Liber Numeri

Glossa ordinaria on Numbers, Anselm of Laon and his school. PL 113 cols 0379B–0446D, 26 chunks, translated in six stints, 2026-09-06. Merged from six stint files. The concatenation is the easy half; what follows is the reconciliation, which is the required step.

Work-wide conventions — what was reconciled at the merge, and what was left alone

⛔ mansio — ONE real drift, and it split the book's own catalogue in two

Numbers' spine is the itinerary: Migne's CAPUT XXXIII enumerates the mansiones from Rameses to the Jordan, and the PRAEFATIO summarizes the same list. Ten chunks rendered mansio "station"; two rendered it "encampment" — chunk 0000 (the praefatio, 13 sites) and chunk 0013 (3 sites).

That is not a synonym choice, it is a broken cross-reference. The praefatio numbers the first, the second, the third, the fourth, the sixth, the seventh encampment; CAPUT XXXIII numbers the thirty-third, the thirty-fifth, the thirty-seventh station. A reader cannot see that these are one series. Swept to "station" at all 16 sites. castra is a separate word and keeps "camp"/"camps" (20 Latin forms, 23 English), so nothing collided.

⚑ One mansio is deliberately NOT "station" and must stay as it is: @0434A–B (chunk 0021; NOT 0433D — the stint report gave that column and it is wrong, which matters because the next reader will grep 0433D and find nothing) mansionis ei meritum deputabitur is the Johannine sense — a dwelling-place, not a stage of a journey — and is rendered "the merit of his dwelling." Different word, different sense; the sweep was checked against the Latin at every site rather than run blind.

✅ Checked for drift and found CONSISTENT — recorded so nobody re-sweeps them

• The exposition tags. A standalone tag (Mystice. @0407, Allegorice, Moraliter) is rendered capitalized and standalone; the same word with a connective particle (Mystice autem @0420, Tropologice vero @0446C) is a mid-sentence adverb and is rendered lowercase, run into the sentence. Every stint made that distinction the same way, unprompted. This is the ALLEGORICALLY./ALLEGORICAL. split of 11063 NOT happening, and the reason it did not is that the distinction is in the Latin, not in the translator's taste.
• ⚠ A raw count of the tags LOOKS like drift and is not: Aliter appears three times in the Latin and "Otherwise" twice in the English, because two of the three (Aliter enim non erit votum acceptabile, aliter sapientes, aliter insipientes) are ordinary adverbs in running prose and not section tags at all. Count the construction, never the word.
• usque ad — the abridgment formula, rendered as far as throughout. 67 occurrences were tested one by one with the etc. splice test in the 0014–0018 range alone, all 67 the formula. The one genuine look-alike is @0428C, where the formula and Mt 5:26's own usque ad novissimum quadrantem stand four words apart — and our TEI does not italicize the second while the plate does, so an italic test would have mis-sorted it too.
• Axis instruments were run before anything was reported: a raw natio grep in the 0019–0022 range returns one hit and it is indignationis; a raw gentes grep returns one and it is legentes. turba absent; sacramentum a checked zero in 0014–0018. daemonium → demon throughout, including Ps 95:5 where the Douay pulls toward "devils".

Apparatus totals as merged

60 markers: 55 [var:] · 2 [sic:] · 2 [ed:] · 1 [cj:]. Plate gate OK — every one stands on a column read at the plate on both printings. 4 supplied em-dashes in the whole work (0.2/chunk against a corpus median of 5.3).

• The single [cj:] @0407C is Pattern 18 the right way round, and was checked on disk rather than taken from the report: the running English carries Migne's word ("the light") and the bracket carries the conjecture, [cj: lux; read lex, "the law"].
• The two [ed:] @0427B and @0437C are Pattern 13a clause 3 — lost non-Latin type that is not recoverable, marking the hole and supplying nothing. The lost Hebrew at @0418B is clause 1 by contrast (plate certain, transliteration standing beside it): restored by TEI patch, no marker. Two clauses, two evidential situations — not a convention drift. ⛔ Neither witness resolves the letters at 0427B/0437C and no higher-resolution capture exists: the archive copy is 600 ppi bitonal JBIG2 and Gallica is ~300 ppi. See data/briefs/PL113-WITNESS-ENCODING.md.

⛔ How to read every "the plate reads X" claim in the sections below

The archive copy is 1-bit JBIG2 — a lossy SYMBOL-SUBSTITUTION codec, not a photograph of heavier ink. Its failure mode is replacing a character with a different, visually similar, well-formed one, so its errors construe. Across this work 105 archive-only candidates were raised and 99 refuted at Gallica; of the 6 that stood, not one was Migne's. Where a section below calls an archive/Gallica disagreement a "press variant", read it as the copies disagree and our TEI follows Gallica unless the evidence is of a class substitution cannot produce.

The book's own conventions, for a later reader

Lemmata are italic; this book has zero guillemets (1,172 spans). Every chapter head spells CAPUT in full. Verse addresses print as VERS. n.--, also as ranges and lists. Migne sets chapter numerals in lowercase small caps (Rom. x, Joan. iv) and the digitization uppercases them — declared on /method, not retrofitted, not a defect.

⭐ The book opens with a display line Corpus Corporum drops in all five Glossa Pentateuch books: Hebraice VAIEDABBER (וידבר), id est : ET LOCUTUS EST. Restored by TEI patch as <head type="book-title"> per Wilson's ruling of 2026-09-06.

Stint 0000–0004. Six plates read inline on both witnesses (archive pp. 196–201, Gallica f191–f195); the reads, including the checked zeroes, are in data/briefs/8963-PLATE-READS-0000.json, and the seven divergences judged to be our own file's are in data/briefs/8963-PENDING-TEI-PATCHES-0000.md.

0. The head of the book — a display line the plate carries and our source does not

At the head of chunk 0000, between the end of Leviticus at 0379A and the <pb n="0379B"/> mark, Migne prints a display line under the LIBER NUMERI. banner giving the book's Hebrew name:

LIBER NUMERI. Hebraice VAIEDABBER (וידבר), id est : ET LOCUTUS EST.

Corpus Corporum drops the whole line, so our TEI body opens straight at <head>PRAEFATIO.</head>. It is two-witness confirmed and written up in data/briefs/8963-PENDING-TEI-PATCHES-HEAD.md; the reading is settled and the encoding is an open ruling that reaches four sibling books, so the patch is not applied and was not applied here.

What this stint did about it: nothing, deliberately. The English of chunk 0000 opens at ## PREFACE. exactly as the Latin opens at ## PRAEFATIO., and no head was invented for the banner. A reader of our page therefore meets this book one line short of Migne's own plate, and that is a known, filed, deliberate state and not an oversight of this stint.

1. The span count — 192 against 192, and both numbers reported

data/briefs/8963-lemmata-0000.txt states 192 spans, of which 0 open-ended, for chunks 0000–0004.

Counted independently in my own Latin, paragraph-bounded (asterisks counted per line, so that an unmatched delimiter would show as an unbalanced paragraph rather than silently swallow a neighbour):

chunk 0000 21 italic spans 0 guillemets 0 unbalanced paragraphs chunk 0001 76 0 0 chunk 0002 14 0 0 chunk 0003 39 0 0 chunk 0004 42 0 0 ------------------------------------------------------------------ TOTAL 192 italic spans 0 guillemets 0 unbalanced paragraphs

Whole-file total: 192. Paragraph-bounded total: 192. They agree with each other and with the brief. The two numbers being equal is the check that matters — a single unbalanced * would have made them differ. This book confirms the launch brief's statement that it has zero guillemets: not one « in 4,682 words.

2. ⚑ MATCHES ONLY ELSEWHERE — all ten in this range, one shared cause

The lemma file flags ten spans in my range (the eleventh grep hit is its own header). They are VERS. 5–15 of Numbers 1, and they are all of one form:

VERS. 5 De tribu Ruben. VERS. 11 De tribu Benjamin. VERS. 6 De tribu Simeon. VERS. 12 De tribu Dan. VERS. 7 De tribu Juda. VERS. 13 De tribu Aser. VERS. 8 De tribu Issachar. VERS. 14 De tribu Gad. VERS. 9 De tribu Zabulon. VERS. 15 De tribu Nephthali.

Adjudicated, and it is one finding, not ten. The Clementine at Nm 1:5–15 reads de Ruben, Elisur filius Sedeur ; de Simeon, Salamiel filius Surisaddai ; de Juda, Nahasson… — i.e. de X with no tribu. Every one of the ten failed the phrase search for the same reason, and the addresses each one did match (1Par 6:62–80, Jos 21:28, Dt 4:43, Ex 31:2) are the Levitical-cities and craftsman lists, where de tribu X is the standard formula. The "elsewhere" is an artefact of a formula that occurs all over the Vulgate, not of a harmony answering for the wrong book.

No [var:] on any of the ten, and I want the reason on the record because ten identical declines look like one habit rather than ten judgments. Three reasons, all needed:

1. These are section rubrics, not quotations. The Glossa's VERS. n.-- De tribu X. is the running label under which the gloss on that verse sits. Migne prints them in a series; the gloss beneath each expounds the name, never the word tribu.
2. Nothing turns on it. de Ruben and de tribu Ruben assert the same thing, and Pattern 14's bar is a divergence "in a way that changes the sense".
3. It is uniform across all ten. A compositor's or a recension's convention repeated ten times in eleven verses is a house style, and marking each instance would put ten maroquin notes on one page to report a supplied noun.

Rendered uniformly as "Of the tribe of X." throughout.

3. 7a″ — the lemma collation, DIVERGENCES NAMED

3a. Divergences material enough to take a [var:] — six, all fired

3b. Divergences checked and DECLINED, with the reason

• 0379C quia omnia in figura contingebant illis — 1Cor 10:11 has Haec autem omnia in figura contingebant illis. Migne drops the connective only; the quoted clause is verbatim. No marker.
• 0383B Sicut exaltavit Moyses serpentem in deserto, sic exaltari oportet filium hominis — Jo 3:14 has Et sicut Moyses exaltavit serpentem in deserto, ita…. Word order and sic for ita; identical sense. No marker.
• 0384A cum plenitudo gentium intraverit, omnis Israel salvus fiet — Rom 11:25–26 has donec plenitudo gentium intraret … et sic omnis Israel salvus fieret. The recension has re-pointed a donec + subjunctive as a cum + future perfect so that the sentence can stand on its own. Tense follows syntax, not sense. No marker.
• 0384C Jam non dicam vos servos, sed amieos — Jo 15:15 has Jam non dicam vos servos … Vos autem dixi amicos. Migne compresses two clauses into one, as this recension does constantly. The compression is not a variant reading. (For amieos itself see §5.) No marker.
• 0386C Nemo enim mundus a sorde, nec si unius diei fuerit vita ejus [n: (Job XV)] — not in the Clementine at all; it is the Old-Latin form of Job 14:4–5. Pattern 14 requires the received reading to be named, and here there is no single clean comparand to name. Declined, and I note that Migne's own locator (Job XV) is off by a chapter.
• 0386C qui peccatum non fecit, nec inventus est dolus in ore ejus [n: (Isa. LIII)] — the words are 1Pt 2:22 verbatim, and Migne's locator points at Isaiah 53:9, which is Peter's own source. His locator, his business; the quotation itself agrees exactly. No marker.
• 0386C In medio populi mei ego habito — 4Rg 4:13 has In medio populi mei habito. An added pronoun, no case change, no sense change. No marker.
• 0387B Ego vos elegi de mundo — Jo 15:19 has sed ego elegi vos de mundo. Word order. No marker.
• 0387C Oportet episcopum irreprehensibilem esse — 1Tim 3:2 has Oportet ergo episcopum. A dropped connective at the head of a clipped citation. No marker.
• 0387D Mihi autem absit gloriari nisi in cruce Domini nostri Jesus Christi — Gal 6:14 verbatim but for the nominative Jesus for Jesu, which is a press variant (§5), not a divergence from the Vulgate.
• 0388B Cor ejus indurabitur quasi lapis — Job 41:15 has tamquam lapis. Synonyms. No marker. (Migne's locators here, Isa. XIV; Job XXI, are again off; not ours to correct.)
• 0391B Nihil enim occultum, quod non reveletur — Mt 10:26 has Nihil enim est opertum, quod non revelabitur. occultum is the word Lk 8:17 uses of the same saying, and the subjunctive follows the Nihil enim frame. Two synonyms and a mood; the assertion is identical. Declined on the sense bar, not because I did not see it — a later reader who thinks this one should have been marked will find the two readings here.

3c. Lemmata collated and found to AGREE with the Clementine exactly

Recorded because a list of findings alone cannot be told apart from a report by an agent that never looked. Each of these was read word by word against sources/vulgate/clementine-flat.txt:

0379C Miserere mei, Deus Ps 50:3 0380D Benedicetis filiis Israel et dicetis Nm 6:23 0383A Ad te levavi oculos meos, qui habitas in coelis Ps 122:1 0383A Castigo corpus meum et in servitutem redigo 1Cor 9:27 0383C Nathanael filius Suar Nm 1:8 0383D Amen dico vobis, non inveni tantam fidem… Mt 8:10 0383D Quae audivi ab eo haec loquor in mundo Jo 8:26 0384A reddidit mihi Deus Jdc 1:7 0384A filius doloris / 0384B filius dextrae Gn 35:18 0384B Sanguis ejus super nos et super filios nostros Mt 27:25 0384C Gallus succinctus lumbos et aries, nec est rex qui resistat ei Pr 30:31 0385A Quisquis externorum Nm 1:51 0385B Unusquisque per turmas Nm 1:52 0385C Per gyrum tabernaculi Nm 1:50, 1:53 0385C Universi qui in castris Judae Nm 2:9 0385C Meridianam plagam erit princeps Nm 2:10 0385D Levabitur autem tabernaculum testimonii Nm 2:17 0385D Ad occidentalem plagam erunt Nm 2:18 0386A Ad aquilonis partem castrametati sunt filii Dan Nm 2:25 0386A Sacerdotum qui uncti sunt Nm 3:3 0386A Applica tribum Nm 3:6 0386B Ego tuli Levitas Nm 3:12 0386D Alienati sunt peccatores a vulva, erraverunt… Ps 57:4 0387A Numera filios Levi Nm 3:15 0387B Levi per nomina sua / Gerson et Caath et Merari Nm 3:17 0387C Hi post tabernaculum metabuntur Nm 3:23 0388A Ad meridianam Nm 3:29 0388A Nisi Dominus aedificaverit domum… Ps 126:1 0388B Ad orientalem plagam Nm 3:38 0388C Quinque siclos Nm 3:47 0388D Tolle summam Nm 4:2 0389A Et operient Nm 4:6 0389C Tunc intrabunt filii Caath / 0389C Nolite perdere Nm 4:15, 4:18 0389D Alii nulla / A triginta Nm 4:20, 4:23 0390A Ad annum quinquagesimum / Omnem leprosum Nm 4:35, 5:2 0390C Reddent ipsum / Excepto ariete Nm 5:7, 5:8 0390D Tu es sacerdos in aeternum secundum ordinem Melchisedech Ps 109:4 0390D Nolite ante tempus judicare quoadusque veniat Dominus 1Cor 4:5 0390D Non fundet / 0390D Aquam sanctam Nm 5:15, 5:17 0391A Scribetque sacerdos in libello / Dabit ei bibere Nm 5:23, 5:24

3d. Lemmata Migne himself frames as ANOTHER reading — collated, and correctly unmarked

Six spans in this range are explicitly introduced as the Septuagint, an alia littera, or a father's own citation. A [var:] on any of them would report as a divergence from the Vulgate something Migne has already told the reader is not the Vulgate:

0382D Alias LXX: Omnis qui procedit in virtute Israel (against Nm 1:3) 0385A [AUG.] Alienigena qui accesserit, moriatur (Nm 1:51: quisquis externorum … occidetur) 0386B [AUG.] Alienigena qui tetigerit, morietur (Nm 3:10: externus qui ad ministrandum accesserit) 0386B [ORIG.] Ego assumpsi Levitas de medio filiorum Israel (Nm 3:12: Ego tuli Levitas a filiis Israel) 0389A [AUG.] Et panes qui semper super eam erunt (Nm 4:7: et panes semper in ea erunt) 0389B Alia littera: et super altare aureum adoperient… (Nm 4:11) 0389C [ORIG.] Nolite exterminare plebem Caath (Nm 4:18: Nolite perdere populum Caath)

⭐ The 0389A one is the case that proves the rule: the entire gloss under it is Augustine arguing about the wording (Ideo dixit, qui semper super eam erunt, id est, quia semper panes, non quia illi semper). Marking it as a divergence would annotate the very thing the paragraph exists to discuss.

4. The one [sic:], and why it is printed as two markers

0381B. After deinde praecepit Migne sets, in italic, an abbreviated run that does not construe and cannot be rendered:

deinde praecepit ne temporis re. p. t. Caute omnia provisa.

Two witnesses, identical: archive p.196 and Gallica f191 both print ne temporis re. p. t. with the same points and the same italic. So the defect is Migne's, the two-witness bar for accusing his type is met, and the marker is fired.

⚑ It is printed as [sic: ne] [0381B] [sic: temporis re. p. t.] — two markers for one defect — and that is deliberate. The column anchor falls inside the run in Migne, between ne and temporis, and the verifier checks a [sic:]'s content as a verbatim substring of the Latin twin with only asterisks and whitespace normalized; an anchor inside the marker would fail that check, and moving the anchor outside it would displace the citation-resolving unit. Splitting a marker mid-construction to match Migne is the 8995 @0605B precedent (non [0605B] recedentes).

