Thus, the formation of {T, C} can also be seen as driven by T(ense)-feature sym-
metry. More generally, head-movement can be regarded as fulfi lling an important
function of fi xing the asymmetry of {X, YP} and forming a feature-equilibrium.
Therefore, our notion of featural symmetry provides an important rationale for
head-movement, reducing it to a special case of symmetry-driven computation.
Exploring these lines of reasoning, we put forward a revised version of Fukui’s
(2011) generalization, where the notion of symmetry/asymmetry is redefi ned
in terms of feature-equilibrium.
(21) SOs asymmetric with respect to some formal feature F are unstable, and
they must be mapped to SOs symmetric with respect to F.
(21) correctly differentiates stable vs. unstable {XP, YP} structures, while suc-
cessfully incorporating head-movement structures into the category of symmetric
structures.
Without further stipulations, Merge can freely combine SOs, generating both
asymmetric and symmetric structures, but we observed that movement opera-
tions typically result in symmetric structures: Subject-raising via IM yields a
φ-equilibrium and wh-movement via IM yields a Q-equilibrium, whereas head-
movement (X^0 -Y^0 -remerge) yields a Cat/T-equilibrium. In contrast, EM, the
fi rst application of Merge for each LI and SO, typically yields F-asymmetric
structures, as we saw above. This pattern seems to be systematic, and we would
like to know why this is almost always the case. To capture this state of affairs,
we propose that syntactic derivation is driven by the need for featural symmetry.
This idea can be stated as in (22):
(22) Dynamic Symmetrization Condition (DSC):
Each formal feature F must form an F-equilibrium in the course of the
derivation.
LIs are associated with their own feature contents. EM freely introduces LIs
and their features (including formal features) into a syntactic derivation, but the
DSC (22) essentially states that once a formal feature F is introduced, the
subsequent derivation must guarantee that F moves to a position where it can
form an F-equilibrium. The DSC (22) thus articulates the view that linguistic
computation is irreversibly directed toward symmetrization, and this idea provides
a natural ordering of operations in (23) below, further refi ning Fukui’s (2011)
generalization (cf. (11), (21)).
(23) For any formal feature F, an application of Merge that creates an
F-asymmetric SO (i.e., an SO that is asymmetric with respect to F) entails
a later application of Merge that yields an F-equilibrium.
This completes our rationalization of Fukui’s hypothesis that syntax is driven
by a need for symmetry.
Feature-equilibria in syntax 19