Then They Understood

• Dictionary of Jesus and the Gospels (IVP, 2nd ed.) — “Feeding of the Multitude,” “Transfiguration,” “Peter,” “Elijah,” “Clean and Unclean,” “Taxation.”
• Anchor Bible Dictionary — “Transfiguration,” “Caesarea Philippi,” “Corban,” “Fringes,” “Temple Tax,” “Tobit, Book of.”
• NIDNTTE — ἐκκλησία, μεταμορφόω, ὀλιγόπιστος, κράσπεδον, ᾅδης, συνίημι.
• Dictionary of New Testament Background (IVP) — “Taxation, Jewish,” “Vows and Oaths,” “Apocrypha and Pseudepigrapha,” “Qumran: Place and History.”
• Encyclopaedia Judaica / Jewish Encyclopedia — “Shekel,” “Tzitzit,” “Nedarim,” “Fiscus Judaicus.”

• W. D. Davies & Dale Allison, Matthew 8–18 (ICC vol. 2) — the fullest technical treatment of this block.
• R. T. France, The Gospel of Matthew (NICNT); Craig Keener, A Commentary on the Gospel of Matthew; John Nolland (NIGTC); Donald Hagner (WBC 33B).
• Ulrich Luz, Matthew 8–20 (Hermeneia) — especially on 16:17–19 and its reception history.
• Carey C. Newman (ed.), The Transfiguration of Christ; A. D. A. Moses, Matthew’s Transfiguration Story and Jewish-Christian Controversy.
• Carey A. Moore, Tobit (Anchor Bible 40A); Joseph A. Fitzmyer, Tobit (Commentaries on Early Jewish Literature) — for the two Greek recensions and the Qumran fragments.
• Patrick W. Skehan & Alexander A. Di Lella, The Wisdom of Ben Sira (Anchor Bible 39) — for Sirach 48 on Elijah.

The crowds at Gennesaret beg to touch τοῦ κρασπέδου τοῦ ἱματίου αὐτοῦ. The word is not “hem” in the sense of a stitched edge; it is the tassel commanded at Numbers 15:38–39, worn on the corners of the outer garment as a memory device — the Greek of Numbers 15:39 reads “and you shall look at them and remember all the commandments of the Lord.” Matthew has already told you Jesus wore them (9:20) and criticised those who enlarged theirs (23:5).

“What you would have gained from me is δῶρον” — a gift, i.e. dedicated to the temple. Mark preserves the Aramaic word and translates it for his readers (Κορβᾶν, Mk 7:11); Matthew does not, because his readers did not need it. The rabbinic tractate Nedarim is largely taken up with such vows and with the grounds for dissolving them — m. Nedarim 9:1 explicitly debates whether a vow may be released “for the honour of father and mother.” A first-century ossuary from the Kidron valley carries the word qorban in a dedication formula, so the practice is attested outside the literature as well.

The dispute is real and both sides are arguing from Scripture. That matters for how the scene is taught: Jesus is not opposing “tradition” in the abstract but naming a specific case in which one rule — the inviolability of a vow (Num 30:2; Deut 23:21–23) — had been allowed to cancel another, the fifth commandment. The Greek verb at 15:6 is ἠκυρώσατε, a legal term: to annul, to make void a contract.

The collectors ask about τὰ δίδραχμα, the two-drachma tax — the annual half-shekel of Exodus 30:11–16, which the Greek Exodus itself renders with this very word: τὸ ἥμισυ τοῦ διδράχμου, “half of the didrachm” (Exod 30:13). Josephus describes the tax being collected from Jews across the empire (Antiquities 18.312), and the Mishnah tractate Shekalim is devoted to it. After AD 70 Rome redirected it to the temple of Jupiter Capitolinus as the fiscus Judaicus (Josephus, War 7.218; Dio Cassius 66.7) — which is why a scene about whether the sons pay would have landed hard on a community reading this Gospel in the 80s.

