I'm interested in what the constraints are on head-directionality and, in particular, which combinations of features are disfavored, unstable, or thought to be impossible.
I came across the final-over-final constraint (FOFC) a while back and am trying to figure out "where" it applies.
I'm going to start out with an intentionally naive statement of the FOFC (henceforth ultra-naive FOFC) and show that it doesn't hold by presenting a single counterexample.
So, in short, my question is:
- Under what circumstances does FOFC apply?
- Is FOFC supposed to be an absolute rule or are FOFC-incompatible forms merely predicted to be diachronically unstable?
- What would constitute a bona fide counterexample to FOFC, one strong enough to suggest that theory is incorrect?
The ultra-naive FOFC is the hypothesis that no FOFC violation occurs at any level in any phrase in any language.
The following English sentence is compatible with the ultra-naive FOFC
I see red cars
I'm not sure whether the subject or verb is considered the head of a clause, so I'll look at just the VP and mark heads like this:
^see [red ^cars]
[...] denote a head-initial phrase and
(...) a head-final one.
[see (red cars)]
However, there are prepositional SOV languages out there. Here's an example sentence from Amharic.
እሱ ወደ ከተማ መጣ Ǝssu wädä kätäma mäṭṭa he to city came
Amharic is pro-drop, so presumably the following sentence is also valid.
ወደ ከተማ መጣ wädä kätäma mäṭṭa to city came
The sentence presumably branches like this, with the verb as the head and the adpositional phrase as a dependent
[[^to city] ^came] ([to city] came)
So, the ultra-naive FOFC makes a bad prediction in this case.
However, if the real FOFC applies only to single phrases at a time then it never actually restricts anything.
So, which kinds of "subtrees" constitute domains in which the FOFC applies?
In order to learn more about FOFC, I read this paper/handout/document. I am not sure how that document is intended to be used, it looks like it is instructional material for a class or material for a conference.
The handout actually mentions the case of prepositional SOV languages, and mentions Tigre, an Ethiopian Semitic language related to Amharic.
Cross-linguistically, adpositions are strong verb patterners (Dryer 1992 i.a.), suggesting that they are [+V]. We therefore predict that PPs embedded under v should respect FOFC. There are a few notable exceptions to this, e.g. West Germanic, Iranian, Sorbian and Tigré (cf. Sheehan 2008a). This suggests that adpositions in these languages might not be [+V], or rather PPs might have a more complex structure. More generally, adpositions might not form a homogeneous class.