現代ギリシア語におけるFactive 'PU'の島について

This paper investigates the patterns of wh-extraction out of factive complement clauses introduced by the factive pu. At least, two extraction patterns are identified as in (1) and (2):
(1) a.*Pjosi lipase [pou sinandise ti to Yani]?
b*Pjoni lipase [pou sinandise o Yanis ti]?
c.*Potei lipase [pou agorases to vivlio ti]?
(2) a. Pjosi lipase [pu prodose ti mistika]?
b. Pjoni lipase [pu pligoses ti]?
c.*Potei lipase [pu efijes ti]?
A long distance wh-phrase passes through the upper Spec(Fin), where it is assigned the feature [-A(naphoric)] or the feature [+A][-A]. The lower Spec(Fin) is occupied by a FACT DP which is assigned the feature [-A] or the feature [+A][-A] after moving from its base TP-adjoined position, as shown in (3):
(3) the structure of the factive pu-Island:
Pjosi lipase [DP [D' [D [FinP ti' [Fin' FACT DPj [Fin pou [TP tj [TP sinandise ti to Yani]]]]]]]? (for (1a))
(4) [FinP Pjosi [Fin' FACT DPj [Fin (pou) ... tj ... ti chain formation of [Pjosi and ti]
a. Pjosi[-A] + FACT DPj[-A] ... tj ... ti → *
b. Pjosi[-A] + FACT DPj[+A][-A] ... tj ... ti → *
c. Pjosi[+A][-A] + FACT DPj[-A] ... tj ... ti → ✓
d. Pjosi[+A][-A] + FACT DPj[+A][-A] ... tj ... ti → *
According to Relativized Minimality based on Full Matching and to some other premises assumed in the text, the chain formation in (4c) alone is designated as legitimate. Thus, in (1), both the argument wh-phrases (pjos and pjon) and the adjunct wh-phrase (pote) are only assigned the feature [-A] in Spec(Fin), leading to the ungrammatical status of (1a-c). In (2), on the other hand, it is the argument wh-phrases alone that can receive the feature [+A][-A], resulting in the legitimate chain formation of type (4c), while there cannot be any well-formed chain formation stipulated, with the adjunct wh-phrase (pote) assigned the feature [-A].