The Register That Does Not Add Up

• Anchor Bible Dictionary — “Ezra, Book of,” “Ezra–Nehemiah, Books of,” “Esdras, First Book of,” “Aramaic Language,” “Persia, History of,” “Beyond the River,” “Sheshbazzar,” “Tattenai,” “Nethinim,” “Urim and Thummim,” “Weights and Measures.”
• Dictionary of the OT: Historical Books (IVP) — “Ezra and Nehemiah, Books of,” “Lists,” “Persian Period,” “Aramaic Language.”
• NIDOTTE — גאל (both of them), יחשׂ, קהל, אֻשַּׁיָּא.
• On the Greek — the NETS introduction to 2 Esdras (Esdras B), and its account of how 2 Esdras and 1 Esdras handle this same list differently.
• Grammars of Biblical Aramaic — Rosenthal, A Grammar of Biblical Aramaic; the Aramaic sections of Bauer–Leander. Named here as the standard tools for the seam measured in stage 3; neither was consulted.

The first two rows below were counted off the files on this disk. The third was not — no ESV Ezra or Nehemiah is held here, so the English row states the ordinary English versification and is the one line in this table that is not a measurement.

Identical, chapter for chapter and verse for verse. No conversion is needed anywhere in Ezra.

What does need saying is the book division and one chapter of Nehemiah. Greek has no book called Ezra: it has 2 Esdras, which is Ezra and Nehemiah run together. Ezra n = 2 Esdras n. Nehemiah n = 2 Esdras n+10. So the Nehemiah 7 this page keeps comparing Ezra 2 with is 2 Esdras 17 — and that chapter has 72 verses in the Masoretic column against 73 in CATSS and in the ESV. The line the Masoretic column does not carry is the horses and the mules. Masoretic Nehemiah 7:68–72 = ESV Nehemiah 7:69–73.

Practical consequence: a commentary discussing “Nehemiah 7:69” may be discussing the camels or the governor’s gift depending on which column it is counting in. Inside Ezra itself, no such care is needed.

The register in Ezra 2 stands again in Nehemiah 7 and a third time in 1 Esdras 5. All three end with the same grand total. Neither of the two summed here reaches it.

The figures below were not copied from a commentary. The Hebrew numerals of Ezra 2:3–60 and Nehemiah 7:8–62 were parsed word by word out of the Masoretic text held here and summed for this page.

The two Hebrew copies differ from each other by 1,271 in their own entries and agree to the digit in their totals. That is the shape of the problem, and it is the same in every printed Bible; nothing here depends on an emendation.

Some individual entries move between the copies, and one of them is famous for another reason.

What is not being claimed. The page does not say which copy is prior, that anyone miscopied anything, or that the total is wrong. It says the entries do not sum to the total in any of the three, and prints both numbers. Explanations offered in the literature — that the total counts people the list does not itemise, that it counts all twelve tribes and not only the families named, that numerals were transmitted in a notation that damages easily — are all arguments, and none of them can be tested from these files.

Ezra 4 names four kings. Searched by name across the chapter, they fall in this order.

Four of the five steps go forward and the fifth goes back about eighty years. 4:24 does not continue 4:23; it resumes 4:5. The two verses share three consecutive words — דריוש מלך פרס, Darius king of Persia — measured Hebrew against Aramaic, since 4:5 is in one language and 4:24 in the other. Everything between them is a dossier about later opposition, filed at the point in the narrative where opposition is the subject.

The regnal dates above come from standard Achaemenid chronology and no cuneiform source was consulted for this page. What was measured here is only the order of the names in the chapter and the three shared words at 4:5 and 4:24.

The book changes language at 4:8 and does not change back until 6:18, and it changes again for 7:12–26. That is not a curiosity; it is the administrative reality of the empire.

Imperial Aramaic was the chancery language of the Achaemenid administration from the Mediterranean to the Indus. Letters between a provincial official and the court were written in it. Ezra 4:7 says so out loud: וּכְתָב הַנִּשְׁתְּוָן כָּתוּב אֲרָמִית וּמְתֻרְגָּם אֲרָמִית — the letter was written in Aramaic and translated. And the Greek row notices it too: ἔγραψεν…γραφὴν Συριστὶ καὶ ἡρμηνευμένην, he wrote a document in Syrian, and translated.

The nearest comparable material outside the Bible is the Elephantine archive — fifth-century Aramaic papyri from a Judean garrison in Egypt, including a petition to the governor of Judah about rebuilding a temple. It is the closest thing there is to Ezra 4–5 in a documentary form. It was not consulted for this page and nothing here rests on it; it is named as the kind of evidence the historical questions are settled from, and as the reading a serious student should go to next.

