of each discipline are grounded and reduced. There are many obstacles for this
to happen, perhaps the largest being the conception of linguistic knowledge as
an innately specified universal grammar that is taken by many to be the explana-
tion for how something as complex as language can be grown by the individual,
instead of just learned. I will leave aside most of these concerns and difficulties
for the purposes of this paper and instead focus on the essentials of linguistic
theory: hierarchical combination. It is my belief that the action of splitting
linguistic concepts in order to get to a lower floor in the causal chain should
start by distinguishing grammar from the process of hierarchical combination
that supports it. As I will show along these lines, connecting hierarchical com-
bination with WM is a relatively straightforward task that seems to be a more
fruitful approach than taking language as a complex object and contrasting it
with the equally complex multicomponent model just to see that nothing
matches. So let’s start from there and expand.
According to Chomsky (2008), “ The most elementary property of language –
and an unusual one in the biological world – is that it is a system of discrete
infinity consisting of hierarchically organized objects.” In most recent theoretical
accounts, the weight of building syntactic structures is carried by the operation
Merge, which takes two lexical items and combines them under a label that is
taken from one of them, thus producing a structure that can be captured by
the notion of hierarchy. Syntax is thus, following the inverted-Y model (Chomsky
1995), a co nnector of three dimensions: a pool of lexical items, and conceptual
and sensorimotor structures that constitute the interfaces, and its mission is to
link the three of them to support linguistic behavior. Regarding discrete infinity,
discreteness refers to the combination of atoms or lexical items, which are
understood as roots of words instead of full-fledged words, morphology also
resulting from combination. Infinity is the property that emerges from recursively
applying Merge over the structures that it builds, a process that would, if left
unconstrained, reach unbounded levels of embeddings that would nonetheless
remain part of the linguistic system.
Of course, there are boundaries that make this just an abstraction. Merge is
implemented in brains, a type of hardware that necessarily imposes multiple
constraints, which have been summarized by the notion of the three factors in
language design (Chomsky 2005): stru ctures that are impossible because of
universal grammar (the “genetic endowment” of the language faculty), which
may arbitrarily restrict combination to favor or deny specific outputs; structures
that are impossible due to constraints on how the interfaces are built (“language-
or even organism-independent” factors), which would crucially include working
memory limitations that make center-embedding a difficult task, and that make
the combinatory process collapse and reset whenever utterances surpass storage
limitations; and, finally, factors related to experience, including, for example,
the acquisition of Spanish over Chinese, which would also determine what gets
merged and when.
This distinction between Merge and constraints on Merge can also be useful
in applying Merge to WM accounts, because it constitutes a dividing line between
Language and working memory 115