Under the Sun

• Anchor Bible Dictionary — “Ecclesiastes, Book of,” “Qoheleth.”
• Dictionary of the OT: Wisdom, Poetry & Writings (IVP) — “Ecclesiastes: Book of,” “Hebel,” “Frame Narrative,” “Carpe Diem.”
• NIDOTTE — הֶבֶל, יִתְרוֹן, עָמָל.

222 and 222, chapter by chapter, with nothing absent and no lettered rows. That is unusual on this shelf. Job carries fourteen lettered labels and a 150-word appendix; Proverbs carries 48 lettered rows, 24 absent numbers and a re-ordered second half. Ecclesiastes has none of it.

What that does and does not mean. It means these two files can be laid side by side without a mapping table. It says nothing about how either text came to be, and this page draws nothing from it beyond the convenience.

The parser used here does detect lettered rows — run against Esther it returns 3:13f, 8:12n and 8:12r. Ecclesiastes has none for it to find.

Counted across all thirty-nine books, the two words תַּחַת and הַשָּׁמֶשׁ stand adjacent 29 times, and all 29 are in Ecclesiastes.

The nine here are 1:3, 1:9, 1:14, 2:11, 2:17, 2:18, 2:19, 2:20 and 2:22. The Greek rows carry ὑπὸ τὸν ἥλιον.

Scope. Counted on the MAM spelling as cleaned, on adjacent tokens under the shelf word rule, in the thirty-nine books on this disk. A different spelling, a different tokenisation, or a search allowing words between them would give a different number. The claim is about this counting, and it is stated so that it can be repeated.

Eight words; five of them carry הבל. No verse in the thirty-nine books carries more, and the set of verses reaching that maximum has exactly one member. Second is 12:8 with three, at the other end of this book.

Ecclesiastes holds 38 of these tokens in 30 of its verses — about half of all of them in the thirty-nine books.

The screen, and why the share is given as a range. A substring search on the skeleton הבל returns 90 tokens across the thirty-nine books, and it over-collects. Named false positives: 8 tokens of the name Abel in Genesis 4, which is the same three consonants; 3 of הבליעל; 1 of הבלעי; and the verb forms ויהבלו and תהבלו. Removing them leaves between 74 and 78 depending on how the verbs are treated, so Ecclesiastes’ share runs between 48.7% and 52.1% — hence “about half” rather than a single figure. Ecclesiastes’ own 38 does not move, because none of the false-positive classes occurs in this book. Narrower whitelists are worse, not better: two mutually incompatible three-skeleton screens both return exactly 33 tokens in 26 verses, so “33” cannot be checked by a reader who is not told which one was used.

And a control that keeps this honest. 371 verses in the thirty-nine books repeat some identical skeleton four times or more; the record-holders are וְאֵת at Joshua 7:24 (ten times in 36 words) and בֶּן in the genealogies. Nor is a book “owning” a word rare — Esther’s הַמֶּלֶךְ runs at more than four times הֶבֶל’s rate here. Repetition alone is not the finding. What is unusual at 1:2 is that the repeated word is the book’s subject, in a verse of eight words.

In the Greek file the match is exact. Row 1:2 runs to ten words. Remove the second ματαιότης ματαιοτήτων — slots six and seven — and what is left is the eight words of row 12:8, form for form, in order, with nothing over.

In the Hebrew the same deletion leaves a difference at exactly one slot — and it is the speaker’s name. 1:2 less its repeated pair gives הבל הבלים אמר קהלת הכל הבל; 12:8 reads הבל הבלים אמר הקוהלת הכל הבל. Six slots, five identical, and at the fourth the Hebrew has קֹהֶלֶת at the front of the book and הַקּוֹהֶלֶת at the end of it.

Compared on the consonantal skeleton after the mandatory cleaner, slot for slot. The page reports that the two verses stand in this relation and says nothing about which was written first or why they differ.

σύν is the ordinary Greek word for with, and in Greek it takes a dative. In the CATSS Greek Qoheleth it stands 34 times across 4,546 words, and 33 of those 34 are immediately followed by an accusative. The thirty-fourth, at label 9:15, is followed by a genitive.

Set that against every other Greek book in the same directory:

This is the one test on this page that could have come back the other way and did not. Every token had four cases available to it. Qoheleth’s 34 could have been ordinary σύν-plus-dative, in which case a high count would have meant only that this book talks about accompaniment a lot; and accusatives could have turned up scattered across the comparison books, making Qoheleth unremarkable in kind. Neither happened.

