Depth(v) in (41b) is not less than Depth(what) in (41a) (since both of them
are 1). We claim that this is why IM of what satisfi es the minimality condition
in (32) in spite of the existence of the apparently intervening elements.^27
The above discussion shows that the generalized minimality condition in (32)
can allow unconstrained applications of Merge (i.e., Sister-formation), while it
can also account for various effects of “relativized minimality” when applied to
non-Sister-relations (Label, Bind, Chain, etc.). As for non-Sister-relations, it is
also important to note that (32) not only sets an account of the locality of
“goal” elements, but also derives the shallowness of “probe” elements in certain
environments, such as the case of binding in (34a) vs. (34b).
It is clear that this unifi cation and extension of the minimal search require-
ment can be achieved only if the traditional notion of probe/goal is eliminated,
and various linguistic relations are unifi ed into generalized M 0 ◦S 0 (WS). Thus,
the above discussion lends further support to our unifi ed approach to various
linguistic relations.^28
6 Conclusion
In this chapter, we pointed out that Merge as traditionally construed is in fact a
composite of two operations, i.e., (i) Search 0 , the selection of n elements,
α 1 ,.. ., αn, from a designated domain of computation (WS), and (ii) Merge 0 , the
formation of an unordered set of the n objects, {α 1 ,.. ., αn}. We argued that
the proposed analysis of Merge as M 0 ◦S 0 is not only required in terms of descrip-
tive preciseness, but also favorable in terms of breadth of empirical coverage.
Specifi cally, we showed that M 0 ◦S 0 can be generalized to cases including Agree(ment),
chain-formation, binding (cf. Kato et al.’s earlier notion of Search) and also labeling
(cf. Chomsky 2005, 2007, 2008). In this manner, we can make a stronger and
more precise sense of the recent claim that Merge is all we need to assume as a
basic operation of syntax (cf. Boeckx 2009, Berwick 2011, Fujita 2013).
Notes
This research is supported in part by AMED-CREST.
1 Space limitations prevent us from providing a full review of how Search works.
We refer the reader to Kato et al. (2014) for details.
2 Kato et al. (2014) argue that the mechanism of feature valuation must be elimi-
nated if a theory of no-tampering syntax is seriously pursued (see Kato et al.
2014 for details). They instead put forward the following hypothesis:
(i) When an Agree(ment) relation is established between an unvalued feature
and a valued feature, the interface systems access it, so that the unvalued
feature will be processed at SEM/PHON in relation to the valued feature.
(Kato et al. 2014:214)
In this chapter, we also assume that feature valuation does not exist and adopt
the hypothesis in (i).
3 It is assumed in Kato et al. (2014) that binding is a sort of agreement (in
φ-features or some independent referential feature [Ref]).
On the primitive operations of syntax 41