“Why do the scribes say that first Elijah must come?” The expectation comes from the last lines of the Book of the Twelve. Sirach 48:10 shows it was already a fixed expectation before Matthew: describing Elijah, Ben Sira says he is ἐπιστρέψαι καρδίαν πατρὸς πρὸς υἱόν, “to turn the heart of a father to a son.”

Tobit is a Jewish novella of the Second-Temple period, probably third or second century BC. Pleasant Springs does not receive it as Scripture, and nothing on this page treats it as such. It matters here for one reason: it is a Greek text of the period, it is in the corpus this shelf measures against, and Stage 4 found the deepest non-citation contact in these four chapters inside it.

Two facts about the book are load-bearing for that finding, so they are stated up front:

• Tobit survives in two substantially different Greek recensions. The shorter (G^I) is the text of Vaticanus and Alexandrinus; the longer (G^II) is the text of Sinaiticus. This shelf holds both — the CATSS files TobitBA and TobitS. They are not minor variants of each other; they are different editions.
• Aramaic and Hebrew fragments were found at Qumran (4Q196–200), and they generally align with the longer Sinaiticus form — which is why most modern editions and translations now follow G^II.

The phrase is not new with Matthew. What is new is what the gates are doing in the sentence. Here is every occurrence in the 30,348 verse units of Greek Old Testament screened here, with the case each time:

• Isaiah 38:10 — ἐν πύλαις ᾅδου καταλείψω, “in the gates of Hades I shall leave the remainder of my years” — dative, after a preposition. Hezekiah’s psalm.
• Wisdom 16:13 — κατάγεις εἰς πύλας ᾅδου καὶ ἀνάγεις, “you bring down to the gates of Hades and bring up” — accusative, after a preposition.
• 3 Maccabees 5:51 — πρὸς πύλαις ᾅδου καθεστῶτας, “already standing at the gates of Hades” — dative, after a preposition.
• Psalms of Solomon 16:2 — σύνεγγυς πυλῶν ᾅδου, “near the gates of Hades” — genitive, after an adverb of place.

⚠️ This count read five until 30 August 2026, and the fifth was a second copy of the first. The list above used to carry one more line: Odes 11:10, annotated — correctly — “the same verse, since Ode 11 is Hezekiah’s psalm.” That parenthesis was the whole problem. The Odes are the liturgical canticles bound after the Greek Psalter, and ten of them are second copies of passages already in the corpus: Ode 1 is Exodus 15, Ode 2 is Deuteronomy 32, Ode 3 is 1 Samuel 2, Ode 4 is Habakkuk 3, Ode 6 is Jonah 2, Ode 11 is Isaiah 38. Counted twice, one occurrence looks like two and a tie band of one looks like a band of two. The Odes stay on this disk; they are out of the screening corpus, so this count is four — and it was never five texts. It was four texts and a duplicate. The old number is printed here rather than swapped in silently.

⚠️ The sharper half of the same trap did not touch this page, and a reader of a measuring shelf should know it is there. Ode 9 is Luke 1 — the Magnificat and the Benedictus — and Ode 13 is Luke 2, the Nunc Dimittis. Screened as Old Testament, Ode 9:48 matches Luke 1:48 at eighteen consecutive words: a Gospel measuring its own text under a Septuagint shelfmark, and a page could print that as a Gospel’s deepest Old Testament contact. Nothing on this page ever landed in Ode 9 or Ode 13 — the one row that moved here came out of the duplication arm, not the Luke arm. It was caught the same way both arms are caught: by asking which book a hit came out of before believing the number.

μετεμορφώθη occurs four times in the Greek New Testament and not once in the 30,348 verse units of Greek Old Testament screened here — not in Exodus 34, where Moses’ face shines, and not in Daniel 10 or 12, where the wise shine like stars. The Septuagint has plenty of glory, plenty of shining, plenty of transformation-language of other kinds; it does not have this verb.