Tattenai. A Babylonian tablet naming a Tattannu, governor of Beyond the River is frequently cited in the commentaries as external attestation of the man in Ezra 5:3. This page has not seen that tablet, has not checked the reading, and does not stake anything on it. It is recorded here as a claim in the literature, not as a finding.

1 Esdras (Ἔσδρας Αʹ) is a separate Greek book covering roughly 2 Chronicles 35–36, Ezra, and Nehemiah 8, in a different order and with a long story about three young bodyguards that Ezra does not have. On this shelf it is a witness and not Scripture, and it is held in the CATSS files, which is why its copy of the register could be checked at all.

Josephus retells this period in Antiquities 11, and follows 1 Esdras rather than Ezra where the two differ. Josephus was not consulted for this page. He is named because a reader comparing the register in a Josephus translation is comparing a fourth text, not a third.

Everyone knows Ezra 4:8–6:18 and 7:12–26 are in Aramaic. This page did not take that on trust; it looked for the seam with a word count.

Two words for the king, forty occurrences between them, and they never once cross. The last Hebrew הַמֶּלֶךְ before the seam is at 4:3; the first Aramaic מַלְכָּא is at 4:8; and in chapter 7 the two alternate at exactly the block edges — Hebrew at 7:11, Aramaic from 7:14, Hebrew again at 7:27.

Two honest limits on this measurement. First, מַלְכָּא resolves the opening seam to 4:8 but דִּי only reaches back to 4:10; the particles cannot by themselves decide whether the change happens at the start of 4:8 or partway through it. Second, the absence of אֲשֶׁר proves less than it looks — it is also absent from the whole of chapter 8, which is Hebrew. Absence of a common word from a short stretch of text is weak evidence; the positive distribution of מַלְכָּא and דִּי is the strong part.

Three times in nine verses of chapter 2 a Hebrew word appears in the Greek row in Greek letters rather than in Greek.

And once the reverse happens. At 2:63 the Persian court title הַתִּרְשָׁתָא, the Tirshatha, is carried across as Αθερσαθα — but אוּרִים and תֻמִּים, which English Bibles leave as Urim and Thummim, are translated: τοῖς φωτίζουσιν καὶ τοῖς τελείοις, the ones that give light and the perfect ones. The row transliterates the Persian title and translates the priestly objects; English does the opposite.

Collected here rather than scattered through the codex. None of them is adjudicated; both readings are printed and neither is called earlier or later.

Those last three are the whole of the numeric disagreement in the register. All forty-two figures of Ezra 2 were parsed out of both columns and compared for this page. Thirty-nine of the forty-two agree exactly. The three that do not push the Greek column’s own itemised sum to 32,008 — which still does not reach 42,360 either.

The measurement first, and then what it is worth.

Ezra 2:13 reads, entire: בְּנֵי אֲדֹנִיקָם שֵׁשׁ מֵאוֹת שִׁשִּׁים וְשִׁשָּׁה — υἱοὶ Αδωνικαμ ἑξακόσιοι ἑξήκοντα ἕξ — the sons of Adonikam, 666. Revelation 13:18 ends …ὁ ἀριθμὸς αὐτοῦ ἑξακόσιοι ἑξήκοντα ἕξ.

Now the counter-example, printed here so the finding above is not read as more than it is. 1 Kings 10:14 and 2 Chronicles 9:13 both say Solomon received 666 talents of gold in one year, and preachers have paired that with Revelation for centuries. Neither of them matches Revelation 13:18 on a contiguous run, and the reason is spelling, not meaning. Greek 1 Kings 10:14 has ἑξακόσια καὶ ἑξήκοντα ἓξ — a conjunction in the middle breaks the run. Greek 2 Chronicles 9:13 has the neuter ἑξακόσια because it is counting talents, where Revelation and Ezra both have the masculine ἑξακόσιοι because they are counting men. Ezra 2:13 is the unique match because of a gender ending and a missing and.

So what does the run of three actually establish? That two verses independently write the same number in the same ordinary way. Nothing on this page says Revelation is quoting Ezra, alluding to Ezra, or has Adonikam in view, and Revelation gives no citation signal — it says let the one who has understanding calculate, which is an instruction to compute, not to look something up. A run of three with a tie band of one is a real measurement and a small one. It is printed because it is true and because it is the kind of thing a reader will meet stated far more strongly elsewhere.

⚠️ A tooling check this finding needed, and passed. Since 30 August 2026 cross-corpus runs apply MorphGNT’s own spelling equivalences, one of which folds ἐξ to ἐκ. That fold also catches the numeral ἕξ, six, so this run is reported by the screen with εκ as its third word. Both underlying texts were read by eye to confirm that each genuinely has ἕξ and not the preposition. They do. Had either side read ἐξ, the run would have been an artefact and this section would not exist.

