18 Note that IM and labeling are formally similar to each other, as shown in (10)
and (12). See also Fujita (2009, 2012) for the view that labeling is a special
case of IM.
19 We assume here that IM can occur not only at phase levels but also at other
derivational points (contra Chomsky 2007, 2008, among others). Thus,
subject-raising to “Spec-T” is assumed to occur before the phase head C is
introduced into the derivation, satisfying the No-tampering Condition (NTC).
See Epstein et al. (2012), Kato et al. (2014), Narita (2014) and Chomsky (2007,
2008, 2013, 2015a) for various proposals to the effect that IM occurs at phase
levels while satisfying the NTC.
20 See also Chomsky (2012:23), where it is suggested that EM, as well as IM,
involves some kind of search.
21 Note that all instances of M 0 ◦S 0 take WS as an input, insofar as the conclusion
we reached in Section 3 is on the right track. Thus, the left-hand side of (22)
and all other equations discussed in the previous sections is to be replaced with
M 0 ◦S 0 (WS). It can reach elements inside vP when vP is a term of WS.
22 Chomsky’s (2000) earlier set-based approach to chain is to assume that each
occurrence is defi ned in terms of its sister (not its mother). This approach has
a much earlier origin in Chomsky (1955/1975), adapted from Quine’s (1940)
notion of “occurrence of a variable.” However, the sister-based defi nition of
occurrence is not readily applicable to occurrences of feature-chains. Therefore,
for the purpose of this chapter, we propose to adopt Chomsky’s (2001)
mother-based version of “occurrence.”
23 Structural prominence is defi ned for SOs, which we take to be LIs and objects
constructed from them via M 0 ◦S 0 in the form of Merge, which serves for
Sister-relation-formation, as we will propose shortly. Crucially, we do not regard
outputs of M 0 ◦S 0 representing non-Sister-relations as SOs.
24 Chomsky (2013) leaves room for cases in which a bundle of features smaller
than an LI (say a collection of φ-features) may participate in labeling. We will
leave this possibility for future research, but the defi nition in (33a) can naturally
incorporate such cases as well.
25 Obviously, (35) leaves many cases of binding failures unaccounted for. In order
to achieve a full-fl edged account of binding, much more than (35) should be
supplied to constrain proper Bind-formation (conditions of binding theory, pos-
sibility of vehicle change, and so on), which falls beyond the scope of this short
chapter.
26 Readers may wonder whether this effect can be attributed to some other factor,
such as the “Phase-Impenetrability Condition” (PIC) (Chomsky 2004 et seq.).
As pointed out by Kato et al. (2014), there is good reason to cast doubts on
the current formulation of the PIC. These doubts pertain to the clear existence
of long-distance (i.e., cross-phasal) dependencies. See Bošković (2007) and refer-
ences cited therein for ample crosslinguistic examples of long-distance Agree(ment)
that clearly violate the PIC. Moreover, binding (say, of pronouns or subject-oriented
anaphors) can no doubt apply in a long-distance fashion as well, and thus bind-
ing is not constrained by the PIC, either. The theory of M 0 ◦S 0 eventually has
to subsume such cases by assumption, and these considerations support the view
that the PIC must be reconsidered.
27 Labeling of vP with v in (29) is not blocked by the existence of, say, read,
though it appears that Depth(read) in (30b) is not less than Depth(v) in (30a).
This is because structural prominence is defi ned only for SOs, and the sets in
(30a-b) are not SOs. See note 23.
28 Note that, given Σ 1 , Σ 2 ∈ WS, M 0 ◦S 0 (WS) as proposed so far can merge Σ 1 and
an element properly contained in Σ 2. Suppose that this type of Merge, known
On the primitive operations of syntax 43