Oblique cases in the Grammar Matrix



Context: creating a test suite for a new GM library.

To what degree are oblique cases supported in the Grammar Matrix, if at all? We can list however many cases we want, but the only thing we are supposed to say about any given verb though is the case of its subject and (direct?) object, is that right?

For example, could I, in principle, model the following phenomenon in Abui using case:

\ref B07.029.1e.20040527Benny6.27.2
\t A raha hookariang?
\m a raha hoo- kariang
\g 2sg.agt chief 3.goal- work.Ipfv
\p pro n pro- v.goal
\f Are you working for the king?

(By saying there is zero marking or underspecified case on the noun “chief” and that the verb “work” requires some “goal” case on its object?)

But I suspect that this is probably stretching it too far since I doubt this is an obligatory argument. Should I just stick to what is clearly obligatory arguments?..


You certainly can model “quirky case” case frames in the GM — define the case values, make sure the lex rules for the nouns for each case are defined too, and then for the verbs choose e.g. transitive case unspecified and then add [ CASE goal ] or whatever as a constraint on the verb’s object.

HOWEVER – that is not what is going on here. The morpheme glossed 3.goal is an affix on the verb. It’s not a case marker. Case is variation it the form of NPs based on the role they play in a sentence.


OK, thanks! In general though, is quirky case in the Matrix supposed to be for obligatory arguments? For direct objects? Or is it wider than that?


As for Abui, I guess it can be seen as a derivational affix – but those are not supported in the questionnaire (or the Matrix), right?


If by derivational affix you mean something that changes the ARG-ST of a verb, yes, now covered, thanks to @curtosis’s library!

@curtosis I wanted to put a link here to MatrixDoc pages for your library, but can’t quickly turn anything up. Still on the todo list?


I meant something that changes the meaning of the verb, actually. Makes something like “cut” mean “cut all the way through” etc.


Ah - change in verb meaning w/o a change in the argument structure might not be covered. On the other hand, if you’re just trying to get at the rest of the phenomena in the sentence, it is allowed to create morphosyntactically and semantically inert lexical rules that just add the affix…


Yes, the MatrixDoc is definitely still on my to-do list. :man_facepalming: :sweat_smile: :alarm_clock:

Notionally, the mechanism is there for an affixing rule to specify a new PRED value, but I don’t think the questionnaire+customization lets you make that the only operation. It’s pretty straightforward, though, so I can try to loop it in to the next update.


Ah, I did not realize valence change allows pred change. I think in this case at least, we may say there is valence change! (Assuming work can be e.g. intransitive).


@olzama Technically it doesn’t allow it, but it needed the capability to specify PRED on an added EP. Because it’s implemented as an individual building-block :slight_smile: it should be fairly simple to reuse it.

That said, your example at the top would be directly supported by the valence change library – it appears to be a straightforward benefactive. work would still be intransitive, and a new benefactive_rel EP would be added that takes chief as its ARG2 and the work evt as its ARG1. You can also specify a case constraint on the added argument if appropriate. The customization system will take care of generating the appropriate lrts to wire everything together.

(In other words, not so much an “optional argument” on the verb, but an optional affix that adds an argument.)