How the accusative-marked element in (12b) is derived is not our concern
here. An important point is that the fact that an accusative case particle can be
attached to the attributive form reinforces the idea that the attributive form
behaves like a nominal.^12
One conclusion we draw from these observations is that the clauses contain-
ing an attributive form had both nominal and clausal properties in Classical
Japanese. Its nominal properties remain to some degree in Modern Japanese.^13
Given this nature of attributive forms, Hiraiwa’s (2001) idea that genitive case
is associated with the attributive form becomes more plausible. However, his
idea does not fully account for the alternation of nominative with genitive case.
We show below how the nominal property of the attributive forms brings about
case alternation from a crosslinguistic perspective. A key to solutions for this
question can be found in Ott’s (2011) proposal for a free relative, a construc-
tion also known as a clause with nominal properties.
3.2.2 Nominative-genitive conversion in the theory of phases
Ott (2011) examines free relatives in comparison to embedded interrogatives and
claims that the categorial duality exhibited by free relatives can be explained in
terms of head Transfer. In a nutshell, he argues that at the stage of derivation,
where what you cook is formed, a head C is removed along with its complement
from the workplace by Transfer on the assumption that no interpretable features
of C remain. This makes it possible for the wh-phrase that remains at the edge
to be the only element visible at the next vP phase, rendering its categorial label
as DP. Therefore, a free relative is a CP up to the derivational stage, where Transfer
of C occurs along its complement, but it becomes a DP after the next vP phase
level. Thus, the dual nature of free relatives is derived. On the other hand, a head
C of the embedded interrogatives cannot undergo Transfer because interpretable
features are present. Consequently, it retains the clausal status throughout the
derivation. In short, the possibility of the head C’s Transfer, which depends on
particular feature specifi cations, can capture the categorial duality of free relatives
and the distinction between the two closely related but structurally different
constructions (free relatives versus embedded interrogatives).
Let us show how the proposed theory of case gives a natural account for the
NGC by adopting the basic idea propose of Ott (2011). We take Japanese
prenominal clauses, including relative clauses like (13a), to be TP, following
Murasugi (1991), Saito (1985), Sakai (1994), and many others. We also assume
that TP can be a phase only in embedded contexts (Zushi 2005).
(13) a. [Taroo-ga kaita] (kizi)
Taro-NOM wrote (article)
‘(the article) which Taro wrote’
b. [ Taroi [ T [vP ti [VP kaita pro ]]]]
Case and predicate-argument relations 59