Changing basic-head-mod-phrase-simple?

For one bit of my analysis, I need to be able to let the user determine whether the modifier in a type that inherits from basic-head-mod-phrase-simple is a NOM or an NP – that is, whether the value of NON-HEAD-DTR...SPR on basic-head-mod-phrase-simple is 1-list or olist. The current definition of basic-head-mod-phrase-simple defines the value of NON-HEAD-DTR...SPR as olist. If I remove this constraint on basic-head-mod-phrase-simple and instead put it on the rules in matrix.tdl that inherit from it (scopal-mod-phrase, isect-mod-phrase, adj-head-phrase, head-adj-phrase), all the regression tests still pass. Are there theoretical, aesthetic, or procedural reasons for avoiding this? I could always define a new type that was identical to basic-head-mod-phrase-simple except for lacking this one constraint, but I’m thinking that avoiding that level of redundancy seems like a good idea.

I’m pretty sure you’ll get broken MRSes if you try to have modifiers attaching to NP instead of NOM. What is configuration that seems to require this?

So, what’s changing is not what the modifier attaches to, but the modifier itself. As far as I can tell, that doesn’t affect the MRSs, except for getting rid of the quantifier in cases where the modifier is a NOM. But the rest seems to work.

That said, the motivating example for including this switch was the construction in Ancient Greek where you get something like “the [the cat’s] paw” – that is, cases where the modifier is an NP. I am not quickly finding any examples of the modifier being a NOM – a lot of the examples I have of modifier-like possessives come from languages where only pronouns can act as modifier-like possessives (like Italian) so the NP/NOM distinction isn’t relevant. Perhaps it’s worth sending a request to lingtyp to see if this happens before trying to accommodate it?

“getting rid of the quantifier” is also a problem, because it means you’ll have an x which is not bound by any quantifier, aka a broken MRS.

Even if the modifier is nominally a NOM, you’ll need to put the quantifier in for it somehow.

But more generally — if this is not an attested combination as far you know, then I wouldn’t try to accommodate it! (Though you could ask Lingtyp…)

I’m not immediately seeing any examples of this in my lit review, so I’ll put this on the list of things to double check on later. Still not sure why I was so very certain I had examples, but I’ll prioritize other things before figuring this out. Thanks!