Two of the four occurrences are Gospel narrative (Mt 17:2; Mk 9:2). The other two are about Christians — Romans 12:2, “be transformed by the renewal of your mind,” and 2 Corinthians 3:18, “we all… are being transformed into the same image from one degree of glory to another.” The New Testament uses this verb of exactly two subjects: Jesus on the mountain, and the people who belong to him.

Matthew has no versification offset. Chapter and verse are the same number in a Greek New Testament and in an English Bible, all the way through, because there is no second column to be offset from — there is no Masoretic Matthew and no Septuagint Matthew.

There is one hole, and it is in this passage. The SBLGNT prints no verse 17:21. Counted out, these four chapters run 36 / 39 / 28 / 26 = 129 verses of Greek text across a numbering that goes to 130. The ESV likewise skips from 17:20 to 17:22 and footnotes the omission. Every count on this page is over the 129 verses that carry text.

But the offset problem is still in the room, three times, because these chapters quote and echo books that carry one. Where a Greek Old Testament reference appears anywhere on this page it is printed in Greek numbering with the English beside it:

• Isaiah 29:13 — no offset. Greek, Hebrew and English all number it 29:13. The one quotation in these chapters that carries a citation formula is also the one with nothing to convert.
• Exodus 21 — the Greek transposes two verses. “Whoever reviles father or mother shall die” is Greek Exodus 21:16 and Hebrew/ESV Exodus 21:17; the law about kidnapping swaps with it. Matthew 15:4 quotes this law, so the reference you cite depends on which Bible you are holding.
• Malachi — ESV 4:5–6 = Hebrew 3:23–24 = Greek 3:22–23. The English has four chapters; the Hebrew and the Greek have three. And the Greek does something further — see the next question.
• Psalm 62:12 — ESV 62:12 = Hebrew 62:13 = Greek Psalm 61:13. This is the verse usually cited behind Matthew 16:27, and it carries a double offset: the Greek psalm number runs one behind the Hebrew from Psalm 9/10 through 147, and the Hebrew counts the superscription as verse 1.

The last three verses of Malachi are the same three verses in Hebrew and in Greek — but they are not in the same order. In the Masoretic text and in every English Bible:

• MT 3:22 = ESV 4:4 — “Remember the law of my servant Moses.”
• MT 3:23 = ESV 4:5 — “Behold, I send you Elijah the prophet.”
• MT 3:24 = ESV 4:6 — “And he will turn the hearts of fathers to children.”

In the Greek text this shelf holds (CATSS, following Rahlfs) the order is:

• Greek 3:22 — “Behold, I send you Elijah the Tishbite.”
• Greek 3:23 — “who will restore the heart of a father to a son.”
• Greek 3:24 — “Remember the law of Moses my servant.”

Scope. Every one of the 129 SBLGNT verses of Matthew 14:1–17:27 against every one of the 30,348 of the 30,637 verse units in the 64 CATSS Greek Old Testament files held here — the canonical books together with the deuterocanonical ones. Greek against Greek, lowercased, accents and breathings dropped, final sigma folded, contiguous runs of whole word forms, never lemmas. Exhaustive and deterministic.

The distribution. All 129 rows reach two; 104 reach three; 39 reach four; 8 reach five; 2 reach six; 1 reaches eight; and nothing reaches nine.

⚠️ Re-screened on 30 August 2026 with the Odes out of the corpus, and this table did not move. Every row below, its run and its tie band, is identical over 30,348 units and over 30,637, and so is the distribution. The eight deepest contacts these four chapters have with the Greek Old Testament were never touching a canticle. The one figure on this page that did move is the gates-of-Hades count in stage 3, five to four.

