Six Words with Samaria, Six with Caesarea

• Anchor Bible Dictionary — “Ephesus,” “Artemis,” “Asiarch,” “Magic (Greco-Roman),” “Agabus,” “Nazirite,” “Miletus.”
• Dictionary of the Later New Testament & Its Developments (IVP) — “Acts of the Apostles,” “Ephesus,” “Church Order,” “Farewell Discourse,” “Magic and Divination.”
• New International Dictionary of NT Theology & Exegesis — ἐκκλησία, ἐπίσκοπος, πρεσβύτερος, ποίμνιον, δέω, ὄνομα.
• New Documents Illustrating Early Christianity (Macquarie) — the standing collection for the Ephesian inscriptions behind νεωκόρος and Ἀσιάρχης.
• None of these were consulted for the measurements below. The measurements were run on the text files named in the footer.

Every page on this shelf states its versification offset, including the pages where the answer is that there is none. On a New Testament book the usual answer is that there is no second column to be offset from. This passage is the exception: it contains a real one-verse difference.

The SBLGNT row and the English row were counted off disk; the ESV row is the printed edition’s chapter shape, since no ESV Acts file is held here. The Greek Acts 19:40 carries two sentences: the town clerk’s warning that the city is in danger of being charged with rioting, and then καὶ ταῦτα εἰπὼν ἀπέλυσεν τὴν ἐκκλησίαν. The English Bibles print that second sentence as verse 41. So on this page Greek 19:40 = English 19:40 + 19:41, and every other reference means the same verse in both columns.

The five-verse difference in the whole-book totals is four omissions plus this one split. Counted off disk against the English text file held here, the verse numbers the English Bibles carry and MorphGNT does not are exactly five: Acts 8:37, 15:34, 19:41, 24:7 and 28:29. Four of them are readings the SBL edition prints without text because they are judged later additions, and none of those four falls in these three chapters. The fifth is 19:41 — the split described above, and the one member of the five that does touch this page.

Why it is worth the paragraph. A reader following a reference from this page into a Greek text would land on the right words either way, because the words are in Greek 19:40. But a reader going the other direction — from an English “Acts 19:41” into MorphGNT — would find no such verse and might conclude the sentence is missing. It is not missing; it is numbered differently. That is precisely the failure mode the offset rule exists to prevent, and it is pleasing to find it in the New Testament.

Acts 19:26 is Demetrius quoting Paul in order to convict him: λέγων ὅτι οὐκ εἰσὶν θεοὶ οἱ διὰ χειρῶν γινόμενοι — saying that those made by hands are not gods.

Measured against the 30,637 verse units of the Greek Old Testament held on this disk, that clause reaches five consecutive words with Epistle of Jeremiah 49, at a tie band of 1. It is the deepest Old Testament run anywhere in these 118 verses.

Counted: the clause οὐκ εἰσὶν θεοί stands in nine verses of the Greek Old Testament held here, and seven of the nine are the Epistle of Jeremiah — vv. 22, 28, 49, 50, 51, 64 and 71. The other two are Jeremiah 2:11 and 16:20. The Epistle of Jeremiah is a single 72-verse tract whose entire argument is that idols are not gods, and it repeats the sentence as a refrain.

⚠️ What this is and is not. It is not a citation. There is no formula, the surrounding sentences are unlike each other, the Epistle of Jeremiah is not quoted anywhere in the New Testament, and the sentence in Acts is not even on the lips of a believer — it is a guild leader’s summary of the preaching he wants stopped. Direction of influence is not asserted. What is reported is that the deepest Old Testament agreement in this passage is a stock polemical formula, that its home in the Greek Old Testament is overwhelmingly one deuterocanonical tract, and that a run of five whose tie band is one can still be a formula rather than a quotation. Reading the band alone would not have told you that; reading the text does.

⚠️ And this particular run rests on a normalisation decision. MorphGNT’s normalised column folds οὐκ and οὐχ into οὐ; CATSS does not. This shelf therefore applies the equivalence οὐκ ≡ οὐχ ≡ οὐ at comparison time, to both sides. Without it this run is 3 instead of 5. Both texts read οὐκ, so the fold is recovering a real agreement that two digitisations had hidden — but it is a decision, it is the decision that makes this the page’s deepest Old Testament run, and Stage 5 returns to it.

