📄 The problematic Hebrew word ‘ועבר’ Nahum 1:12: The study of ancient translations

Summary

Chia investigates a single verb in Nahum 1:12, ועבר ("and he/it will pass"), on which the ancient versions disagree. The LXX leaves it out. The Vulgate and the Masoretic text keep a singular verb. The Peshitta and the Targum read a plural with the people of Nineveh as subject, and the Targum even adds "over the Euphrates." Using reasoned eclecticism, the author weighs external evidence (age, reliability and spread of manuscripts) and internal evidence (scribal probability, authorial probability, lectio difficilior, lectio brevior, harmonisation). He also adds witnesses missing from the BHS and BHQ apparatuses. Several of these turn out to be unusable for this verse because they are fragmentary or silent. The surviving testimony, led by Mur 88 (c. 132–135 CE), supports the singular consonantal text. Contextual analysis then tracks grammatical persons across Nahum 1:9–14 and identifies the subject as the "wicked counsellor" of 1:11. The verse thus pairs a collective judgement (the Ninevites cut off) with an individual one (the counsellor passing away), framed by imagery of receding waters. This matters because it settles a disputed reading and its referent.

Key points

• The versions fall into four camps: omission (LXX), singular with the Lord as subject (Vulgate et pertransibit, MT), plural with Nineveh as subject (Peshitta), and a plural expanded with "the Euphrates" (Targum).
• Witnesses added beyond BHS/BHQ include 8Hev XII gr (Rahlfs 943), Mur 88, the Hexapla and 4QXII g.
• The LXX, Hexapla, 8Hev XII gr and 4QXII g are set aside for this verse. 4QXII g preserves only Nahum 1:7–9, 2:9–11 and parts of chapter 3.
• Mur 88 is the earliest witness supporting ועבר. Its fragmentary text agrees with the MT.
• Longman explains the plural emendation ורבעו as haplography or faulty word division before the next word, which begins with ו. The author also argues that Gelston (BHQ) wrongly assimilates the LXX to the Peshitta and Targum.

Interesting ideas

• I1. Witnesses with nothing to say about a reading are removed from the evidence rather than counted as neutral or dissenting, so only informative sources are weighed. — For Mova (learn): In silver-data agreement, a witness parser that failed on a span could abstain instead of voting, so its broken output does not dilute or distort consensus.
• I2. The ambiguous subject is resolved by tabulating every person and number marker in the surrounding verses and seeing which referent fits the singular slot. — For Mova (frames): Coreference for null or ambiguous subjects could build a passage-wide table of person/number/gender agreement and pick the antecedent from it, giving an explainable trace.
• I3. Divergent readings are explained by a concrete copying mechanism, such as a letter lost at a word boundary, rather than just being listed as variants. — For Mova (tokens): Tokenizer repair rules could model specific boundary-error mechanisms (merged or split words, a dropped letter) to propose and justify alternative segmentations.
• I4. Lectio difficilior and harmonisation treat smoother, context-conforming readings as suspect, because copyists tend to normalise. — For Mova (learn): When an LLM teacher's analysis is "smoother" than what several parsers agree on, that fluency could be flagged as possible normalisation bias rather than trusted.
• I5. Each ancient translation is read as evidence of how its translator parsed the source, for example which subject they assumed. — For Mova (back): Translations of a sentence could serve as extra witnesses for disambiguating its UD tree and frames, since a target-language rendering often commits to one reading.