📄 Transitivity and intransitivity in the grammar of Bar Hebraeus

Summary

Bohas examines how the Syriac scholar Bar Hebræus (BH) took the transitive/intransitive distinction from the Arabic grammarian Zamaḫšarī. He also shows how BH adapted it to Syriac. Earlier Syriac grammar had ignored the distinction because Dionysius Thrax, its model, does not discuss it. The method is philological. Bohas translates and annotates BH's chapter on verb valency and compares it point by point with Zamaḫšarī's Mufaṣṣal. BH keeps the Arabic idea of "causes of transitivity": derived verb patterns that turn an intransitive verb into a transitive one, or a transitive verb into a doubly transitive one. Syriac lacks case endings, so BH does not need Zamaḫšarī's main concern, ensuring every complement is accusative. He therefore drops the direct/indirect split. Instead he adds an analysis of where these causes fail, either giving no transitivity or giving an unrelated meaning. He also gives a detailed account of labile verbs, which are both transitive and intransitive. The study shows a borrowed grammatical concept being reworked to fit a different language's structure.

Key points

• BH's chapter has four sections:
  • examples of intransitive and transitive verbs;
  • causes of transitivity;
  • failure of those causes;
  • verbs that are both transitive and intransitive.
• BH matches two Syriac derivations to Arabic patterns:
  • prefixing ʾ/m/n/t corresponds to Arabic form IV ʾafʿala (e.g. škeb "lie" → ʾaškeb "make lie down");
  • vowelling the first consonant with a corresponds to form II faʿʿala (e.g. bṣar "be small" → baṣar "diminish").
• Transitive verbs may take one, two or three objects. A trailing purpose clause is treated as explaining an existing complement, not as a fourth one.
• Syriac often marks the direct object with the preposition l-, much like Hebrew ʾet. This is one reason BH does not separate direct from indirect transitivity.
• BH works from written letters, not sounds, so he treats baṣar as a vowelled trilateral regardless of Eastern gemination. He fills his chapter with scriptural examples, far more than the Arabic grammars do. Zamaḫšarī and BH lived more than a century apart.

Interesting ideas

• I1. Valency-changing morphology is modelled as a "cause" with a predicted effect, and the cases where it does nothing or yields an unrelated sense are listed explicitly. This guards against assuming derivations are always compositional.
• I2. A preposition is counted as a legitimate route to transitivity, so a PP-marked participant can be the true object. In Arabic, "go out with him" is treated as equivalent to the causative "brought him out."
• I3. The direct/indirect object distinction is dropped where the language's surface marking does not support it. Categories are tailored to the language rather than copied from the source tradition.
• I4. When counting a verb's complements, BH separates true arguments from material that merely explains an argument.
• I5. Labile verbs, which work both transitively and intransitively, are given a dedicated section rather than folded into either class.

For Mova

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

• I1 (world): Verb-class entries for derived or causative forms could carry an explicit "does not alternate / idiosyncratic sense" flag, so the knowledge layer never infers a causative reading by rule alone.
• I2 (frames): Role assignment could map certain verb + preposition pairs onto the same patient/theme slot and causative frame as the bare transitive, rather than demoting the PP to a peripheral oblique.
• I3 (tree): When converting UD dialects or importing treebanks, the obj/obl/iobj decisions for preposition-marked objects should follow language-specific evidence, and these conversion choices should be documented.
• I4 (frames): Frame building could test whether a purpose or relative clause elaborates an existing role before opening a new role slot, which avoids inflated argument structures.
• I5 (world): Alternating verbs (e.g. ergative pairs like break/open) could be stored as one entry with two linked frames, so the world model tracks the same state change whichever syntax is used.