📄 Robert R. Ratcliffe: The "Broken plural" problem in Arabic and comparative Semitic - Allomorphy and analogy in non-concatenative morphology

Summary

This is not a research article but a short German-language book review by Rainer Voigt, published in Aethiopica 3 (2000), pp. 255–257, not 2013 as given in the title line. It assesses Robert R. Ratcliffe's monograph on Semitic "broken" (internal) plurals, which grew out of his 1992 Yale dissertation. The book classifies Arabic broken plural patterns. More than half of it compares plural formation across Semitic, especially the former "South Semitic" group, and it attempts a staged reconstruction of Proto-Semitic and Proto-South-Semitic plural systems.

Voigt credits Ratcliffe with structuring Arabic plurals more systematically than earlier work and with bringing out the shared plural morphology of Ethio-Semitic, Arabic, Old South Arabian and Modern South Arabian. His criticisms are mostly methodological:

• transcriptions that hide inflectional endings and junctures;
• a questionable definition of the stem;
• derivations that rely on implausible glide segments or unattested intermediate forms;
• reconstructions that are internal rather than truly historical.

He nonetheless calls the data-driven book stimulating reading.

Key points

• The reviewed book is Ratcliffe (1998), John Benjamins, Current Issues in Linguistic Theory vol. 168, xi + 261 pages, priced Hfl 168 / $84.
• Transcription: Writing Arabic forms as unsegmented strings like kalbun hides the case or mood ending. Voigt prefers marking the stem boundary, and he argues that stem prefix vowels belong to the stem.
• Specific derivations:
  • The odd form ʔajrin "young dogs" is better explained by paradigm leveling or analogy with weak-final roots than by an unattested pattern.
  • In biibaan (plural of baab), n is not a believable hiatus filler; in Arabic, glottal stop fills hiatus after a long vowel.
• Syllable templates: The final case-vowel syllable should be ignored. Suffixed forms such as fu.ʕuu.lu-ka show that the weight of this syllable is irrelevant to the pattern.
• Other objections:
  • The Ge'ez transcription, which follows Lambdin, writes the central vowel as e and thereby misleads readers.
  • Some reconstructions, such as ʔaCCuC derived via dissimilation (p. 226), seem far-fetched.
  • The strict ordering of "level I" (internal) before "level II" (external) morphology is not justified.

Interesting ideas

• I1. Represent words with explicit stem and ending boundaries rather than flat strings, so that inflection is visible and junctures can be analysed.
• I2. Prefer analogy or paradigm leveling over inventing unattested intermediate forms when explaining an irregular form.
• I3. Match shape templates on the stem only, excluding inflectional material whose properties do not affect the pattern.
• I4. Do not assume an ordering between rule levels; justify why one layer must apply before another.
• I5. Notation should not suggest false relations, as when a symbol makes one sound look like the short counterpart of another.

For Mova

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

• I1 (tokens): Lemmatizer output could keep a marked stem+affix segmentation, making every morphological decision inspectable rather than implicit.
• I2 (tokens): For irregular English forms, the morphology module could record an explicit "analogical/lexical exception" reason instead of chaining ad hoc rewrite steps.
• I3 (tokens): Pattern-based morphology rules should first strip inflection and then match, which avoids spurious template variants.
• I4 (learn): Self-induced repair rules could have their application order tested against held-out data with negative controls, rather than fixed by assumption.
• I5 (back): Labels in the meaning graph and generated glosses should be chosen so that surface similarity of symbols never implies a relation that the graph does not contain.