is a biunique relation, while Merge is not. The presence of multiple occurrences
of identical case and case alternation in Japanese, and their absence in agreeing
languages, refl ects this difference between the two operations.
It was also claimed that case features of nominal expressions and φ-features
of verbal expressions play an important role in determining SO labels. This
claim enables us to capture the three-way typological distinction of languages
proposed by Nichols (1986). Dependent-marking languages, such as Japanese,
make use of case features to determine labels, whereas head-marking languages
like Mohawk utilize φ-features to provide labels to SOs. The third type of lan-
guage, such as English, uses both features for this purpose. Thus, in dependent-
marking languages, case serves as a key feature in establishing the system of
predicate-argument relations that is a central component of human language of
thought. Therefore, the investigation of the nature of case as well as φ-agreement,
as they pertain to predicate-argument relations, continues to be one of the
major topics in theoretical linguistics, and as such to make a signifi cant contri-
bution to understanding the human language.
Notes
This research is supported in part by AMED-CREST.
1 Zushi (2014a, b) considers the mechanism (1) to be an algorithm that applies
in syntax. Whether the algorithm is reduced to a more basic operation deserves
further investigation. The mechanism could be reformulated as a case licensing.
We could argue that a nominal expression is merged with a case marker to form
KP, and that the KP is case licensed when it is merged with a particular element.
For example, a nominal combined with an accusative marker (AccP) is licensed
when it is merged with a verb. We leave further elaboration on this line of
approach for future research.
2 Space limitations do not allow us to discuss dative case. See Zushi (2014a, b)
for discussion.
3 The structural environment in which genitive case valuation occurs is basically
analogous to valuation of nominative case. This is made possible if we assume
that a noun, unlike a verb, cannot directly take its argument and that it must
be combined with a functional head, as proposed by Baker (2003) and Kayne
(2009). See Zushi (2014a, b) for discussion.
4 Note that the covariance relation does not hold in modifi cation relations in
Japanese. For example, attributive adjectives cannot be merged directly with
nouns in this language, but instead they constitute (reduced) relative clauses
when they modify nouns. Therefore, a covariance relation is not formed between
an adjective and a noun.
5 Note that the operation involving labeling defi ned by Kato et al. (this volume) is
not ambiguous. The operation determining labels, M 0 ◦ S 0 (∑) = {∑, LI}, applies
to a syntactic object (∑) and selects it together with a LI, and forms a set of them.
Subsequently, the LI is interpreted as a head of the object. The question that was
addressed in the text is how one can recognize a particular expression as a LI.
6 This idea can be extended to particles other than case markers, such as the
topic marker wa. It is reasonable to argue that wa identifi es the nominal to
which it is attached as an argument. When XP-wa is merged with an element
in the CP phase, for example CP, wa makes it possible for C to be selected
as a head. Thus, case markers basically determine SO labels in the vP domain,
62 Mihoko Zushi