Neuroscience and Biobehavioral Reviews 81 (2017) 213–224

Contents lists available at ScienceDirect

Neuroscience and Biobehavioral Reviews

journal homepage: www.elsevier.com/locate/neubiorev

Review article

A revival of Homo loquens as a builder of labeled structures:

Neurocognitive considerations
T. Goucha a,b,∗ , E. Zaccarella a , A.D. Friederici a,b
a
Max Planck Institute for Human Cognitive and Brain Sciences, Department of Neuropsychology, Leipzig, Germany
b
Berlin School of Mind and Brain, Humboldt University, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The core capacity of human language is described as the faculty to combine words into hierarchical struc-
Received 2 March 2016 tures. This review aims to isolate the fundamental computation behind the language faculty together
Received in revised form 24 January 2017 with its neural implementation. First, we present our central hypothesis by confronting recent linguis-
Accepted 26 January 2017
tic theory with evolutionary arguments: linguistic humaniqueness is reflected in the labeling of word
Available online 17 March 2017
combinations forming asymmetric hierarchical structures. Second, we review the neurolinguistic litera-
ture, especially focusing on dual-stream connectivity models. We put forward that the dorsal pathway,
Keywords:
especially the arcuate fascicle, is responsible for the rule-based combinatorial system, implementing
Labeling
Language evolution labeling and giving rise to hierarchical structures. Conversely, the ventral stream is rather responsible for
Syntax semantic associative operations. We further present evolutionary neuroanatomical evidence grounding
Hierarchical processing our hypothesis. We conclude by suggesting further avenues of research as well as open questions to be
Arcuate fascicle addressed. With the aim to expand our knowledge on the neurobiology of language, we hope to provide
a testable hypothesis for the origin of language syntax bringing together evidence from different fields.
© 2017 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2. The Merge-only hypothesis and why we need labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3. Neural implementation of the language faculty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.1. From hierarchical processing in the prefrontal cortex to the language network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2. The neural implementation of labeling in the language network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
3.3. Integrating the hypothesis with evolutionary evidence in neuroanatomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
3.4. Implications for future research: Which predictions can we make? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4. Conclusions: Is there really something special about language? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

1. Introduction
support the emergence of language as a species-specific faculty.
Language is as a powerful communicative tool which enables While this quest has stimulated psycholinguistic and neurolin-
humans to exchange complex states of affairs with each other. The guistic studies (Friederici and Singer, 2015), comparative research
language system constitutes a rather recent evolutionary innova- has also provided important insights about the commonalities
tion, and appears to be unique to the human species. Given this and differences between humans and nonhuman primates (NHPs)
uniqueness, a central aim for current cognitive research is to iso- (Berwick et al., 2013) In parallel, linguistic theory has accord-
late some specific trait in humans, which would most plausibly ingly addressed the new challenges arising from the findings of
other fields (Bolhuis et al., 2014; Everaert et al., 2015). How-
ever, these attempts to isolate the language faculty are yet to
∗ Corresponding author at: Max Planck Institute for Human Cognitive and Brain achieve success. This apparent insufficiency might arise from the
Sciences, Stephanstraße 1A, 04103, Leipzig, Germany. difficulty to integrate insights from these very disparate disci-
E-mail address: goucha@cbs.mpg.de (T. Goucha). plines. Here, we try to overcome these shortcomings and provide