Here are all eight rows that reach five or more, with the tie band on every one. This is the whole list — nothing has been selected out:

• Mt 15:8 — run 8, band 1 — Isaiah 29:13. ἡ δὲ καρδία αὐτῶν πόρρω ἀπέχει ἀπ᾽ ἐμοῦ. A citation, and Matthew says so: “well did Isaiah prophesy of you.”
• Mt 17:6 — run 6, band 1 — Tobit 12:16 (Sinaiticus recension). Not a citation. Discussed below.
• Mt 15:9 — run 5, band 1 — Isaiah 29:13 again. μάτην δὲ σέβονταί με διδάσκοντες. The same citation, continued.
• Mt 15:4 — run 5, band 2 — Deuteronomy 21:13 and 1 Esdras 4:25. This is a citation of the fifth commandment that does not measure against the fifth commandment. Discussed below.
• Mt 14:2 — run 5, band 2 — Greek 3 Kingdoms 21:12 and 21:31 (= ESV 1 Kings 20:12, 20:31). καὶ εἶπεν τοῖς παισὶν αὐτοῦ, “and he said to his servants” — a narrative formula, and nothing more.
• Mt 14:35 — run 5, band 1 — Genesis 26:7. αὐτὸν οἱ ἄνδρες τοῦ τόπου. Five words shared with the men of Gerar asking Isaac about his wife. A coincidence.
• Mt 15:11 — run 5, band 1 — Numbers 32:24. τὸ ἐκπορευόμενον ἐκ τοῦ στόματος. In Numbers it is a vow you must keep; in Matthew it is what defiles a person. A coincidence — and a good one, because the words are identical and the subject is opposite.
• Mt 17:27 — run 5, band 2 — Greek Daniel 3:25 in both the Old Greek and Theodotion. καὶ ἀνοίξας τὸ στόμα αὐτοῦ. In Daniel, Azariah opens his mouth in the furnace to pray; in Matthew, Peter opens a fish’s. A coincidence.

Matthew 17:6 — the disciples’ reaction to the voice from the cloud — carries a run of six consecutive words with a tie band of 1. It is the deepest contact in these four chapters that is not the Isaiah citation. Its partner is Tobit 12:16, in the Sinaiticus recension:

καὶ ἐταράχθησαν οἱ δύο καὶ ἔπεσαν ἐπὶ πρόσωπον αὐτῶν καὶ ἐφοβήθησανTobit 12:16, Sinaiticus (G^II) — “and the two were troubled, and they fell on their face and were afraid”

καὶ ἀκούσαντες οἱ μαθηταὶ ἔπεσαν ἐπὶ πρόσωπον αὐτῶν καὶ ἐφοβήθησαν σφόδραMatthew 17:6 — “and when the disciples heard it, they fell on their face and were exceedingly afraid”

And in both books, the very next verse is the same reassurance: Tobit 12:17, μὴ φοβεῖσθε; Matthew 17:7, ἐγέρθητε καὶ μὴ φοβεῖσθε.

At Matthew 15:4 Jesus quotes the fifth commandment with an explicit formula: ὁ γὰρ θεὸς εἶπεν, “for God said.” If a run-length screen were a reliable citation-detector, its deepest hit should be Exodus 20:12 or Deuteronomy 5:16. It is neither.

• Mt 15:4 against Greek Exodus 20:12 — run 3. τίμα τὸν πατέρα and then it stops.
• Mt 15:4 against Greek Deuteronomy 5:16 — run 3. Same three words, same stop.
• Mt 15:4 against Deuteronomy 21:13 — run 5. A captive bride mourning her father and mother.
• Mt 15:4 against 1 Esdras 4:25 — run 5. A courtier arguing that a man loves his wife more than his father and mother.

Scope. The 129 verses against the other 7,798 verses of the Greek New Testament. Same normalisation, same run engine.

The distribution runs twice as deep as Screen 1. All 129 reach two; 120 reach three; 91 reach four; 64 reach five; 39 reach six; 23 reach seven; 19 reach eight; 13 reach nine; 12 reach ten; 9 reach eleven; 8 reach twelve; 6 reach fourteen; 4 reach fifteen; 3 reach sixteen; and nothing reaches seventeen.

