Causative verbal forms

[jheinecke]

I've been working for a while on Chechen (an ergative language). Chechen has conjugated causative forms for verbs: e.g. охьадожо "to drop [something]" --> охьадожадайта "make [somebody] drop [something]. The enforcer (Causer) of the action is marked by the Ergative case, the Causee in a Locative case, all depend from the verb

Ӏумара   Аслане    куьзга     охьадожадайтира.
Omar-ERG Aslan-LOC mirror-ABS make-drop-PAST
"Omar made Aslan drop the mirror"

I wonder which dependency relation is appropriate for the causee, Aslan: obl does not seem appropriate

1   Ӏумара      Ӏумар       PROPN   _   Case=Erg    4   nsubj   _   Gloss=Omar.erg.sg
2   Аслане      Аслан       PROPN   _   Case=All    4   obl _   Gloss=Islam.all.sg
3   куьзга      куьзга      NOUN    _   Case=Abs|Number=Sing    4   obj _   Gloss=mirror.abs.sg
4   охьадожадайтира охьадожо    VERB    _   Tense=Past  0   root    _   Gloss=pverb.fall.caus.rpst

Another idea is to use obj:agent for the causee, and keeping obj for the object of the caused action

1   Ӏумара      Ӏумар       PROPN   _   Case=Erg    4   nsubj   
2   Аслане      Аслан       PROPN   _   Case=All    4   obj:agent   
3   куьзга      куьзга      NOUN    _   Case=Abs|Number=Sing    4   obj
4   охьадожадайтира охьадожо    VERB    _   Tense=Past  0   root    

I couldn't find many examples from other languages (a part from the French faire faire quelqu'un quelque chose). What is your opinion?

[Stormur]

If it's in the locative, I think Aslan well deserves an obl relation... why do you think that it is not appropriate? Of course, if the locative is not used for clear direct objects elsewhere in the language, or if Aslane does not behave as a true secondary object (then iobj). What happens with passivization?

You could always specifiy this particular argument with a subrelation like obl:cau. At the moment it is registered for Telugu, but there's no documentation.

[jheinecke]

The reason why I'd prefer not to use obl is that I think this argument (Aslan) is more "core" than not. In French-GSD (even htouhg French uses a faire + infintive for the cuasative) the Causee is annotated as obj:agent (cf: UD_French-GSD/fr-gsd-train.conllu, sentence fr-ud-dev_01246) There is passivization (at least not as found in Indo-Eurpoean languages)

[ftyers]

This also exists in Turkic, but in Turkic the causee is usually in Dative, see discussion here regarding iobj:caus and obj:caus. But I think that obl:caus would also work fine too.

[Stormur]

The reason why I'd prefer not to use obl is that I think this argument (Aslan) is more "core" than not. In French-GSD (even htouhg French uses a faire + infintive for the cuasative) the Causee is annotated as obj:agent (cf: UD_French-GSD/fr-gsd-train.conllu, sentence fr-ud-dev_01246) There is passivization (at least not as found in Indo-Eurpoean languages)

The impression that I get from this example is that this argument is actually ousted from the core, if it ever was there, in virtue of the fact that it is made assume an oblique case different from ergative, absolutive, nominative or accusative. This of course may depend on the base verb.

Honestly, I don't at all agree with the obj:agent interpretation you mention, if the situation in French is vaguely reminding of that in Italian (the causative is expressed through the verb fare 'to do/make') : generally speaking, if the base verb is intransitive, than the causee appears as an object (i.e. no preposition), and indeed it can be passivized:

• cadere 'to fall (intr.)':
  • cado 'he/she/it falls'
  • faccio cadere Aslan 'I make/let Aslan fall'
  • Aslan viene fatto cadere (da me) literally 'Aslan is made fall (by me)'

But with a transitive verb, the causee becomes oblique, the verb keeps its object, and no passivization of the causee is possible:

• scrivere 'to write':
  • scrivo una lettera 'I wrote a letter'
  • faccio scrivere una lettera ad Aslan 'I make/let Aslan write a letter', literally 'I make write a letter to Aslan'
  • *Aslan viene fatto scrivere una lettera (da me) 'Aslan is made write a letter (by me)' is impossible

Italian does not admit double objects. So, a subrelation like obj:agent cannot be generalized to all causative constructions, because indeed, the causee is not always an object. The situation might be analogous in Chechen, and as @ftyers point out, in Turkish too, where the dative is oblique. However, I think that iobj is extremely controversial as it is described now and should be really motivated, so if obl is possible, it should be favoured.

[jheinecke]

I agree, iobj is out of question here (by the way Chechen as a dative case, with similar semantics in other languages, so here we could use iobj. I'll go for obl or obl:caus then. Thanks for your input!

[Stormur]

Ah, of course there is also the more general and more widespread obl:arg, which might be as good as obl:caus, if this is deemed to be too specific!

[dan-zeman]

+1 for attaching the causee as obl if it is in the allative case. BTW if the causative form is analyzed as an inflection of the basic (active) form, as your CoNLL-U excerpt suggest, you should annotate the verb with Voice=Cau.