http://dx.doi.org/10.1016/j.neubiorev.2017.01.036
0149-7634/© 2017 Elsevier Ltd. All rights reserved.
214 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
a testable framework based on data collected by different disci-
plines, always bearing in mind a coherent big picture. The present
framework will especially draw on evidence from cognitive neu-
roscience, which will accompany the argumentation along this
paper.
A multidisciplinary approach to the question of language evo-
lution and its brain implementation is the most adequate strategy
to provide us with helpful empirically testable frameworks that
can contribute to the progress of the field of biolinguistics (Boeckx,
2013; Bickerton, 2014b). A clear advantage of such a multidisci-
plinary perspective is the possibility to combine the constraints of
the different fields to reduce the scope of hypotheses in considera-
tion (Fitch, 2014). If we consider the language faculty to be a recent
adaptation of our species, there are several aspects we must take
into account. First of all, the fundamental language computation in
focus should not only be in agreement with linguistic theory, but
also with comparative cognitive research. This evolutionary differ-
ence should thus constitute a cognitive novelty in our species, in
particular the cognitive process in question should not be present
in our closest relatives. Otherwise, it would be difficult to explain
how language developed exclusively in humans. Furthermore, we
should carefully consider the neural implementation of this cogni-
tive difference that originated language. This means that we must
be able to provide a plausible neuroanatomical substrate for this
computation that shows some novelty in humans in comparison
with NHPs.
A striking fact about language is the possibility to combine
words in a way that new relationships are established between
them, and novel meanings arise, being conveyed by the way they
are combined. Because of this, a fine-grained description of the
algorithm holding this combinatorial process takes on a crucial
role. One of the dominating views in the field is the hypothe-
sis of merge being the crucial computation, put forward by the
linguist Noam Chomsky (Chomsky, 2007; Bolhuis et al., 2014;
Everaert et al., 2015). In this framework, the language faculty
consists of a single computation—Merge—that combines two ele-
ments and gives rise to an unordered symmetrical set. Merge is
therefore considered the basic hierarchy building computation that
enabled humans to master language within the Minimalist Pro-
gramme (Berwick and Chomsky, 2011; Bolhuis et al., 2014; but
see also Boeckx and Benítez-Burraco, 2014). While we support
the idea of a simple and parsimonious combinatorial computa-
tion at the basis of language, we will nonetheless try to elucidate
why Merge might need expansion in order to provide explana-
tory adequacy in both cognitive and evolutionary domains. Here,
we put forward that the labeling of the outcome of the opera-
tion Merge is a necessary cognitive prerequisite for a complete
account of the uniqueness of human language processing. It is
through labeling that asymmetric hierarchical structures can orig-
inate, thus distinguishing language from other communication
systems (Murphy, 2015a, 2015b). In the following sections we
will start by explaining our proposal from a language theoretical
perspective. We will then integrate the hypothesis with current
neuroscientific findings and discuss its concrete neural implemen-
tation. Subsequently we will provide comparative data in support
of our view. In the final part, we will suggest possible avenues for
future research.
2. The Merge-only hypothesis and why we need labels
We start by discussing the assumption that the evolutionary
step that allowed humans to acquire language is a novel combina-
torial faculty that generates hierarchical structures unique to the
human communication system. At the center of the Minimalist Pro-
gram (Chomsky, 1995) stands the computation Merge, defined as
the combinatorial mechanism that brings two elements together
to form an unordered set (Chomsky, 2013), as in:
␣ˇ → {˛ˇ} (1)
The expression above can be read as “take two elements ␣
and ␤, and string them together to form a new set containing
both”. Merge can create a new set including words from any lex-
ical categories (e.g., nouns, verbs, determiners, which correspond
to terminal nodes), or sets that had already been formed by Merge
itself. This possibility of a computation to take its own output as an
input again is what defines recursion, a cornerstone in the discus-
sion about the language faculty (Hauser et al., 2002). For the output
of this computation to be correctly interpreted at the interfaces
(either the sensorimotor interface or the conceptual intentional
interface), the issuing sets need to be labeled. The label of the new
set is attributed according to the labels of its elements and rela-
tionship established between their categories. Labeling determines
a local asymmetry between the two items in the set, since one of
the two elements assigns its label (let it be ␣) to the new merged
object, in this case the label ␥␣ (2):
␣␤ → ␥␣ {␣␤} (2)
In linguistic theory, the element assigning the label to the
merged object is said to project and it is called the selector or head
of the structure formed by the two merged elements. Narita (2014)
addresses how the labeling of the set resulting from Merge is closely
related to the categories of the elements of the set. In order to illus-
trate how labeling arises, let us consider the example of one of
the most essential asymmetries between categories in language –
the asymmetry between nouns and verbs. In this case, the verb (V,
“eat”) is usually the selector, taking the noun (N, “apples”) as its
object. The verb hence imposes its label onto the other component
with which it was merged, e.g. a verb phrase (VP, “eat apples”).
In more abstract terms, this shows different elements in language
play different roles when combines. Some elements in language
