Wh-ques-phrase using unslashed head daughter

I just realized that my current version of the customization system is generally doing this:


The tree on the right I don’t want because there is no extraction happening. The top S is licensed by the wh-ques (filler gap) rule in both trees.

The question is why is the tree on the right happening. Something should be ruling it out, and I am not sure how deep in the grammar. I am not actually seeing any relevant differences in the matrix core (between my version and the trunk), so, might be in the wh-ques rule itself? But it does say that the head daughter should be slashed. So I am a bit confused. The head daughter in this case is head-subject.

What should be ruling out the tree on the right, in principle? I would think it is a SLASH <! !> somewhere (up the hierarchy of the decl-head-subj?..)

This looks to me like you have an underspecified SLASH value somewhere, so the S without the extraction rule is nonetheless compatible with the constraints on the head daughter of head-filler. Because of the way lexical threading of SLASH values + diff lists appends works, an underspecified SLASH value anywhere will make the whole thing underspecified.

But also, what rule is licensing VP -> V there? Opt-comp, I guess? (That could be completely orthogonal.)

Ah!! It is actually ex-comp! So all is well. I am so used to looking at Russian grammars at this point that I assume any V->VP is morphology and am completely blind to it :). Thanks!

1 Like

The Parse | Compare functionality in the LKB can be helpful here, as can just hovering over the nodes to see which rules are involved when something puzzling is happening…