resist further computations. For example, once a φ-symmetric SO Σ = {α, β}
in (25a) is constructed, then neither Σ nor its constituents may be subjected
to further movement. (25b-d) show that α and β cannot undergo further
φ-symmetry-related operations. (25e) further shows that the φ-symmetric SO
Σ itself is also inaccessible to further φ-related computation.
(25) a. __ seems that [Σ [α John]i [β is believed ti to be a great linguist]].
b. ∗ [α John]i seems that [Σ ti [β is believed ti to be a great linguist]].
c. ∗ [β is believed ti to be a great linguist]j seems that [Σ [α John]i tj].
d. ∗ There seems that [Σ [α John]i [β is believed ti to be a great linguist]].
e. ∗ [Σ [α John]i [β is believed ti to be a great linguist]] seem(s) (that) t.
f. It seems that [Σ [α John]i [β is believed ti to be a great linguist]].
The same sort of observation can be made regarding Q-equilibrium in (10).
For example, (26b-d) show that wh-movement of constituents in a Q-equilibrium
results in much stronger deviance than the “weak island” effect in (26e).
(26) a. (Guess) ___ C he wonders [Σ [α which boy]i [β C ti read which book]].
b. ∗ Guess [α which boy]i C he wonders [Σ ti [β C ti read which book]].
c. ∗ Guess [β C ti read which book]j C he wonders [Σ [α which boy]i tj].
d. ∗ Guess [Σ [α which boy]i [β C ti read which book]] C he wonders t.
e. ?? Guess [which book]j C he wonders [Σ [α which boy]i [β C ti read tj]].
These data suggest that constituents of an F-equilibrium cannot enter into
further F-related operations. We will refer to this restriction as the Equilibrium
Intactness Condition (EIC).
(27) Equilibrium Intactness Condition (EIC):
Constituents of an F-equilibrium are stable and invisible for further
F-related computation.
Note that examples like (25) and (26) have been attributed to the traditional
idea that Move/IM is a “costly” operation (unlike EM) and can apply only
when it yields otherwise impossible feature-checking (such as checking of abstract
Case-features). As pointed out by Chomsky (2007, 2008), this “last resort”
conception of Move runs directly afoul of the free and unconstrained concep-
tion of Merge, hence this stipulation should be eliminated. Our proposal is that
this elimination can in fact be naturally achieved while maintaining unconstrained
Merge, if the results of its free application are subject to general – possibly
third-factor – conditions such as the DSC (22) and the EIC (27) on formal
features.6,7
In the next section, we will further argue that the DSC and the EIC provide
room for certain parametric variation concerning the distribution of argument
nPs across languages, focusing on English and Japanese as representative
examples.
Feature-equilibria in syntax 21