📄 [rev. of] S. BETTEGA, L. D’ANNA, "Gender and Number Agreement in Arabic", Brill, 2023

Summary

Giuseppe Petrantoni reviews (RiCognizioni 10, 2023) the Brill monograph by Bettega and D'Anna, which explains how gender and number agreement works across written and spoken Arabic and how that system developed over time. The book has five chapters:

• a critical survey of earlier scholarship;
• a synchronic account of the morphological markers and agreement patterns found across the dialect continuum;
• a Semitic-comparative history of agreement in written Arabic;
• the treatment of agreement by Arab grammarians from Sībawayh to al-Šidyāq;
• the history of agreement in the spoken dialects.

The authors sort nouns into three agreement classes using both form and meaning. They group dialects into Types A, B and C, plus a set of "divergent" varieties. They argue that feminine singular agreement with plural nouns probably arose together with the spread of broken plurals, and that in the dialects it is an old retention rather than an innovation. In contrast, they attribute the loss of feminine plural agreement to external, non-linguistic pressures. The reviewer praises the clarity, the tables and the multilingual bibliography, and recommends the volume.

Key points

• The book is Studies in Semitic Languages and Linguistics vol. 109 (Brill, 2023, 407 + XXI pages). It covers Old Arabic (Classical, as codified by early grammarians) and New Arabic (spoken varieties).
• The authors use two representative varieties for side-by-side paradigm tables (2.8–2.15):
  • Najdi stands for gender-distinguishing dialects.
  • Damascus stands for non-distinguishing dialects.
• The three agreement classes are:
  • Class I: male humans, which take masculine plural -īn.
  • Class II: female humans, nonhuman animates and inanimates, which take -āt.
  • Class III: masculine nonhumans and inanimates, which also take -āt.
  • In all three classes, feminine singular agreement remains an optional alternative.
• The dialect types are:
  • Type A (table 2.20).
  • Type B (2.21): no gender in the plural, but F.SG agreement is kept.
  • Type C (2.22): mostly a single "common plural".
  • Divergent varieties, such as Ḥassāniyya and contact-shaped Darfur Arabic.
• The survey finds two things:
  • Every gender-distinguishing dialect shows F.PL agreement with nonhuman plurals.
  • F.SG agreement with plurals is possible in all of them.
• The human/non-human split appears in the earliest grammars. However, the obligatory F.SG rule for non-human plurals was only fixed in the 19th century.

Interesting ideas

• I1. Agreement classes are defined by crossing a formal property (grammatical gender) with a semantic one (humanness/animacy), rather than by either alone. This captures patterns that neither criterion explains by itself. — For Mova (tokens): The English morphology layer could carry an animacy/humanness feature alongside Number. This would let it predict which/who and singular/plural concord with nouns like "staff" or "police".
• I2. Some controllers allow more than one legitimate agreement pattern, so an "alternative option" is a valid outcome, not an error. — For Mova (tree): The parser's agreement checks could accept both singular and plural verbs with collective subjects. They would record which option was taken instead of penalising one of them.
• I3. Dialects are profiled as discrete agreement types (A/B/C/divergent), each documented by one representative variety with comparable tables. This turns messy variation into a small, checkable typology. — For Mova (tree): UD dialect conversion could keep a compact profile table per treebank describing how it annotates agreement features. Conversions could then be validated against that profile.
• I4. The study contrasts when a rule was codified by grammarians with what texts and corpora actually show. It reveals that a supposedly ancient norm was formalised late. — For Mova (learn): Evaluation could separate prescriptive expectations from observed corpus usage. This avoids training repair rules toward a norm the data does not follow.
• I5. Agreement on the target (pronoun, verb, adjective) is treated as information inherited from a controller noun, with humanness affecting the form. — For Mova (frames): Coreference could weigh pronoun number and humanness against candidate antecedents. For example, it could allow plural "they" for a singular collective, but only when the antecedent is human.