At the derivational stage illustrated in (2b), C initiates Search, so that an
agreement relation is established between the Q feature ([Q]) of C (which
Kato et al. take to be unvalued) and the Q feature of what (which they take to
be valued) at the edge of vP.^2 After what is internally merged to the edge of
CP from the edge of vP as in (2c), it also initiates Search from the higher
position and a chain is created between the two copies of the wh-phrase. Agree
and chain-formation are thus unifi ed under a single operation Search.^3
Kato et al. (2014) present Search as one of the basic operations of syntax (or
a version of the basic operation of syntax, Merge^4 ). It is implicitly assumed there
that Search (as well as Merge) is a primitive operation, which cannot be decom-
posed any further. As can be seen in (1), however, Search clearly consists of
two parts: one that picks out two elements (that is, a feature or feature complex
which the Search initiator contains and a feature or feature complex identical
to it contained in the c-command domain of the Search initiator) and one that
establishes a relation between them. Thus, it seems quite plausible to regard
Search as being decomposable, or a composite operation.
In this chapter, we propose that Search is a composition of two more primi-
tive operations, which we call 0-Search (Search 0 ) and 0-Merge (Merge 0 ) and
formulate as in (3) and (4), respectively (the formulation of Search 0 will be
slightly changed in the next section).
(3) Search 0 :
Search 0 is an operation such that given an SO Σ, it picks out n elements
contained in Σ (i.e. Search 0 (Σ) = α 1 ,.. ., αn, where α 1 ,.. ., αn are ele-
ments contained in Σ).5,6
(4) Merge 0 :
Merge 0 is an operation such that given n objects, it forms the set of these
objects (i.e. Merge 0 (α 1 ,.. ., αn) = {α 1 ,.. ., αn}).
We claim that what is called Search in Kato et al. (2014) is Merge 0 (Search 0 (Σ))
(henceforth, M 0 ◦S 0 (Σ)), and that when the output of M 0 ◦S 0 (Σ) is a two-
membered set {X, Y}, it is interpreted as “there is a relation between X and Y”
at SEM and/or PHON.^7
Let us see how agreement, chain-formation and binding, which Kato et al.
(2014) unify under Search, are recaptured under M 0 ◦S 0 (Σ).^8 First, consider
φ-agreement between T and the subject nP.^9 Suppose that the derivation has
reached the stage where T and vP have been merged to form {T, vP}. First,
Search 0 applies to this SO and picks out the (unvalued) φ-features ([φ]) of T
and the (valued) φ-features of the subject nP, located at the edge of vP. Then,
Merge 0 applies to these sets of features and forms a set containing them. Thus,
M 0 ◦S 0 (Σ) takes {T, vP} as an input and yields {[φ] (of T), [φ] (of the subject
nP)} as the output, as shown in (5).10,11
(5) M 0 ◦S 0 ({T, vP}) = {[φ] (of T), [φ] (of the subject nP)}
On the primitive operations of syntax 31