The town clerk’s speech (19:35–40) is the most technically Roman paragraph in Acts, and the measurement can only confirm how unusual its vocabulary is, not what it meant.

νεωκόρος is literally temple-sweeper and was used as an honorific civic title: the city, not a person, is the goddess’s warden. διοπετής is fallen from Zeus. Acts 19:35 is the only verse in the Greek New Testament, or in the 30,637 verse units of the Greek Old Testament held here, that contains either word. Its deepest run against the whole Greek New Testament is 2, at a tie band of 261 — which is another way of saying the sentence has no verbal relatives at all.

What was NOT checked. The Ephesian inscriptions, the coinage that carries νεωκόρος, the Artemision’s standing in the first century, the identity of the Asiarchs, the meteoritic or aniconic object behind διοπετής, and the ‘Ephesian letters’ associated with magical practice. None of that is on this disk and none of it was consulted. The reference works name it; this page measured only the Greek wording.

Acts 20:18–35 is the only speech in Acts that Paul addresses to Christians — every other set piece of his is to a synagogue, a crowd, a council or a governor. Peter and James do address believers, at 1:16–22, 11:5–17 and 15:7–21; Paul, in this book, never does anywhere else.

Ancient literature has a recognised shape for this: the departing leader gathers his own, reviews his conduct, foretells what is coming, warns of trouble after his death, and commits them to God. Jacob does it in Genesis 49, Moses in Deuteronomy 31–33, Joshua in Joshua 23–24, Samuel in 1 Kingdoms 12, and Jesus in John 13–17. The genre is real; the verbal contact is not. Measured against the Greek Old Testament, no verse of the Miletus speech reaches five, and its deepest single run is 4 — a 4 reached by seven of its twenty-two rows, among them τὸ πνεῦμα τὸ ἅγιον at 20:28, shared with Isaiah 63:10, Isaiah 63:11, Psalm 50:13 = ESV 51:11 and Susanna (Theodotion) 45 — tie band 4.

The honest form of the observation. A shared genre produces shared moves, not shared words, and a screen that only finds verbatim sequences will always report a farewell discourse as unrelated to every other farewell discourse. That is a limit of the instrument, not a finding about the text, and it is stated here so that the zeroes elsewhere on the page are read correctly.

James asks Paul to pay the expenses of four men under a vow and to purify himself with them, so that everyone will know he keeps the law.

The vocabulary is priestly: ἁγνίζω (21:24, 26), ἁγνισμός (21:26 — once in the New Testament), ἐκπλήρωσις (21:26 — once in the New Testament), and the shaving of the head. Measured against all 26 verses of Septuagint Numbers 6, the Nazirite chapter, no verse of Acts 21:20–26 reaches beyond 2 — and the two rows that do reach 2 are 21:24 (τὴν κεφαλήν, tie band 6) and 21:26 (τοῦ ἁγνισμοῦ, tie band 1, with Numbers 6:5). The first is furniture; the second is the article plus the very noun this paragraph singled out — and two words is still only two words.

⚠️ NOT VERIFIABLE FROM LOCAL CORPORA. Whether this is a Nazirite vow, how a diaspora Jew discharged one, what the seven days were, and how the payment worked are questions for the Mishnah tractates Nazir and Sheqalim, for Josephus, and for the archaeology of the temple courts. None of those are held here. The theological question that commentators argue over — whether Paul was right to do it — is not a question a word count can touch, and the page does not pretend otherwise.

Every repeated word in the diagram above was counted from the MorphGNT lemma column across the 118 rows, not taken from an outline.