are rather functional and dynamic, and usually refer to events.
These establish the relationships between elements in a phrase and
between phrases in a sentence, thus determining the hierarchical
structure. In turn, other elements are mainly content-bearing and
static, and rather refer to sorts. The constituents involved in (2)
operate accordingly. The head ␣ stands in an asymmetrical rela-
tionship with its complement ␤, based on the internal grammatical
relationship between the two elements. The same analysis applies
to the phrase “which apples”. Formally, the concrete bare syntac-
tic set {which, apple} resulting from the Merge of “which” and
“apple” is labeled Determiner Phrase (DP) because it is a property
of the determiner (D) to require a noun (N) as a complement (3).
Therefore, the label received by the newly formed set expresses
the asymmetric hierarchical nature of this same set. This happens
because one of its elements, the determiner (D), projects its label
to the newly formed phrase, a determiner phrase (DP):
which, apple → DP {whichapple} (3)
In the Minimalist Program, for the sake of theoretical ele-
gance, Merge is kept rather undefined. Consequently, the labeling
algorithm is excluded from the basic language computations and
labels are only established for interface requirements (Berwick
and Chomsky, 2016). The definition of Merge is therefore left open
regarding both the properties of its inputs (e.g., its category) and
of the structure that constitutes its output (e.g., which label the
merged set takes). This, however, appears to conflict with the
desired explanatory adequacy of the language faculty from a biolin-
guistic perspective. In particular, it does not seem to be able to
directly account for the evolutionary mechanisms of the emer-
gence of language. When considering evolution, we must equate
which could have been the difference to our ancestors, which could
 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
eventually have led to the fixation of that specific trait. The human
language has the advantage of allowing individuals to exchange
information about complex affairs and mental states. Humans are
able transmit combinations of meaningful units that establish rela-
tions of dominance between themselves, i.e. form hierarchical
structures. Yet, in most cases of animal communication, the chan-
nel only allows information to be transmitted in the form of linear
strings (i.e., sequences where one element follows the previous one
without any apparent hierarchical structure). Without any further
information about a generative rule, a communication channel only
allowing linear strings makes it difficult to transmit information
organized in a hierarchical fashion. However, in humans, hierar-
chical structures can be conveyed due to successful linearization.
This is achieved by flattening hierarchical structures during produc-
tion or by building hierarchical structures from a string of words
during perception. This process constitutes one of the most cru-
cial issues in linguistic theory, at least since the seminal work of
Kayne (1994).1 In general, the presence of asymmetric structures
in language has long been debated from a biolinguistic perspec-
tive (Sciullo and Jenkins, 2016). Kayne’s view was later extended to
include the generation of structures exhibiting points of symme-
try in the course of derivation, for which a linearization ordering
cannot be determined (Dynamic Antisymmetry; Moro, 2000).2
Moving from these theoretical considerations, it is not straight-
forward to understand how Merge can account for the generation of
a hierarchical structure in and of itself. Crucially, from an unordered
set without labels it is not possible to retrieve any relation of dom-
ination between the elements within the set, which would in turn
be a necessary condition for hierarchical structure. Labeling consti-
tutes a crucial performative advantage for dealing with hierarchies,
especially for sharing them when hidden in linearized sequences. In
language use, participants share conventions concerning the label-
ing of the elements present in linear sequences. In particular, these
conventions allow them to recognize which labels the single ele-
ments carry and which combinatorial rules apply between them.
Specifically, they know which of the merged elements selects the
other, thus being the head of the new merged set at each com-
bination. Eventually, this allows the language user to retrieve the
hierarchical relationships between the elements in the sequence.
1
According to Kayne’s Linear Correspondence Axiom (LCA), there is a direct map-
ping between hierarchical syntactic structure and the linear order in a sequence,
such that asymmetric relations between hierarchical nodes map onto precedence
relations. Formally, a lexical entry ␣ will linearly precede a lexical entry ␤ if the
corresponding non-terminal node of ␣ will asymmetrically c-command the corre-
sponding non-terminal node of ␤. C-command, in turn, is defined as a structural
relation between two nodes, where one node is said to c-command another one in
case the two nodes do not dominate each other, but the first node up in the tree that
branches dominates both of them (Reinhart, 1976). When both nodes c-command
each other, there is symmetric c-command. When one node c-commands the other,
but the latter does not c-command the former an asymmetric c-command config-
uration is established and linear ordering follows. The sentence John likes apples,
for example, can be structurally represented with brackets’ notation (with reduced
formalism for simplicity) as [[John][likes [apples]]], where the node containing John
asymmetrically c-commands the nodes containing likes and apples, while the node
containing likes asymmetrically c-commands the node containing apples. From this,
it follows that John must precede likes, which in turn must precede apples.
2
Mirror structures, like copular sentences in Italian for example (“Questa foto sul
muro è la causa della rivolta”, i.e., “This picture on the wall is the cause of the riot”)
