Using CASE to limit overgeneration with possessives


Hello all (and Emily in particular),

Now that I’m using preexisting phrase types for possessive phrases, I need a way to make sure that 1) a marked possessor never appears without its possessum, and 2) a marked possessum never appears without its possessor.

I’ve got some ideas for #2, but in order to accomplish #1, Emily proposed using CASE in a slightly novel way. I’m going to outline that approach here with the goal of making sure I’ve understood it fully before I implement it:

  1. In languages that have case, rather than having all cases inherit directly from case, have them inherit from some intermediate type (something like ‘real-case’) that contrasts with something like ‘poss-case.’
  2. Then give all marked possessors the value [ CASE poss-case ], which is incompatible with all other ‘real’ cases.
  3. In languages that have case, all verbal arguments will already have a required ‘real’ case value, so possessors will not show up as verbal arguments without their possessum. (This is the part I’m least sure on. Does some extra work have to be done here?)

Emily: Am I getting this right?
Everyone else: Do you see any ways to improve this approach or extend it to work for #2 as well as #1?


Yes, with a couple of caveats / clarifications:

  1. If the language uses genitive case for possessors, you don’t need to go through all of this, right? (Though you’d need the poss-rel to come from somewhere other than possessor at that point.)
  2. The additional work is that you’ll have to look at the case frame requirements — if the user has said [ CASE nom ] there’s nothing to add. It’s when the language has a case system but a certain position is underspecified for case that you need [ CASE real-case ].
  3. You’ll need to worry about the complement position of case-marking adpositions as well as the subject of predicative adjectives and the subject of auxiliaries as well.


Thanks! That clears a lot of things up. I’ll work on implementing this and ask further questions if they arise.

Re: genitive case for possessors: in that scenario, the documentation will recommend using a juxtaposition construction where the possessor is constrained to be [ CASE gen ]. That means there won’t be any marked possessors as such, and so this problem won’t arise. The case lib should already be doing the work of keeping genitive case things from showing up in the wrong places, I assume.