I have a question about how to get case agreement between a possessor pronoun and its possessum, like the kind you see in Russian:
mo-j drug my-NOM friend.NOM u mo-ego drug-a by my-GEN friend-GEN
Agreement in person, number, and gender works like this: the lexical entry or inflection rule that builds the possessor pronoun puts the agreement features at HEAD.POSSESSOR.POSS-AGR, which is of type png. The possessor pronoun then is the daughter of a unary phrase rule, which does the work of identifying the PNG features of the possessum with the POSS-AGR features of the possessor.
The problem with case is that it doesn’t fall under PNG, so I can’t put it at POSS-AGR. I also can’t give the possessor pronoun an inherent case, because I’m already using case sneakily for another purpose. In order to keep possessors from running around loose in the grammar, I’m marking all of them with a CASE feature called poss-case, which contrasts with real-case, which is the ancestor of all the actual cases. So poss-case is incompatible with nom, acc, gen, etc, so (in the Russian example above) I can’t put the feature [ CASE gen ] on the possessor pronoun. That means that doing the trick where the unary rule identifies something between its daughter and the possessum won’t work, because its daughter doesn’t have a place to store the case feature in the first place.
I can get around this in the modifier case (like in Italian or Greek), by having the genitive form of the possessor pronoun simply constrain the thing on its MOD list to be [CASE gen]. That won’t work on spec-like possessors because they have to go through a unary rule before acting as specifiers.
Any ideas on how to deal with this?