(16) [sono kizi-{o/∗no} kaita] kisya-ga uttae-rare-ta.
the article-{ACC/∗GEN} wrote journalist-NOM accuse-PASS-PAST
‘The journalist who wrote the article was accused’
Unlike nominative case valuation, accusative case valuation occurs within the
vP phase and is not affected by the nature of the higher T/CP phase. Therefore,
the object never receives a genitive case value.
The fourth property (9d) – the relevance of attributive forms to NGC – was
discussed above. The fi fth property (9e) refl ects the fact that a nominative NP
with an adverbial particle such as saemo ‘even’ cannot be a target of conversion,
if we assume that such a particle requires a clausal structure throughout the
derivation for licensing purposes (see Miyagawa 2013, Nakai 1980).
(17) [Taroo-saemo-{ga/∗no} katta] hon
Taro-even-{NOM/∗GEN} bought book
‘a book that even Taro bought’
The nominative subject featuring saemo ‘even’ in (17) is allowed, as it occurs
in the clausal structure throughout the derivation. On the other hand, the
prenominal clause becomes NP at some point of derivation after Transfer of T,
and the licensing condition for adverbial particles attached to the genitive subject
cannot be met at that level. Therefore, the genitive subject with the adverbial
particle is disallowed, exactly like the genitive element in regular noun phrases,
∗Taroo-saemo-no hantai ‘(even) Taro’s objection.’ In this manner, the central
properties of NGC follow straightforwardly from the proposed case system and
the theory of phases.
4 Conclusion
This chapter investigated what role formal features, such as case and φ features,
play in narrow syntax, with a special focus on case. It was proposed that case
valuation takes place based on a covariance relation between the two elements
with respect to their φ-feature specifi cations. Depending on the feature specifi ca-
tions of lexical items in a given language, either Merge or Agree is employed
for case valuation. In non-agreeing languages, such as Japanese, a covariance
relation is established by Merge, and thus Merge is chosen as a device for case
valuation, rendering Agree unnecessary. In agreeing languages, on the other
hand, a covariance relation is established by Agree; therefore Agree is chosen
for case valuation. It was also claimed that by incorporating the idea proposed
by Kato et al. (this volume) that Agree and Merge are essentially the same
operations, the two modes of case valuation can be reduced to a single
operation.
Although case valuation based on covariance relation is a single system, dif-
ferent case arrays emerge between agreeing languages and non-agreeing lan-
guages because of the difference between the two operations. Namely, Agree
Case and predicate-argument relations 61