Jeremiah 10 is not in the same order in all three
The Masoretic chapter runs straight through. The Greek has no verses 6 to 8 and no verse 10, and it prints verse 9 before verse 5. The scroll does the same, and splits verse 5 around it.
The arrangement in the middle row is an editor's, and the section below is about whether it can be believed. The two outer rows are not: they are what the Masoretic Text and Rahlfs's Greek actually print.
Two editors say the same thing about verses 6, 7, 8 and 10 independently. Rahlfs simply has no such verses. Stipp, setting the two editions side by side in columns, leaves the alexandrian column of all four of them empty, which is his statement that the Old Greek has nothing answering to them at all. And his apparatus names 4Q71 at 10:6 and 10:10 — he read this manuscript here too.
The Masoretic chapter is too long for the fragment
Only 114 of the 1114 letter-places of this fragment are extant, about 10% of it, and what survives falls near the end of each line. That is little to read, but enough to measure: it fixes where particular words stood, and the lines between them have a known width.
Each bar is one line of the fragment, drawn to its length; the dark marks are the extant text. The two outlined lines are 5 and 6.
One line, and two texts asking to go into it
Line 5 ends with הב ייפהו במקבות, from Jeremiah 10:4. Line 6 ends with תכלת וארגמן, from 10:9. They are consecutive lines of one fragment, so whatever stood between those two words occupied a single line.
The Masoretic arrangement asks for two and a half lines of writing in the space of one. It will not go in. That is the argument the editors make for this scroll, and it does not depend on a single reconstructed letter: it depends on where the extant text stands and how wide the lines are.
What this does not show
It shows that 10:5–8 as the Masoretic Text has them were not there. It does not by itself show that verse 9 stood before verse 5, or that verse 10 was missing: the fragment has room for the shorter text in several arrangements, and the one printed above is the editor's, chosen because the Greek has exactly that order. Between the extant text of 10:9 and that of 10:11 the sums do not decide anything — the Masoretic text with verse 10 and the scroll's text with the second half of verse 5 come to nearly the same length.
What the manuscript reads
Two adjacent words that are both extant are a join, and a join is testimony: whatever the Masoretic Text has between them did not stand in this copy. It is the only kind of omission a manuscript can attest, since an omission within a lacuna is the modern editor's and not the scribe's. Below, the words in blue are extant; the words in ochre are what the Masoretic Text sets between them.
נבוזרדן רב־טבחים את
Masoretic only, glossed chief, butcher.
אחיקם בן־שׁפן ואת
Masoretic only, glossed son, Shaphan.
[ו]י̇באו̇ עד תחפחס
Masoretic only, glossed unto.
בפתח בית־פרעה בתחפנחס
Masoretic only, glossed house, pharaoh. The Greek has these words, so here the scroll is shorter than the Masoretic Text and the Greek alike.
אשר
This word is extant, and the Greek has nothing answering to it, so here the manuscript agrees with the Masoretic Text.
Three of the five go with the Greek, one with the Masoretic Text, and one with neither. That is the texture of the thing: 4QJerᵇ's neighbour 4QJerᵈ is close to the text behind the Greek without being a copy of it.
4QJerb has no such join at all
Its extant text is 114 letters in 15 discrete runs, and no two of its words are adjacent in the manuscript. Everything above about Jeremiah 10 rests on the extent of the lacunae and nothing on two words standing next to each other. The two manuscripts are evidence of different kinds, and worth keeping apart.
The same test where it ought to find nothing
Three omissions in one manuscript is three data points, and no way to tell a finding from a habit of the method. The other four Jeremiah manuscripts from the Judaean Desert are of the Masoretic type, so they are the null: if the join test is reading the shape of a text rather than the shape of the damage, they should carry the bracketed pluses that 4QJerd omits.
| manuscript | extant | verses | decidable pluses | carries | omits | omitted |
|---|---|---|---|---|---|---|
| 4QJerᵈ4Q72a | 45% | 9 | 4 | 1 | 3 | 75% |
| 4QJerᵇ4Q71 | 10% | 22 | 0 | 0 | 0 | — |
| 4QJerᶜ4Q72 | 34% | 116 | 21 | 21 | 0 | 0% |
| 4QJerᵃ4Q70 | 33% | 140 | 17 | 17 | 0 | 0% |
| 2QJer2Q13 | 25% | 47 | 3 | 3 | 0 | 0% |
| 4QJerᵉ4Q72b | 33% | 3 | 1 | 1 | 0 | 0% |
A plus inside a lacuna is not evidence either way, so the denominator is the pluses a manuscript can actually answer for.