⚑ 7a⁗ check, run explicitly. With both markers deleted the clause reads "then he commanded … All things were cautiously provided for." — (1) it keeps its verb, praecepit; (2) the only negation in the run is ne, the negative-purpose particle, and everything it governs is inside the corrupt run, so there is no translated text left for it to negate and nothing has been parked in the bracket; (3) there is no interrogative and no conjunction outside it; and (4) — the fourth question the brief says 7a⁗ does not ask — praecepit has lost its object, and that loss is Migne's, not the marker's: the object is precisely what his type has destroyed. The English therefore supplies nothing.

⭐ The one candidate on this page that was NOT fired, and the reason is the whole method: the archive copy prints Cauta omnia provisa four words later, against our twin's Caute. On one witness that is a clean, sound Latin word and would have made a confident [var:] or [sic:]. Gallica prints Caute and our twin is right. Same page, same line, one refuted and one confirmed.

5. Press variants — five, no marker, no patch, and one of them costs the reader

The two copies of PL 113 are two printings in this range, not two photographs. Five sites differ in a way ink cannot produce, and our twin follows the Gallica printing at every one:

col Gallica (= our TEI) archive copy note 0384C lumbos lumbo Pr 30:31 reads lumbos 0384C amieos amicos Jo 15:15 0384D dexteram dextram 0386A Applica tribum (LEMMA) Accipe tribum Nm 3:6 reads Applica tribum 0387D Jesus Christi Jesu Christi Gal 6:14

⛔ 0384C amieos is the one to argue about, so here is the argument. Our Latin prints a non-word inside a scripture lemma — Jam non dicam vos servos, sed amieos — and a [sic:] is the obvious move. It is refused. The archive copy prints a clean c; the two witnesses disagree; and 8950's measured bar is that a claim about Migne's type needs two witnesses agreeing. What is true is only that one printing of Migne carries a broken sort here, which is not a sentence our apparatus has a marker for. The English renders the sense, friends, per the 2026-09-05 amendment to Pattern 7 — except that with no marker standing there is nothing to disclose the defect to a reader, so this crux is the only disclosure, which is exactly why it is written out at this length.

⛔ 0386A is a whole WORD, not a letter, and it is the trap of the set. The archive's lemma reads Accipe tribum where Gallica, our twin and Nm 3:6 read Applica tribum. The archive's reading is sound Latin and locally plausible — Augustine's own citation four words later begins Accipe tribum Levi — so a single-witness read would have found a divergence, argued it well from the immediate context, and been wrong. Recorded here as the largest press variant yet seen on this volume.

⛔ 0387D Jesus Christi stands in the English as "our Lord Jesus Christ". The nominative where the genitive is wanted is Migne's (Gallica's printing, which our twin follows), and a Latin case ending has no English exponent — Pattern 9. There is no emendation to make and none to undo; the facing Latin column shows the form.

6. Refuted archive-only candidates — nine, our twin right at every one

Recorded as findings, because a page of refutations is what a second witness is for. Every one of these read as a defect on the archive copy alone and every one was overturned by a single Gallica fetch:

0381B Cauta → Caute (twin right) 0382B Licer → Liber (twin right) 0382B poctores → doctores (twin right) 0382C ctc. → etc. (twin right) 0385B Quid → Qui (twin right) 0385A de illis Israel → de illis filiis Israel (twin right; a whole word) 0388A ub Eleazaro, Sid est → Sub Eleazaro, id est (twin right) 0388C adjiciatur → abjiciatur (twin right) 0386A Accipe tribum → Applica tribum (press variant, not ink — see §5)

⚑ ub Eleazaro, Sid est at 0388A is the one worth remembering. A displaced capital S is exactly the shape of 8955 @0629D's ulit for Tulit, which is a [sic:] that STANDS. Same shape, opposite verdict, and only the second copy separates them.

⚑ Score for this range: 9 archive-only readings raised, 9 refuted, 0 confirmed. The archive copy of PL 113 is the worse witness here by nine to nothing.

7. Unexpressed subjects and supplied words — every one named

The dominant defect class, so each supplied pronoun or noun is listed with what licensed it.

• 0382D usque ad per fidem evangelicam et apostolicam traditur → "it is handed down". A resumption tail with no expressed subject. The nearest candidate in the printed words is qui procedit, but "he is handed down" is not what the tail can mean. "It" is mine and nothing in the printed words fixes it — flagged rather than chosen silently.
• 0383A Simeon generaliter omnes obedientes → "Simeon generally [signifies] all the obedient". The verb is elided by the abridgment; significat stands in the exactly parallel sentence eleven words above (Princeps tribus Ruben eos generaliter significat). Bracketed.
• 0383D Ephraim minor … gentilem populum praelatum → "[signifies] the Gentile people". Same elision, verb carried from Christum significat in the same sentence. Bracketed.
• 0384A cum Christus … in regno suo gentem obliviosam invenit — the subject is expressed (Christus); noted only because the long parenthesis makes it easy to lose.
• 0386D nec pecudis caput, scilicet, stultitiae sequaris imaginem → "nor follow the head of a beast, that is, the image of folly". Second person singular throughout this paragraph, as Et tu ergo fixes.
• 0389D Alii nulla, etc. → "Others with no curiosity, etc." ⭐ The negation here is the trap. nulla is feminine ablative and attaches to the elided curiositate of Nm 4:20 (Alii nulla curiositate videant quae sunt in sanctuario), not to alii. Rendering it "Let no others…" would have moved the negative onto the quantifier and made the verse forbid a class of persons rather than a manner of looking — Test 2a's exact failure. The noun is supplied per convention 3 (a fragment is rendered as the phrase it stands for).
• 0390D ne forte judicet rem incertam → "lest perchance he judge an uncertain matter". The subject is the one reserving all things to Christ's judgment, i.e. the minister of the preceding clause. Named here because the sentence has three possible antecedents.
• 0390D damnanda aut liberanda reservatur → "she is reserved either to be condemned or to be set free". The gerundives are feminine, and the subject is the suspected wife of the Nm 5 ordeal, who is the referent of the whole chapter. The morphology decides it, not the context.
• 0382B ut ducibus … perveniamus → "that with [such] leaders we may come". A truncated Isidore tail; the demonstrative is supplied and bracketed because English cannot leave a bare instrumental ablative standing.

8. usque ad — the etc. splice test, run per occurrence

usque ad stands 61 times in this range. 56 are the abridgment formula — every one of them an italic usque ad standing immediately after an etc., and naming where the recension resumes — and all 56 are rendered "as far as" (56 in the Latin, 56 in the English). The remaining 5 are ordinary Latin and are rendered by sense:

• 0379D, 0380B mansiones a Ramesse usque ad Jordanem — spatial, "as far as the Jordan". Unitalicised, and no etc. precedes it.
• 0382A omnium mansionum ab Aegypto usque ad terram promissionis — spatial, "as far as the land of promise".
• 0383D Factus est obediens Patri usque ad mortem — ⭐ inside a scripture lemma (Php 2:8), rendered "even unto death". This is precisely addendum D's trap: the words are the verse's own, they sit inside an italic span, and no italic test could separate them from the formula. The etc. test does: no etc. precedes it, and what follows is not a resumption point but the rest of the verse.
• 0386A multitudinem gentium usque ad Christum frigore infidelitatis torpentem — temporal, "until Christ". Unitalicised, mid-clause.

⚑ Two further look-alikes checked and correctly left alone: 0390D quoadusque veniat Dominus (1Cor 4:5, a different word) and 0388D ad domum summi sacerdotis (no usque).

9. Vocabulary axes — the calls made, including the ones the axes did not rule on

• gentes family. gentilium spurcitia (0379D), gentilem populum (0383D, 0384B), plenitudo gentium (0384A), multitudinem gentium (0386A) → Gentiles in every case: all five sit in an explicit Judaei / gentes opposition (vel crudelitate Judaica; gentilem populum Judaico praelatum; omnis Israel salvus fiet; repulsis Judaeis). doctores gentium (0382C), Isidore's mansion tail, also → Gentiles, the Pauline title.
• ⚑ gentem obliviosam (0384A) → "a forgetful NATION." Singular gens of the Jewish people, which axis 4 rules is nation, not Gentile — and here the axis is doing real work, because the sentence's other people-word (populum Judaicum) is four words away and the two must not collapse. natio itself does not occur in this range.
• multitudo → multitude (0386A). ⚑ turba does not occur in this range, so the turba/multitudo pair is not exercised; checked, not assumed. (The \bturba grep is the one the brief warns manufactures hits off turbare; it returned nothing here either way.)
• mysterium → mystery / sacramentum → sacrament, kept 1:1 and they stand in adjacent sentences at 0382A (multiplici mysteriorum obscuritate … secundum convenientiam sacramentorum), which is where a single English word would have shown.
• diabolus → devil (0383B destruxit diabolum, 0384B diabolus inique tentavit). daemonium does not occur.
• The commentator's voice is not archaised: ait and dicitur → "says", "it is said" (0383D, 0384B, 0391B). -eth forms appear only inside italic scripture (who dwellest, there standeth one, thou art a priest, there is no king that resisteth him).
• Thou / you (Pattern 17). Singular tu/tibi → thou at 0381B–0381C (Vide ergo quanta tibi transeunda), 0386D (Et tu ergo si), 0387A (invenies), 0390D (Tu es sacerdos). Plural vos/vobis → you at 0380D (Benedicetis), 0383D (Amen dico vobis), 0384C (non dicam vos servos), 0386C (quem vos nescitis), 0387B (Ego vos elegi), 0391B (locutus sum vobis, ille vos judicabit). The morphology decides each one.

10. Migne's own printing, left alone

• 0382A Numerorum libri multiplici mysteriorum obscuritate involutum Patres ostendunt. Both plates print the genitive libri where the syntax of involutum … ostendunt wants an accusative librum. Rendered "The Fathers show the book of Numbers wrapped in a manifold obscurity of mysteries" — a Latin case ending has no English exponent (Pattern 9), so there is nothing to emend and nothing to mark. Recorded so a later sweep does not churn it.
• 0386A Quatuor fili Aaron — fili for filii, on both plates. The -ii > -i contraction is standard medieval orthography and standard in nineteenth-century printing of it: Pattern 9, attested spelling, no marker, no italics.
• 0389B Nec non et altare — rendered "And likewise the altar". ⚑ Flagged explicitly because the negation rule is strict and this looks like two dropped negatives: nec non is a fixed affirmative idiom in Vulgate and medieval Latin ("and moreover"), and Nm 4:11 itself reads Nec non et altare aureum involvent hyacinthino vestimento. Rendering it with two English negatives would be the mistranslation, not the fidelity.
• 0390C VERS. 9. Ad sacerdotem. — Migne prints this one address with a bare full stop where every other in the range prints .--. Reproduced exactly as printed; not normalized.
• 0388D qui est oboli 20 — Migne sets an arabic numeral in a text that elsewhere uses Roman. Reproduced as "which is 20 oboli".
• Chapter heads. CAPUT PRIMUM. → CHAPTER ONE. (spelled, as Migne spells it); CAPUT II., CAPUT III., CAPUT IV., CAPUT V. → CHAPTER II./III./IV./V. (numerals, as Migne sets them). The numerals mirror the plate, per the measured corpus rule.

11. Sections read clean

• Punctuation, band by band. Question marks: 0 / 0 / 1 / 3 / 0 across chunks 0000–0004, matching the Latin exactly. The one in 0002 (quanto magis reliqui homines propria signa … gerunt?) and the three in 0003 (de quo ergo medio assumuntur Levitae?; Quomodo enim quis errare potuit …?; Aut quomodo potuit falsa loqui …?) are all direct questions, so Pattern 8b's colon recast was not needed anywhere. ⚑ Two indirect questions in this range (Merito quaeritur quare 0384D, Quaeritur quomodo in Exodo legitur 0390C) carry no mark on the plate and none was supplied.
• Column anchors: 13 / 11 / 3 / 11 / 10, verbatim, in order, and each checked against the word it stands beside in the Latin rather than merely for presence (test 2b). Two are split mid-construction to match Migne: [0381B] inside the corrupt run at §4, and [0389C] inside ut ipsum [0389C] vestimentum hyacinthinum.
• Notes: 24 / 30 / 6 / 19 / 23, verbatim and untranslated. ⚑ Chunk 0001's frontmatter declares noteCount: 29 against 30 [n:] in its own body; the frontmatter is copied verbatim as required, and the discrepancy is the chunker's, not the English's.
• No [ed:], no [d:], no [nt:], no [cj:], no [cn:] in this range. Nothing in the digitization is missing here (the seven defects are all substitutions, and all of them are filed as patches); Migne sets nothing twice; every [n:] in the range is a bare citation or siglum, none is editorial prose; and no printed word is a sound Latin form whose faithful English misleads.
• Em-dashes: zero supplied in all five chunks.
• Paragraphing follows the Latin twin line for line in all five chunks.

12. Candidates raised and rejected, with reasons

• 0389B Posset ista locutio videri, quam velut absurdam … looked at first like a sentence with its predicate missing — and it is not. videri takes its complement from the tanquam debuerit dici clause five lines down; the long relative and the quoted alia littera stand between them. Rendered without supplying a word. ⚑ Raised here because the first reading of it produced an English "might seem so" with an invented "so", which was withdrawn.
• 0390C reconciliat (Job XV)** — the italic parenthetical citation sits in the middle of the clause and Migne's locator is plainly wrong for the sentence it annotates. It is his, it is a citation, and it rides as printed inside its italic span.
• 0382C Sexta est mansio — the archive's poctores gentium argued briefly for a defect in Isidore's list; refuted at Gallica (§6). The reason it is recorded is that a corpus-frequency argument was available and would have pointed the right way here for the wrong reason, which is the 9002 nomo lesson.
• 0384D VERS. 39 — a verse-address change is the class most recently withdrawn as a scanning artefact, so it was nearly declined on principle. It is made because it has an external control the withdrawn class never had (Numbers 1 has 54 verses; there is no v. 59; Nm 1:39 is Dan's 62,700, and the lemma is Septingenti). The reasoning is set out in data/briefs/8963-PENDING-TEI-PATCHES-0000.md.

Anselm of Laon and his school, Glossa ordinaria on Numbers, PL 113. Chunks 0005–0009 cover CAPUT VI–XIV, 4,979 Latin words, 246 marked spans.

0. What was checked, and how

• Span count. My lemma brief (data/briefs/8963-lemmata-0005.txt) states 246 spans for chunks 0005–0009. Counted paragraph-bounded in my own Latin: 0005 = 49, 0006 = 36, 0007 = 54, 0008 = 53, 0009 = 54, total 246, and 0 guillemets (this book has none). Every paragraph's asterisk count is even, so no unmatched delimiter is masking a mis-pairing in either direction. The totals agree.
• ⚑ MATCHES ONLY ELSEWHERE: none in this range. The work carries 13; none falls in chunks 0005–0009.
• Plates. All seven pages of the range read at the plate, on two witnesses, corner numbers read first on every one: archive.org patrologiaecurs04migngoog pp. 201, 202, 203, 204, 205, 206, 207 (= cols 391–404, map confirmed at p.201) and Gallica bpt6k5505319w ff. 196–202 (map confirmed at f196). No foot-of-page apparatus on any page — seven checked zeroes. Full ledger: data/briefs/8963-PLATE-READS-0005.json.
• Markers fired: 15 [var:], and nothing else. Zero [sic:], zero [cj:], zero [ed:], zero [d:] — see §3 for why each candidate was declined.

1. ⭐ The finding that governed every other call: our TEI's exemplar is the GALLICA printing

Thirteen sites in cols 0391–0404 divide the two copies. In twelve of them our TEI sides with Gallica against the archive copy — including in four places where Gallica carries a non-word and the archive copy prints the sound Latin. The thirteenth (0401D) runs the other way, which is what makes this a fact about two printings rather than about one bad photograph.

What follows, and it decided four calls below: a non-word our TEI shares with Gallica is Migne's own type in the exemplar Corpus Corporum transcribed. It is not our corruption (so it takes no TEI patch) and it is not on both printings (so it may take no [sic:], which would accuse a plate that in one copy is clean). Each is rendered by sense and logged here, with both witnesses named. Addendum A's "a non-word is our corruption" heuristic would have produced four false findings in this range.

⚑ The single most dangerous archive-only reading was 0396A non erudit for nos erudit — over-inking that manufactures a negation in the sentence "or because through both Testaments with a harmonious voice he instructs us unto the faith of the truth." Read at 1800 dpi on Gallica f198: nos. Had it stood, the English would have said the two Testaments do not instruct us.

2. The [var:] markers — 15, all against the Clementine, all on read columns

Fired where Migne's printed words differ inside the quotation. A quotation merely clipped (stopping short, or opening mid-verse) takes no marker — otherwise nearly every lemma in an abridgment would carry one. That line is stated here so a later reader can see it was a rule and not a mood.

