generated by (set-forming) simplest Merge (cf. Richards 2008). F or example,
the orders of VO and OV, which were described by the Head-Parameter in the
P&P framework, are not currently specifi ed in the narrow syntax because the
simplest Merge generates only the linearly unordered set {V, O}. After Transfer,
the unordered set must be linearly externalized for the SM systems to ‘execute’
the representation (as temporally ordered successive articulatory gestures or
hand shapes) as either VO (e.g. in English) or OV (e.g. in Japanese). That is,
Merge generates order-indeterminacies or order-underspecifi ed representations
in the narrow syntax. (The revolutionary empirical claim is then that word-order
is not syntactic, but rather phonological, see e.g. Kayne 1994, Chomsky 1995)
Linea rization of {V, O} can apply in either of the two logically possible ways:
V-O or O-V. Word order variation is ‘fi xed’ by the learner one way or the other,
in the process of exposure to externalized PLD. (See Richards 2008 for a m ore
detailed account based on a third-factor interpretation of the Precedence Reso-
lution Principle proposed in Epstein et al. 1998.) Under the Borer-Chomsky
conjecture, syntax is no longer equipped with parameters but rather variation
in I-languages is attributed to variant features in the lexicon, internalized on
the basis of linguistic experience (externalized acoustic or hand shape) input
into UG.
The lexicon and the morpho-phonological component can be the loci of
variation, so that externalized input plays the crucial role in triggering variation
among I-languages (the strong interpretation of the Berwick and Chomsky
2 011 hypothesis excerpted above). It is important to note that, if variation is
restricted to externalization, it sounds like only the second factor (experience)
is involved. But this is not correct, in that it is experience input into UG that
determines a particular I-language/knowledge state. With respect to this point,
Epstein (2014) propos es:
if the organism is held constant (two human infants) and the exposure is
varied (Tagalog vs. Japanese acoustic disturbances in the environment), then
any differences in the development of the two infants, must be due to a
species-level property by which these variant exposures are mapped to those
particular developmental trajectories. In this sense, “language variation” (in
humans) is, contra. much standard locution, innate (biologically constrained).
That is, it is a defi ning property of the species that the possible class of vari-
ant developmental trajectories is determinable by variant experience. Contrary
to the implication of the standard nature-nurture dichotomy, “nurture” is
then itself, defi nable only in terms of nature, and “human language variation”
is a species property or capacity (and not “that which is not innate”).
(Epstein in press)In the next section, we present some cases showing that the third factor can
also be implicated in crosslinguistic variation (as well as externalization) and
demonstrate how exactly the third factor can contribute to variation.
Eliminating parameters from narrow syntax 131