📄 Textual Commentary on the Ethiopic Text of Ezekiel 1–11

Summary

Knibb provides a verse-by-verse textual commentary on the Classical Ethiopic (Gǝʿǝz) version of Ezekiel 1–11, supporting his critical edition and Schweich Lectures. The underlying problem is that the translation was made from Greek sometime between the mid-4th and early 7th century, but the oldest known manuscripts are about a millennium later. The surviving text therefore carries several layers of later revision. Drawing on 15 collated manuscripts, he sorts the witnesses into an older group (Eth I) and a younger, Hebrew-corrected group (Eth II). Shared errors then divide Eth I into four families. Each note sets the Ethiopic against the Göttingen Septuagint, the Masoretic Hebrew, the Syriac Peshiṭta and a Syriac-based Arabic version. The notes record agreements, departures, habits of word choice and outright misreadings. The commentary confirms that the version sits with the Alexandrian Greek text. It also shows that Syro-Arabic influence was occasional rather than systematic, and that the Hebraizing revisions probably date to the 16th century. The work matters because it separates the original translation from later layers, so each reading can be traced to its source.

Key points

• 37 Ethiopic manuscripts of Ezekiel are known, dating from the early 15th to the 20th century. Ṭānāsee 9 was the oldest until Ted Erho reported a Leiden fragment in 2017. That fragment covers Ezek 37:23–48:13 and may date to the late 13th or early 14th century.
• 15 manuscripts were collated: 11 Eth I and 4 Eth II (Abbadianus 35, BL Add. 24,991, BL Or. 502, BL Or. 504).
• Eth I splits into four families:
  • Family 1: Ṭānāsee 9, EMML 26 and EMML 25.
  • Family 2: Berlin Petermann II Nachtr. 42, BL Or. 501, EMML 2080 and EMML 2082.
  • Family 3: Abbadianus 55, EMML 1768 and the Aksum Ṣǝyon manuscript, a smoothed text from around 1500.
  • Family 4: Cambridge Add. 1570 (1588/89) on its own, revised against the Hebrew.
• The Ethiopic follows the Alexandrian A-text closely. Among Greek minuscules it most often agrees with 106–410, and also with 534/130, which Ziegler judged the closest relative.
• Eth I is heavily abbreviated in chapters 42–48. Eth II fills these gaps from the Hebrew and builds on the Family 3 text.

Interesting ideas

• I1. Manuscript families are defined by errors they share, not by correct readings, since a common mistake is strong evidence of common descent. — For Mova (learn): When several witness parsers agree on silver data, check whether they share the same characteristic mistakes; correlated witnesses should count as one vote, not several.
• I2. "Double readings", where two alternative renderings sit side by side, reveal that a second source was blended into the text. — For Mova (learn): Flag outputs or training labels that look like two analyses merged together, and use them as markers of mixed provenance in silver data.
• I3. For each source word, Knibb tracks its usual equivalent and the exceptions. For example, ἀλλότριος is usually ነኪር፡ but sometimes ፀር፡ "enemies", and αἰχμαλωσία is sometimes rendered with a verbal construction. This separates contextual choices from errors. — For Mova (back): Keep a per-lemma table of expected renderings in translation through the meta-world, and log every deviation with a reason, so lexical choices stay explainable.
• I4. Some errors come from misreading an ambiguous form in the source. The Greek εἶπον was taken as a first-person indicative, which then produced a wrong translation. — For Mova (tokens): Keep competing morphological readings of ambiguous forms available to later stages rather than committing early, and record which reading was chosen.
• I5. How close the transliterations are to the Hebrew argues against an Arabic intermediary such as Saʿadya. Fine surface detail is used to rule out an indirect source. — For Mova (world): When the knowledge graph combines facts from several sources, use small surface fingerprints such as exact spellings or name forms to infer which source a fact really came from.