1. 0391B Mt 10:22 — Qui autem perseveraverit, salvus erit · Vulg. qui autem perseveraverit usque in finem, hic salvus erit. An internal omission, and of exactly the words the gloss has just supplied for itself (Sed innocentiam teneat usque ad finem vitae).
2. 0391C 2 Tim 2 — Non enim qui inchoat, sed qui consummat, coronabitur · Vulg. 2 Tim 2:5 Nam et qui certat in agone, non coronatur nisi legitime certaverit. No word in common but coronat-; printed as scripture with a chapter citation, so a reader will look for it.
3. 0391C Eph 4:30 — Nolite contristare Spiritum sanctum, in quo signati estis omnes · Vulg. Spiritum Sanctum Dei … in quo signati estis in diem redemptionis. Both witnesses print omnes.
4. 0392A Ez 18:24 — Cum conversus fuerit justus a justitia sua et fecerit iniquitatem, omnes justitiae ejus oblivioni tradentur · Vulg. Si autem averterit se justus … Omnes justitiae ejus quas fecerat, non recordabuntur.
5. 0392B Jo 1:29 — peccata mundi · Vulg. peccatum mundi. Plural on both printings.
6. 0393A Phil 2:13 — Ipse enim est qui operatur in nobis et velle et perficere · Vulg. Deus est enim, qui operatur in vobis et velle, et perficere pro bona voluntate. A pronoun change, which is the class this project treats as load-bearing.
7. 0393D–0394A Jo 10:18 — Nemo tollit animam meam a me, sed ego pono eam, et iterum sumo eam · Vulg. Nemo tollit eam a me: sed ego pono eam a meipso, et potestatem habeo ponendi eam, et potestatem habeo iterum sumendi eam. Migne supplies animam meam out of v. 17 and drops both potestatem habeo clauses.
8. 0394A 1 Cor 9:11 — Si enim seminant spiritualia, non est magnum si metunt temporalia · Vulg. Si nos vobis spiritualia seminavimus, magnum est si nos carnalia vestra metamus? Migne turns an affirmative rhetorical question into a negated statement and changes carnalia to temporalia — and the gloss is arguing precisely that temporal support is no great return. Rendered as Migne prints it, negation and all.
9. 0396B Is 11:2 — super quem requievit spiritus sapientiae et intellectus · Vulg. Et requiescet super eum spiritus Domini: spiritus sapientiae… Future to perfect, and spiritus Domini dropped; the tense is the whole christological point of the gloss.
10. 0396C Nm 8:6 — lemma Tolle Levitas de me · Vulg. Tolle Levitas de medio filiorum Israel. ⚑ Both printings read de me., with a stop. It is very likely the abridger's clipping of de me[dio], but de me is complete Latin and I will not conform a lemma to a guess: rendered "Take the Levites from me," with the Vulgate in the marker.
11. 0396D–0397A 1 Pet 3:18 — ut nos offerret Deo Mortificatos quidem carne, vivificatos autem spiritu · Vulg. mortificatus … vivificatus, of Christ. Accusative plural for nominative singular moves the dying and quickening from Christ to us. Rendered as printed.
12. 0397A 1 Cor 6:11 — Sed abluti estis, sed sanctificati estis in nomine Domini Jesu · Vulg. adds a third member, sed justificati estis, and reads Domini nostri Jesu Christi. The omission is load-bearing: the gloss's own sentence argues that the one passion obtained two things, expiationem … et sanctificationem.
13. 0398A Mc 13:37 — Quod uni dico, omnibus dico · Vulg. Quod autem vobis dico, omnibus dico: Vigilate. The one/all contrast is the point of the abridged sentence.
14. 0399C Mt 10:5–6 — In viam gentium ne abieritis, sed ite ad oves… · Vulg. inserts et in civitates Samaritanorum ne intraveritis and reads sed potius ite.
15. 0401B Jer 8:4 — Nunquid qui cadit non adjiciat ut resurgat? · Vulg. Numquid qui cadit non resurget? An Old-Latin-flavoured form; the double negative is carried into the English ("Shall he that falleth not add that he should rise again?").

2a. Lemmata and quotations collated and found to AGREE with the Clementine

Recorded because a list of findings alone cannot be told from a list by someone who never looked. Checked against sources/vulgate/clementine-flat.txt, verse by verse: Nm 6:3, 6:5, 6:6, 6:7 (×2), 6:9 (×2), 6:11, 6:12 (×3), 6:13, 6:14, 6:15 (×2), 6:16, 6:17, 6:19 (×2), 6:24, 7:2, 7:3 (×2), 7:10 (×2), 7:11, 7:13, 7:14, 7:84/86, 7:89, 8:2, 8:7, 8:9, 8:10, 8:12, 8:24, 9:6, 9:8, 9:10, 9:12, 9:14, 9:23, 10:2, 10:4, 10:8, 10:33, 10:34, 10:35, 11:1, 11:5, 11:16, 11:17 (×2), 11:21, 11:24, 11:25, 11:29, 12:1, 12:10 (×2), 12:14, 12:15, 13:3, 13:24 (×2), 13:26, 13:27, 14:12 (×2), 14:34; and Ps 35:9, Ps 126:1, Rom 8:26, Jo 1:14, Jo 15:5, Gal 4:19, Mal 1:10, 2 Tim 4:2, Apc 21:5, 2 Cor 6:2 — all verbatim.

2b. Divergences collated and deliberately given NO marker, with the reason

• 0392A 1 Jo 2:1 Advocatus noster apud Patrem · Vulg. advocatum habemus apud Patrem. A recasting into a nominal phrase; nothing is asserted that the Vulgate does not assert.
• 0393C Ps 43:24 Quare avertis faciem tuam, et oblivisceris inopiae nostrae? · Vulg. Quare faciem tuam avertis? oblivisceris inopiae nostrae et tribulationis nostrae? Word order, one sentence for two, and a terminal clipping.
• 0393C Ps 118 Respice in me et miserere mei. ⚑ Worth a reader's eye: those exact words are Ps 24:16 (and Ps 85:16); at Migne's own address, Ps 118:132, the Clementine reads Aspice in me. Migne's citation rides as printed — citations are refDisplay, not ours to resolve in the text.
• 0393C Ps 79:4 Ostende faciem tuam, et salvi erimus, cited Psal. LXXVII. Verbatim from Ps 79:4, clipped of its opening Deus, converte nos.
• 0394B 1 Tim 5:18 / Dt 25:4 Non alligabis os bovis triturantis · Vulg. 1 Tim 5:18 os bovi trituranti (dative). Case only.
• 0395A Act 4:32 Quia multitudinis credentium erat cor unum · Vulg. Multitudinis autem credentium. A connective adapted to the host sentence, which is quotation grammar, not a divergence.
• 0396A Gal 3:19 Lex ordinata per angelos in manu mediatoris. Verbatim from the verse's second half; Lex is lifted from its own opening Quid igitur lex?
• 0399C Mt 28:19 Euntes docete omnes gentes, cited (Matth. XVIII). The wording drops only ergo. The citation number is Migne's own error (the verse is Mt 28:19) and prints xviii on both copies. Left exactly as printed.
• 0400A 2 Tim 4:2 Praedica verbum, insta opportune importune, cited (I Tim. IV). Verbatim; the book number is Migne's, on both copies.
• 0401A/B Phil 3:20 Nostra autem conversatio in coelis est, cited (II Cor. X). Verbatim; the citation is Migne's, on both copies.
• 0400D, 0402A — three quotations Migne himself labels Alia littera (Et requievit super eos spiritus…, Et respexit Aaron ad Mariam…, Et proclamavit Moyses…). A reading the plate declares to be another reading is not a divergence; it is the apparatus doing its job.
• 0402B Is 40:15 Omnes gentes sicut gutta situlae, et sicut sputum reputatae sunt, and Migne prints its own witness — (Isa. XL, secundum LXX). Same reason. The Clementine has quasi stilla situlae … quasi momentum staterae.
• 0391C Nm 6:8 lemma Omnes dies separationis · Vulg. Omnibus diebus separationis suae. Case and number of the same words; no assertion moves.
• 0401A Nm 11:26 lemma Remanserant (plural) · Vulg. Remanserat (singular, with a plural subject — the Vulgate's own oddity). No English exponent for the difference either way.
• 0403C Nm 14:9 lemma quia sicut panem ita possumus · Vulg. …ita eos possumus devorare. A terminal clipping; the lemma simply stops.
• 0403A Nm 13:18 lemma Ascendite ad meridianam plagam · Vulg. Ascendite per.
• 0402A Nm 12:13 lemma runs on into v. 14's opening (Cui respondit Dominus). Migne's span, not a divergence.

3. Candidates raised at the plate and DECLINED, with the reason for each

Every one of these is a non-word or a broken construction that a corpus-frequency argument would have convicted. All were carried to the second witness first.

1. 0397C si agenda for sit agenda (ut sive poenitentia si agenda, sive studia virtutum exercenda, efficaciter perficiant). Gallica f199 and our TEI print si; the archive copy prints sit, which construes. Divided witness → no [sic:] (it would accuse a plate that is clean in one copy) and no TEI patch (our file agrees with a printing). Rendered by the sense the archive's sit gives: "whether penance is to be done, or the pursuits of the virtues to be exercised."
2. 0400A exnortantis for exhortantis (inde se verbis exnortantis inclinaret). Gallica f200 + TEI exnortantis, archive exhortantis. Same treatment; rendered "by the words of one exhorting him."
3. 0403C commmatio for comminatio (Fit commmatio haec a Domino). Gallica f202 + TEI carry the three-m form; the archive prints comminatio, and Comminatio prints correctly six lines above on both copies — which is exactly the internal corroboration that has been measured to be worthless about whose an error is. Same treatment; rendered "This threat is made by the Lord."
4. 0403D vermanseris for permanseris (usque ad si vermanseris in bonitate). Gallica f202 + TEI vermanseris (a clean italic v, no descender); archive permanseris. Same treatment; rendered "if thou shalt remain in goodness." (Pattern 17: singular addressee, so thou.)
5. 0401A praecepit / praecipit for praecepti — Non negligentia praecepit Domini, sed devotione humilitatis indignos se judicantes perceptione tanti muneris. Both printings set a finite verb (Gallica + TEI praecepit, archive praecipit), and neither construes: the participle judicantes is left with no main verb and Domini with no head noun. The sense plainly wants the genitive praecepti Domini — "not from negligence of the Lord's precept, but from the devotion of humility" — and the non … sed devotione correlative fixes the negation on the ablative of cause, not on the verb, so the English carries it there. [cj:] declined on two grounds: 18a scopes out anomalies that merely fail to construe, and printing our conjecture in the running text would invert the marker (the pattern-18 failure measured on 8967). Rendered literally, with the stranded verb visible: "Not by negligence did the Lord command, but by the devotion of humility…"
6. 0401C non solum Judaei haeretici — Mystice, non solum Judaei haeretici, qui non recipiunt legem et prophetas, detrahunt Moysi. The non solum has no answering sed etiam on either copy. Rendered literally ("Mystically, not only Jews, heretics, who do not receive the law and the prophets, detract from Moses"), which keeps the negation on solum where the correlative puts it and leaves the missing member visible. [cj:] declined: 11535's admitted class is a correlative whose answering member is printed and whose own is not, and this is the inverse.
7. 0394B digne / digni operantur and 0394A seminant / seminent — press variants, both construing; our TEI = Gallica. No marker, no patch, per the standing rule that a press variant is not a defect.
8. 0396 head CAPUT VIII (Gallica, TEI) / CAPUT VIII. (archive) — press variant. The English head mirrors the Latin twin and prints CHAPTER VIII with no period, against CHAPTER IX. etc. That asymmetry is Migne's, not ours.
9. 0399A VERS. 15. (Gallica, TEI) / VERS. 13. (archive) — a verse digit, which is the one thing over-inking alters most readily and the one thing with no internal control. Nm 9:15 is Nubes operuit tabernaculum, so the TEI is right; left alone regardless, per the withdrawn verse-address rule.
10. 0398B/0404 various commas and periods differing between the copies — recorded in the plate ledger, not adjudicated. Migne's punctuation is his editors' and the two printings differ freely.

4. Migne's own printing, reproduced and not tidied

• 0402A: a VERS. with no number. Migne prints VERS.-- Clamavitque Moyses ad Dominum dicens… on both copies, between VERS. 10 and VERS. 14. Reproduced as printed.
• 0403A/B: VERS. 24. twice in succession, the second glossing De malis quoque granatis (Nm 13:24 is a long verse). On both copies. Order and numbering reproduced.
• 0397D–0398A: the chapter's verse addresses run 12, then 6, 8, 10, 11, 12, 13, 14, 15, 23. CAPUT IX opens on VERS. 12 and then goes back to VERS. 6. That is Migne's ordering; not reordered.
• 0402D: usque ad usque ad tricesimam secundam — the abridgment formula immediately followed by an ordinary temporal usque ad. Both are on both copies. Rendered "as far as these things are contained up to the thirty-second."

5. usque ad — the etc. splice test, applied per occurrence

The formula (etc., usque ad …) occurs 70 times in these five chunks and is rendered "as far as" throughout. Seven occurrences are ordinary Latin and are rendered by sense; they are named here because a silent departure from the axis is indistinguishable from drift:

• 0391B Sed innocentiam teneat usque ad finem vitae → "unto the end of life".
• 0391D et usque ad consensum descenderit → "and shall have gone down as far as consent" (spatial-metaphorical, not the splice).
• 0397B et usque ad finem perseveraverint → "and have persevered unto the end".
• 0398A testes esse usque ad ultimum terrae (Act 1:8) → "unto the uttermost part of the earth" — inside a resumption tail, but the words are the verse's own.
• 0402D usque ad finem septimanae mundi, id est usque ad resurrectionem → "until the end of the week of the world, that is, until the resurrection" (two occurrences).
• 0402D the second half of usque ad usque ad tricesimam secundam → "up to the thirty-second".

6. Unexpressed subjects supplied, and what decided each

• 0392D praesentans … perducit ("he brings him to the entrance of the heavenly life"). The subject is the priest, from the lemma's own verse: Nm 6:16 Quae offeret sacerdos coram Domino. Both readings construe grammatically; only the verse decides.
• 0392A Et deprecabitur … faciet nos deprecari — the subject through the paragraph is Christ, named two sentences earlier (offerat vero sacerdoti, id est, Christo), and confirmed by the two proof-texts (1 Jo 2:1, Rom 8:26).
• 0393A facit summus sacerdos sanctum suum offerre — subject printed; no supply needed.
• 0401A judicantes … judicavit — the subject of judicantes is the two men who remained in the camp (Eldad and Medad, Nm 11:26), from the lemma; the subject of judicavit is printed (judex cordium Deus).
• 0402D transiit, et venit ad eum quem Verbum carnem factum ante non credidit — the subject is the people of the preceding clause (erga illum populum), and eum quem … non credidit is Christ. Rendered "then he passed over, and came to him who had not before believed the Word made flesh," which keeps the printed order and the printed negation.
• 0396C quomodo rite provehatur ad dignitatem sacerdotii — impersonal passive; rendered "how one is duly advanced", not "how he is advanced", because no antecedent is printed.

7. Vocabulary axes as applied here

• gentes. Gentiles in the typological opposition — 0399C Mt 10:5 (In viam gentium), 0399D (post de cunctis orbis nationibus gentes ad fidem convocari, where nationibus = nations and gentes = Gentiles in one sentence), 0402B Rom 11:25 (plenitudo gentium), 0398C Gentiles = "the Gentiles". Nations where generic — 0399C Mt 28:19 (omnes gentes, the axis's own named case), 0402B Is 40:15 (Omnes gentes sicut gutta situlae) and the caeterae gentes four words later, which the quotation governs.
• multitudo → multitude (0395A multitudinis credentium; 0401A Coturnicum multitudinem). turba → crowd (0398B relictis turbis). Both occur; neither borrows the other's word.
• daemonibus → demons (0398C). No diabolus/satanas in this range.
• mysterium → mystery (0394D legis mysteria; 0397A talibus mysteriis; 0399A ecclesiasticis mysteriis) against sacramentum → sacrament (0397A Sacramentum passionis; 0399A Christi quoque sacramentum). 1:1 throughout.
• natio → nation (0399D nationibus; 0401C locorum et gentium nomina, rendered "of places and of nations").
• Pattern 17. Three second-person singulars, all rendered thou: 0393C Benedictionem quam percepisti ("which thou hast received"); 0402B Si enim consideres… ("if thou shouldst consider"); 0403D si vermanseris in bonitate ("if thou shalt remain"). One plural, rendered you: 0402A Gal 4:19 donec formetur Christus in vobis; and Eph 4:30 in quo signati estis omnes likewise plural.

8. Mechanical checks run on the finished files

• Column anchors 1:1, verbatim, in order, and each checked against the word it stands beside in the Latin (test 2b), not merely for presence. 0005 = 10, 0006 = 10, 0007 = 12, 0008 = 8, 0009 = 11.
• [n: …] notes 1:1 and byte-identical to the Latin twin's, in order: 29 / 23 / 32 / 27 / 34. None was converted to [nt:]: every one is a citation or a siglum, none is editorial prose.
• Italic spans 1:1 with the Latin twin after marker-stripping: 49 / 36 / 54 / 53 / 54.
• Question marks at parity: 2 / 0 / 0 / 2 / 3.
• Colons and semicolons compared BAND BY BAND, not by total (test 3). Twelve mismatches were found this way and all twelve repaired. One deliberate excess remains and is declared here: 0400D carries one English colon more than the Latin, because Pattern 8b requires it — Quaeritur utrum hoc Moses diffidendo dixerit an quaerendo? is recast as "It is asked: did Moses say this out of distrust or out of inquiry?" so the printed ? survives in a sentence English can hold.
• Em-dashes: zero. Four were written and all four were replaced by the comma, colon or full stop the Latin prints at that point.
• Word ratio 1.51 / 1.49 / 1.51 / 1.50 / 1.38.
• Marker-strip read-back (test 1): every [var:] deleted, and each host sentence read aloud. No stranded article, preposition or doubled mark; no marker stands in place of a rendering, because [var:] never wraps Migne's words in the first place.
• Negation attachment (tests 2 and 2a): each printed non/nec/neque/nisi/nunquam/ nullus/nihil was traced to the single Latin word it attaches to and checked on that word in the English. Two sites are the reason the test exists here — 0401A (non negligentia … sed devotione, §3.5) and 0401C (non solum with no answering member, §3.6) — and both have a quantifier or correlative beside a grammatically plausible verb, which is the named trigger condition.