The three rows at sixteen, each with a tie band of 1:

• Mt 16:24 — sixteen words with Mark 8:34: εἴ τις θέλει ὀπίσω μου ἐλθεῖν… καὶ ἀκολουθείτω μοι.
• Mt 16:25 — sixteen words with Luke 9:24.
• Mt 16:4 — sixteen words with Matthew 12:39. The third of the three deepest ties this passage has with anything at all is Matthew with himself, eleven chapters earlier, on the sign of Jonah.

Matthew 17:5 is routinely taught as a fusion of Old Testament texts: Psalm 2:7 (“you are my Son”), Isaiah 42:1 (“my chosen, in whom my soul delights”), Deuteronomy 18:15 (“listen to him”), sometimes Genesis 22:2 (“your beloved son”). Here is what those four comparisons actually measure:

• Mt 17:5 against Greek Psalm 2:7 — run 2 (υἱός μου).
• Mt 17:5 against Greek Isaiah 42:1 — run 1.
• Mt 17:5 against Greek Deuteronomy 18:15 — run 1.
• Mt 17:5 against Greek Genesis 22:2 — run 1.
• Mt 17:5 against Greek Exodus 24:16 — run 2 (τῆς νεφέλης) — the cloud on Sinai, six days, the voice from the cloud. Two words.
• Mt 17:5 against Matthew 3:17 — run 11, band 1. λέγουσα οὗτός ἐστιν ὁ υἱός μου ὁ ἀγαπητός, ἐν ᾧ εὐδόκησα.

Scope. Each of the 129 verses against the other 128, and separately against the other 939 verses of Matthew.

Inside the passage: 125 rows reach two, 78 reach three, 39 reach four, 18 reach five, 6 reach six, 2 reach seven, and nothing reaches eight. The two rows at seven are 16:6 and 16:11 — the warning about the leaven of the Pharisees and Sadducees, given and then repeated within five verses. The rows at six include the pair this study cares about most: 14:20 and 15:37, the two feedings, καὶ ἔφαγον πάντες καὶ ἐχορτάσθησαν καί.

Against the rest of Matthew: the deepest is 16:4 with 12:39 at sixteen, then 17:5 with 3:17 at eleven, then 16:16 with 26:63 at eight — Peter’s confession and the high priest’s question share the clause σὺ εἶ ὁ χριστὸς ὁ υἱὸς τοῦ θεοῦ. The same eight words are a confession in chapter 16 and an accusation in chapter 26.

Two Greek editions and one English translation, all named: the SBL Greek New Testament as digitised by MorphGNT; Rahlfs’ Septuaginta as digitised by CATSS; the Masoretic Hebrew (Sefaria’s Miqra according to the Masorah) for the Malachi and Exodus ordering checks only; and the ESV for the English column. The commentaries listed in Stage 1 informed the framing, and where this page disagrees with a standard reading it says so out loud rather than quietly.

• No critical apparatus of any kind. Not Nestle-Aland, not the UBS, not Swete, not the Göttingen Septuagint. Every “the text reads” on this page means these two digital editions read. Where a variant is known to matter it is named — 17:21, and the two Tobit recensions — but no variant has been adjudicated here and none should be taken as adjudicated.
• The Göttingen edition of Tobit was not consulted, and it is the edition that matters for the Tobit finding. This page’s run of six rests on the CATSS transcription of the Sinaiticus recension. Hanhart’s critical text could read differently at exactly the words that produce the run. This is the single largest hole in the page, and the finding is stated conditionally throughout because of it.
• No Qumran text was read directly. The remarks about 4Q196–200 are reported from the secondary literature (Fitzmyer, Moore), not from photographs or transcriptions on this disk.
• No rabbinic text was read in the original. Nedarim, Shekalim and the Corban discussion are reported from standard reference works. The dates involved are late and the page says so.
• No archaeological report was read. The Corban ossuary is cited as it is reported in the dictionaries, not from the excavation publication.
• Josephus and Dio Cassius were not read in the original languages here. The temple-tax references are standard and easily checked, but they were not verified against a critical edition on this disk.