Seventy verses measured Hebrew against Hebrew, contiguous word forms only.

Distribution across all seventy verses: 2 verses share only one word; 4 share two; 18 share three; 13 share four; 10 share five; 5 share six; 7 share seven; 3 share eight; 1 shares nine; 5 share ten; one shares fourteen and one shares sixteen. Every verse of Ezra 2 has a partner in Nehemiah 7.

The deepest overlap in the whole comparison is the verse about the family whose record was lost. Ezra 2:61 and Nehemiah 7:63 hold sixteen consecutive words in common. The verse that follows in both copies says these men sought their registration among those enrolled in the genealogies, but they were not found there. The line about a document that could not be produced is the line the two documents agree on most exactly. The page reports that and does not build a doctrine on it.

⚠️ One thing to know before reading the distribution. The Masoretic edition on this disk prints the scribal section markers ס and פ inside the verse, and the shelf’s frozen normaliser counts them as tokens. Ezra 2 and Nehemiah 7 both carry one at the end of nearly every entry, so 34 of the 70 rows above end their shared run on a section marker rather than on a word and are one longer than a word-only count would make them. The four deepest rows — 16, 14, 10 and 10 — are not among them, and neither is the single-word row at 2:22.

Direction is not asserted. Nothing here says Nehemiah 7 copies Ezra 2, that Ezra 2 copies Nehemiah 7, or that both copy an archived original — all three are argued in the literature and none of them can be settled by a run screen.

124 CATSS rows — 2 Esdras 2–5 — against all 7,927 verses of the Greek New Testament. 84 reach two, 47 reach three, 16 reach four, 5 reach five, and nothing reaches six.

Two of those five are worth something and two are formulae, and the page says which.

Ezra 3:2 and 3:9 are the counter-examples. And his brothers and and and his sons and are five-word runs made entirely of the commonest words in Greek. They are the deepest runs in these four chapters and they mean nothing at all. Print them beside the others or the others get over-read.

Ezra 5:14 is the one to look at twice. τὰ χρυσᾶ καὶ τὰ ἀργυρᾶ stands in five Greek Old Testament units and every one is about the same objects — the vessels Nebuchadnezzar carried out of the temple. Here they are being carried back; in Daniel 5 Belshazzar is drinking from them. In the New Testament the five words stand once, at Revelation 9:20, and they are not vessels at all: they are idols of gold and silver and bronze and stone and wood, which cannot see or hear or walk. The same five words, and the temple’s treasure has become the thing men worship instead. That is an observation about vocabulary. The page does not call it a quotation, and Revelation gives no signal that it is one.

Ezra 5:11 is the tightest of the three. τοῦ οὐρανοῦ καὶ τῆς γῆς stands in four Greek Old Testament units, and three of them are deuterocanonical — Epistle of Jeremiah 53, Tobit 7:17 and Tobit 10:14 — which are witnesses on this shelf and not Scripture. Ezra 5:11 is the only Scripture verse in the Greek Old Testament with the sequence, and it is the only verse in the Hebrew Bible that says God of heaven and earth in Aramaic. In the New Testament it stands twice, and the two are the same saying in two Gospels: I thank you, Father, Lord of heaven and earth. A tie band of 2 that is really a tie band of 1, and a phrase any two writers could reach independently.

Screened separately, the three chapters this page quotes come out in the opposite order to their length.

Seventy verses of names and numbers do not reach four anywhere in the New Testament — and forty of those seventy do not reach even two. Seventeen verses of official Aramaic reach four ten times. The reason is not theological. Ezra 5 is built out of phrases the Greek Bible uses constantly — the house of God, the God of heaven, gold and silver — and Ezra 2 is built out of proper names and numerals, which are the least repeatable material there is. A run screen measures how ordinary a verse’s vocabulary is at least as much as it measures contact, and that is worth remembering every time one of these tables is read.

The arithmetic is the strongest thing on this page and it is also the most easily misread. Two numeral parsers were written for this study — one for Hebrew, one for Greek — and they reproduce the two standard published sums for Ezra 2 and Nehemiah 7 exactly (29,818 and 31,089), which is the check that they work. But a numeral parser is a piece of software written for one page, not a frozen part of this shelf’s tooling, and it is not covered by the regression lock that guards every other figure here. The two totals agreeing with the standard figures is good evidence and it is not the same as a proof.

