2.2 Generalized case valuation
We have seen a recent claim made by Zushi (2014a, b) that case valuation in
terms of Merge should be available in addition to Agree. The fundamental
theoretical assumption that underlies this proposal is that Merge and Agree are
among the elementary computational operations in human language. This is
the widely held view shared by most of the studies within the minimalist pro-
gram. However, more recent studies claim that Agree, as well as other notions
including chains (formed by internal Merge) and binding, ought to be subsumed
under a more general operation of minimal Search. This operation seeks two
elements that have an identical feature complex and establishes a relation between
them (Kato et al. 2014). Developing this line of research further, Kato et al.
(in this volume) argue that the operation of Search defined by Kato et al. (2014)
is decomposed into more primitive operations, 0-Search (Search 0 ) and 0-Merge
(Merge 0 ). The former searches certain features and selects n (typically two)
elements from a designated domain of computation, while the latter forms a
set of the selected elements. The set formed through this operation is interpreted
as a certain linguistic relation at the CI interface. Kato et al. also propose that
the standard notion of Merge should be decomposed into the two elementary
operations, Search 0 and Merge 0 , suggesting that Search and Merge are essentially
the same operations of a composite of Search 0 and Merge 0 (see Kato et al. for
details). This is a promising line of research that enables us to eliminate addi-
tional devices that have been stipulated in various aspects of grammar.
Now, if we follow Kato et al. (this volume) in assuming that Merge and
Search (hence Agree) are the same operations, it turns out that the two modes
of case valuation proposed by Zushi (2014a, b) can be consolidated into a single
operation. Let us first consider how a covariance (case) relation based on Agree
can be reformed. According to the proposal made by Kato et al., φ-feature
agreement takes place in the following way. Suppose that T and vP are merged,
forming an SO {T, vP} at some point of the derivation. When applied to this
SO, Search 0 searches (unvalued) φ-features of the subject nP and (unvalued)
φ-features of T, and picks out these two elements (n and T). Merge 0 applies
these sets of features and forms a set of them. It is plausible to say, then that
based on the set formed by Merge 0 , a covariance relation defined in (3i) is
established between the two elements. Therefore, the case feature of the subject
is valued as nominative.
External Merge can also be decomposed into the two primitive operations
of Search 0 and Merge 0 based on the proposal made by Kato et al. In this case,
Search 0 applies to the workspace, a domain that the computational system can
operate in, including SOs and the lexicon. Suppose that Search 0 picks out two
elements, for example hon ‘books’ and yomu ‘read,’ from the workspace. Then,
Merge 0 forms a set {{hon}, yomu}. A covariance relation defined in (3ii) is formed
based on this set, since neither of the two elements have φ-features. Accordingly,
the case feature of the nominal is valued as accusative. In this way, incorporating
the proposal made by Kato et al. into Zushi’s (2014a, b) system of case
50 Mihoko Zushi