linear order, like “suffix an emphatic particle nog to the third orthographic
word” (Sm ith and Tsimpli 1995). This clearly shows that languages do not use
simple computations based on linear order or fixed positions in a sentence.
Of much relevance is a recent study on the acquisition of word order in an
artificial language. Cu lbertson and Adger (2014) show in an elegant way that
language learners prefer structural knowledge to distributional knowledge of
their native language when learning a new language (an artificial language based
on their native language). The fact that learners’ inference for the word order
in a new language rests on abstract structure more than linear ordering of words
in their language clearly shows that learners are biased toward abstract hierarchi-
cal structure in language acquisition, contrary to the claim in Fra nk et al. (2012).
Neuroimaging data also disconfirm Frank et al.’s claim. The development
and availability of neuroimaging techniques have provided us with new windows
into where exactly within the brain specific language processing (syntactic,
semantic, or phonological) is localized. Major current models of language pro-
cessing refer to anatomical regions (e.g., Broca’s area), pathways (e.g., dorsal
or ventral), or networks of interconnected brain regions (see Fried erici 2011
and Poepp el 2012 for the review). To examine the neuroimaging evidence that
Broca’s area responds differently to structure-dependent rules and structure-
independent rules (Musso et al. 2003), I will limit myself to Broca’s region.^1
In Andrea Moro’s words (2008: 161), “[in] the same way as impossible sentences
have led to significant advancement in the understanding of the formal mecha-
nisms behind linguistic competence, languages may lead to significant advance-
ment in the understanding of the neurobiological nature of the limits of variation
across languages.” This motivated Andrea Moro and his colleagues to find
neuroimaging data for structure dependence in the brain. They investigated the
neural correlates of Germans acquiring new linguistic knowledge of two para-
metrically different languages (Italian or Japanese) by examining whether the
brain would distinguish between structure-dependent and structure-independent
operations. Universal Grammar (UG)-consistent, structure-dependent rules, for
example, include a rule such as “dropping the subject pronoun in Italian,” while
UG-inconsistent, structure-independent rules include a rule such as “forming
negation by inserting nai after the third of the sentence in Japanese.” Partici-
pants learned the rules quite accurately possibly due to ceiling effects of learning.
They carried out fMRI experiments to compare the acquisition of syntactic rules
based on hierarchical structure to ones based on fixed positions in the sentence.
What they found is that syntactic rules following structure dependence selectively
activate Broca’s area, including the pars triangularis (BA 45), while syntactic
rules following structure-independent rules do not. Broca’s area is involved in
the acquisition of new linguistic knowledge respecting the principle of UG.
Whatever property in the brain it is that leads to the use of structural informa-
tion in language processing, this is one of the most exciting discoveries in
cognitive neuroscience research based on linguistic theory and has far-reaching
consequences for research on language faculty. To strengthen their conclusion,
I would like to play the devil’s advocate. Musso et al. (2003) report that during
Syntax in the brain 221