In the structure (13b), the subject Taro is assumed to be raised to the edge of
T. This is possible, since T, as a phase head, allows unbounded Merge to be
applied to its edge, and the NP at its edge can be interpreted in terms of “about-
ness” relation (Saito 1985, Rizzi 2010). Regarding the label for the structure
(13b) as a whole, there is an option in Japanese whereby category determines the
label. Suppose that the subject receives a nominative case value within vP, and
moves to the edge of T. The label of the structure (13b) is determined by T,
according to the proposal of labeling presented in Section 2.3. When T determines
the label, the structure retains clausal properties throughout the derivation.
Suppose that the case feature of the subject is not valued within vP. This is
possible because no φ-agreement relation between the subject and v is assumed
under the proposed case theory.^14 Since T has no role in this case, it is trans-
ferred along with its complement, making it possible for the element that remains
at the edge to determine the label. As a result, the structure turns out to be
an NP, yielding the nominal character of the clause.
(14) [nP [NP Taroi [T [vP ti [VP kaita pro]]]] n ]
Transfer of T enables the subject Taro to satisfy the condition of genitive
case valuation, since it becomes a sister to the external head n. Therefore, the
case feature of the subject is valued as genitive. In this way, genitive case mark-
ings on the subjects of the prenominal clauses are derived.
Let us briefl y demonstrate how the basic properties of NGC listed in (9)
follow from the proposed analysis. The fi rst property deals with the optionality
of NGC (9a). In the proposed analysis, there is no optionality in case valuation
per se. Alternation of nominative with genitive case is caused by the different
structures determined by the choice of how the prenominal clauses are labeled.
The label of the prenominal structure can be determined either by T or by the
edge element. In the former instance, the subject of the clause receives a nomi-
native case value; in the latter, a genitive case value.
The second property (9b) is that NGC occurs in prenominal clauses, but not
in embedded clauses headed by an overt complementizer, as shown in (15).
(15) Taroo-ga [Hanako-{ga/∗no} kenkoo-da to] itta
Taro-NOM Hanako-{NOM/GEN} is-healthy that said
‘Taro said that Hanako is healthy’
Unlike prenominal clauses, clauses headed by an overt complementizer are
regarded as CPs. Given the standard assumption that the element merged with
C should be an operator-like element, an argument NP cannot be raised to the
edge of C in the embedded CP clauses in (15). Thus, CPs of this sort cannot
be nominalized in the way that TPs can be in prenominal clauses. Therefore,
no NGC is possible in these clauses.
The third property (9c) is that there is no conversion of accusative to geni-
tive, which also follows directly from our proposed case mechanism.
60 Mihoko Zushi