We propose that the primitive operation Search 0 makes use of two kinds of
features to determine SO labels: (i) case features of nominal expressions and
(ii) φ-features of verbal expressions. Languages employ at least one type of
feature to provide labels to SOs. This claim enables us to capture the three-way
typological distinction of languages proposed by Nichols (1986). In dependent-
marking languages, such as Japanese, the former means of labeling based on
case features is employed, whereas in head-marking languages, such as Mohawk,
the latter way of labeling based on φ-features is utilized. This claim allows for
the third type of language that makes use of both means to provide labels to
SOs. English seems to be an example of this type. In its CP domain, the
φ-features of C/T determines SO labels, while a case feature of nominal may
play this role in the VP domain where no obvious agreement is found in this
language. On the other hand, the system we are proposing here predicts that
if neither of these two features were employed, then the result would be ruled
illegitimate, since there would then be no way of labeling SOs.^7
3 Consequences for peculiar case phenomena in Japanese
3.1 Multiple occurrences of identical case
Having laid out the theory of case valuation, let us now examine how it accounts
for case peculiarities in Japanese. First, we deal with multiple occurrences of
identical case. This phenomenon has challenged Agree-based case theory, since
Agree (or feature checking) is generally assumed to be a biunique relation
between a probe and a goal. To apply Agree/case checking to Japanese, one
must modify the theory to allow more than one nominal to enter an Agree
relation. Ura (2000) takes this direction. He proposes, couched in the frame-
work of Chomsky (1995), that UG provides a [±multiple] feature with a
functional head. A [+multiple] feature allows more than one element to enter
into feature checking, whereas a [−multiple] feature permits only one element
to involve feature checking. Thus, it is argued that in Japanese, unlike in Eng-
lish, T has a [+multiple] feature, which allows more than one nominal to occur
in Spec of TP and enter into feature checking with it. As a result, more than
one nominal receives a nominative case. Hiraiwa (2001) adopts and elaborates
upon this idea within the framework of Chomsky (2001) regarding the notion
of multiple Agree.
On the other hand, researchers that adopt a structure-based view of case
assignment approach this phenomenon from a more general and crosslinguistic
perspective. They relate this phenomenon to other syntactic peculiarities (e.g.,
scrambling) in Japanese, in contradistinction to English, to see what accounts
uniformly for typological differences between the two types of languages. On
this view, the phenomenon of multiple nominative/genitive is not explained by
the case theory per se, but by a more general principle that accounts for the
lack of biuniqueness effects in other domains of grammar as well.^8 Thus, Kuroda
(1988) argues that Agreement is forced in English but not in Japanese, which
Case and predicate-argument relations 53