📄 Final ka(-mā), a Shared Innovation of (Traditional) South Semitic

Summary

The paper challenges the widely adopted "Central Semitic" grouping, which places Arabic with Northwest Semitic instead of with Epigraphic South Arabian and Ethiopic. Corriente first restates his earlier objections. The features cited for Central Semitic are, he argues, either minor, contradicted by dialect evidence, or artifacts of the prescriptive canon of Arab grammarians. He then answers the theory's supporters on their own terms. They stress shared innovations over retentions, so he offers a shared innovation of the traditional South Semitic branch: the purpose conjunction ka(-mā) "in order to". The method is comparative-philological. He gathers attestations from grammars, lexicons and texts, including Dillmann, Chaine, Reckendorf, Bauer, Johnstone, Raz and Andalusi Arabic sources. He traces how the conjunction's meaning shifted from comparison to time to purpose. The result is that all three South Semitic languages built their purpose clauses the same way, and North and East Semitic did not. He also argues this is not an accidental parallel change. The paper therefore supports the older Brockelmann-style classification of Semitic.

Key points

• The paper restates five rebuttals to the arguments for Central Semitic:
  • glottalisation of emphatics;
  • the supposed absence of geminate imperfectives in Arabic;
  • generalized prefix vowels, countered by the taltalah evidence;
  • the -t- perfect suffixes;
  • the alleged negative bal.
• The purpose conjunction appears in Epigraphic South Arabian as ky, in Ge'ez as kämä, and in Classical Arabic as kay/li-kay, with rare kamā. Kamā is more common in Andalusi Arabic, which the author links to South Arabian influence.
• The meaning developed from pan-Semitic comparative kv- to a temporal sense and then to purpose. The temporal sense survives in South Semitic: ESA bkn, Mehri k(ə), Ge'ez ʔəmkämä, Tigre kəm(sal), and Andalusi kamá "when". Andalusi kamá can also be causal.
• As supporting evidence, Arabic and Ethiopic both put the verb in the subjunctive after these conjunctions. The ESA script is unvocalized, so this cannot be checked there.
• Brockelmann observed that complex syntax arose separately in each Semitic language. A shared subordinator is therefore strong evidence for classification.

Interesting ideas

• I1. Weight shared innovations above shared retentions, because keeping an old feature proves little about close kinship. — For Mova (learn): When several witness parsers agree, treat agreement on unusual analyses as stronger evidence than agreement on default ones, and weight silver labels accordingly.
• I2. Hoenigswald's "overlapping innovations" warning says two sisters can make the same change independently, so a shared feature needs a test against accidental parallelism. — For Mova (learn): Check whether witness parsers share training data or architecture before counting their agreement, since correlated errors can pass for consensus; add negative controls for this.
• I3. Normative descriptions can hide real variation, as when Arab grammarians' analogy principle left out dialect forms and made features seem absent. — For Mova (tree): When converting between UD dialects or treebank conventions, record constructions the standard guidelines filter out, so the parser is not trained on a cleaned-up "mirage" of English.
• I4. The author confirms the conjunction finding with an independent grammatical correlate: the subjunctive mood required after it. — For Mova (frames): Accept a purpose or temporal reading of a clause only when a second cue agrees, such as the modal or infinitival form of the verb (e.g. "so that … could"), which keeps the choice explainable.
• I5. A single marker can move along a path from comparison to time to purpose to cause. — For Mova (world): Store this path in the compiled knowledge for English connectives such as "as" and "so", so event-order and causal links can be ranked by context rather than fixed to one sense.