📄 Typology of case alignments in NENA dialects

Summary

The paper describes how North-Eastern Neo-Aramaic (NENA) dialects mark subjects and objects in the perfective. Aramaic was originally nominative-accusative. Through contact with Iranian languages, which show tense/aspect-conditioned split ergativity, NENA came to use ergative-style marking (the so-called L-suffixes) on verbs built from the past stem.

Barotto surveys descriptive grammars of Jewish and Christian dialects (Sanandaj, Urmi, Barwar, Hertevin and others). Following Doron and Khan, she groups the dialects into three types, according to how far the ergative suffix has spread to intransitive verbs: Split-S, Dynamic-Stative and Extended Ergative.

She explains which intransitives receive ergative marking using Hopper and Thompson's graded notion of transitivity, with subject volitionality and animacy as the main factors. Her central claim concerns the Extended Ergative type. It looks ergative in markedness, but S and A are coded the same way, so it is really a transitional stage on the way back to accusative alignment rather than a stable system.

This matters because extended ergativity is cross-linguistically rare. The NENA cases also differ from African marked-nominative systems, so they are useful evidence for studies of how alignment changes.

Key points

• Split-S dialects (e.g. Jewish Sanandaj): ergative marking covers transitive and unergative verbs, while unaccusatives keep absolutive marking. Examples are "kalba nwəxle" ('the dog barked', ergative) and "brati qima" ('my daughter rose', absolutive).
• Dynamic-Stative dialects (e.g. Jewish Urmi):
  • All intransitives take ergative marking in the preterite.
  • Absolutive marking survives only in a perfect-stative form, so "brati qima" means 'has risen' and "brati qəmla" means 'rose'.
• Extended Ergative dialects (e.g. Christian Barwar): the absolutive is no longer used for intransitive subjects. The marking pattern looks ergative, but S and A are treated alike, which is a nominative-type alignment.
• Lexical variation across dialects:
  • Verbs such as 'bark', 'sneeze', 'dance' and 'convert to Islam' are marked differently from dialect to dialect.
  • Jewish Urmi is the only dialect that treats 'bark' as intransitive.
  • Verbs that take an object, such as 'see' and 'hear', are always coded transitively, even when their subject lacks agency.
• Further splits are mentioned only briefly:
  • New accusative markers, which in some dialects are obligatory for 1st/2nd person objects.
  • The "Hertevin paradigm", in which subject endings change when a pronominal object is explicit.

Interesting ideas

• I1. Transitivity is treated as a scale of semantic properties (agent control, instigation, punctuality, dynamism) rather than a yes/no question about whether there is an object, which explains why 'bark' and 'sneeze' pattern differently.
• I2. The relationship is asymmetric: having an object is enough to force transitive coding, but lacking one does not guarantee intransitive coding, so experiencer subjects of 'see' and 'hear' are still coded as transitive.
• I3. How marked a form is and which arguments it groups together are kept as separate dimensions, so a system can look ergative in its morphology while behaving nominatively.
• I4. The same verb can be marked differently across varieties, so the paper states its generalisations as tendencies with item-level exceptions (e.g. Urmi 'bark').
• I5. The paper contrasts a resultant state ('has risen') with a past event ('rose') of the same verb, each with its own marking.

For Mova

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

• I1 (frames): Score each intransitive subject on these features to choose between an agent-like and an undergoer-like role, instead of using one undifferentiated subject role.
• I2 (frames): Let a syntactic obj trigger a transitive frame while still labelling the subject as experiencer rather than agent.
• I3 (tree): During UD dialect or treebank conversion, keep case/agreement features separate from deprel assignment, and do not derive grammatical relations from surface marking alone.
• I4 (world): Store compiled verb-class knowledge as defeasible defaults with per-lemma overrides, rather than as hard rules.
• I5 (world): Model perfect forms as a state that holds at reference time and simple past forms as an event in the timeline, so questions about current states can be answered.