The honest weakness, stated up front. Seventeen of the 33 are neuter forms — τά, πάντα, πᾶν, ὅσα, τοῦτο — in which nominative and accusative look identical, so the case there is the tagger’s reading rather than a fact about the form. Sixteen are forms that cannot be anything but accusative — τήν (4), τόν (8), τούς, πάσας, πᾶσαν, πάντας. On those sixteen alone the contrast is still sixteen against nothing.

And where it is not. Only three of the 34 — 1:14, 2:17, 2:18 — fall inside this study, which holds about a fifth of the book’s words. That shortfall is not distinguishable from chance and no claim is made that these two chapters differ in kind from the rest of the book. This is a fact about Ecclesiastes, reported on the page that opens it.

1. “Every one of the 29 rows carrying σύν answers a Hebrew verse containing אֶת.” The arithmetic is right — 29 of 29, with no exceptions. It is still not evidence. Run the identical test on ordinary word pairs in the same two files and they all pass: θεός/אלהים 36 of 36, ἥλιος/שמש 32 of 33, ματαιότης/הבל 30 of 31, μόχθος/עמל 24 of 24. In two files that agree 222 numbers to 222, a co-occurrence test has no other answer available. And σύν/אֶת is the weakest of the five: 55 Hebrew verses carry an אֶת and only 29 of them have a σύν in the matching row.

2. “σύν stands at the same point in the verse that אֶת does — positions correlate at r = 0.997.” Also withdrawn. Every ordinary pair in these files lands at about the same figure: πᾶς/כל 0.993, καρδία/לב 0.996, ἄνθρωπος/אדם 0.992. The correlation measures that these two texts run in the same order, which is known already, and says nothing about σύν.

And there is a reason underneath both failures that is worth stating on its own. The CATSS file carries no Hebrew alignment column. It records, for each Greek word, that word’s own form, part of speech, parse and lemma — and nothing about which Hebrew word it answers. So a test of that kind cannot be run on these files at all; what was actually being measured both times was co-occurrence inside a shared verse number.

What survives is the case, not the correspondence. The page publishes the construction, on the Greek side alone, and does not claim word-for-word alignment with anything.

The CATSS Greek Qoheleth carries δέ exactly zero times in 4,546 words. Against Job at 53.7% of rows and Proverbs at 55.8%, that is striking.

But it is not unique, and an earlier draft of this page said it was. Greek Song of Songs is also exactly zero, in 2,025 words, and Sirach is near-zero at 51 tokens in 3.6% of its rows. The count stands; the word “alone” does not, and has been removed.

The first attempt to show that 1:2 is exceptional used a crude root-finder that stripped a leading ה as though it were the definite article. On הֶבֶל that is wrong — the ה is part of the root — so the verse’s five tokens were split across two buckets and 1:2 did not appear in the results at all.

The tables in stage 3 use no root-finder — only exact spellings, with the screen printed.

Paul writes that the creation was subjected to ματαιότης, and it is tempting to run that back to this verse. The measurement does not support it. The longest shared run of word-forms between Romans 8:20 and Ecclesiastes 1:2 is zero; the two verses share no form at all. Romans 8:20’s best run against any of the 222 rows is one word. And the dative form Romans carries, ματαιότητι, stands in this file at row 6:4 — not at 1:2.

They share a lexeme. That is worth saying, and it is all that is worth saying.

No Hebrew or Greek critical apparatus. No Qumran fragment (4QQoh^a exists and was not consulted). No Targum, Syriac, Vulgate, Aquila, Symmachus or Theodotion — which matters most here, because the σύν construction is exactly the sort of evidence those witnesses bear on, and this page reaches none of them.

And nothing on date. The relative שֶׁ־, the Persian-period loanwords and the σύν usage are all cited in the literature as dating evidence. This page counts what is on disk and draws no conclusion about when anything was written, in either language, or about who wrote it.

Chapter 2 spends twenty-three verses building houses, vineyards, gardens and pools, and then says I hated life. Only after that does 2:24 arrive: eat, drink, find enjoyment in your toil — and this is from the hand of God.

The order is not incidental. The comfort in this book is never offered instead of the hard sentence; it is offered on the far side of it.

And the counting has a limit worth naming. Twenty-nine occurrences of a phrase, five tokens in eight words, and thirty-three accusatives are facts a script can establish. Whether הֶבֶל means vanity, or breath, or something a reader only recognises after a long enough life, is not.