📄 An ‘Old Amharic’ Commentary on the Nicene Creed

Summary

Appleyard edits and translates a short religious text: an Amharic commentary on the Nicene Creed, copied at the end of a Psalter and Wəddase Maryam manuscript held by the Ethiopian church of Däbrä Gännät in Jerusalem. He could not see the manuscript, so he works from Roger Cowley's private transcription.

The text has two parts:

• a long historical preamble on Arius and the Council of Nicaea, which the author checks against the known chronology;
• a phrase-by-phrase treatment of the Creed, giving each Geʿez phrase with an Amharic rendering and, in places, theological discussion framed as a dispute.

The central linguistic question is what "Old Amharic" means here. The author shows that this is a later commentary register, not the most archaic stage of the language. It mixes conservative traits, such as retained laryngals, with modern forms. He also separates spelling quirks from genuine grammatical differences.

The edition gives scholars a dated example of the andəmta commentary style together with a catalogue of its non-standard features.

Key points

• The manuscript used carries the shelfmark JE 48 E, also catalogued as Dabra Gannat MS 186. Related preamble versions appear in d'Abbadie 101, EMML 2468 and EMML 7007. The article appeared in Aethiopica 6 (2003), pp. 111–136, not 2013.
• The preamble runs to more than 800 words. The commentary also covers the additions to the Creed prompted by Macedonius' heresy.
• Recurring formulas organise the commentary:
  • "it means" introduces a translation;
  • "if one says", "if I say to you" and "if you say to me" open question-and-answer disputation.
• The text lacks the copula -tt, a key marker of genuinely Old Amharic.
• It does show several older traits:
  • consistent retention of laryngals;
  • a short 1st-plural prefix n-;
  • fully inflected auxiliaries in some 3rd-plural compound verbs, alongside standard forms;
  • distinct verbal extensions and older lexical items.

Interesting ideas

• I1. The author refuses to read every odd spelling as a linguistic fact: he separates scribal ambiguity (such as confusable vowel orders) and morpheme-by-morpheme spelling from real pronunciation or grammar.
• I2. Each archaic feature is listed with its counter-examples, so the variety is described as a gradient mix rather than a yes/no label.
• I3. Fixed formulas ("it means", "if one says") reliably mark whether a passage is a gloss or an objection-and-reply move.
• I4. The author rebuilds the true historical timeline in order to show where the preamble's account of Arius is garbled, and flags dates that do not add up.
• I5. Provenance is kept visible throughout: the edition relies on a transcription rather than the original, marks later insertions in superscript, and notes page breaks.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tokens): the normalizer could mark each non-standard token as "orthographic" or "grammatical" before morphology, so lemmas are not polluted by spelling noise.
• I2 (tree): dialect or register detection before UD conversion could record per-feature ratios instead of one register tag, so conversion rules fire only where evidence is strong.
• I3 (frames): discourse-marker lexicons could map such formulas to speech acts (definition, hypothetical objection, rebuttal), giving explainable structure to argumentative texts.
• I4 (world): checking a text's event order against compiled knowledge (e.g. Wikidata dates) could flag inconsistencies explicitly instead of silently trusting the narrative.
• I5 (learn): silver training data could carry similar provenance flags (source layer, later edits), so evaluation can exclude or down-weight second-hand material.