Two of these are worth saying out loud. Every occurrence of “crowd” and every occurrence of “cry out” in these three chapters falls inside one of the two riots, which is what makes the frame in the diagram above a counted observation rather than an impression. And “Asia” is the thread that ties the two ends together: the province is named seven times, the last of them at 21:27, where the men who seize Paul in the temple are Jews ἀπὸ τῆς Ἀσίας. Caveat: these are lemma counts over 118 verses. Small numbers move easily, and a different lemmatiser would shift a row or two — the δέω / δεῖ merge above is the clearest example, and it is exactly the kind of thing that turns a real thread into an inflated one if nobody checks.

All 118 verses of Acts 19:1–21:40 were screened three times: against the other 7,809 verses of the Greek New Testament; against the Greek New Testament with the whole of Acts removed (6,925 verses); and against the 30,637 verse units of the Greek Old Testament held on this disk. Contiguous runs of whole word forms, never lemmas, with the tie band beside every run.

That middle row is the finding. Taking one book out of a 7,927-verse corpus leaves three of the ten rows that reached five or more — and both of the six-word runs are among the seven that fall. The three that survive are Acts 19:1 with Luke 5:1, Acts 19:8 with Luke 9:11, and Acts 21:11 with John 11:44 and Hebrews 3:7 — and the first two are Lukan sentence furniture (ἐγένετο δὲ ἐν τῷ τόν; περὶ τῆς βασιλείας τοῦ θεοῦ).

Read the bands. Acts 20:28’s run of four is shared with nineteen other New Testament verses — it is τὸ πνεῦμα τὸ ἅγιον, and it means nothing about relationship. Acts 19:35 reaches two at a band of 261 and Acts 21:14 at a band of 196: those are grammar, not contact. A run reported without its band would have made all three look like discoveries.

1. Acts 19:5 and Acts 8:16 — six words, band one.

2. Acts 19:6 and Acts 11:15 — six words, band one.

The exact six-word sequence in each case occurs in exactly two verses of the whole Greek New Testament — which is what a tie band of one means when the source row is excluded from its own target corpus. Two consecutive verses, two runs of six, and the two partners are the only other scenes in Acts where Luke narrates the Holy Spirit falling on a group outside Jerusalem. Not the only place outside Jerusalem where a group is filled with the Spirit — Acts 13:52 is one — but the only two Luke tells with this verb.

⚠️ What the page will not say. Not that Acts 19 is modelled on Acts 8 and 11, nor the reverse. Not that the three episodes are a single redactional pattern. Not that the sequence of baptism, hands and Spirit in these verses settles any question about how the Spirit is given. Direction of influence is never asserted on this shelf, and inside a single book a word count cannot even pose the question properly. What is measured is that these two sequences are exclusive to these four verses, and that removing Acts from the corpus removes both.

1. The Miletus speech and Ezekiel’s watchman: run of 1.

Every commentary on the Miletus speech hears Ezekiel, and the theological link is real. Verbally there is nothing there at all — and the row’s actual deepest partner is Pilate. This is the clearest demonstration on the page that a certain connection can be verbally invisible, and that the instrument measures one thing only.

2. The one saying of Jesus quoted in Acts that stands in no Gospel has no verbal home either. Acts quotes the Lord with a remembering formula twice — here and at 11:16 — and the words remembered at 11:16 are already printed at Acts 1:5, so this is the only one that stands nowhere else.

Μακάριόν ἐστιν μᾶλλον διδόναι ἢ λαμβάνειν appears in no Gospel. On the Greek-side convention used here a 0 means the two verses share no single word at all, which is what Acts 20:35 and Matthew 10:8 do. A saying can be remembered and reported and still leave no trace a sequence-matcher can find.

3. ‘The will of the Lord be done’ and Gethsemane: run of 2.

Acts 21:14 is τοῦ κυρίου τὸ θέλημα γινέσθω. Against Luke 22:42, Matthew 26:42 and Matthew 6:10 the run is 2 in each case — τὸ θέλημα. Its deepest run against the entire Greek New Testament is 2, at a tie band of 196. The sentence everybody hears as an echo of Gethsemane is, to this instrument, ordinary Greek.

