D, 1182B), and roughly forty spans were checked and found to agree with the Clementine at the verse being glossed. The remaining declines are itemized in §3.

⭐ Two divergences are internal to the book and would be invisible to a per-lemma check: 2 Ptr 2:9 at 1169D (Deus) against 1180C (Dominus), and Rom 1:21 at 1170A (in pravis cogitationibus) against 1177A (in cogitationibus). Both are Migne quoting himself two ways within seventeen columns.

§2 · SECTIONS READ CLEAN, BY COLUMN BAND

Read against the Latin twin, marker-stripped, asking only does it assert what he asserts?

• 1167B–1167D (Prologue, Sap 1:1–1:7) — clean. The Prologue's own claim that the book is absent from the Hebrew canon and that its style Graecam magis eloquentiam redolet is the standing reason this work's lemmata diverge more freely than Ruth's, and is carried plainly.
• 1168B–1168C (Sap 1:7–1:16) — clean apart from the Rom 1:29 [var:].
• 1169A–1169B (Sap 2:17–3:8) — clean apart from the Dan 7:27 [var:].
• 1169C–1170A (Sap 3:18–4:19) — clean apart from the 2 Ptr 2:9 and Rom 1:21 [var:].
• 1170B–1170C (Sap 5:14–6:7) — clean. The non solum … sed at 1170C was checked under 2a and the negative sits on solum/opera, confirmed by the sed member.
• 1171B–1171D (Sap 6:22–7:24) — clean apart from Sap 7:15 and the Homo note.
• 1172A–1172D (Sap 8:1–8:20) — clean apart from Jo 1:18 and Sap 8:7.
• 1173A–1173D (Sap 9:7–10:8) — clean apart from Sap 10:5.
• 1174A–1174D (Sap 10:13–11:21) — clean apart from Sap 11:1, 2 Ptr 1:19, Ps 77:34.
• 1175C–1176B (Sap 12:12–12:26) — clean apart from Ps 144:17 and Mal 3:6. The out-of-sequence VERS. 16 / VERS. 15 is Migne's and is reproduced.
• 1177B–1177D (Sap 14:1–14:5) — clean apart from 1 Ptr 3:21.
• 1178B–1179A (Sap 14:11–14:27) — clean apart from Ps 11:4 and Ex 20:7. The Quintilian quotation at 1178D is set by Migne without quotation marks; none were added (Pattern 6).
• 1179B–1180B (Sap 15:1–15:18) — clean apart from the four [var:] named above.
• 1180C–1181B (Sap 16:2–16:28) — clean apart from Sap 16:16.
• 1181D–1182B (Sap 17:6–17:20) — clean.
• 1182C–1184B (Sap 18:9–19:20) — clean apart from Sap 19:10. The sicut … ita at 1184A and the non … sed in the same sentence were both re-checked under 2a (see §3).

§3 · CANDIDATES RAISED AND REJECTED, AND THE CALLS MADE UNDER DOUBT

The [cj:] — 1177A ex invisibilibus, and why [sic:] and [var:] both lost

Migne prints Ut enim Deus invisibilis ex invisibilibus cognosceretur, opus ab eo factum est, quod opificem manifestaret. As printed the sentence says the invisible God was to be known from invisible things — and its own next clause says a work was made to show the workman, which is Rom 1:20's argument from the visible. The English asserts what the author is arguing against: Pattern 18's founding shape.

• [sic:] lost because nothing is broken: invisibilibus is a correct, well-formed Latin word occupying its slot. Pattern 12 excludes exactly this.
• [var:] lost because the words are the gloss's own prose, not a quoted scripture. 18a's separating rule is where the divergence sits, not how it reads.
• So: [cj: invisibilibus; read visibilibus, "from visible things"], additive, with Migne's word keeping its English. This is the only [cj:] in the work.

Rejected [cj:] candidates

• 1170C flagellum collectionis — "the scourge of gathering," in a sentence about hearts that preaching does not soften. collectio in medical Latin is an abscess, and correptionis is an obvious guess. Declined: the sentence still asserts what the gloss asserts (a scourge breaks what preaching cannot), so it fails Pattern 18's bar, which is falsity and not obscurity. Rendered literally.
• 1179D ad notandum Deum vivum — one expects cognoscendum. Declined: notare = "to mark, take note of" renders in place and asserts nothing false.
• 1170B qua praesentia mensura tantum putant — rough past construing. Rendered as closely as the words allow ("whereby they reckon present things only by measure"). Declined: a crux, not a conjecture; nothing here is a real word making a false claim.
• 1173D Gn 28:13's dropped Dominus — the closest [var:] decline in the work. Rejected because the gloss's subject is Jacob's ladder and the divine name is not what it turns on.

