that what matters is not the surface linear proximity of the attractor to the verb
but structural closeness of the attractor to the verb in the derivation of the sen-
tence. The attraction errors reviewed so far univocally show that it is not linear
distance between the attractor and the verb that elicits attraction errors, but the
hierarchical structure involved. Readers are referred to Le wis and Phillips (2015)
for more evidence that hierarchical structure matters in agreement attraction. In
the next section, structure dependence is discussed in relation to the brain.
3 Structure-dependence in the brain
It has been observed that our linguistic knowledge makes use of structure-
dependent rules that require an analysis of a sentence into an abstract hierarchical
structure and operates on abstract grammatical constructs such as phrases or
clauses, instead of structure-independent rules that just operate on the rigid-
order of words (Ch omsky 1957). Structure dependence is so ubiquitous in
language that it is difficult to appreciate its importance in the design of language.
Structure dependence is interesting for several reasons. First, the only physical
aspect of syntax, i.e., linearity, does not play a role in syntactic rules (Mo ro
2008, Yu sa 2012a, b) , suggesting that core properties of language are not
transparent to perceptual systems. Structure dependence is an abstract principle
with no overt realization in any language, making it difficult to receive enough
information of structure dependence. It is highly possible for a person to go
through life without encountering any relevant examples that would help to
choose between structure-dependent rules and structure-independent rules (Be r-
wick et al. 2011). This poses challenging questions that characterize the factors
underling language acquisition, since evidence bearing on structure dependence
is almost non-existent (Ya ng 2002). Second, structure dependence is a coun-
terintuitive principle: the structure-dependent rule based on abstract hierarchical
structure is computationally more complex than the structure-independent rule
relying on linear order (Ch omsky 2013). Third, the principle puts severe limits
on logically possible rules that human languages exploit (Mo ro 2008). This is
a matter of the poverty of the stimulus problem in language acquisition. Struc-
ture dependence is derived from Merge, which applies to X, Y, forming Z=
{X, Y} and closeness is derived from laws of nature (Ch omsky 2013). Structure-
dependence is a deep-rooted principle of human language that is not readily
accessible from observation. Readers are referred to Be rwick et al. (2011) and
Ch omsky (2013) for detailed arguments about structure dependence.
There are two relevant behavioral studies that tested the principle of structure
dependence. Cr ain and Nakayama (1987) report that English-speaking kids aged
3 to 5 do not produce yes/no questions that violate the principle of structure-
dependence in their mother tongue. This result suggests that the principle is a
deep-rooted innate property. More interesting is the case of Christopher, an
unprecedented linguistic savant, who can “read, write and communicate in any
of fifteen to twenty languages” (Sm ith and Tsimpli 1995: 1). However, he could
not acquire an invented language that involves a number of rules referring to
220 Noriaki Yusa