The Greek sum is reported with less confidence than the Hebrew, and deliberately so. Greek numerals share their spelling with common words — ἕξ with ἐξ, εἷς with εἰς — so an automatic parser over Greek is noisier than over Hebrew. The Greek figure quoted in stage 3 (32,008) was therefore not taken from the parser directly; it is the verified Hebrew sum adjusted by the three entries the two columns actually disagree about, each of which was read by eye.

The Aramaic is being screened by a Hebrew normaliser. The Masoretic corpus here is treated as one text; the tooling has no notion of a language boundary and folds Aramaic consonants exactly as it folds Hebrew ones. That is fine for the counting done in stage 3, which is lexical, and it means no morphological claim on this page rests on the tooling — the determined state, the construct, the participles are all read by eye from printed forms.

1 Esdras was used for three numbers and nothing else. Its copy of the register was read for Adonikam, the singers and the grand total. Its own entries were not summed and it was not screened against the New Testament, and it is in any case a witness on this shelf and not Scripture.

No Hebrew or Greek critical apparatus — not BHS, not BHQ, not the Göttingen Septuagint, which for this book would matter, since the numerals are exactly the sort of reading that varies between manuscripts and Rahlfs prints one line of the tradition.

No Qumran material, and for this passage that is a named gap rather than a general one. 4QEzra (4Q117) preserves fragments of Ezra 4 and Ezra 5 — the very chapters this page measures. None of it was consulted. Any statement here about what the Hebrew and Aramaic of chapters 4 and 5 read is a statement about the Masoretic text as printed in the edition on this disk.

No Syriac, no Vulgate, no Old Latin, no Aquila, Symmachus or Theodotion. No Targum — and no Targum on Ezra–Nehemiah is held here in any case. No Elephantine papyri, no Persepolis tablets, no Babylonian tablet of Tattannu, no Achaemenid inscription. The regnal dates in stage 2 are the standard ones and were not derived here.

And one thing this page deliberately does not attempt. It does not reconcile 42,360 with 29,818. Every reconciliation on offer — that the total counts women, or the northern tribes, or people the list summarises rather than itemises, or that a numeral notation was damaged in transmission — is an argument from outside the text, and this shelf’s rule is that a page which cannot check something says so. The two numbers are printed side by side and left there.

• H. G. M. Williamson, Ezra, Nehemiah (Word Biblical Commentary 16).
• Joseph Blenkinsopp, Ezra–Nehemiah (Old Testament Library).
• Sara Japhet, From the Rivers of Babylon to the Highlands of Judah — collected essays on the returns and the lists.
• Lester L. Grabbe, A History of the Jews and Judaism in the Second Temple Period, vol. 1 — on what the Persian-period sources will and will not support.
• Franz Rosenthal, A Grammar of Biblical Aramaic — the standard tool for the language of 4:8–6:18.
• Bezalel Porten and Ada Yardeni, Textbook of Aramaic Documents from Ancient Egypt — the Elephantine material in edition.
• None of these were consulted for the measurements on this page, which were run on the text files named in the footer.

Seventy verses of names. Parosh, Shephatiah, Arah, Pahath-moab, Elam, Zattu, Zaccai, Bani, Bebai, Azgad, Adonikam. Then the men of Bethlehem, of Netophah, of Anathoth, of Kiriath-arim; then the priests, the Levites, the singers, the gatekeepers, the temple servants; then the ones who came up out of Tel-melah and Tel-harsha and could not prove who their fathers were.

The figures do not balance, and the book keeps them. Nobody smoothed them. The entries sum one way in Ezra and another way in Nehemiah and the grand total is the same in both, and the compiler who set the two copies down in the same collection let that stand. It is not the behaviour of someone inventing a document. It is the behaviour of someone copying one.

And what the list is for is stated inside it. Three priestly families sought their registration among those enrolled in the genealogies, but they were not found there — and were put out of the priesthood until a priest should arise with Urim and Thummim. That priest is not recorded as arising. The register exists because in this community a name on a list was the difference between serving at the altar and standing at the back. Which is why the deepest agreement between the two copies falling on precisely that verse is worth a second look, even though it proves nothing.

Then the work stops for sixteen years, and the passage ends inside a letter. The last thing on this page is not a promise or a victory. It is a provincial governor quoting the builders back to the emperor: from that time until now it has been in building, and it is not yet finished — and then asking the archives whether these people are telling the truth. The answer is on the next page. What is on this one is the waiting.

This page closes the last dark chapters in Ezra. With chapters 2, 4 and 5 now quoted, and alongside Ezra 1:1–6:22 and Ezra 7:1–10:44, every one of the book’s ten chapters now has a verse quoted from it on this shelf. Ezra is complete.