Rejected [sic:] candidates

• 1176C ipse est sapientia (Job 28:28) against Vulg. ipsa. Declined under Pattern 9: Latin grammatical gender has no English exponent; there is nothing to render either way. Marked here so a later sweep does not churn it. It is also unitalicized — the gloss's own quotation, not a lemma.
• 1170D Melior est sapientia printed with no VERS. address, and 1175B O Quam bonus the same. Not defects; Migne's own layout, reproduced.
• 1171C's missing terminal stop after [n: (Joan. I)]. Pattern 8: never supply. The English ends without one.
• 1181C quod autem dicit. — Migne prints a full stop in the middle of an accusative-and-infinitive construction. Preserved, and the English breaks where he breaks. Declined as a defect to mark: Pattern 8 governs, and no marker in this corpus means "his punctuation will not construe."
• 1176B facerent, ea, quae non conveniunt — an intrusive comma after facerent. The English drops it, because English syntax has no exponent for a comma in that slot. Logged rather than marked; the Latin column keeps it.

2a trigger sites: every one, and what fired

Nine sites carry a printed negative with solus/omnis/omnino/tantum/non solum … sed/ sicut … ita adjacent. All nine were re-derived from the sentence's own structure:

1. 1170C non solum opera, sed cognitiones — negative on solum/opera, sed member proves it. ✓
2. 1175D nisi indebitam mortem susciperet, non nos… liberaret — on liberaret; the omnia clauses are separate sentences. ✓
3. 1176A contra dominatorem omnium se repugnare non posse — on posse; omnium qualifies dominatorem and is affirmed. ✓
4. 1177A ut Deus omnium intelligeretur… quod non possunt alii — on possunt, subject alii. ✓
5. 1175A per quem omnia facta sunt, et sine quo factum est nihil — omnia affirmed, nihil under sine quo. ✓
6. 1178A Stultus factus est omnis homo… nec est spiritus in eis — omnis affirmed, nec on est. ✓
7. 1171D cui non datur spiritus ad mensuram — on datur; omnia is the next sentence. ✓
8. 1182B sicut… corpora ibi nullo bono operi utilia: ita et hic animae — nullo on bono operi, not on the pivot. ✓
9. ⚠ 1184A non enim creaturam sibi condidit contrariam, sed per omnia voluntati suae consentaneam — this one was caught by the test and repaired. The first draft read "he did not found a creature contrary to himself", floating the negative onto the verb. The sed member pivots on contrariam/consentaneam, so the negative belongs on contrariam: "he founded a creature not contrary to himself, but agreeable in all things to his own will." Exactly 8990 @0614B's shape, and it read better wrong.

Also checked: 1184B non suae tantum conditioni, sed Creatoris… voluntati — "served not only their own condition, but the will of the Creator." ✓ And 1183A non Deum sed hominem purum credidit — the sed fixes the host as Deum: "believed him to be not God but mere man." ✓

Unexpressed subjects supplied, each one named

• 1167B reprimitis — "you": the qui judicatis terram of Sap 1:1, plural, so you and not thou (Pattern 17).
• 1167C invehitur, 1169C ait, 1170D convertit, 1176C dicit — "he": the writer of the book, established in the Prologue. Consistent throughout.
• 1168C celare non possunt — the unexpressed object is errorem suum, from the clause before; rendered "hide it".
• 1172A asserit, exponit — "she": Sapientia, subject of Attingit in the lemma (Sap 8:1). Held for the whole of CAPUT VIII, including glorificat and repraesentat.
• 1173D perduxit — "it": Christi misericordia, the subject of liberavit in the same sentence. Not Christ, and not the martyrs.
• 1177C praeparat — "it": Dei potentia, subject of traduxit and divisit.
• ⚠ 1183A praenoscebat, viderat, credidit — the subject is inside the part the abridgment cut, and no candidate can be proved. Rendered with a bare "it", which is not wrong; ⚑ the blind read asked for the candidates by name (E4), and here they are, in descending order of warrant: (1) the high priest — pontifex, standing behind in figura pontificali, the phrase the abridgment ends on, and the only person in view; (2) the synagogue (synagoga, fem. sg.), the Glossa's habitual subject for foreknowing Christ and then refusing him; (3) the Hebrew people, the collective the Byssum terrae deputant Hebraei opening names. Nothing decides among them, so nothing is chosen in the text. ⛔ The negation non Deum sed hominem purum is correctly scoped on Deum and was left alone.
• ⚑ 1172C exstitit, qui stat, super quem, ipsum — repaired (E1), and the blind read's framing was half right. It reported an English "he" with no antecedent, which was true: the sentence had rendered Partus Virginis as "The Virgin's childbearing", an English it. But the choice is not he-versus-it; it is what partus means. The clause chain ends ipsum gentes deprecabuntur — a childbearing is not entreated — and qui stat in signum populorum / super quem continebunt reges os suum are Is 11:10 and 52:15, where the referent is a person. Partus also means the offspring, and on that sense masculine qui, quem, ipsum and admirandus all construe with the head. Now "The Virgin's offspring was to all men one to be wondered at, because he came forth…" — nothing imported, the antecedent supplied from the Latin word itself.
• ⚑ 1176D qui … distribuebant — repaired (E3). The plural verb agrees with alii, the worshippers, but English word order bound "who" to "the stars" standing next to it, and stars distributing their own positions reads plausibly enough to pass unnoticed. Now "others the circle of the stars, men who distributed…".
• ⚑ 1182D qui … sanctificavit — repaired (E2). Masculine singular qui refers to the Son (Omnipotens sermo tuus … hic tunc … fecit), but the English bound it to "the Egyptians". The sentence is now broken at a semicolon: "…upon the firstborn of the Egyptians; and he afterwards sanctified to himself the firstborn of the Hebrews."