Chunks 0010, 0011, 0012, 0013. Heads: CAPUT XV · XVI · XVII · XVIII · XIX · XIX (cont.) · XX.

Span count. My lemma file (data/briefs/8963-lemmata-0010.txt) states 190 spans. Counted in my own Latin: 190, flat, and 190 again counted paragraph-bounded, with zero paragraphs carrying an odd number of *. This book has no guillemets and my four chunks contain none. The totals agree.

Plates. Every column in this range was read at the plate, on both witnesses, corner numbers first. Six archive pages (207–212), six Gallica leaves (f202–f207). Reads, including the checked zeroes, in data/briefs/8963-PLATE-READS-0010.json; TEI divergences in data/briefs/8963-PENDING-TEI-PATCHES-0010.md. No foot-of-page apparatus of either layer was found on any of the six pages — six checked zeroes, which extends the work's sample from three pages to nine.

1. The three ⚑ MATCHES ONLY ELSEWHERE spans in my range — all adjudicated

@0407A VERS. 8.-- Locutus est Dominus, — Clementine Nm 18:8 opens Locutusque est Dominus ad Aaron. Migne prints the lemma without the enclitic -que, which is why it matches only the Locutus est Dominus of 1 Rg 3:17 etc. Read on both witnesses; it is Migne's, and it is the ordinary clipping of a citation's opening word. No marker; the same author writes Locutusque est Dominus at 0409D for Nm 19:1, so his practice is not uniform. Rendered "The Lord spoke".

@0413A VERS. 8.-- Usque ad vesperam. — the Clementine reads usque ad vesperum at Nm 19:7, 8, 10, 19, 21 and 22, without exception in this chapter. Migne prints vesperam, the feminine form, and his own gloss beneath it expounds Ad vesperam sacerdos intrat. Confirmed on both witnesses. No [var:]: the two forms are the same noun in two declensions and no sense turns on it (Pattern 14 requires a divergence "in a way that changes the sense"). Rendered "Until the evening", and the gloss's own usque ad vesperam rendered "until the evening" to match.

@0413D VERS. 11.-- Qui tetigerit ea. — Clementine Nm 19:11 reads Qui tetigerit cadaver hominis; Migne's words match Ex 30:29 exactly and nothing in Numbers. Confirmed on the plate, and the Alia littera Migne prints in the very next breath is a third reading again (Qui tetigerit morticinum, cf. Nm 19:13). This one did take a [var:]: the object is different, so what the reader is told the man touched is different. Rendered "He that toucheth these things", with the Clementine's cadaver hominis in the marker.

2. Markers fired, with the column

Two [var:] in chunk 0010, one [var:] in 0011, one [cj:] in 0011, four [var:] in 0012, one [var:] in 0013. No [sic:], no [d:], no [ed:] anywhere in the range.

The one [cj:], and which word is where — @0407C. Migne prints Non enim mandatum, sed lux umbram habere futurorum dicitur bonorum. The running text of my English prints MIGNE'S word — "the light" — and the conjecture lex is inside the bracket, per Pattern 18 as amended: the light [cj: lux; read lex, "the law"]. lux is a real, well-formed Latin word, so there is nothing for a [sic:] to wrap; and the faithful English misleads, which is exactly the gap Pattern 18 was ruled for. The evidence that lex is meant is threefold and all of it is on the same plate: the sentence's own antithesis is mandatum against lex (Origen's whole distinction), Heb 10:1 which is being quoted reads umbram habens futurorum bonorum of the law, and the very next paragraph opens Notandum vero quod lex dupliciter dicitur. Both witnesses print lux. The rival conjecture I weighed and rejected was that lux is sound and means the Gospel's light standing in for mandatum's opposite — rejected because umbram habere futurorum bonorum is predicated of the shadow-bearing thing, and light does not bear the shadow of what is to come; the law does.

⚑ The trap next door. Twenty lines earlier, at 0407B, the archive copy also appears to print sacerdotibus lux mandat offerri — and Gallica prints lex, as our twin does. The same substitution, false in one column and real in the next. Nothing but the second witness separates them, and I raised both.

3. 7a″ — the lemma collation. What I checked, including what agreed.

I walked all 190 spans. The 60 usque ad formula spans are not verse lemmata and are treated in §4. Of the remainder:

Divergent from the Clementine, marked — the nine [var:] of §2.

Divergent from the Clementine, NOT marked, and why. Every one of these is rendered as Migne prints it, and every one is logged here rather than smoothed:

• @0404C Vitulum offeret. — Nm 15:24 reads offeret vitulum de armento. Migne inverts. English word order is fixed by grammar so the inversion has no English exponent (Pattern 9's logic); Migne's own gloss beneath restores the Vulgate order (vitulum de armento offerunt).
• @0404D usque ad justi fulgeant sicut sol — Mt 13:43 reads fulgebunt*. This sits inside an abridgment tail (Augustine's sentence, not a lemma), and the subjunctive may be Augustine's syntax; rendered "may shine". No marker: a [var:]` on a spliced tail would accuse the wrong text.
• @0407C Non veni solvere legem, sed adimplere — Mt 5:17 reads non veni solvere, sed adimplere, with legem in the preceding clause. Migne conflates. Not material; the gloss's whole point is legem.
• @0407D Plenitudo legis est dilectio — Rom 13:10 reads Plenitudo ergo legis. A dropped connective in a citation.
• @0407D Haec autem sunt allegorica cited to (Gal. IV) — Gal 4:24 reads Quae sunt per allegoriam dicta. Migne quotes loosely; rendered as printed ("But these things are an allegory").
• @0409A Gratis accepistis, gratis date. Nolite possidere aurum neque argentum, neque pecuniam — Mt 10:8-9 reads neque argentum, neque pecuniam after Nolite possidere aurum, so this agrees; the join of 10:8 to 10:9 across a full stop is Migne's abridgment, not a variant.
• @0409C Sapientiam loquimur inter perfectos, sapientiam non hujus saeculi — 1 Cor 2:6 reads Sapientiam autem loquimur … sapientiam vero non hujus saeculi. Two dropped connectives.
• @0412D Vos autem estis genus electum, regale sacerdotium — 1 Pt 2:9 has no estis. A supplied copula; no sense turns on it.
• @0414A In tribus et quatuor iniquitatibus non adversabor, ascribed only to "the prophet" — this is Amos 1:3 etc. in an Old-Latin form; the Clementine reads Super tribus sceleribus … et super quatuor non convertam eum. No [var:] fired because Migne names no address, and a [var:] must cite one. Rendered as printed.

Checked and found to AGREE with the Clementine exactly (this list is the half of 7a″ that is easy to omit): Nm 15:8 Quando vero de bobus feceris holocaustum @0404B · Nm 15:18 Cum veneritis in terram @0404B · Nm 15:22 Quod si per @0404C · Nm 15:30 Anima vero @0404C · Nm 16:12 Vocaret Dathan @0404D · Nm 16:26 Recedite a tabernaculis @0404D · Nm 16:37 Praecipe Eleazaro and Tollat thuribula @0405A · Nm 16:40 Sicut passus est @0405B · Nm 16:43 Moyses et Aaron and Operuit nubes @0405C · Nm 16:48 Et stans inter and Et plaga cessavit @0405C-D · Nm 17:2 A cunctis principibus @0406A · Nm 17:5 Germinabit virga @0406B · Nm 17:8 Invenit germinasse virgam and Turgentibus @0406B-C · Nm 18:2/7 Tu et Filii, Tu autem et filii (tui) @0406C-D · Nm 18:7 Si quis externus @0406D · Nm 18:9 Omnis oblatio @0407D · Nm 18:11 Jure perpetuo @0408B · Nm 18:12 Omnem medullam @0408C · Nm 18:15 Quidquid primum @0409A · Nm 18:19 Pactum salis @0409B · Lv 2:13 Quidquid obtuleris sacrificii sale condies @0409B · Nm 19:2 Nulla sit macula and Nec portaverit @0410A · Lv 4:12 In locum mundum @0410C · Nm 19:4 Et tingens digitum in sanguine ejus, Asperget contra fores tabernaculi, Septem vicibus @0410C-0411B · Nm 19:5 Comburetque eam cunctis videntibus, Tam pelle et carnibus ejus @0411B-D · Nm 19:6 Lignum quoque cedrinum, et hyssopum, coccumque bis tinctum @0412A · Nm 19:7 Et tunc demum lotis vestibus (et corpore suo) @0412D · Nm 19:9 Colliget autem, Effundet eos, ut sint multitudini filiorum Israel in custodiam, Multitudini filiorum @0413A-C · Nm 19:10 Habebunt hoc @0413D · Nm 19:12 Si die tertio @0414A · Nm 19:13 Polluet tabernaculum @0414A · Nm 19:15 Vas quod @0414A · Nm 19:16 Si quis in agro @0414B · Nm 19:19 Lavabit et se @0414B · Nm 20:1 Veneruntque filii @0414B · Nm 20:11 Percutiens virga @0414C · Nm 20:12 Quia non credidistis @0414C · Nm 20:22 In montem Hor, qui est in finibus @0414C · Nm 20:30 Omnis autem multitudo @0414D · Ct 5:10 Dilectus meus candidus et rubicundus @0409D · Ps 115:16-17 Dirupisti vincula mea, tibi sacrificabo hostiam laudis @0410B · Mt 5:13 Vos estis sal terrae @0409C · Mt 5:22 Qui irascitur fratri suo, reus erit judicio @0408A · Dt 6:5 Diliges Dominum Deum tuum ex toto corde tuo… @0409B · 2 Cor 4:18 Quae enim videntur temporalia sunt @0408B · Lc 2:34 In signum cui contradicetur @0414C · 1 Th 5:19 Spiritum nolite exstinguere @0414A.

Fragmentary lemmata expanded under convention 3, with the verse that decided each. Named here because expansion is a judgment and a blind reader will look: Quod si per. → "But if through ignorance" (Nm 15:22 per ignorantiam praeterieritis) · Turgentibus, → "The buds swelling" (Nm 17:8 turgentibus gemmis) · Quidquid primum. → "Whatsoever first cometh forth" (Nm 18:15 erumpit e vulva) · Eo quod. → "Forasmuch as" (Nm 16:38) · Qui ejectam extra castra. → "Who, when she is cast out without the camp" (Nm 19:3). Complete one-word lemmata were NOT expanded and are rendered as they stand — Sacerdos. · Fimbrias. · Tulit. · Recedite. · Descenderuntque. · Immolabit. · Asperget. · Primitiarum. · Purissimo. · Omne. · Hyssopum — because they elide nothing (the addendum on naming the control's class).

4. usque ad — the etc. splice test, applied per occurrence

usque ad / usque in occurs 70 times in my four chunks. 66 are the abridgment formula and are rendered "as far as"; four are ordinary Latin and are rendered by sense. Each of the four is named here with the Latin that forced it, because a silent departure from the axis is drift:

• @0406B sed nostra adhuc corpora usque in finem mundi differunt a gloria resurrectionis — the resumption tail of a spliced sentence, standing four words after a real formula in the same line → "until the end of the world".
• @0411A plebs Judaea … humilitatem Domini usque ad passionem persecuta est — no etc. precedes it, and passionem is the term of the persecution, not a resumption point → "even unto the passion".
• @0413A qui usque ad vesperam, id est, finem mundi, extra mandata vitae … positus est — the gloss expounding its own lemma Usque ad vesperam, temporally → "until the evening".
• @0414B usque ad immunda erit usque ad vesperam — the two uses stand five words apart in one sentence: the first follows etc. and is the formula, the second is Nm 19:19's own words → "as far as she shall be unclean until the evening".

⚑ One formula does NOT follow an etc. and would be missed by a mechanical splice test: @0412B Mittatur in flammam, usque ad ut in ipsa discat opera recta quae faciat. Migne sets it italic and it names a resumption point, so it is the formula; it is counted among the 66. This is why the test was applied per occurrence and not by a grep in either direction.

5. Migne's own printing, reproduced and not tidied

• Verse addresses out of sequence @0405B–C: VERS. 43, then VERS. 41, then VERS. 48. Confirmed on both witnesses. Reproduced in that order. (VERS. 41's lemma Recedite is in fact Nm 16:45's word; no address was altered — nothing machine-readable consumes a VERS. number.)
• VERS. II.-- @0408B, a Roman numeral for verse 11 among Arabic ones, and VERS 7.-- @0412D, missing the period after VERS, and VERS. 12-- and VERS. 37-- without a terminal period. All confirmed on the plate (the first three) or in our twin against a Gallica read (the last two). Carried verbatim.
• CAPUT XX. opens at VERS. 20.-- over the matter of Nm 20:1. His numbering; reproduced.
• Two sentences that simply stop, with no terminal mark: @0407C …futurorum dicitur bonorum, @0411A …humilitatis ejus imitata, @0412B …opera recta quae faciat, and @0413C …in mentis munitionem vertuntur. Our twin carries no mark at any of the four. Two of them (@0411A, @0413C) were the ones where the archive copy appeared to show a terminal period, and Gallica refutes both; the other two were not separately re-checked at Gallica because our twin and the archive agree. Pattern 8 forbids supplying a mark, and the English carries none.
• Sigla with Migne's own internal spacing — [n: (RAB., in Num. )], [n: (AUG., ubi supra. )], [n: (ISID., in Num. )] — reproduced verbatim.
• Citations that name the wrong place are reproduced as printed and not corrected: (Hebr. V) on Heb 9:12 @0404C · (I Cor. V) on 2 Cor 4:18 @0408B · (Rom. II) on 1 Cor 5:8's azyma sinceritatis et veritatis @0408C · (II Cor. IV) on Rom 2:28-29 @0408C · (Psal. LXXXVII) on the Ps 87/140 conflation @0405D. Each is Migne's, on the plate.

6. Passages read clean, and candidates rejected

Read clean and not marked. The whole of CAPUT XVII (0406A–0406C), the angel-firstfruits sequence of Origen at 0408C–0409A, Gregory's Septem vicibus exposition at 0411B, the cedar/hyssop/scarlet triad at 0412A–0412C, and the whole of CAPUT XX at 0414B–0414D collate word for word against both witnesses with nothing to report beyond §5.

Candidates rejected, with reasons. Sixteen letter-level candidates were raised off the archive copy; fourteen were refuted by Gallica and two stood (the two TEI patches). The full list is in data/briefs/8963-PENDING-TEI-PATCHES-0010.md Part C. Two are worth naming here because the argument for them was good:

• @0409B cognito for cognitio — the archive's letterform is clean, and creaturae et creatoris cognito comprehenditur can be read as an ablative absolute. Gallica prints cognitio.
• @0411C supensione for suspensione — a non-word, which the addendum says is the class worth raising; and Augustine's etymology of cremare really does depend on hanging. Gallica prints suspensione. ⚑ This is the corrected addendum A working as written: the non-word was a reason to fetch the second witness, not a substitute for it.

Declined calls, said plainly.

• @0413C–D per hoc nobis immunditiam providemus reads oddly — the sense wants cleanness, not uncleanness. Both our twin and the archive plate carry immunditiam, and it construes: providere + accusative can mean to provide against a thing. Rendered "we provide against uncleanness for ourselves". A real, well-formed word reading oddly is Migne's until a plate says otherwise, and here the plate says it is his. No marker.
• @0406A reconditus sit thesaurus is a subjunctive in a clause parallel to the indicative largitur. Rendered as an indicative statement ("there lies hidden") because English has no natural exponent for the shift and no sense turns on it. Logged rather than smoothed silently.
• @0405D occurrit quassationi — Isidore's word for the plague. Rendered "ran to meet the shattering", literally; "the plague" would have been the smoother and less honest choice.

7. Vocabulary axes, as applied — with the instrument checked

Counts below are from a case-insensitive pass over my own Latin, with the compound noise stripped by hand; the raw greps the brief warns about all fired here and all were false.

• turba → crowd. One occurrence, @0406D (Caetera autem turba meditetur). A raw \bturba grep also returns nothing else in range, but the neighbouring perturb- family does not occur, so the count is clean.
• multitudo → multitude. Three occurrences: multitudini @0413B and Multitudini @0413C (both inside lemmata of Nm 19:9), multitudo @0414D. ⚑ A case-sensitive grep returns two and misses the capitalised one at the head of a lemma.
• natio → nation. ZERO real occurrences. A raw natio grep returns three hits in this range and every one is a false positive — incarnationis, praedicationem, contemplationis class endings. Reported as a checked zero rather than as three renderings.
• The gentes family. gentibus ×2, gentium ×2, Gentilis ×2, Gentilium ×1, gentilium ×1, gentilitate ×1, Gentilitas ×2 — all of them in the tight block at 0413B–0413C where Rabanus and Gregory are working the Judaeus / Gentilis opposition explicitly (infirmitatem Domini Judaeus mactavit, subtilia ejus mysteria Gentilis intelligit). So: gentes/gentibus/gentium → Gentiles (never "nations" — the typology is express here, not generic), gentilis → Gentile, gentilitas / gentilitate → the Gentile world. ⚑ The raw gent grep also returns diligenter, intelligentiae and argentum; those were discarded.
• mysterium → mystery, sacramentum → sacrament, 1:1. Six mysteri- and four sacrament-. Tested where the two families stand a column apart on the same argument: sanguinis Dominici sacramentum @0412C against subtilia ejus mysteria @0413B, and secreta Dominicae incarnationis / subtiliora sacramenta @0413C–D.
• daemonium, diabolus, satanas, ethnicus — zero occurrences. Checked, not assumed.

