📄 Robert Hetzron (ed.): The Semitic Languages

Summary

This is a critical review, written in German, of the reference volume The Semitic Languages edited by Robert Hetzron (Routledge Language Family Descriptions, 1997). The volume appeared shortly after Hetzron's death. The reviewer, Rainer Voigt, goes through the chapters most relevant to Ethiopian and South Semitic studies and checks them against his own comparative work.

On classification, he argues that Old South Arabian lacks the iparras-type imperfective. That points to Central Semitic affiliation, not to a South Semitic grouping with Ethiopic. He also names exclusive Ethiopic innovations that could define the branch. On scripts, he corrects claims about:

• Ge'ez and Amharic orthography
• vowel notation
• Latin-script Semitic languages
• the supposed non-uniqueness of Semitic root-and-pattern word structure

He also flags omissions, such as Syriac and the Tigrinya script reforms, outdated facts about post-1991 Ethiopia, and a faulty language map. His verdict is that, despite these errors, no other single work surveys the Semitic family so completely.

Key points

• The reviewed book (Routledge, 1997; xx + 572 pp.) has 23 chapters by 24 authors in three parts:
  • classification, scripts and grammatical traditions (pp. 3–65)
  • sketches of nine ancient languages (pp. 69–260)
  • modern languages
• Voigt cites his 1987 Journal of Semitic Studies paper. There he places Old South Arabian with Central Semitic because the iparras form is missing.
• He proposes these exclusive Ethiopic innovations:
  • the auxiliary hlw
  • strong conjugation of middle-weak verbs (e.g. Ge'ez yǝmäwwǝt), which survives in Amharic yǝmot
• Gaps he points out:
  • no chapter on Syriac or on the Syriac grammatical tradition
  • the extra Tigrinya letters and the orthography made official in 1993 are missing
  • Gafat (now extinct) is placed in an obsolete province
• The opening map is wrong in several places. It treats all of Ethiopia and whole Near Eastern states as Semitic- or Arabic-speaking, even though state borders rarely match language borders.

Interesting ideas

• I1. When sorting related varieties into groups, give verbal-morphology features more weight than features like a definite article, which can arise independently in unrelated branches.
• I2. A branch should be justified by an innovation that only its members share, not by overall resemblance.
• I3. A spelling that looks etymological may really be morphological: Amharic writes a contracted auxiliary so that it parallels the form carrying an object suffix.
• I4. The consonantal root is a real unit for speakers, letting them see different words as one family; the consonant-only script grew out of that awareness.
• I5. A political boundary is not a language boundary, so mapping states directly to languages produces errors.

For Mova

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

• I1 (tree): When detecting or converting UD dialects of a treebank, use systematic differences in verb-morphology annotation as the main signal and treat surface particles as weak evidence.
• I2 (learn): When grouping witness parsers or error clusters, define a cluster by the errors only its members make, so shared common mistakes do not create false families.
• I3 (tokens): The tokenizer and lemmatizer should first try to explain odd spellings by paradigm consistency before assuming archaic or etymological forms.
• I4 (tokens): Lemma tables could store a shared derivational base linking related English words (e.g. agent noun and verb), so later stages can relate them explicitly instead of by string similarity.
• I5 (world): In the Wikidata-based ontology, keep country, region and language-area entities separate, and never infer a place's language from its state alone.