4. The growth notice that does not match its own family.

The Acts 5–6 page on this shelf reports Acts 6:7 and Acts 12:24 agreeing for six words at a tie band of one. Acts 19:20, the third notice, measures 2 against Acts 12:24, 2 against Acts 6:7, and reaches only 3 anywhere in the New Testament. Two members of a “refrain” can be verbatim twins while the third shares nothing with either — and a page that reported only the pair would have implied a formula that does not exist.

5. The second-deepest Old Testament run in the passage is a calendar phrase.

Acts 21:15 is Μετὰ δὲ τὰς ἡμέρας ταύτας ἐπισκευασάμενοι ἀνεβαίνομεν εἰς Ἱεροσόλυμα. Its deepest Greek Old Testament run is 5, at a tie band of 2 — with Daniel 1:18 and Judith 16:21. The five words are Μετὰ δὲ τὰς ἡμέρας ταύτας, and after these days. It means nothing. It is here because it is the number the screen returned, and because a page that printed only the runs it liked would not be a measurement.

Acts circulates in two substantially different forms, and this page measured one.

The so-called Western text of Acts — Codex Bezae and its allies — is roughly a tenth longer than the text printed in critical editions, and several of its best known expansions fall inside these three chapters, including an addition at 19:9 that gives the hours of Paul’s teaching in the hall of Tyrannus. None of that is on this disk. The Greek here is the SBL edition as digitised by MorphGNT: a single printed text, no apparatus, no variants, no witnesses attached.

And this passage contains the best-known variant in the book. Acts 20:28 reads τὴν ἐκκλησίαν τοῦ θεοῦ in the edition measured here; τοῦ κυρίου is widely attested and is printed by other editions, and the closing words διὰ τοῦ αἵματος τοῦ ἰδίου will bear more than one construal. The page cannot adjudicate either question and does not try. What it can say is what it measured: the row’s deepest New Testament run is four words at a tie band of nineteen, and none of the disputed words is inside it.

Acts 19:26 against Epistle of Jeremiah 49 is a run of 5 with the fold and a run of 3 without it. That is stated here rather than buried.

Both printed texts read οὐκ. The difference is that MorphGNT’s normalised column folds οὐκ and οὐχ into οὐ and CATSS does not, so across the two corpora a Greek Old Testament οὐκ could never match a New Testament οὐ and every shared run reaching a negated verb stopped dead there. This shelf applies the equivalence οὐκ ≡ οὐχ ≡ οὐ at comparison time, to both sides alike, which is true of the language and false of nothing. The fold can only lengthen runs, never invent them — it makes two identical words look identical — but it is a decision, and on this page it is the decision that produces the headline Old Testament figure.

Audited across the whole passage, and the two halves of the audit are kept apart because they are two different rules. The negative fold on its own changes the maximum of exactly one row out of 118 — Acts 19:26, Greek OT screen, 3 → 5. The shelf also applies the editions’ other declared equivalences (ἐκ ≡ ἐξ, the movable nu) to both sides, and those change exactly one more row — Acts 20:30, Greek OT screen, 2 → 3, and that one is ἐκ ≡ ἐξ, not the negative. Two rows out of 118 between them. Neither of the two headline runs of six is affected, and the New Testament distribution is identical with every equivalence applied and with none of them.

No commentaries were used for any figure on this page. Not Bruce, not Barrett (ICC), not Fitzmyer (Anchor), not Keener’s four volumes, not Pervo, not Haenchen, not Witherington, not Peterson. Their names are in the attribution below as the literature a reader should go to next, and as an honest statement that the page did not go there first.

No inscriptions, no coins, no papyri, no Josephus, no Mishnah. Which means the Ephesian νεωκόρος title could not be dated, the office of Asiarch could not be described, the διοπετής could not be identified, the magical texts behind 19:19 could not be examined, and the mechanics of the vow at 21:23–26 could not be checked. Everything measured on this page was measured against the text files named in the footer and nothing else.

No ESV Acts file is held on this disk. The English column was typed from the printed ESV rather than read off a file; the chapter shape in the Stage 1 table is the printed edition’s for the ESV row, and a disk count for the Greek row and for the English row counted beside it.