Register. The commentator's own voice is not archaised: dicit / ait / inquit → "says". Archaic second-person forms appear only where the Latin is second-person singular, per Pattern 17 — inside quoted scripture (Diliges Dominum Deum tuum @0409B, Quidquid obtuleris @0409B, Non occides @0407C) and in Origen's singular addresses to his reader (sumpseris @0407A, invenies @0407C, adjunxeris @0409B). Where the Latin is plural the English is "you" (Cum veneritis in terram @0404B, Quid vultis … ad vos @0406A, Gratis accepistis, gratis date @0409A, Vos estis sal terrae @0409C, Tradetisque @0410B).

8. Supplied subjects — every "he/she/it/they" the Latin does not print

Named here per the brief's dominant-defect rule. In a gloss the antecedent is usually in the lemma above or in the Vulgate verse the lemma comes from, not in the sentence.

• @0405C Operuit nubes. → "The cloud covered." The subject is nubes, and it is nominative in Migne's own lemma; the gloss beneath confirms it (obtexerit nubes tabernaculum).
• @0408C videtur … oblatus @0408C → "he seems to be offered". The antecedent is the man whose firstfruits are being offered in Origen's tail, not the angel; per angelum is the instrument. Both readings construe; I took the one the ablative per angelum forces and say so here.
• @0411A In electis vero operationem humilitatis ejus imitata → "But in the elect it imitated the working of his humility." The subject is plebs Judaea from the previous clause, which the participle imitata agrees with in gender and number; ejus is Christ. The clause has no expressed verb (imitata est) and no terminal mark, and both are reproduced.
• @0411A Asperget, id est, resistenti plebi … exempla monstrabit → "he shall show". The subject is quisquis passionis ejus vias imitatur, from the preceding paragraph's own gloss on the same lemma; it is not the priest and not Eleazar.
• @0412D Potest iste sacerdos … significare → "This priest … can signify". iste sacerdos is expressed; noted only because the antecedent shifts here from Christ (the priest of the paragraph above, Omnes qui in Christo credimus sacerdotes sumus) to the Jewish priest, which is the point of the sentence and is easy to carry over wrongly.
• @0413B Quia Gentilis sacramenta quae agnovit, praedicare non cessat → "the Gentile ceases not". Subject expressed; flagged because the sentence before it makes vir mundus the actor and a reader may bind Effundet eos to him.
• @0414C Hic Moyses Dominum offendit and @0414D Aaron plangitur, Jesus non plangitur — all subjects expressed; no supply anywhere in CAPUT XX.

Glossa ordinaria on Numbers 21–24. Heads: CAPUT XXI., CAPUT XXI. (cont.), CAPUT XXII., CAPUT XXIII., CAPUT XXIV.

0. Span count against the brief — DONE, and it agrees

data/briefs/8963-lemmata-0014.txt states 258 spans for chunks 0014–0018. Counted in my own Latin, body only (frontmatter excluded), paragraph-bounded:

0014 = 56 · 0015 = 24 · 0016 = 39 · 0017 = 59 · 0018 = 80 → 258

Every body line in all five chunks carries an even number of asterisks, so no span runs across a paragraph and the total is not the product of two cancelling errors. Guillemets: zero in all five chunks, as the launch brief says for this book. 258 = 258.

⚑ No span in my range carries ⚑ MATCHES ONLY ELSEWHERE. The thirteen in this work fall outside cols 0415A–0427B. Two spans carry a ✓ at an address that is not the one Migne's head names, and both are read below (§2.1, §2.2).

1. Markers fired — FOUR, all [var:], no others

No [sic:], no [cj:], no [d:], no [ed:], no [nt:] in 5,237 Latin words.

All four columns were read at the plate on both witnesses (pp. 215, 216, 217 of the archive PDF; Gallica f210, f211, f212). None is an archive-only reading.

Why each cleared the Pattern 14 bar (the divergence changes what the sentence asserts):

• 0419C — present indicative est becomes future erit, and augurium becomes auguratio. Migne's own text prints the Clementine's Non est augurium in Jacob four columns later at 0423A, as the lemma of VERS. 23 — so the two forms stand in the same work, in the same book, and the difference is the printing's, not our reading of it. That internal control is what decided this one.
• 0420B — filii Beor for ex Bosor imports the Numbers patronymic into the Petrine citation, in a commentary on Numbers, where a reader will take the agreement as the Apostle's.
• 0421B — second person becomes third and de caelo is dropped: Isaiah's apostrophe to Lucifer becomes a report about him, and the gloss builds on the report (illuminatus ab illo Lucifero).
• 0423B — perfect quid operatus sit becomes future quod perficiet, and Migne's gloss expounds exactly the future (id est cum oportet, cum expedit).

2. The 7a″ collation — every lemma walked, agreements included

2.1 Verse addresses that do not match the words

• 0417D VERS. 20.-- In Bamoth. — in Bamoth stands at Nm 21:19 (de Nahaliel in Bamoth); Nm 21:20 begins De Bamoth, vallis est. Migne's head is one verse ahead of his own lemma. ⛔ Not altered. A verse address may not be touched on any evidence short of an external control, and nothing machine-readable consumes it. Recorded, not repaired.
• 0416D VERS. 16.-- Foderunt illum principes et excuderunt reges. — printed after VERS. 17, out of sequence, on both witnesses; and its words are Nm 21:18, not 21:16. Both preserved as printed.

2.2 The lemma the whole of CAPUT XXI turns on, and why it took NO marker

0416D–0417B, Foderunt illum principes et excuderunt reges. The Clementine (Nm 21:18) reads quem foderunt principes et paraverunt duces multitudinis. Migne's excuderunt reges is a different verb and a different subject, and it is not incidental: three quarters of the column is an etymology of excudere (quod ab excudendo, non excidendo dictum est), and the kings-versus- princes contrast that follows exists only because of it.

I declined a [var:] here, and the reason is that Migne prints the Vulgate reading himself. Eight lines below, at 0417C, he sets Et paraverunt duces. as its own lemma with its own gloss. A reader meets both readings on the same page in Migne's own voice; a [var:] would tell him a third time and would imply we had found something the edition had not. This is the same test Pattern 18a applies to a conjecture the edition already glosses, and it decides this the same way. The rival was weighed and lost on that ground alone — had Migne not printed paraverunt duces, the marker would have fired.

2.3 Divergences from the Clementine, collated and NOT marked

Each was checked against sources/vulgate/clementine-flat.txt and rendered as Migne prints it.

• 0415B Nm 21:12 ad torrentem Zareth for Zared — a name-form, no sense at stake.
• 0415D Sap 8:2 Hanc quaesivi mihi ducere uxorem for quaesivi sponsam mihi eam assumere. ⚑ Rendered "to take to me for a wife", not "for a spouse": the gloss around it is built on uxor (sanctorum uxores, in conjugium sumere, hujusmodi conjugem), so the familiar word would have cut the argument from its proof-text. Declined a marker only because the Migne form is itself a well-attested Old-Latin shape and the sense is unchanged.
• 0416A Nm 21:16 Congrega populum et dabo ei aquam de puteo — the verse ends at aquam; de puteo is Migne's addition, and it is the word the whole gloss is about. Recorded.
• 0416D–0417A 1 Cor 2:10 etiam alta Dei for etiam profunda Dei. Rendered "the deep things of God" — altum bears "deep" and the immediate context is a well and abyssi venas, so the English is faithful to Migne's word and the divergence is lexical only. No marker: nothing the reader would take differently.
• 0417A Mt 28:19 docentes for docete, absorbed into the gloss's own participial sentence (Impleverunt enim mandatum Dei, docentes...). Not a divergence in the quotation but a construction.
• 0417B Mt 5:19 Qui fecerit et docuerit sic homines, magnus vocabitur — a conflation: sic homines belongs to the verse's first limb, which ends minimus vocabitur, while magnus vocabitur is the second. Migne's sentence nevertheless asserts what the second limb asserts, so no marker; recorded because a reader collating will see the seam.
• 0417D Munera mea data mea cited to (Levit. II) — in no Clementine verse in that form. Rendered literally ("My gifts, my givings"). No marker: nothing to compare it to.
• 0418A Jn 12:31 Ecce ... mittetur foras for Nunc ... ejicietur foras.
• 0418A 1 Jn 5:19 Quia omnis mundus in maligno positus es for mundus totus ... positus est, and cited to (Joan. V). See §3.2 for the es/est split between the printings.
• 0418B Rom 6:4 Consepulti ... illi per baptismum in morte for cum illo ... in mortem.
• 0420B Jn 7:52 quia propheta a Galilaea non surgit — Migne's word order for a Galilaea propheta.
• 0420C Ex 32:34 Angelus meus ibit tecum — Ex 32:34 has praecedet te; ibit angelus meus is Ex 23:23's shape. Migne's siglum names XXXII. Recorded, not altered.
• 0421B 1 Cor 1:27 Quia stulta mundi elegit Deus, ut confunderet sapientes for quae stulta sunt mundi ... ut confundat.
• 0421B Nm 22:38 Ecce adsum for En adsum, with aliud loqui transposed.
• 0422B Gen 3:14 Maledictus tu ab omnibus bestiis terrae for maledictus es inter omnia animantia et bestias terrae.
• 0422C Mt 5:14 non potest abscondi civitas supra montem posita — Migne's word order.
• 0423A Nm 23:22 Deus eduxit eum for illum.
• 0424B Os 11:1 Ex Aegypto vocavi Filium — meum dropped.
• 0425B Nm 24:13 the lemma stops mid-clause at ut vel boni; rendered as the phrase it stands for (convention 3), "to utter either good".
• 0425C Nm 24:14 Dabo tibi consilium — tibi is Migne's addition to dabo consilium.
• 0426C 1 Reg 15:6 Discede de medio Amalec et non percutiam te, quia fecisti misericordiam Israel — 1 Sam 15:6 reads Recedite ... ne forte involvam te cum eo : tu enim fecisti misericordiam cum omnibus filiis Israel. Far apart, and Migne's is a paraphrase rather than a quotation; recorded, no marker.
• 0426D Sir 24:31 elucidant me, vitam aeternam habebunt — a mid-verse fragment of Qui elucidant me; nothing diverges.

Checked and found to agree with the Clementine EXACTLY (recorded because a findings-only list is indistinguishable from no list): Ps 30:20 Quam magna multitudo dulcedinis tuae @0417C · Ps 71:2 Deus, judicium tuum regi da @0418B · Jn 14:6 Ego sum via, veritas et vita @0418B · Jn 14:30 Venit princeps mundi hujus, et in me non habet quidquam @0418A · Jn 19:7 Nos legem habemus, et secundum legem debet mori, quia Filium Dei se fecit @0420B · Ps 73:19 Ne tradas bestiis animas confitentes tibi @0419A · Nm 23:19 Non est Deus quasi homo, ut mentiatur @0422D · Nm 23:21 Non est idolum in Jacob @0422D · Nm 24:9 Accubans dormivit ut leo @0424D · Nm 24:11 Dominus privavit te honore disposito @0425B · Nm 24:16 Qui cadens apertos habet oculos @0425D · Nm 24:17 Videbo eum, sed non modo and Orietur stella ex Jacob @0425D–0426A · Nm 21:8 Fac serpentem @0415A · Nm 21:10 Profectique filii Israel castrametati @0415B · Nm 21:22 Via regia @0418A · Nm 22:5 Misit ergo nuntios ad Balaam filium Beor @0419B · Nm 22:16 Sic dicit Balac filius Sephor @0420A · Nm 22:22 Stetitque angelus Domini @0420C · Nm 22:23 Cernens asina angelum stantem @0420C · Nm 23:8 Quomodo maledicam @0421C · Nm 23:13 Videre non possis @0422C · Nm 23:15 Sta hic juxta @0422C · Nm 24:5 Quam pulchra @0424A · Nm 24:6 Ut valles nemorosae @0424A · Nm 24:18 Israel vero fortiter aget @0426B · Nm 24:21 Vidit quoque @0426C · Gal 4:4 ubi venit plenitudo temporis misit Deus Filium suum @0426A · Gal 5:6 sed fides quae per dilectionem operatur @0422A · 2 Tim 2:11 cui si commorimur, etiam convivemus @0426B · Rom 12:14 Benedicite et nolite maledicere @0421C · 1 Reg 1:18 Vultus ejus non sunt amplius in diversa mutati @0422A (only Vultus ejus for vultusque illius).

3. What the plates settled — 7 pages, 2 witnesses, 18 candidates, 1 stood

Full record in data/briefs/8963-PLATE-READS-0014.json. No foot-of-page apparatus of either layer on any of the seven pages — seven checked zeroes.

3.0 ⭐ THE ONE FINDING THAT STOOD: a lost Hebrew word @0418B

Our twin reads In Jasa, Hebraice iahtsah quod interpretatur Mandati impletio. Both witnesses print a pointed Hebrew word after the transliteration — יַהְצָה — read at 600 dpi on the archive (p. 214) and at 1800 px on Gallica (f209). Corpus Corporum has dropped it.

This is the one structural class 8950's two-witness audit left standing: a whole non-Latin word, which ink neither adds nor removes, and which both copies carry. Filed as a TEI patch (data/briefs/8963-PENDING-TEI-PATCHES-0014.md), not as a marker: the loss leaves the sentence sound and Migne's own transliteration stands beside it, so Pattern 13a clause 1 governs and an [ed:] would put our voice on the page for a word an English reader could not read anyway. Same treatment as 8950 @0167D (ad portam Enaiim עינים).

⚑ Our Latin for cols 0415A–0427B contains no Hebrew or Greek character at all, so any non-Latin type on these plates is by definition a loss. Both columns of all seven pages were checked and this is the only occurrence in my range.

3.1 Every archive-only candidate fell

Seventeen readings in which the archive copy differs from our TEI were raised and taken to Gallica; all seventeen were refuted, our twin agreeing with Gallica every time. (The Hebrew of §3.0 is the eighteenth candidate and the only one that stood — and it is the only one where the two copies AGREE against our twin, which is exactly why it stands.) The table is in data/briefs/8963-PENDING-TEI-PATCHES-0014.md. Four of them construe as sound Latin (praecationis, Populus totus, Balaam for Balam, regni/regui aside), and Populus totus habitabit @0421D would have looked like a real find: it reads well, and only Nm 23:9's solus and the second witness say otherwise.

⚠ One candidate was raised against GALLICA and withdrawn by re-reading Gallica. At 0423D Gallica at 1400 px appears to print super illam; at 1800 px it prints super illum, with the archive and our twin. Same trap the launch brief records for Gallica's digits, in a letter. Never settle a Gallica letter at the page-width fetch.

3.2 FOUR sites where our exemplar (Gallica) carries the defect and the archive is clean

This is the class the mid-flight correction named, and this range has four of it. None takes a patch, a [sic:] or a [cj:]: the two printings disagree, so the bar for a public claim about Migne's type is not met in either direction.

• 0417A — Gallica and our twin print the non-word putenm; the archive prints a clean puteum. Rendered "this well". ⛔ Under the pre-correction non-word heuristic this would have been filed as our corruption and patched, and it would have been wrong.
• 0418A — Gallica and our twin print in maligno positus es; the archive prints est. Rendered by sense, "the whole world is set in wickedness". A [sic:] was considered under Pattern 12's "no grammatical slot" test — a second-person es with mundus as subject has none — and declined, because the archive prints the sound form and a marker would accuse a type the other printing does not carry.
• 0419A — Gallica and our twin print a full stop inside the phrase, ut de eorum. correctione et profectu; the archive prints a comma. Preserved in the English under Pattern 8 (never delete a mark the plate prints); the English reads "that from their. correction and progress", which is exactly as odd as the page.
• 0425D — Gallica and our twin open the head with an unclosed parenthesis, (VERS. 16.-- Qui cadens apertos habet oculos., which is never closed; the archive prints no parenthesis. Preserved.

3.3 Migne's own punctuation, confirmed on both witnesses and carried through

• 0416C quod est esca et nutrimentum ignis Si gustaveris — no full stop after ignis, on both copies. Carried; the English runs on in the same place.
• 0422B quam praedixerat Balaam in Israel orituram — no terminal stop, on Gallica and in our twin (the archive has one). The paragraph ends without a mark and so does the English.
• 0422D usque ad haec vero a perfectioribus secretius advertenda sunt? — a question mark closing an abridgment tail, on both copies. Kept, rendered as a direct question ("but are these things to be observed more secretly by the more perfect?"), no 8b recasting needed.
• 0425A — the siglum (Joan. 9) with an arabic numeral among roman ones, on both copies. Passed through verbatim.
• 0426B — (Id.) in lowercase where every neighbour is (ID.), on both copies. Verbatim.
• 0420B / 0425C — Balam (not Balaam) at both sites, on both copies at 0425C. Rendered "Balam".

