📄 Elias Yemane: Amharic and Ethiopic Onomastics. a Classical Ethiopian Legacy, Concept and Ingenuity

Summary

This is a book review by Stefan Weninger in Aethiopica 8 (2005, pp. 251–257), not a research paper. It assesses Elias Yemane's monograph on Ethiopian personal names. Weninger first shows that onomastics has been neglected in Ethiopian studies, unlike the study of Asian Semitic languages, so a broad treatment fills a real gap. The book has two parts. The first covers religious history, outside influences, the meaning given to names, and name changes. The second covers name structure: names built on a finite verb, nominal names, names grouped by use (throne names, names for horses and slaves), and very productive Ge'ez construct-state compounds. The reviewer points out that Ethiopian practice differs from older Semitic patterns. A saint or Mary can fill the "sacred" slot, and these names express veneration rather than literal kinship. He calls the book pioneering but methodologically weak. It mixes formal and semantic categories, gives no attestations or distribution data, and repeats outdated claims about historical influences. In his view it is a foundation for future research rather than a finished account.

Key points

• Bibliographies show how little earlier work existed. Leslau's 1995 survey cites only two works. Lockot's 1982 bibliography lists three and files them under calendar, heraldry and numismatics. Dillmann's 1865 name index was long the only compilation.
• The book is Studies in Onomastics 7 (Edwin Mellen Press, 2004, about 300 pages). It deals mainly with personal names, and animal names get only pp. 174–178.
• The structural part covers only names formed with Amharic or Ge'ez material, so biblical names are left out. Its examples include the verbal name Abärra ("he enlightened") and construct-state names such as Gäbrä Mädhən, Haylä Maryam and Gäbrä Giyorgis.
• The reviewer's criticisms:
  • The book cites no sources showing how old the names are. He suggests indices such as Huntingford (1969).
  • It does not separate established names from individual coinages.
  • It ignores names from pre-Aksumite Sabaic and Aksumite inscriptions.
  • It overstates Jewish and Syrian influence on early Ethiopia.
• He recommends two directions for future work: studies of how names spread, past and present, and detailed analyses of single name types, citing Kleiner (2004) and Revol-Tissot (2003) as models.

Interesting ideas

• I1. Many Amharic names are whole clauses built on a finite verb, such as "he enlightened", so a name's internal grammar can look like a sentence even though it works as a single referent.
• I2. A name's literal meaning should not be read as a claim about the person; "child of Mary" signals devotion, not parentage.
• I3. The reviewer's main methodological complaint is that formal and content categories are mixed in one classification, which should keep two separate axes.
• I4. Claims that something is old or widespread need attested sources and frequency data, not just plausible narrative.
• I5. Words can move between name classes, for example the toponym Aksum used as a personal name, or Kasa meaning "compensation" used as a replacement name.

For Mova

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

• I1 (tree): Treat a recognised multi-word name as one flat proper-noun unit (UD flat/compound). That stops the parser from building a clause around a verb hidden inside the name.
• I2 (world): Never derive kinship or holdings from the etymology of a name. Keep name meaning as optional metadata, separate from facts about the entity.
• I3 (tokens): Keep morphological or structural features of a name apart from semantic or type labels in the lexicon. Each axis can then be checked and explained on its own.
• I4 (learn): Each compiled knowledge item or induced repair rule should cite its evidence and corpus frequency. Unattested items should be flagged as individual coinages, not general rules.
• I5 (world): Let an entity's type come from context rather than from a fixed lexicon entry. A known place name can then be resolved as a person when the frame requires it.