4 It is suggested in Kato et al (2014:Section 4) that Search could ultimately be
reduced to Merge. Below we will suggest a different way of unifying the two
operations.
5 Given what it actually does, a more appropriate – though cumbersome – term
for this operation might be Search/Select.
6 In what follows we will not discuss cases where more than two elements are
obtained as a result of Search 0 , but leave open the possibility that such cases exist.
7 It is suggested in Kato et al. (2014:218) that the output of Search, which func-
tions as a sort of generalized probe-goal mechanism, should be represented as
an ordered pair, not a(n unordered) set, because a probe and a goal have to be
distinguished. Departing from Kato et al. (2014), we claim that the output of
M 0 ◦S 0 (Σ) is represented as a set, not an ordered pair. This departure is reason-
able to the extent that the traditional notion of probe/goal is to be abandoned
under the current proposal. In this connection, it may be worth noting that
Chomsky has recently suggested the possibility that there is no probe in the
application of Agree, which should be reduced to some search procedure along
with labeling (see Chomsky 2015b).
Obviously, a set {X, Y} itself does not represent a relation R holding between
X and Y. Our claim is that the interface systems somehow derive R from {X, Y}
(and other information available to them). How this is done is an issue full
investigation of which we leave for future research. See Section 5 for some
related discussion.
8 Kato et al. (2014) hypothesize that what initiates Search are all and only ele-
ments at phase-edges, but we do not adopt a similar hypothesis for M 0 ◦S 0 (Σ)
here: we assume that M 0 ◦S 0 (Σ) can apply at any point of the derivation in
principle and its application is not forced by phase-edge elements (see note 17
below).
9 In this chapter, we use “nP” to refer to nominal phrases.
10 We will discuss how the outputs of M 0 ◦S 0 (Σ) should be represented in Section 4.
Until then, we will represent them in an informal way as in (5).
11 Here we tentatively assume that subject-T agreement occurs as soon as T is
introduced into the derivation (see Kato et al. 2014 for the argument that the
mechanism of feature inheritance should be eliminated). If it turns out that
subject-T agreement occurs after C is introduced into the derivation (as sug-
gested, for example, by Chomsky 2007, 2008, and Richards 2007), M 0 ◦S 0 (Σ)
would take CP, rather than TP, as an input for the agreement.
12 Note that the distinction between valued vs. unvalued features plays little role
in M 0 ◦S 0 (Σ). Thus, it can in principle establish agreement relations not only
between a “higher” unvalued feature and a “lower” valued feature (as in tradi-
tional theories of Agree; see Chomsky 2000 et seq.), but also between a “higher”
valued feature and a “lower” unvalued feature (see Baker 2008), and in fact
between two valued features or between two unvalued features as well.
13 Kato et al. (2014) do not discuss labeling.
14 In Section 5, we will discuss how M 0 ◦S 0 (Σ) picks out a particular element within
Σ as its label.
15 The tentative assumption here is that labeling occurs every time a new SO is
created. If it turns out that labeling occurs at phase levels, as suggested in
Chomsky (2013), we would claim that what Search 0 in M 0 ◦S 0 (Σ) for labeling
picks out is an SO contained in Σ and its label.
16 A similar point applies to Search in Kato et al. (2014).
17 Since External Merge (EM) is also an instance of M 0 ◦S 0 under our proposal, it
is assumed here that M 0 ◦S 0 in principle can apply at any point of the derivation
(see note 8).
42 Takaomi Kato et al.