4. Construals declined, and the reasoning a later reader cannot reconstruct

• 0419C — nec daemones scit benedicere Isaac, et Jacob, et omnes sancti. Two parses. (a) ars (from Ars enim magica nescit benedicere) remains the subject of the singular scit, and daemones, Isaac, Jacob, omnes sancti are all objects: "nor does it know how to bless demons, and Isaac, and Jacob, and all the saints." (b) daemones is the subject, with the verb attracted to the singular, and the saints are an elliptical nominative ("Isaac and Jacob and all the saints [knew how]"). I took (a), because it is the only one that accounts for the singular scit without supplying a verb the page does not print, and because it needs nothing added. (b) is the likelier sense of Origen's argument and is recorded here for that reason. Both witnesses print scit.
• 0423B — sed verbum quod DEUM posuit in ore ejus. The parallel two columns earlier (0421C, dicit verbum Dei positum in ore ejus) and the following sentence (Nisi enim verbum a Deo esset) both want Deus as the subject. Both witnesses print Deum, a real word in a real slot. Rendered literally, "the word which set God in his mouth". No [cj:]: Pattern 18's bar is a printed word that makes the sentence assert what the author is arguing against, and 18a scopes the extension to negations and correlatives. A case-slip is neither, and the sentence is odd rather than false.
• 0423C — Erigendum dicit populum futurum, de eo enim quae videbat, dixisset, erectus est. Does not construe as printed (quae against eo, dixisset with no protasis). Both witnesses agree with our twin, letter for letter. Rendered as closely as the words allow and left; no conjecture is defensible here and none is offered.

4.1 Unexpressed subjects supplied, and what fixed each

Every "he/she/it/they/none" the English supplies that the Latin does not print, with its warrant:

• 0415A inivit praelium, et captivum duxit Israel → "he joined battle" — the Canaanite, named as the subject of Audivit in the sentence before, and the subject of Nm 21:1.
• 0417B quem non possunt fodere vel excudere, nisi reges et principes → "which none can dig or hammer out but kings and princes." The Latin's possunt has an indefinite subject which the nisi clause then restricts; "none … but" carries the non and the nisi together without adding a third negative. The alternative ("which they cannot dig … save kings and princes") leaves they pointing at the princes of the lemma, who are precisely the men who can.
• 0421A sed ista angelum vidit … et vidit, et detulit, et locuta est → "she saw … she saw, and reported, and spoke." ista is the ass (asina), feminine, and the sentence's whole point is the contrast with homo (quem homo videre non potuit) in the same clause.
• 0424B eduxit illum Pater ad semetipsum → "the Father brought him out" — both printed.
• 0426B Perdet liberatum de civitate → "He will destroy the man delivered out of the city" — the subject is the Israel of VERS. 18's lemma, two lines above, not Edom.

5. Vocabulary — the axes, and the two calls the axes did not rule on

• gentes — "nations" where the word is the quoted verse's own or its immediate re-quotation (Nm 23:9 @0421D, Nm 24:8 @0424C, Mt 28:19 @0417A); "Gentiles" in the gloss's own voice where it stands opposed to the people of God (ex gentibus credentes @0420C, quae sacrificant gentes @0423C, initium gentium @0426B–C). ⚑ The one place the rule is uncomfortable is 0417A, where reges gentium and gentes congregaverunt sit three words from Mt 28:19's omnes gentes; the axis rules that verse "nations", so all three read "nations" there rather than break the sentence across two English words. Flagged rather than decided silently.
• gentilis → Gentile, gentilitas → the Gentile world (bruta Gentilitas @0421A, gentilem plebem @0421A, fierent gentiles @0426C, gentilis efficitur Amalec @0426C) — the 2026-09-04 addition, applied as written.
• daemonium/daemones → demon(s); diabolus → devil; satanas does not occur in my range. Checked with the brief's own warning in view: \bdaemoni misses daemones, daemonum, daemone, all of which occur here.
• turba and multitudo — turba does not occur in cols 0415A–0427B; multitudo occurs twice (@0417C, @0422B) and reads "multitude" both times. The axis is not exercised in this range.
• natio — does not occur. (The brief's warning about damnatio-/inclinatio- is why this is stated as a checked zero rather than left out.)
• mysterium → mystery, sacramentum → sacrament: sacramentum does not occur here; mysteria/mysticum read "mysteries"/"mystical" throughout.
• usque ad: 67 occurrences in my five chunks, and all 67 are the abridgment formula — every one follows an etc., (or Alia littera,) and names the point at which the father's sentence resumes, so every one reads "as far as". There is not one ordinary-Latin use in this range, temporal or otherwise. Counted per chunk: 12 · 8 · 8 · 18 · 21.
• The commentator's own voice is not archaised: dicit/ait → "says". thou/thee appear only where the Latin is grammatically singular — the second-person address of the Origen homilies (Si ergo intellexisti @0415D, Si autem veneris @0416C, convocat te lex @0416A, Si intellexisti ... noli dissimulare @0424D) — and you where it is plural (Nisi occulte audieritis, plorabit anima vestra @0417C, Date initium cordi vestro @0416B, Incedite in lumine ignis vestri @0416D).

6. Structure and the mechanical checks

• Column anchors 1:1 and in position, checked against the word each stands beside in the Latin, not merely for presence (test 2b): 14 · 3 · 9 · 9 · 15. Two are deliberately mid-construction because Migne's are — rationalem [0421B] naturam and nuper geniti infantes rationabiles [0423C] sine dolo lac accipiunt.
• Notes 1:1 and in order, contents untranslated, including the sigla that sit inside an italic lemma span (@0415A, @0415C, @0418D, @0420A, @0422A, @0423D, @0427B). No [n:] in my range is editorial prose, so no [nt:] was created; every one is a citation or an attribution siglum.
• Question marks by column band, not by total: 0014 = 1 (@0416C) · 0015 = 0 · 0016 = 1 (@0421B) · 0017 = 3 (@0422B, @0422D, @0423A) · 0018 = 2 (@0425A, @0424D). Latin and English agree band by band, not merely in sum.
• ## heads 1:1. CAPUT XXI. (cont.) → CHAPTER XXI. (cont.) — the (cont.) is our chunk division and is left exactly as it stands, untranslated and unexplained, for the builder to merge. Numerals mirror the plate (CHAPTER XXI., not "CHAPTER TWENTY-ONE").
• Paragraph structure 1:1 with the Latin twin in all five chunks (27 · 14 · 27 · 37 · 49 lines).
• Em-dashes: 3 in 5,237 Latin words (corpus median 5.3 per chunk; this is 0.6 per chunk) — 0014 @0416D, 0016 @0420B and @0421A. Each carries an appositive the Latin marks with its own break; none was introduced for rhythm, and none was removed from a faithful sentence to avoid one.
• The marker-strip test (strip every marker, read what is left) was run on all five chunks: no stranded article or preposition, no doubled mark, every clause keeps its verb, its negation, its conjunction and its object. This is trivially clean here because nothing is quarantined — the four [var:] are notes beside a complete rendering, never in place of one.

7. Read clean

Cols 0415A–0427B were read line by line against the Latin twin with the plate open. Beyond what is recorded above — the four [var:], the Hebrew loss of §3.0, and the four Gallica-side defects of §3.2 — I found nothing to raise: no dittography, no split or fused type in our twin, no lost Hebrew or Greek (the range prints one Hebrew word, iahtsah @0418B, transliterated by Migne himself and carried through), no empty markup, no place where the digitization has visibly tidied a rough passage. The two paragraphs I looked at hardest and left alone are §4's nec daemones scit and Erigendum dicit populum futurum — both are Migne's on two witnesses and both are odd.

Chunks 0019, 0020, 0021, 0022. Heads: CAPUT XXV · XXVI · XXVII · XXVIII · XXIX · XXX · XXXI · XXXII · XXXIII.

Span count. My lemma file (data/briefs/8963-lemmata-0019.txt) states 184 spans. Counted in my own Latin: 184, flat, and 184 again counted paragraph-bounded, with zero paragraphs carrying an odd number of *. This book has no guillemets and my four chunks contain none (« count = 0). The totals agree.

⚑ MATCHES ONLY ELSEWHERE: my file carries none. Zero spans in my range are so marked, so none was owed an adjudication under that head. What my file does carry in quantity is the bare ✓ with an address, and §1 below reads those addresses.

Plates. Every column in this range was read at the plate, on both witnesses, corner numbers first. Six archive pages (219–224), six Gallica leaves (f214–f219). Reads, including the checked zeroes, in data/briefs/8963-PLATE-READS-0019.json; TEI divergences in data/briefs/8963-PENDING-TEI-PATCHES-0019.md. No foot-of-page apparatus of either layer was found on any of the six pages — six checked zeroes, which extends the work's sample by six pages.

The ratio the launch brief predicted, measured again on my range: 27 archive-only readings raised, 27 refuted at Gallica, 0 stood. The two findings that stood are of the other class entirely — whole Hebrew words present on both printings and lost by our TEI.

1. The lemmata — 7a″, every span walked

1a. The lemmata that agree with the Clementine exactly, checked and recorded

These are recorded because a list of findings alone cannot be told from a report by an agent who never looked. Checked word by word against sources/vulgate/clementine-flat.txt:

Nm 25:2 Et adoraverunt deos earum · Nm 25:7 Quod cum and Arrepto pugione · Nm 25:9 Occisi sunt viginti quatuor millia (Migne clips the Clementine's closing hominum) · Nm 26:1 Postquam noxiorum · 26:2 Numerate omnem · 26:53 Istis dividetur terra · 26:65 Nisi Caleb filius Jephone, et Josue filius Nun · Nm 27:8 Homo cum mortuus fuerit absque filio · 27:12 Ascende in montem · 27:14 Quia offendistis · 27:16 Provideat Dominus Deus · 27:20 Et dabis, Et partem gloriae, Ut audiat · 27:21 Pro hoc · Nm 28:3 Haec sunt sacrificia · 28:9 Die autem Sabbati and Et duas decimas similae oleo conspersae · 28:11 Vitulos de armento, Arietem unum, Agnos anniculos septem · 28:12 Et tres decimas similae, Et duas decimas · 28:13 Et decimam decimae, per agnos singulos · 28:17 Quintadecima die · 28:26 Dies etiam primitivorum quando offeretis novas fruges · Nm 29:1 Mensis etiam septimi · Nm 30:3 Si quis · 30:4 Mulier si, Et juramentum · 30:6 Nec obnoxia tenebitur · 30:7 Si maritum · 30:11 Uxor in domo · Nm 31:2 Ulciscere prius · 31:4 Mille viri de singulis · 31:8 Et reges eorum, Balaam quoque · 31:15 Cur feminas · 31:16 Ad suggestionem · 31:21 Eleazar · 31:22 Aurum · 31:24 Et lavabitis · 31:27 Qui pugnaverunt · 31:50 Depreceris · 31:51 Susceperuntque · Nm 32:1 Filii autem Ruben · 32:4 Terra quam percussit Dominus, Jumenta plurima · 32:29 Si transierint filii Gad et filii Ruben · Nm 33:1 Hae sunt mansiones, Per turmas suas, In manu Moysi et Aaron quas descripsit · 33:2 Moyses juxta castrorum loca, quae Domini jussione · 33:3 Profecti igitur, Manu excelsa · 33:5 In Soccoth · 33:6 De Soccoth, In Etham · 33:8 In Mara · 33:9 In Elim ubi erant · 33:10 Sed et inde egressi, super mare Rubrum · 33:11 In deserto Sin · 33:12 In Daphca · 33:15 In deserto Sinai · 33:17 In Haseroth.

And, among the inline quotations: Ps 109:4 Tu es sacerdos in aeternum secundum ordinem Melchisedech — exact · Mt 20:16 Multi enim sunt vocati, pauci vero electi — exact, and the printed address (Matth. XX) is the right one of its two homes · 1 Cor 13:10 Cum autem venerit quod perfectum est, evacuabitur quod ex parte est — exact · Lv 16:29 Affligetis animas vestras — exact · Ps 104:37 non erat in tribubus eorum infirmus — exact · Ps 95:5 Omnes dii gentium daemonia — exact (Migne clips the opening Quoniam) · Jos 5:13 Noster es, an adversariorum? — exact · 1 Cor 15:7 Deinde apostolis omnibus — exact · Ps 127:2 Labores manuum tuarum quia manducabis, beatus es et bene tibi erit — exact but for Migne's comma where the Clementine sets a colon.

1b. Divergences that took a [var:] — twelve, with the column

(Thirteen rows; the count in the marker tally below is by marker, and 0435D carries two.)

1c. Divergences examined and declined — named, with the reason

• @0427D Initiatus est Israel for Nm 25:3 Initiatusque est Israel Beelphegor. The dropped enclitic -que is the ordinary clipping of a citation's opening word, adjudicated for this work at 0407A by the 0010 stint. No marker. The second lemma four lines down, Et initiatus est Beelphegor, adds Et and drops Israel: same class, and it is Origen's own re-quotation.
• @0429D Da eis … Dabis ei possessionem (Nm 27:6). The second is the alia littera Origen expounds, standing inside his own siglum. Migne does not label it, but the surrounding paragraph is Origen's lemma and nothing in the gloss turns on da eis against dabis ei. Declined; recorded here instead.
• @0429C Volo, Pater, ut ubi ego sum, illic et isti mecum sint (Jo 17:24) and Sicut tu in me, Pater, et ego in te, ita et isti in nobis unum sint (Jo 17:21). Word order and isti/ipsi; the claim is unchanged in both. Declined under Pattern 14's "in a way that changes the sense".
• @0435A Qui enim descendit, ipse est qui ascendit (Eph 4:10) — an added enim, a dropped et. Declined. Tecum descendam in Aegyptum (Gn 46:4, Ego descendam tecum illuc) — Migne names Egypt where the Vulgate is deictic, but Gn 46:3 has already named it. Declined.
• @0435A qui est Dei virtus et Dei potentia — 1 Cor 1:24 reads Dei virtutem et Dei sapientiam. Migne does not italicize or cite it, so it is an allusion in the commentator's own voice and not a scripture citation; Pattern 14 does not reach it. Rendered "the power of God and the might of God".
• @0432D Mundabit eum sacerdos, et mundatione non mundabis reum (Levit. XIII). No such sentence stands at Lv 13. But the whole paragraph is Augustine expounding the LXX (its two preceding readings are printed under an express LXX:), so the comparand is not the Clementine. Declined; the divergence is recorded and the English renders what is printed.
• The four expressly-labelled alternative readings take no marker, because Migne labels them himself: LXX: Et definitiones quas definivit adversus animam suam @0432C · LXX: Dominus mundabit eam, quia abnuit pater ejus @0432D · Alia littera: Et accepit Moyses et Eleazar sacerdos aurum… @0434C · Septuaginta sic: Cum virtute sua @0435A. Also Et sic colligeris. Novissime inquit: Apponeris ad populum tuum @0433B, where the second half is a second reading introduced by Novissime inquit.
• Spellings that are the same word: In Kalendis for the Clementine's In calendis @0431C · in Setim for in Settim @0427B (inside the [var:] above, and not separately marked) · sepulcra for sepulchra @0438B. No sense turns on any of them.
• @0436D the gloss's own prose paraphrases its lemma as ubi sunt duodecim fontes aquarum, et septuaginta palmae erant — present tense, and the order of fountains and palms reversed against Nm 33:9. It is not an italic span and it is not offered as a quotation, so no marker; rendered as printed, tense shift and all.

1d. Two spans the pre-scan marked ⚠ NOT in Clementine verbatim for a mechanical reason

• @0436A Et in diis eorum. Omnes dii gentium daemonia. — one italic span carrying two quotations, Nm 33:4 and Ps 95:5. Each is exact on its own; only the concatenation is unfindable.
• @0428C (ORIG., hom. 21 in Num) and @0433B (ORIG., hom. 25 in Num.) — sigla that our TEI carries as italic spans rather than as [n:] notes. The plate sets both as ordinary parenthetical sigla, exactly like their neighbours. They are reproduced verbatim and untranslated in the English, italics and enclosed column anchor and all; the inconsistency is our TEI's markup, not Migne's page, and it is small enough not to be worth a patch.

2. Markers fired, with the column

Twelve [var:] — @0427B, @0429A (×2), @0429C, @0432D, @0433A, @0435C (×2), @0435D (×2), @0437C, @0437D, @0438A. Distribution: 4 in chunk 0019, 0 in chunk 0020, 2 in 0021, 7 in 0022.

Two [ed:] — @0427B and @0437C, both for a Hebrew word Migne prints and our TEI has lost.

Zero [sic:], zero [cj:], zero [d:], zero [nt:], zero [cn:] in the whole range.

2a. Why zero [sic:], when 27 candidates were raised

Every candidate this range produced was an archive-copy reading refuted at Gallica, and a [sic:] on any of them would have accused Migne of the photograph's defect. The two findings that survived both witnesses are losses in our file, which is a patch and an [ed:], never a [sic:]. Of the readings confirmed on both copies, none is broken type: Orabatur and cultu are real Latin words in real cases (Pattern 12's boundary: a real word merely wrong is rendered, not marked), and the two unconstruable places at 0434B and 0434D are Migne's syntax, not his sorts.

2b. Why the one [var:] on a non-scripture-looking word is nevertheless a [var:], not a [cj:]

@0427B Orabatur. It has the Pattern 18 shape — a real word whose faithful English says something false — and it would have been tempting to fire [cj: Orabatur; read Morabatur…]. It is not a [cj:], because of Pattern 18's own boundary: a divergence inside a QUOTED SCRIPTURE is Pattern 14's, and this is the verse lemma of Nm 25:1. So Migne's word is rendered in the running text ("He was entreated") and the Clementine's reading stands in a [var:] beside it. Which word is where: Orabatur is Englished in the text; Morabatur is inside the bracket, named as the Vulgate's. Nothing in this range fired a [cj:], so Pattern 18's inversion hazard does not arise.

