Four Words and Five

• Anchor Bible Dictionary — “Hosea, Book of,” “Jacob,” “Admah,” “Sheol,” “Ephraim.”
• Dictionary of the OT: Prophets (IVP) — “Hosea: Book of,” “Israelite History,” “Death and the Afterlife.”
• Dictionary of NT Use of the OT (Beale & Carson) — on Matthew 2:15, Luke 23:30, 1 Corinthians 15:55, Hebrews 13:15, Revelation 6:16.
• NIDOTTE — שְׁאוֹל, שׁוּב, חֶסֶד, פַּר.

The previous Hosea study had no offset at all and the one before it turned on one. This study has two.

Chapter counts — Masoretic and CATSS: 17 / 15 / 11 / 15 / 15 / 10. ESV: 17 / 15 / 12 / 14 / 16 / 9. Eighty-three verses either way, and the Hebrew book has a 14:10 where the English book stops at 14:9. Established from content at each seam rather than assumed: Hebrew 12:1 and English 11:12 both begin Ephraim has surrounded me with lies; Hebrew 14:1 and English 13:16 both begin Samaria shall bear her guilt.

Hosea carries three offsets in all. The first, at Hebrew 2:1, was the subject of the first study on this book; these are the other two. Seven of the twelve Minor Prophets carry an offset of this kind.

All 83 CATSS rows of this study against all 7,927 verses of the Greek New Testament.

Every row reaches two, so the two-band carries no information at all; only the upper bands separate anything. The single row reaching five is Hosea 12:6, and on the New Testament side it is a three-way tie the screen does not separate — Revelation 4:8, 19:6 and 21:22 each carry the identical five forms, which the CATSS row reads as κύριος ὁ θεὸς ὁ παντοκράτωρ.

⚠️ That five is a maximum within this section, and the boundary is doing work. Run the identical screen over all 197 rows of Hosea and it reaches 17, at Hosea 2:1 against Romans 9:26 — the subject of the first study on this book. Five is the ceiling for chapters 9–14, not for Hosea.

A run’s depth and a run’s rarity are not the same thing, and here they come apart cleanly.

⚠️ The denominator in that column moved on 30 August 2026 and the numerators did not. All three phrase counts were re-run with the Odes withdrawn from this shelf’s screening corpus, and no canticle carries any of the three: thirteen, one and one, exactly as before. What changed is what the screen is allowed to look at: the 64 files HOLD 30,637 verse units and the screen now SEES 30,348. The 289 units of the Odes stay on the disk. Stage 4 says why they come out.

The thirteen CATSS units carrying the five forms of 12:6 are Hosea 12:6 itself, ten verses of Amos (3:13, 4:13, 5:8, 5:14, 5:15, 5:16, 5:27, 9:5, 9:6, 9:15), Nahum 3:5 and Zechariah 10:3. The section’s deepest run is built from words that recur across the Twelve. Reporting the maximum alone would have made it the headline.

Hosea 10:8 against Luke 23:30 is the only pair in this study that wins its screen from both ends.

From the Hosea side nothing else in the New Testament reaches even three — the distribution drops straight from one verse at four to 162 at two. From the Luke side, twenty-three CATSS units reach three and exactly one reaches four. Each is the other’s single deepest partner across both corpora.

⚠️ This screen has now been re-run twice, against two corpus expansions, and the finding has survived both. The shelf has gone 35 files → 44 → 64, and 18,183 verse units → 23,061 → 30,637. On 19 August 2026 Isaiah, Jeremiah, Lamentations, Ezekiel and both Greek Daniels arrived; on 30 August 2026 Exodus, Leviticus, Numbers, Joshua (A and B) and the whole deuterocanon — Judith, Tobit, Wisdom, Baruch, the Epistle of Jeremiah, 2–4 Maccabees, Susanna, Bel, Odes and the Psalms of Solomon — arrived with them. The finding held both times: Hosea 10:8 is still the only verse unit in the whole of it that reaches four against Luke 23:30. What moved, both times, is the shoulder of the distribution.

⚠️ And then it was re-run a third time, on the same day, because one of the twenty new files should never have been in the screen — and this time a published figure moved. The Odes are the liturgical canticles appended to the Greek Psalter. Two of them are New Testament text — Ode 9 is Luke 1, the Magnificat and the Benedictus, and Ode 13 is Luke 2, the Nunc Dimittis — and ten more are second copies of passages the corpus already holds (Ode 1 = Exodus 15, 2 = Deuteronomy 32, 3 = 1 Samuel 2, 4 = Habakkuk 3, 5 = Isaiah 26, 6 = Jonah 2, 7 and 8 = Daniel Th 3, 10 = Isaiah 5, 11 = Isaiah 38). The files stay on the disk and come out of the screen: 30,637 held, 30,348 screened.