Not one of the 42 bracketed pluses these four could answer for is missing from them, against 3 of 4 in the pair. Fisher's exact test gives p = 0.00026. The instrument is reading the edition.
Why this is a null and not silence
In the whole book they produce one join between them, 4QJerᶜ at 21:7, where the manuscript sets החרב against ביד and the Masoretic Text has ומן־הרעב between them. Stipp does not bracket it, so the Greek has it too: it is this manuscript’s own omission, most likely an eye slipping from one מן to the next. The test does find omissions here. It never finds one over a Masoretic plus.
One rule had to be added to get here. A single letter is not an anchor: the ETCBC writes a prefixed particle as a word of its own, and at 4QJera 14:4 the alignment paired the bet of the manuscript’s בארץ with the bet of the Masoretic בעבור ten words earlier, then reported the ten words between as excluded. Both anchors and the gap must now be two letters or more. It removes three of the four joins the Masoretic-type manuscripts had and none of 4QJerd’s, so it makes the contrast above narrower to claim, not wider.
The passages in full
The Masoretic verse, the scroll beneath it, and Rahlfs's Greek beneath that. In the Masoretic line, ochre marks a word the scroll does not carry; a tinted ochre marks one the extant text excludes rather than the editor. In the manuscript line, grey letters in brackets are the editor's restoration, a dot above a letter marks an uncertain reading, and ° is a letter surviving only as traces.
4QJerᵇ — Jeremiah 9:22 to 10:22
no part of this verse is in the scroll
the Greek has no such verse
no part of this verse is in the scroll
the Greek has no such verse
no part of this verse is in the scroll
the Greek has no such verse
no part of this verse is in the scroll
the Greek has no such verse
no part of this verse is in the scroll
no part of this verse is in the scroll
no part of this verse is in the scroll
no part of this verse is in the scroll
4QJerᵈ — Jeremiah 43:2 to 43:10
How this was measured
The Masoretic Text is the ETCBC BHSA database. The scrolls are the ETCBC dss dataset, Martin Abegg's transcription, in which every sign carries a flag saying whether it is extant or restored; nothing on this page treats a restored sign as a witness. The Greek is Rahlfs, from the LXX dataset. Greek Jeremiah runs on its own versification: Greek 50 is Masoretic 43, while in chapter 10 the numbers agree and it is the verses themselves that differ.
For the question of which Masoretic words are absent from the Greek, the page follows Hermann-Josef Stipp, Textkritische Synopse zum Jeremiabuch, 15. korrigierte interne Auflage (2021), as parsed in the companion study. Every reading it is relied on for here was checked against Rahlfs directly.
Why the Greek is Rahlfs and not the Synopse's own panel
The Synopse prints the Greek beside the Hebrew sentence by sentence, which is a finer alignment than the verse-level one used here, and it would be the natural text to set. Checked against Rahlfs verse by verse over both passages, it agrees on 32 of 34. The exceptions are 10:5 and 10:9, and they fail in one particular way: each carries the other's Greek. Stipp's panel for 10:5 is 29% similar to Rahlfs at 10:5 and 99% similar to Rahlfs at 10:9, and the panel for 10:9 is the mirror of that.
That is where the two editions are transposed, and it is the fault of reading the panel rather than of the edition: Stipp marks the transposition himself, with his star and a margin cross-reference. But a Greek panel laid out against a Hebrew column running in Masoretic order cannot show the Greek's own order, and the Greek's own order is what the first section of this page is about. Rahlfs read as a text keeps it. So the Greek here is Rahlfs, and the Synopse is used for what only it can give: which Masoretic words the Old Greek lacks.
One caution about the apparatus citations quoted above. The parse takes Stipp's margin apart into spans, so the manuscript siglum and the place it stands at are reliable, while the wording around them is not necessarily whole. Nothing on this page rests on the wording of a marginal note.
The measurement is in src/collate.py and the page in
src/build.py. No figure or reading on this page is typed in by
hand; all of them are computed and can be reproduced by running those two.