3. Migne's own printing, reproduced and not tidied

• @0434B the plate — both plates — prints Probantur tamen haec omnia, alia per ignem, alia per aquam uniuscujusque opus quale sit ignis probabit, with no punctuation whatever between aquam and uniuscujusque. Two clauses are run together. Pattern 8 forbids supplying a mark, so the English runs them together too: "…some by fire, some by water the fire shall try of what sort each man's work is." It reads as an error and it is faithfully his.
• @0434D usque ad qui est terra promissionis intrare — on both copies, and it does not construe (a nominative terra with intrare). Rendered as closely as the printed words allow: "as far as who is the land of promise to enter." Neither qui est terrae promissionis intrare nor ut terram promissionis intraret is offered in the text; the crux is the whole answer.
• @0431A cultu Dei, ablative, where the sense wants the accusative cultum. Both copies. Rendered as an ablative of respect — "and in the worship of God has carried men over from things visible to things invisible" — with the object of transtulit supplied as "men", which is named in §5 below. No marker: Pattern 9 (a case that English cannot expone) and Pattern 12's real-word boundary both point the same way.
• @0433C Cur feminas. is printed with a full stop, though Nm 31:15 is a question (Cur feminas reservastis?). The English keeps the full stop. Pattern 8: never supply a mark he does not print. Question-mark parity for chunk 0021 is 0 against 0.
• Chapter head CAPUT XXXI, no full stop. Our twin has none; Gallica has none; the archive prints one. Two printings, not two photographs — a press variant, which takes neither a patch nor a marker. The English head is CHAPTER XXXI, mirroring our twin, and it is the only head in the range without a stop.
• Numerals mirror the plate throughout: CAPUT XXV. → CHAPTER XXV., and so on to CHAPTER XXXIII. No head is spelled out.

4. The usque ad occurrences, decided one by one

The etc. splice test was run on every occurrence. All but one are the abridgment formula and are rendered as far as, matching the shipped chunks of this work.

⭐ The exception is @0428C, and both are in one sentence: …etc., usque ad ubi cruciatur die ac nocte usque ad novissimum quadrantem. The first follows an etc. and names where Rabanus's sentence resumes — formula, "as far as". The second is Mt 5:26's own words inside the resumed tail and is rendered by sense, "even to the last farthing". This is precisely addendum D's trap, and note that our TEI does not italicize the second while the plate does — so an italic test would have got it wrong in this instance as well.

One near-miss worth recording: @0431B usque ad qui sacerdos est in aeternum secundum ordinem Melchisedech [n: (Hebr. V)] — the formula, whose resumption tail happens to be a quotation of Heb 5:6. Rendered as the formula plus the tail; the tail is not an italic span and takes no [var:].

5. Unexpressed subjects and objects supplied, each one named

Per the brief's dominant-defect rule, every pronoun or subject I supplied that the Latin does not print, with what fixed it:

• @0427C deos earum → "their gods", feminine. The antecedent is filiabus Moab in Nm 25:1, the verse the lemma comes from — not the Madianites, who enter only at 25:6.
• @0427D Consecratus est → "He was consecrated". The subject is Israel, from the lemma immediately above (Initiatus est Israel), not the individual Israelite of 25:8.
• @0429B video ipsummet Moysen … puto → "I see" / "I think". First person, expressed; the speaker is Origen, whose siglum governs the paragraph, and the English keeps "my Lord" at Christus Dominus meus rather than neutralizing it.
• @0430A dicit Dominus … non sanctificaverunt eum → "they did not sanctify him". eum is God, not Moses: Nm 27:14 reads nec sanctificare me voluistis. Both readings construe in Latin; the verse decides.
• @0430C non elegit filios, non nepotes, nec rogat → the subject is Moses, from Provideat Dominus Deus in the verse he is quoted as speaking. Note the English carries all three negatives, on the three words the Latin puts them on (non elegit, non [elegit] nepotes, nec rogat); I repeated "he did not choose" rather than let one non cover two objects, because English elision there would have moved the second negative onto the verb.
• @0431A transtulit has no expressed object → "has carried men over". The object is the worshippers implied by cultu Dei; the alternative is to make cultu the object (i.e. read cultum), which is the conjecture named in §3 and is not taken in the text.
• @0432D noceant animae suae → "they harm their own soul". The subject is the women of the lemma (Mulier si…), carried over from talibus votis; it is not the vows.
• @0433C redit ad viros vel parentes and hic velut auctor sceleris interficitur → "he is killed": Balaam, named in the sentence before.
• @0434A Spectat nos angelorum chorus → expressed. But quis plus afferat auri and quis lapides pretiosos curiose intuetur are indefinite quis — "who brings back", "who looks" — and take no supplied subject.
• @0435B praeparemus nos … consumamus … expleamus … deficiamus → first person plural throughout, expressed in the verbs; the "we" is the readers (legentes), which is why the English says "so that we, reading…".
• @0436A Facit et nunc → "He does so even now": the subject is the Lord, from faciet Dominus vindictam in the sentence before, not the demons.
• @0437A haec gratia suscipiat tentationem — see §6.
• @0438A Una de pluribus, sed major omnibus, non separatur in ordine → "it is not separated in order". The feminine subject is mansio (the twelfth station, Sinai), from Sinai duodecima mansio in the line above — not Scriptura and not the apostles' number.
• @0438C lepra perfunditur → "the people of the Jews is overspread": expressed (populus Judaeorum). in typum murmurantis is a genitive singular participle with no noun; I rendered it "as a type of one murmuring against the Church", not "of the people murmuring", because populus Judaeorum is the main clause's own subject and making it the genitive too would collapse the type into the thing typified.
• @0438D venies … venies → second person singular, expressed, addressed to the viator. Pattern 17: thou throughout that paragraph, and throughout @0436D–0437A, because every verb and pronoun there is singular (Vides, te, venisses, pertulisses, superasses, suscipiat te, discernas, Intuere, sepelieris, mortificaveris). There is no plural address anywhere in my four chunks.

6. Passages where two readings construe, and I say which I took

• @0437A Haec te amoenitas post amaritudinem, haec requies post laborem, haec gratia suscipiat tentationem. Two construals. (a) te is the object of suscipiat and tentationem is a second object without a conjunction; (b) te belongs with the first two subjects and tentationem is the object of the third, the sentence eliding the verb. I took (b) — "May this pleasantness after bitterness receive thee, this rest after labour, this grace the temptation" — because it is the only reading on which every printed word has a slot, and because Origen's argument is that grace takes up the temptation that the refreshment interrupts. The likeliest emendation, haec gratia post tentationem, is not taken into the text: it is a conjecture, it is not on either plate, and the printed accusative is intelligible without it.
• @0436B ipsius poenitentiae lacrymis uritur daemon → "the demon is burned by the tears of that very penitence". ipsius could be possessive of the fornicator ("of his own penitence"); I took the intensive with poenitentiae because the sentence's whole point is the penitence, not the penitent.
• @0429C optima quaeque virtutum merito percipiunt → "receive by merit whatsoever things of the several kinds are best". virtutum could be "of the virtues" (the reward being virtues) or partitive with optima quaeque; I took the partitive, because the sentence is about dividing spoils and merito already carries the virtue.
• @0434D pensanda sunt … ex persona dicentis → "from the person of the speaker". persona is not the theological persona here and is not capitalized.

7. Vocabulary axes — checked, with the instrument checked first

• daemonium → demon, diabolus → devil. ⭐ The pressure point is @0436A, where Migne quotes Ps 95:5 Omnes dii gentium daemonia and the Douay reads "devils". Rendered "demons", per axis 3, and the same word is kept eight lines later at uritur daemon. diabolus occurs twice (@0433B diabolica castra, @0438B tentabatur a diabolo) and is "devil" in both. There is no satanas in the range.
• gentes family. gentium @0436A is Ps 95:5's dii gentium → "of the Gentiles", the standing English of that verse and the typological sense here (idolaters against the God of Israel). @0438C Ecclesiam de gentibus congregatam and plenitudinis gentium are squarely the Judaei/gentes opposition — "the Church gathered out of the Gentiles", "the fulness of the Gentiles". ⚑ Instrument checked, and it fired exactly as the brief warns: a raw natio grep returns one hit in my four chunks and it is indignationis @0430C — no natio stands anywhere in my range, so no "nation" appears in my English and none should. A bare gentes grep likewise returns one hit, legentes @0435B; the real gent- population is gentium @0436A and @0438C, and gentibus @0438C.
• turba/multitudo: turba does not occur in my range at all (and \bturba returns nothing here, not even a turbare form). multitudo occurs once, @0429A Est ergo multitudo laudabilis → "There is therefore a praiseworthy multitude". Axis 2 holds trivially, since only one of the pair is present.
• mysterium/sacramentum: both occur and are kept 1:1. mysteriis idoli @0427D → "the mysteries of the idol"; mysteria coelestia @0428C → "the heavenly mysteries"; sacramenta consecrationis @0430C → "the sacraments of consecration"; sacramenta pulsare @0437C → "to knock at the sacraments".
• Commentator's voice not archaised. dicit / ait / inquit → "says", never "saith"; -eth forms appear only inside quoted scripture (healeth, hath), where the italic span and the [n:] citation both mark the quotation.
• mansio → station. Fixed against the work's own usage for the numbered stations: the formula "the nth station" already stands 12 times in the shipped chunks (0005 ×1, 0008 ×5, 0009 ×2, 0014 ×3, 0015 ×1) and 14 more times in 0023 and 0024, the chunks that continue this very catalogue. "Encampment" is used 16 times, but only in the ARGUMENTUM (0000, 13) and at 0013 (3), where it renders the journey's stages in summary rather than a numbered station. My range carries ten, including Nm 33:1's lemma Hae sunt mansiones → "These are the stations". ⚑ The one place this costs something is @0435A sed per multas mansiones, which echoes Joan. 14:2's mansiones multae; "mansions" would have caught the echo and broken the chapter's own word. Within a work, one voice: "through many stations", and the echo is recorded here.
• profectus/profectio. profectiones et mansiones @0435B → "goings forth and stations"; ordo profectuum @0437C, @0438D → "the order of the goings forward". Kept distinct from proficisci → "set forward", which is the catalogue's own verb.

8. The three tests, and the fourth

1. Marker-strip read-back. Every [var:] and both [ed:] deleted, and each host sentence read aloud. Nothing is stranded: [var:] never wraps Migne's words, and neither [ed:] stands in for a rendering — the sentence at 0427B reads "By the interpretation of the Hebrew names, we find that Sittim is said in our tongue to be answer or refutation" with the marker gone, and the sentence at 0437C reads "In Hebrew Daphquah, that is, a knocking." Both are complete clauses with their verbs, their objects and their conjunctions.
2. Negation attachment, host by host. The dense sites are @0430C (non elegit … non … nec rogat, three hosts, three negatives, handled in §5), @0430D (non enim res hujusmodi quasi partiliter divisae minuuntur — the non is on the verb, and the correlative sed totae sunt proves it), @0433A (ut non nisi rationi motio illa consentiat — both words carried, "consent to nothing but reason"), @0433B (qui negant … nec fidem … recipiunt — carried as "nor receive"), @0435B (nec segniter … nec spatium viae expleamus — carried), @0436B (Non enim sine ipsis consummatur peccatum — the negative governs the prepositional phrase, not the verb, and the English keeps it there: "For sin is not accomplished without them"), and @0435C (Non dixit in hac nocte, nec in hac die, sed hac nocte — the whole point is the preposition, and the negatives sit on it). ⚑ Test 2a's trigger (a negative on a verb with a quantifier or correlative nearby) fires at @0428A non solum idololatriam, sed et omnem carnis affectum — the non governs solum, and the English reads "not only idolatry, but also every affection", with the sed et pivot intact.
3. Punctuation by column band, not by total. Question marks: 1/1 in 0019 (@0427C), 1/1 in 0020 (@0430D), 0/0 in 0021, 5/5 in 0022 (@0435D ×3 counting the Coloss. quotation, @0436B, @0437C) — band for band, no mismatch. Column anchors 10/10, 10/10, 10/10, 16/16, verbatim and in order, and each checked against the word it stands beside, not merely for presence (test 2b): the two that fall mid-construction are there was not among [0435B] their tribes one that was feeble (Latin: in [0435B] tribubus) and he keeps a woman for spoil [0433D] (Latin: in praedam [0433D] de bello). [n:] notes 20/20, 21/21, 21/21, 27/27. Paragraph counts 26/26, 35/35, 34/34, 33/33.
4. Does it ASSERT what he asserts? Read back paragraph by paragraph with the counts already clean. Three places where the smooth English was the suspect and was changed back: the 0434B run-on (§3), where an inserted colon read better and was removed; @0437A, where "let this grace receive thee after temptation" read beautifully and silently supplied a post (§6); and @0436D promptior reddatur ad aliquos labores, where "readier for further labours" is the obvious sense and the Latin says some — kept as "some labours", because Origen's next station is Alus, labores, and aliquos is doing work.

Zero em-dashes in all four chunks.

9. Sections read clean, and candidates rejected

• Chunk 0020 is the clean one: zero markers of any kind in 1,000 Latin words, cols 0430A–0432B. It was read twice for that reason. Its lemmata are Nm 27–29's sacrificial calendar and every one of them matches the Clementine exactly (§1a); its two dense paragraphs (the Augustinian quaest. 53 at 0430A–C and Origen at 0430D–0431A) carry the range's other unexpressed-subject sites, both named in §5.
• Candidates rejected: the twenty-seven archive readings in data/briefs/8963-PENDING-TEI-PATCHES-0019.md Part B, each refuted at Gallica and each named there with both readings. Also rejected, with reasons, at §1c (eight declined divergences), §2b (a [cj:]), and §3 (a patch on CAPUT XXXI, which is a press variant).
• Not rejected, escalated instead: the two Hebrew words. They are Migne's, they are lost from our Latin, and the codepoints are not transcribed by me — the glyph count at 0427B is three where Shittim would want four, and reading non-Latin type by eye is exactly what reference_paleography-gt-ingest forbids. The [ed:] says a word is missing and supplies nothing; the pending-patches file records the two leaves and asks for a 1200 dpi render.

Anselm of Laon and his school, Glossa ordinaria on Numbers. This stint closes the book: chunk 0025 ends at 0446D, where LIBER DEUTERONOMII begins on the same plate.

0. Span count against the lemma brief

data/briefs/8963-lemmata-0023.txt states 102 spans for chunks 0023–0025. Counted paragraph-bounded in my own Latin: 0023 = 50 · 0024 = 38 · 0025 = 14 = 102. No unmatched * delimiter in any paragraph of any of the three files. Guillemets: 0 in my range, as the launch brief says for this book. The totals agree.

⚑ MATCHES ONLY ELSEWHERE: zero spans in my range. The string occurs in my brief only in the legend that explains the mark; no inventory line carries it. Nothing to adjudicate.

1. Plates read — pp. 225, 226, 227 (unread before this stint) and p. 228 (re-read)

Corner numbers read FIRST on every page, before anything else on it:

p.225 → 439 / 440 p.226 → 441 / 442 p.227 → 443 / 444 p.228 → 445 / 446

confirming page = (column + 11) / 2 for patrologiaecurs04migngoog.pdf. Gallica calibrated the same way: f220 corners 439/440, so f = (column + 1) / 2 holds. Both maps verified, not assumed.

⛔ FOOT-OF-PAGE APPARATUS: NONE, ON ALL FOUR PAGES, BOTH COLUMNS. Checked zeroes, recorded in data/briefs/8963-PLATE-READS-0023.json. ⭐ p.228 is the page where the work division falls mid-page — the shape CLAUDE.md warns sets notes mid-page — and there is no note there either: the rule under the last line of Numbers is followed directly by the LIBER DEUTERONOMII head.

2. ⭐ Second-witness tally — 15 archive-only candidates, 14 refuted at Gallica

Every letter- and word-level candidate the archive copy produced was taken to Gallica ark:/12148/bpt6k5505319w, leaves f220–f223. Fourteen fell. One stood, and it was ours.

⭐ This range corroborates the run's governing finding from the other side. Our TEI agrees with Gallica at every one of the fourteen refuted sites — and at 0444B, where the two printings really differ, our TEI again follows Gallica. The one site where our TEI stands alone (0440C) is the one where the two printings agree with each other.

⛔ And it reproduces the "never let one adjudication settle the next" warning exactly. Thabath for Thahath was raised at 0440B and 0440C on the same column, from the same archive page, on the same evidence. The first is ink and the second is real. Had I settled 0440C from the 0440B refutation I would have silently kept a reading neither plate carries.

⚑ Three of the fourteen refutations were words the archive lost, not letters it changed (in @0442D, atrio @0445B, and the i of XVI). That is the direction the withdrawn "direction rule" once called safe on one witness; here it was wrong three times out of three.

3. The one divergence I judge to be OURS — 0440C Thahath for Thabath

Filed in data/briefs/8963-PENDING-TEI-PATCHES-0023.md.

Migne sets the VERS. 26 lemma Thahath (both witnesses; = Vulg. Nm 33:26) and then, eleven lines below in the Origen gloss, Thabath (both witnesses). Our TEI prints Thahath in both places — the digitization tidied Migne's second form into agreement with his first and with the Vulgate. This is brief §4's normalization class, caught only because the plate was open.

Under translation-style.md "THE PLATE WHERE READ", the English renders Thabath at 0440C — so the English chunk deliberately reads Thahath at the lemma and Thabath in the gloss, exactly as Migne prints. No [sic:]: the type is not defective (it is a variant transliteration of a Hebrew name), and a [sic:] would in any case fail the verbatim-substring guard, since the word it would accuse is not in our Latin.

4. Markers fired — eight, all [var:], all on read columns

⛔ No [sic:], no [cj:], no [ed:], no [d:] in this stint. Fourteen of my fifteen candidates were the archive copy's ink; the fifteenth is our file's and takes a patch, not a marker. Zero [sic:] in 3,109 Latin words is what this class looks like when it is measured.