The Epistle of Jeremiah is not in the Protestant canon this church reads. It is held on this disk because CATSS ships the whole Greek Old Testament collection, and it is reported here as a witness to Greek usage, not as Scripture. Excluding it entirely would drop the passage’s deepest Old Testament run from 5 to the 5 at Acts 21:15 — which is a date formula — and then to 4. The page names the corpus boundary rather than quietly choosing one.

The headline is a claim about a corpus boundary and about two exact sequences, so anything that moves either moves it.

Four things would overturn it. One: a different edition of Acts. If the SBL editors had read εἰς τὸ ὄνομα Ἰησοῦ Χριστοῦ at 19:5, as some witnesses do elsewhere in Acts, the run of six would not exist. Two: a different definition of a word — the screen compares forms, not lemmas, so ἐβαπτίσθησαν and βεβαπτισμένοι are different words to it, and a lemma-based screen would produce a longer and quite different set of runs everywhere. Three: counting non-contiguous agreement. A screen allowing one word of slack would find far more, and would find it in the Gospels too. Four: a different corpus. Include the Apostolic Fathers, or Josephus, or the papyri, and phrases that look exclusive here would very likely stop looking exclusive.

And one thing that would not break it, but should temper it. Acts is a long book — 1,002 verses in this edition, about an eighth of the Greek New Testament — and its subject matter is repetitive by design. A passage has more chances to match its own book than any other single book, simply because Acts is large and Luke retells the same events. Acts 11:15 is Peter recounting Acts 10:44; the two are a report and its source, so finding the passage agreeing with 11:15 rather than 10:44 is partly an accident of which retelling got the preposition. The page reports the effect; it does not claim the effect is surprising in kind. What is worth printing is the size of it: two runs of six with tie bands of one, in consecutive verses, both gone the moment one book leaves the corpus.

1. The word for ‘church’ is also the word for a riot. ἐκκλησία stands five times in these three chapters and three of them are a mob in a theatre that mostly did not know why it had come (19:32). Luke does not change the noun and does not explain himself. The word means assembly; what an assembly is depends on who called it.

2. The passage is held together by two sentences that are the only two of their kind. “Some cried one thing and some another” occurs twice in the whole Greek New Testament — a theatre in Ephesus and a temple court in Jerusalem — and the two riots are what the journey runs between. The measured run is only two words, and the distribution is exhaustive. Both numbers are true and they say different things.

3. The famous lines are not the deep ones. “It is more blessed to give than to receive” measures nothing. “The will of the Lord be done” measures two. “Innocent of the blood of all” measures one against the Ezekiel everyone hears in it. Depth of verbal agreement and weight of meaning are different quantities, and this page has one instrument, which measures the first.

4. Paul goes to Jerusalem against the Spirit-inspired advice of people who love him. The disciples at Tyre tell him διὰ τοῦ πνεύματος not to go (21:4); Agabus says what will happen and the whole company begs him not to go (21:12); and he goes. Luke reports both the warning and the refusal without resolving them, and closes the scene with the company saying τοῦ κυρίου τὸ θέλημα γινέσθω. The page notes what the text does and leaves the theology to the reader.

5. The itinerary is stated once and then obeyed. “I must also see Rome” (19:21) is the hinge, and the verb of binding answers it four times: bound in the Spirit (20:22), bound in Agabus’ sign (21:11), willing to be bound (21:13), bound with two chains (21:33). Two chains is the same phrase as Peter’s in Acts 12:6, where an angel made them fall off. Nobody comes for Paul.

Everything above is reproducible from the substrate this shelf runs on. The screens are exhaustive and deterministic; nothing was sampled, so no seed applies. Every count in Stage 3 comes from the MorphGNT lemma column; every run comes from the normalised word column with the negative fold applied to both corpora.

The passage ends with a man in chains on a flight of steps, motioning for silence, about to speak in the language the crowd was not expecting. Whether the six words Acts 19:5 shares with Samaria and the six Acts 19:6 shares with Caesarea were meant to bind those three cities together, this page does not know and does not say. It says that the sequences are exclusive to those four verses, and that removing one book from the corpus removes them both.