Plate reads requested — four, and one is a set

raw/scans/pl113/patrologiaecurs04migngoog.pdf, PDF page = (column + 11) / 2:

⭐ Record the negatives too. Three forms that looked odd and came back clean on a corpus-frequency check, so no read is asked for them: 1179D notandum (113 occurrences corpus-wide — ordinary Latin), 1170D exigetur (a normal future passive), and 1172A pertingit (a normal compound). All three are rendered as printed with no marker.

What was NOT found

Not one dropped or intruded negation, in either direction. Every non, nec, neque, nisi, nunquam, nullus, nihil and ne the Latin prints is in the English, and none is in the English that the Latin does not print. That was checked band by band and then again site by site under 2a — and the one real defect the 2a pass found (1184A) was a relocation, which no count would have seen. Ruth's profile repeats here: the danger in a Glossa book is not polarity, it is the lemma quietly conformed — which is why §1 is the long section and this one is the short one.

§4 · THE BLIND READ — SITES FOUND, REPAIRED, DECLINED

An independent blind reader read the finished English against the Latin. This section records what it raised, what was repaired, and what was declined and why. A logged column is not a checked column (7a′), so the declines are argued, not asserted.

Repaired — 16 edits, every one traceable to a numbered finding

D1's candidate, named as the finding asks: the elided verb is almost certainly a healing verb from Sap 16:7 (sanabatur) — Rabanus' point being that it was not the bronze serpent that healed but the one it prefigured, which the surviving tail confirms (quia mortuus per humanitatem, quasi aeneus est per Divinitatem). It stays a conjecture and stays here.

Declined, and what the reader got right

• ⭐ The reader was right about far more than it was wrong about, and two of its sharpest arguments were refuted by the plate, not by us — see §5. Recording that is the point of a blind read: the argument for Homo at 1171D was excellent and the answer is still no.
• ❌ 1168A/1168B convenientem as a transcription error — declined on the plate. The corpus-frequency argument was good (convenientem habet once corpus-wide against nine convenientiam) and the plate refutes it: it prints conve-|nientem, hyphenated, exactly as our file. Not a transcription question.
• ❌ A [cj:] at 1168B for the same word — declined, reasoned. With the plate settled, the question is Pattern 18: does the faithful English assert something false? It does not. The printed accusative has no head noun, which makes the clause elliptical, not false, and Pattern 18 explicitly declines "a merely awkward or elliptical sentence"; 18a further scopes the marker to negations and correlatives and leaves anomalies that merely fail to construe declined. Convenientem and the conjectured convenientiam mean nearly the same thing, so a gloss would tell the reader nothing he does not already have. What was genuinely wrong was Pattern 7, not Pattern 18: "an agreement" rendered a word not on the page. Repaired by rendering the printed accusative, with the participle carried by an English indefinite — the same licence this work already used at 1167C (in malevolam → "into a malicious soul", where English cannot carry a bare adjective). The reading is now Migne's and visible.
• ❌ 1183A's "it" — declined as an edit, granted as a disclosure. No candidate can be proved; all three are now named above (E4).
• ⛔ The three un-italicized [sic:] markers were re-checked and NOT touched. All three sit inside an italic scripture lemma where Pattern 10 forbids nested italics; the six italicized ones are in plain prose. 9/9 consistent, and it is not an inconsistency.
• ⛔ Haec venditum justum, → "She forsook not the just man that was sold" stands. The reader examined the supplied negation and let it stand; so do we. It is convention 3 expanding a lemma from its own verse (non dereliquit), and it is this work's uniform practice. Recorded as examined, not missed.
• ⛔ 1176C's affirmative potuerunt cognoscere bonum artificem where Sap 13:1 has non potuerunt: the English correctly did not import the Vulgate's negative. Checked negative; left.
• ⛔ Chunk 0002 (1173A–1175B) read clean end to end and was not edited, including flagellum collectionis frangit corda pravorum quae praedicatio non emollit, the Qui eum mercede sua… privare voluit naming Laban, and non tam hi… sed et omnes Chananaei scoped on tam hi.
• ⛔ All 9 [sic:] pass the 7a⁗ swallow test — every stripped clause keeps its verb, negation, conjunction and object. Zero 7a⁗ defects. The [cj:] at 1177A is aimed at the right word (invisibilibus), in the correct form, verbatim in the twin. All 65 anchors sit beside the word they stand for. Diligite justitiam,, Justorum autem animae., Proditio, and the Sap 8:1 usque ad trap all survived the conformation hunt and are untouched, as are all five ordinary usque ads.