Each [var:] stands beside a faithful rendering of what Migne prints; none replaces one. Columns: 0439B · 0441B · 0441D · 0442A · 0442B · 0444A · 0444C · 0444C — all four pages read.

1. @0439B si fuerint peccata vestra sicut Phoenicium, ut nivem dealbabo (Is 1:18). Clementine: Si fuerint peccata vestra ut coccinum, quasi nix dealbabuntur. Migne prints an Old Latin form: Phoenicium (Phoenician purple, not "scarlet") and an active dealbabo, "I will whiten", against the Vulgate's passive. ⚠ This is the exact site of one of Ruth's four measured silent repairs (Phoenicium rendered "as scarlet"); it is deliberately not conformed here.
2. @0441B Si spiritus potestatem habenti ascenderit super te, locum tuum ne dimittas. Clementine Ecl 10:4: potestatem habentis … locum tuum ne demiseris. Migne's habenti is a real Latin form (dative participle) with no construction in the sentence, so Pattern 7 governs and it is rendered flat — "for him that hath power" — not silently repaired to the genitive the Vulgate has. Both witnesses print habenti. Migne's own siglum reads (Eccli. X) (Ecclesiasticus) for a verse of Ecclesiastes; that is his, and rides as printed.
3. @0441D Maledictus qui facit opus Dei negligenter et prohibet gladium suum a sanguine (Jer 48:10). Clementine: opus Domini fraudulenter, et maledictus qui prohibet… Two substantive changes (Dei for Domini, negligenter for fraudulenter) plus the loss of the second maledictus qui. negligenter is the material one and is what the marker names. Both witnesses.
4. @0442A Pastor enim, bonus ponit animam suam pro ovibus suis (Jo 10:11). Clementine: Bonus pastor animam suam dat pro ovibus suis. ponit for dat. ⚑ Migne's comma after enim is his and is carried into the English ("a shepherd, a good one, layeth down…"); it looks like a compositor's slip and both witnesses print it.
5. @0442A–B Transiens, videbo visionem hanc magnam (Ex 3:3). Clementine: Vadam, et videbo… — and the divergence is load-bearing*: the gloss is expounding transitus and calls the speaker verus Hebraeus, id est transitor*. Conforming this lemma would have destroyed the argument it is quoted for. Both witnesses.
6. @0444A Venerunt ad montem Abarim contra Nabo (Nm 33:47). Clementine: montes Abarim. Migne's singular sits against his own gloss two lines down (vertitur in montes transeuntium, mansio transeuntium in montibus sita est). Both witnesses print the singular.
7. @0444C Versatus sum in miseria, dum mihi configitur spina (Ps 31:4). Clementine: conversus sum in aerumna mea, dum configitur spina — an Old Latin / Roman-Psalter shape, with mihi added. Both witnesses.
8. @0444C Beatus qui tenebit et allidet parvulos suos ad petram (Ps 136:9). Clementine: parvulos tuos. Material: the gloss has just made the little ones belong to the malignant spirits, which is what suos gives and tuos does not. Both witnesses print suos.

5. 7a″ — every marked span collated, INCLUDING the ones that agree

⭐ The list below is the deliverable, not the markers. Spans that agree with the Clementine verbatim are recorded as such; a list of findings alone cannot be told apart from a report by an agent that never looked.

Agreeing exactly with the Clementine (checked, no marker, no crux owed): Nive dealbabuntur in Selmon (Ps 67:15) · Domine, quis habitabit in tabernaculo tuo? (Ps 14:1) · Idoneos nos fecit ministros Novi Testamenti (2Cor 3:6) · Mihi mundus crucifixus est (Gal 6:14) · Virga tua et baculus tuus, ipsa me consolata sunt (Ps 22:4) · Quam dulcia faucibus meis eloquia tua (Ps 118:103) · Dirupisti vincula mea (Ps 115) · Nolite locum dare diabolo (Eph 4:27) · Afferte Domino filios arietum (Ps 28:1) · In matutino interficiebam omnes peccatores terrae (Ps 100:8) · Et non dixerunt qui praeteribant, Benedictio Domini super vos (Ps 128:8) · Quando autem factus sum vir, evacuavi quae erant parvuli (1Cor 13:11) · Commovebit Dominus desertum Cades (Ps 28:8) · Qui parce seminat, parce et metet… (2Cor 9:6) · Ne polluatis terram habitationis vestrae (Nm 35:33) · Attendite ne justitiam vestram faciatis coram hominibus (Mt 6:1) · Cui vult nubat, tantum in Domino (1Cor 7:39) · Praeterit enim figura hujus mundi (1Cor 7:31) · the Numbers station lemmata In Rethma · In Remmomphares · In Lebna · In Ressa · In monte Sepher · In Arada · In Maceloth · In Thahath · Castrametati sunt in Thare · In Salmona · Transieritis Jordanem · Disperdite cunctos · Pluribus dabitis latiorem · Ibuntque confinia · Haec erit terra · Decernite quae urbes · Postquam autem ille obierit · Accesserunt autem et principes.

Diverging, but NOT materially — crux only, no [var:] (a [var:] is a public claim and Pattern 14 reserves it for divergence that changes the sense):

• @0439A Multitudinis credentium erat cor unum et anima una — Act 4:32 has Multitudinis autem credentium. A dropped connective.
• @0440A Ecce quam bonum et quam jucundum, habitare, fratres, in unum — Ps 132:1 has habitare fratres in unum, i.e. only commas differ, and our TEI does not carry them at all (Gallica prints fratres, with the comma, the archive without: press variant, §2).
• @0440B Noli altum sapere, sed sapere ad sobrietatem [Rom. XII] — a conflation, not a variant: Noli altum sapere is Rom 11:20 (which continues sed time) and sapere ad sobrietatem is Rom 12:3. Migne's own siglum says Rom. XII. There is no single received verse to set against it, so no [var:] is possible; the note is the whole record. ⚑ The brackets round [Rom. XII] are Migne's own, printed in the running line, and pass through as printed.
• @0440C Ego enim ostendam ei quanta oporteat eum pati pro nomine meo — Act 9:16 has ostendam illi quanta oporteat eum pro nomine meo pati. Pronoun and word order; sense unmoved.
• @0441A Sinistra ejus sub capite meo et dextera ejus amplexabitur me — Ct 2:6 has Laeva ejus … dextera illius. Sinistra for Laeva is a genuine Old Latin lexical divergence, but both words are "left hand" and nothing in the gloss turns on it. Both witnesses print it.
• @0442A Domine, in te speravi non confundar in aeternum — Ps 30:2 has In te, Domine, speravi; non confundar… Word order only.
• @0442C Aqua alta consilium in corde viri — Pr 20:5 has Sicut aqua profunda, sic consilium in corde viri. alta for profunda is the same sense; the correlative is abridged.
• @0443B Corde creditur ad justitiam — Rom 10:10 has Corde enim creditur.
• @0443D Inter nos et vos magnum chaos firmatum est — Lc 16:26 has inter nos et vos chaos magnum firmatum est. Word order.
• @0445C–D Si quis scandalizaverit unum ex pusillis istis qui in me credunt — Mt 18:6 has qui autem scandalizaverit unum de pusillis istis.
• @0445D Omnis qui oderit fratrem suum, homicida est, nec habet vitam aeternam — 1Jo 3:15 has omnis qui odit fratrem suum, homicida est. Et scitis quoniam omnis homicida non habet vitam aeternam in semetipso manentem. The second clause is the recension's own abridgment, not a variant reading; oderit for odit has no English exponent (Pattern 9's class).
• @0446C Omnes enim fratres sumus. ⚠ Declined deliberately, and here is why. The nearest received text is Mt 23:8 omnes autem vos fratres estis — second person, not first — but Migne cites three places at once (Joan. X; Matth. XXIII; II Cor. XI), and sumus stands in none of them. A [var:] would have to assert which verse is being diverged from, and I cannot show that. Addendum B: the crux is always available and always safe. The first person is the clincher of Rabanus' argument about schism and is rendered as printed.

Non-scriptural spans checked in place (the 19 italic usque ad, Montem. · Habitatores. · Jetebathae. · Reus, · Nubant. · Audivitque. · Locutus est ergo Dominus ad Moysen · omnis qui procedit in virtute Israel · contra fontem): none is a scripture quotation to collate. All 19 italic usque ad in my range follow an etc. and are the abridgment formula, rendered "as far as" without exception — the etc. splice test returns 19 for 19, and no look-alike hides inside an italic span. The instrument was checked before this was reported: a raw usque` grep over the three files returns 29, of which only 19 are the formula. The other ten are ad septimum usque pervenias @0441B (postposed usque, "as far as the seventh"), pervenientes usque ad montem altissimum @0445A and tendunt fines usque ad Sephama @0445B — three ordinary, unitalicised uses rendered by sense — plus eousque @0440B, usquequo @0441C and @0442B, usque eo @0441D, and Hucusque @0442B and @0443C, which are different words and must not be swept in with the formula.

6. Renderings, subjects supplied, and things a blind reader will stop at

• The station lemmata. In X is rendered "At X" throughout (At Rethma, At Remmomphares, At Lebna, At Ressa, At mount Sepher, At Arada, At Salmona, At Dibongad, At Asiongaber), In finibus as "In the borders" and In desertum as "Into the desert". ⚑ Flagged for the merge: chunks 0022 (another stint) carries the same construction from VERS. 5 onward and may have chosen differently; this note exists so the merge can normalize rather than guess.
• @0441B habenti. Rendered "for him that hath power" — a flat dative, which does not construe and is not meant to. The Vulgate genitive is in the [var:] and nowhere in the running text.
• @0442A Jetebathae. Migne prints the lemma in a genitive/dative form where the station is Jetebatha (Nm 33:33) and his own gloss two words later reads Jetebatha interpretatur bona. A proper name's case ending has no English exponent (Pattern 9's class), so it is rendered "Jetebatha." and logged here rather than marked.
• @0441A Seman. The lemma is Hesmona (VERS. 29) and the Origen gloss beneath it says Inde ad Seman venitur. Migne's; rendered as printed, no reconciliation attempted.
• @0440D Methcha against the Vulgate's Methca (Nm 33:28); Migne's spelling, both witnesses, rendered as printed.
• @0442B decimas octavas continet mansiones. Rendered literally, "contains the eighteenth stations". It does not read well; decimas octavas is an ordinal where the sense wants eighteen, and "contains eighteen stations" would be a silent repair of Migne's own Latin. Both witnesses.
• @0442D–0443A Alius enim in valle, alius in campis, alius moritur in monte…. The verb moritur stands only in the third member and governs all four. English cannot open a series with no verb, so "dies" is placed in the first member and the printed position is recorded here.
• @0443A Aaron, plangitur,. Migne prints a comma between subject and verb. An English comma in that slot is simply an error rather than a preserved oddity, so it is not carried; the printed comma is recorded here instead. Every sentence-level mark in the range is carried 1:1 (colons 33/33, semicolons 13/13, question marks 2/2, exclamation 1/1 on chunk 0023; 12/12, 8/8 on 0024; 4/4, 3/3 on 0025 — counted with markers stripped).
• @0446A VERS. 28. printed BEFORE VERS. 26. — verified on the plate, both columns of p.228. Migne's ordering, reproduced as printed and not reordered (and VERS. 26. in fact addresses Nm 35:33). Nothing machine-readable consumes a VERS. number.
• @0446A quia humanum genus peccando sibi mortem intulit post mortem veri sacerdotis … accepit — two finite verbs with no connective and no mark between them (plate confirms: no comma after intulit). The English carries the asyndeton rather than supplying a semicolon.
• @0445D — a full stop our TEI lacks. The archive plate prints … nec habet vitam aeternam (I Joan. III). Hic secundum Exodum …; our TEI runs the two sentences together with no mark. The English carries the stop, because English prose cannot run them together, and the divergence is filed in the pending-patch file — ⚠ marked there as resting on one witness, since I did not fetch Gallica for a full stop and a period is exactly the mark ink adds and removes most readily.
• @0446D complacet; sibi solemnitatem agit. Migne's semicolon between the two halves of one clause; carried, per Pattern 8, though it reads oddly.
• Subjects supplied. venitur / applicatur impersonal passives are rendered "one comes", never "he comes", except where Migne himself has a personal subject (venit in Gadgad @0441D, applicuerunt @0442B). At 0444A–B the subject of transit … abscedit … non invenitur is anima from the Ubi enim … anima transierit clause above, so the English carries "it", not "he" — the same slot that produced Ruth's worst site.
• Vocabulary axes. turba → crowd and multitudo → multitude, and they stand four words apart at 0439A (omnem turbam credentium … Multitudinis credentium), which is exactly the pair the axis exists for. gentes → nations at 0442B (multitudinem gentium, generic, no Judaei/gentes opposition in view) but → Gentiles at 0446C, where Rabanus sets credentium Judaeorum against doctori de gentibus in one sentence; gentilis → Gentile @0446B per the axis. mysterium → mystery (0441C, 0443B) and sacramentum → sacrament (0445D), 1:1. No natio, no daemonium, no ethnicus in the range. ⚑ Instrument checked before reporting: a raw \bturba grep in my range also catches nothing spurious, and natio matches only interpretatur-type strings, i.e. none.
• Register. The commentator's own verbs are not archaised (dicit → "says", ait → "says"); -eth forms appear only inside quoted scripture (layeth, hateth, soweth, doth). thou/thee/thy renders singular address throughout (0439C Vides, 0440B–D Venisti · ascendisti · confiteris, 0441B super te · locum tuum, 0441C fueris eruditus · scieris, 0441D–0442A transieris · applicabis, 0442B venias · te); you renders the plurals at 0444C (Transieritis · dabitis), 0446A (Ne polluatis) and 0446B (Attendite … faciatis).

7. Negations — named host by host (test 2 / 2a)

Every printed negative in the range, with the single Latin word it attaches to:

non superbiat @0440B → superbiat ("may not be proud") · nec … vidit, nec … audivit @0439D → the two verbs · nec ab isto unquam recedit @0440A → recedit, and both negatives are carried ("nor does he ever depart"), not collapsed into one English "never" · non debemus esse otiosi @0440D–0441A → esse otiosi · ne … se ingerat @0440C → ingerat · nec minimum quidem … praetereat and nec unum iota, aut unus apex … otiosus @0441C → praetereat and habeatur · Non enim possumus … nisi … interfecerimus, nec parcat manus nostra @0441C–D → possumus, then parcat · nec decorae esse nec plenae @0441D → the two predicates, not the verb · non nisi post tentationum experimenta venitur @0442A → venitur · non confundar @0442A → confundar · Et non dixerunt @0442B → dixerunt · non ponuntur @0442B → ponuntur · nec aliter expedit @0442C → expedit · non desperemus salutem, sed iterum pugnemus @0443A → desperemus, and the sed correlative confirms the host · quae non licet homini loqui @0443B → licet · ne audiamus @0443C → audiamus · quamvis non invenerimus … non dubitamus @0443C → invenerimus, then dubitamus (⚠ the archive plate prints nou for the first of these; refuted at Gallica, §2 — had it been believed, the English would have lost a negative in a concessive clause where nothing downstream would have caught it) · nisi contempta fuerint … non possumus @0444A → contempta fuerint, then possumus · qui non inveniebatur and non invenitur in ullo saeculari actu @0444A–B → inveniebatur, invenitur · non invenitur ejus memoria @0444B → invenitur · sine fine perseverat @0444B → perseverat · Nihil enim tam periculosum @0444B → periculosum · ne … pervadantur @0446B → pervadantur · ne … perdatur @0446B → perdatur · ne … detrimentum patiatur @0446C → patiatur · Non debet esse schisma @0446C → esse schisma · nec habet vitam aeternam and nec revocatur … nisi @0445D–0446A → habet, revocatur · Et nisi … occidantur … nullo modo mundabitur, nec … habitabit @0446A–B → occidantur, mundabitur, habitabit.

⚑ Test 2a trigger population (a negative on a verb with a quantifier or correlative adjacent) fires at three sites in the range, and all three were proved from the structure, not from what read well: nec decorae esse nec plenae @0441D (the correlative nec … nec puts both negatives on the predicates, not on videantur); nec minimum quidem mandatum praetereat @0441C (the negative is on the verb, and quidem intensifies minimum — "not even the least"); non nisi post tentationum experimenta venitur @0442A (the non nisi is a unit governing the whole temporal phrase, not the impersonal verb alone, hence "not save after").

8. Test 4 — does the English ASSERT what he asserts?

Read back paragraph by paragraph with the counts already clean. Two places where a smooth English sentence would have said something Migne does not, both left rough on purpose:

• @0441B the Ecclesiastes citation (§4.2): the smooth reading is the Vulgate's genitive, and the smooth reading is the one that is not on the page.
• @0442B decimas octavas … mansiones (§6): "eighteen stations" is the smooth reading and is a repair of Migne's ordinal.

And one place where the printed text is odd and is not a defect: @0442D alius in monte, alius in monte montis — "on the mountain of the mountain" is Origen's own figure for a further ascent, not broken type; both witnesses, no marker.

9. What I did NOT do

• No [sic:] anywhere. Fourteen of fifteen candidates were the archive copy's ink and the fifteenth is our own file's; neither class licenses a public claim against Migne's type.
• No conjecture ([cj:]). No site in the range meets Pattern 18's bar: nowhere does a real printed word give faithful English that asserts something the author did not. habenti @0441B comes closest and is handled by [var:], because the rival reading is a received text, not a conjecture of ours.
• No [ed:]. Nothing in my range is text the digitization lost that the plate carries, apart from the two punctuation sites in the pending-patch file, which are patch questions, not holes.