This page is a clean demonstration of the second reason, and it is worth reading slowly. The Luke 23:30 two-band fell from 495 to 492, and the three rows that left are Odes 2:41, Odes 2:43 and Odes 7:33 — which are Deuteronomy 32:41, Deuteronomy 32:43 and Daniel (Th) 3:33, all three of which were already in the band under their own shelfmarks. The screen was counting three verses twice. Not a wrong reading; a doubled one. The 495 printed here until 30 August 2026 was inflated by exactly that much, and the honest figure is 492. Nothing else on this page moved: the four-band is still Hosea 10:8 alone, the three-band is still twenty-three and still carries the same members, and the three phrase counts are still thirteen, one and one.

The first reason did not bite here, but the reader should know the instrument can do it. Screened with the canticles in, Ode 9:48 matches Luke 1:48 at eighteen consecutive words — deeper than anything on this page by a factor of four, and it is Luke measured against Luke under a Greek Old Testament shelfmark. A page reporting that as a Gospel’s deepest Old Testament contact would be measuring its own text. It was caught by asking which file each new hit came out of, not by any number looking wrong. A run of eighteen looks like a discovery until you open the file.

The three-band ran 13 → 19 → 23. The 2026-08-19 additions were Ezekiel 6:3, 36:4 and 36:6, Daniel (OG) 9:12, and Daniel (Th) 9:12 and 9:13. The four added on 2026-08-30 are 2 Maccabees 8:18, Baruch 2:1, Psalms of Solomon 8:27 and Tobit 11:14 (S recension) — every one of them a witness, not Scripture, and not one of them in the books of Moses that arrived the same day. The two-band ran 266 → 347 → 495 → 492 — the last step is the Odes coming back out, and it is the only figure on this page that the withdrawal moved. Every figure in the table above is measured over the screened 30,348, not the held 30,637.

What the measurement is, and is not. It counts contiguous runs of normalised word forms, Greek against Greek, on the files held here — 30,637 CATSS verse units across 64 files, of which 30,348 are screened (the 289 units of the Odes are held and not screened), and 7,927 verses of the SBL Greek New Testament. It says nothing about which text is prior, nothing about citation, and nothing at the level of lemmas or synonyms. Isaiah, Jeremiah, Lamentations, Ezekiel and both Greek Daniels joined the shelf on 19 August 2026 and every Greek-side count on this page has been re-run across them; what the shelf holds is 39 of the Septuagint’s books, not all of them, so no count on this page is a statement about the Greek Old Testament as a whole.

These files give both a surface form and the edition’s own normalised form, and the difference between them is enormous before this screen touches it.

Nearly all of the raw difference is accentuation, which the screen removes anyway. What survives the fold is 2,547 word rows in 124 shapes — led by οὐκ/οὐ at 825, then a class of restored elisions (ἀλλ’/ἀλλά 213, δι’/διά 149, ἐπ’/ἐπί 143), with ἐξ/ἐκ at 234.

On the axis this page reports — how deep each of the 83 Hosea rows gets — nothing moves.

The same 48 rows reach three, the same 10 reach four, the same single row reaches five, and Hosea 12:6’s three-way New Testament tie is the same three verses on either column.

Counted any other way, it does move, and the page says so rather than claiming a stability it did not measure. Of 478,516 row-verse pairs, 5,292 carry a different run on the two columns — 2,091 deeper on the surface column, 3,201 deeper on the normalised one. Fourteen of those movers touch a run of three. Counted by New Testament verse, 357 verses reach three on the surface column against 348 on the normalised one.

One figure printed on this page is among the movers. Hosea 11:1 against Matthew 2:15 is 2 on the surface column and 1 on the normalised one, because the CATSS row reads ἐξ and the normalised column reads ἐκ. It is neither the only mover nor the largest: Hosea 11:3 against Matthew 24:39 and against Luke 2:43 each fall from three to one, and Hosea 11:7 against Revelation 22:18 rises from two to three. This page reports the surface column throughout, which is the column that answers the question being asked — what word forms stand in the text.

Scope, twice: this is measured on the 64 CATSS files and the 27 Greek New Testament files held here and on those only, comparing two columns of the same edition’s own rows. The normalised column is that edition’s editorial normalisation. Nothing here is a statement about what any writer wrote, or about which reading is earlier.

The whole sentence stands on both sides, and the run still stops at four. The columns show why.

Both address mountains first and hills second. The two imperative clauses stand in opposite orders to each other. And only one of the two imperatives is the same word form on both sides: καλύψατε is identical, while the other reads πέσατε in the CATSS form column and πέσετε in the New Testament surface column — same lemma πίπτω, same parse, different form. So the second clause could not have joined a run in any case.

The run of four, ἡμᾶς καὶ τοῖς βουνοῖς, straddles the join: the last word of one clause, the connective, and the two words opening the next. Four words, and no pair in either corpus does better from either side.

Revelation 6:16 carries the sentence too, and shares two. Three different runs tie for it — τοῖς ὄρεσιν, ἐφ' ἡμᾶς and ἡμᾶς καί — and it adds καὶ ταῖς πέτραις, and to the rocks, which stands in neither Hosea column.

This is the most-discussed verse in the study, and the three readings can simply be laid side by side.