§5 · THE PLATE

raw/scans/pl113/patrologiaecurs04migngoog.pdf, PDF page = (column + 11) / 2, confirmed against the printed corner numbers (p. 589 heads 1167 left / 1168 right).

One patch — and it is a NORMALIZATION, not a corruption

data/tei-patches/8969.json, col. 1168B: our file read id est Verbi Dei; the plate prints a bare id Verbi Dei, no est. Read at 1200 dpi on a tight crop and confirmed by eye. The contrast is on the same line twice, since id est ejusdem est scientiae cum Verbo follows immediately with both forms spelled out. Corpus Corporum supplied the copula, completing an ellipsis Migne left short — the same class as 8948 convivum, 8990 concilio, and 8992's leges→legis found in this batch. ⭐ The digitization does not only corrupt; it silently CORRECTS, and a corrected form reads perfectly, so nothing downstream can see it.

Rendering, and the marker question, reasoned. The English now reads "hath knowledge of the voice, that of the Word of God, that is, he is of the same knowledge with the Word." That is Pattern 7: the printed words and no more. The English keeps the ellipsis because English can keep it — the copula drops out of "that of the Word of God" exactly as est drops out of id Verbi Dei, and the reader meets the contrast with the spelled-out id est in the next clause just as Migne's reader does. Supplying "that is" would have been Pattern 8's forbidden supply performed on a word instead of a mark.

No marker, and this is the deliberate part. Pattern 12 marks type that is the plate's defect. An ellipsis Migne himself printed is not a defect — it is his text, and a nineteenth- century editor's compression of a formula he had already spelled out twice on the same line is about as far from broken type as a page gets. Pattern 12's own operational test settles it: the crux verb here is "rendered as printed", so the words on the page are English, and ours — no marker. Nor is it Pattern 18: an elliptical sentence is the pattern's named ❌ class, and nothing false is asserted. ⚑ What is worth recording instead is that our Latin was wrong and now is not — the patch, not a marker, is the honest instrument, because the defect was never Migne's.

⚑ Also observed on that line and deliberately not patched: the plate prints Verbo. Vel. with a full stop where our transcription has Vel,. A comma-for-period in the source transcription is a corpus-wide punctuation question, not a single-site claim, and sweeping it is out of scope. Recorded so the observation is not lost.

Two checked negatives — the plate said no, twice

⚑ The batch-wide pattern, and it is the most useful thing this read produced

Across three works this session:

• Every suspicion aimed at a NON-WORD was CONFIRMED by the plate as ours — interpetratur, quisbusdam, per verse, nortales, Qui dam, quorum dam.
• Every suspicion aimed at a REAL, WELL-FORMED WORD was REFUTED — credulitatem, Homo, convenientem, dat offensum: the plate had Migne printing it.

⭐ Corpus frequency is decisive for non-words and worthless for real words. A non-word has no defence: if the corpus has never seen it, the digitization made it. A real word has the best defence there is — Migne printed it — and its rarity is evidence about Latin usage, not about this plate. Weight future plate requests accordingly: non-words first, real words only where a marker actually depends on the answer.

Two reads OWED, not run — and why shipping is still honest

⛔ Neither fires a marker. Both are rendered as printed, as ordinary English, with no apparatus claiming anything about the plate. The principle, stated plainly: an unread candidate is safe to ship if and only if it fires no marker. A marker is an assertion about Migne's page and may not be made from an unread plate; a literal rendering asserts nothing beyond our Latin, which is disclosed and sits in the facing column. These two are owed, not checked, and this table is the record of the debt.