• 1. The Tobit run is edition-dependent, and this page knows it. Six words in the Sinaiticus recension, two in the Vaticanus/Alexandrinus one. Anyone quoting the number must quote the recension with it, or they are quoting a number that does not exist on its own.
• 2. Matthew’s Bible may not have contained Tobit at all. Whether a Jewish-Christian community in the 80s counted Tobit as Scripture is exactly the question the page cannot settle, and it does not need to: the finding is a fact about two Greek texts, not a claim about a canon. But a reader who wants the finding to mean something will have to settle that question first, and this page has not.
• 3. The Malachi ordering point is a description of two editions, not an argument. The Greek manuscript tradition of Malachi is not uniform, and Rahlfs’ text is an edited text. That the Twelve ends with Moses in the Greek held here is true of the Greek held here. A different edition could order it differently, and the page has not checked one.
• 4. The negative counts are corpus-bounded and nothing more. “Zero occurrences of μεταμορφόω” means zero across 64 CATSS files. “Zero occurrences of ὀλιγόπιστος” means the same. Neither is a claim about Greek at large, about Philo, about Josephus, or about any papyrus.

On the transfiguration. The strongest objection to Stage 4’s handling of 17:5 is that run-length is simply the wrong instrument for allusion, and that a scholar who reads Psalm 2 and Isaiah 42 there is not making a claim the run engine is competent to test. That objection is correct, and the page grants it in the section itself. What is offered is not a refutation but a measurement placed beside the reading, so that a reader knows the difference between “the words are the same” and “the theology is the same.”

On Matthew 16:18. The grammatical count is solid and the exegesis it supports is deliberately thin. Whether πέτρα is Peter, Peter’s confession, or Christ himself is a question this page does not touch, because no measurement available here bears on it. Readers should be suspicious of anyone who claims that a word count settles it.

On the two feedings. Most critical scholarship treats 14:13–21 and 15:32–39 as two tellings of one tradition. The evidence assembled here — the six-word run, the differing numbers, and 16:9–10 — establishes that Matthew presents them as two events. It does not establish that they were two events, and the page does not claim it does.

• Does the Göttingen Tobit preserve the six words? Not checkable here.
• Is the Rahlfs ordering of Malachi 3:22–24 supported across the Greek manuscript tradition, or is it one witness family? Not checkable here.
• Do any of Matthew’s deep synoptic runs favour one solution to the Synoptic problem over another? Not from run lengths. A run is symmetrical and carries no direction.
• Was the temple tax of 17:24 the biblical half-shekel or the post-70 fiscus Judaicus in the mind of Matthew’s first readers? Both are defensible; nothing on this disk decides it.

Read straight through, Matthew 14–17 is a story about people who are present at everything and understand almost nothing. They distribute the bread themselves and then panic about bread twice more (14:19; 15:36; 16:7). They watch a man walk on water and confess him Son of God, and two chapters later he is the one asking them who they think he is (14:33; 16:13). The ὀλιγοπιστ- word lands on them three times in four chapters — Peter alone at 14:31 (ὀλιγόπιστε, singular), the disciples together at 16:8 (ὀλιγόπιστοι), and their ὀλιγοπιστία at 17:20.

And twice — only twice in the whole Greek New Testament, and both of them here — Matthew writes τότε συνῆκαν, “then they understood” (16:12; 17:13, where he adds “the disciples”). Both times it comes after a rebuke, not instead of one. Understanding, in this section, is not the reward for being clever. It is what happens on the far side of being told you have missed it.

The deepest verbal tie the transfiguration has is with the baptism — eleven words, said twice, the second time with two Greek words added: “listen to him.” The Father says the same thing at the beginning of the ministry and near its turn, and the only difference the second time is a command to the people standing there.

If the sentence had not changed, what would you have concluded from hearing it again? And what does it mean that the one thing added was not new information about the Son, but an instruction to the listeners?