The Matthew ending matches the Hebrew in number and person, and differs from the CATSS row in both. The nouns differ lexically and in gender as well — בֵּן, neuter τέκνον, masculine υἱός — while the case is accusative in both Greek rows. The contrast in person sits on the possessive alone: both Greek finite verbs are first-person singular.

How each column was read. The two Greek statements rest on the files’ own tagged parse and lemma columns — the CATSS object is tagged accusative plural, the Matthew object accusative singular. The Hebrew number and person are read off Masoretic pointing, because no tagged Hebrew morphology corpus is held here. And what follows from the three readings — which text anyone was working from, and why the row stands as it does — is not something these files can settle. The page reports the columns and stops.

Hosea 13:14 and 1 Corinthians 15:55 — longest run two, and two runs tie for it. Both ask ποῦ twice; both name θάνατε; both carry κέντρον. Three things break the runs: the possessive stands before the noun in one and after it in the other; the second address is ᾅδη in the row and θάνατε again in Paul; and where the row has δίκη, Paul has νῖκος. σου θάνατε and τὸ κέντρον both reach two, and the page names both rather than picking one.

Hosea 14:3 and Hebrews 13:15 — longest run two, and it runs the other way. The Hebrew ends וּנְשַׁלְּמָה פָרִים שְׂפָתֵינוּ, which the ESV renders we will pay with bulls the vows of our lips. The CATSS row ends ἀνταποδώσομεν καρπὸν χειλέων ἡμῶν. Hebrews 13:15 has καρπὸν χειλέων. At 11:1 the New Testament wording sat with the Hebrew; here it sits with the Greek row. Across the five passages examined the pattern does not run one way, and the page does not tidy that into a trend.

It finds runs of identical word forms and nothing else. A passage that inflects a word differently, reorders a phrase, substitutes a synonym, or works from a Hebrew text registers as a break or as nothing. All four of those happen in the five passages examined above, which is why the numbers are small — and small numbers here are a fact about the test, not a verdict on the passages.

The maxima on this page are local, and both boundaries are load-bearing. Five is the ceiling for these 83 rows; the same screen over all 197 rows of Hosea reaches 17. And the “unique in both directions” result for Hosea 10:8 and Luke 23:30 is scoped to the 30,348 CATSS verse units screened here and the 7,927 New Testament verses. That scope was widened on 19 August 2026 and the result was re-run against the wider one. When this page was first published the Greek Old Testament here did not include Isaiah, Jeremiah, Ezekiel or Daniel, and the page said so. All of them are on the shelf now, with Lamentations and a second Greek Daniel besides, and Hosea 10:8 is still the only one of 30,348 screened verse units reaching four against Luke 23:30 — though twenty-three now reach three where nineteen did after the first expansion and thirteen before it.

⚠️ That scope narrowed again on 30 August 2026, by 289 units, and one number on this page moved with it. The Odes — the canticles appended to the Greek Psalter — were withdrawn from the screening corpus the day after they arrived on it. Two of them are New Testament text and ten are second copies of passages already held. Luke 23:30’s two-band fell from 495 to 492 because Odes 2:41, 2:43 and 7:33 are Deuteronomy 32:41, Deuteronomy 32:43 and Daniel (Th) 3:33, and were being counted twice. The four-band, the three-band and all three phrase counts are unchanged. 30,637 held, 30,348 screened, and this page now says which it means.

And a hand-picked pair can only ever return one number. Four of the five passages here were chosen because they are discussed, not because a screen surfaced them. That is why each is reported against its whole distribution — Matthew 2:15 is one of 145 New Testament verses tied at two against Hosea 11:1, and 1 Corinthians 15:55 one of 71 tied at two against Hosea 13:14. Only Luke 23:30 stands alone at its depth.

No Hebrew or Greek critical apparatus. No Qumran material — the Minor Prophets scroll from Nahal Hever bears directly on the Greek of these books and was not consulted. No Targum, Syriac, Vulgate, Aquila, Symmachus or Theodotion. For Hosea 11:1 in particular the scholarly literature is very large and none of it was consulted; this page reports three columns and stops.

And nothing about the history. The fall of Samaria in 722, the identity of Admah and Zeboiim, and the date of chapter 14 are questions a word count cannot touch.

One row in these six chapters reaches a run of five against the Greek New Testament. Its five words are κύριος ὁ θεὸς ὁ παντοκράτωρ — and the same five stand in ten verses of Amos, in Nahum, in Zechariah, and in three places in Revelation. Thirteen CATSS verse units carry them.

One word shorter, Hosea 10:8 and Luke 23:30 share four: ἡμᾶς καὶ τοῖς βουνοῖς — us, and to the hills. Screened from the Hosea side, no other New Testament verse reaches even three. Screened from the Luke side, no other CATSS verse unit reaches four. Each is the other’s single deepest partner in twenty-three thousand rows and eight thousand verses.

Had this page reported its maximum and stopped, it would have printed the wrong verse — and a phrase carried across half the Twelve would have been the headline instead of four words about mountains and hills.

Third study on the Minor Prophets shelf, and the last on Hosea — fourteen chapters, 197 verses, three verse offsets, and two studies whose results point in opposite directions.