have been indeed suggested to represent such a case where two Noun Phrases (NPs;
“the cause of the riot”, “this picture on the wall”) are first both included in the same
node, [è [questa foto sul muro, la causa della rivolta]]. According to Dynamic Anti-
symmetry, instability due to points of symmetry within a structure is then rescued
by the movement of one of the two NPs to a preverbal position. As such, if neither
NP is able to project a label, one of the two elements moves, forcing the other to
project and establish the asymmetry (Moro, 2014). Similarly, movement – a perva-
sive phenomenon of language – is taken to be the computational source preserving
phrase structure asymmetries when symmetric structures occur and linearization
cannot be determined.
215
Only this way is it possible to go back and forth between a hier-
archical mental structure and a linear string produced in vocal
communication. According to Berwick et al. (2013) the interchange
between the mental structure and the linear string is supported by
the sensory-motor interface. Within the Minimalist Program, label-
ing plays a secondary role because it constitutes an additional step
in combining two lexical items. In this account, labeling is only
needed in language externalization, playing an interpretive role at
the interfaces and providing the basis for the linearization (Everaert
et al., 2015; but see Hornstein 2009 for a different account). How-
ever, with the aforementioned performative advantage, we put
labeling forward as plausibly contributing for the emergence of
language along with Merge.
Although we pinpoint labeling as a fundamental linguistic com-
putation, labeling is indeed also necessary for more superficial
linguistic phenomena, like word order. Most languages exhibit a
preferred word order between two elements in a phrase, called
headedness (Comrie, 1989), which determines whether the head or
the complement come first in a linear order. We do agree that lin-
earization is a mere requirement of the material constraints of the
sensorimotor interface (i.e., the necessity to convey one element at
a time, hence producing a string). But note that the same labeling
can give rise to different linear orders and the same linear orders can
also correspond to a different labeling. An example of the former
would be languages with different headedness, which we would
see in the French and German translations of “blue car”, “voiture
bleue” and “blaues Auto”, respectively with the noun correspond-
ing to car in the first or last position, but without any consequences
for interpretation. The latter can be illustrated by the two possibili-
ties of interpretation of “green tea cup” as either {green, {tea, cup}}
or {{green, tea}, cup} (Fujita, 2014). Statistical learning could point
a more likely interpretation, but only a fully hierarchical labeling
procedure would allow the speaker to retrieve both interpretations.
This example illustrates how labeling is not only responsible for
surface phenomena, but is in fact a fundamental computation in
language processing.
3. Neural implementation of the language faculty
3.1. From hierarchical processing in the prefrontal cortex to the
language network
Given the fundamental hierarchical nature of human language,
neurobiological investigations into the neural implementation of
the language faculty should be able to describe which brain areas
and which neural circuits are consistently involved in the pro-
cessing of hierarchy. We know that the prefrontal cortex (PFC)
in general is involved in such processes regardless of the cogni-
tive domain (e.g. Koechlin et al., 1999). For the dorsolateral PFC,
there is an established rostrocaudal gradient regarding the degree
of abstraction of mental representations (Badre, 2008), with more
anterior areas being responsible for more abstract processing. This
gradient has been interpreted as reflecting a certain hierarchy of
processing complexity between the subcomponents of the dorso-
lateral PFC along this gradient (Badre and D’Esposito, 2009) as well
as the corresponding degree of automaticity of these processes in
the opposite direction (Jeon and Friederici, 2013). Importantly, in
the specific case of language processing, the more abstract complex
hierarchies seem to rather engage posterior areas of the inferior
frontal gyrus (IFG; Jeon and Friederici, 2015). In this respect, the
involvement of the IFG within the language system seems to be
governed by other principles of organization, more relevant for
the language dimensions, like phonology, syntax or semantics (e.g.,
Uddén and Bahlmann, 2012). That’s why we will henceforth focus
our attention in the ventral part of the PFC and its subcomponents.
216 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
Fig. 1. Dual stream model of language processing. This model integrates data from functional and anatomical neuroimaging into two main dorsal pathways involved in
syntactic processing and articulation respectively and two main ventral pathways involved in local combinations and semantic processing in general.
In the IFG, we find both Broca’s area and the ventrally adjacent
frontal operculum (FOP), and the anterior insula (Fig. 1). Broca’s
area is the part of the left IFG spanning between the precentral
sulcus and the horizontal branch of the sylvian fissure. Within
this area, there is a well-established gradient between more ante-
rior and more posterior areas (Hagoort, 2005), regarding function
(Dapretto and Bookheimer, 1999), cytoarchitecture (Amunts et al.,
1999) and anatomical connectivity (Friederici, 2012). In terms of
brain function, the more posterior pars opercularis (roughly corre-
sponding to Brodmann Area (BA) 44) is involved syntax processing,
as shown in meta-analytical studies (Hagoort and Indefrey, 2014;
Rodd et al., 2015), literature reviews (Friederici, 2011), and patient
studies (Wilson et al., 2010). The more anterior pars triangularis
(BA 45) (often extending to pars orbitalis (BA 47)), in turn, is
rather engaged by sentence-level semantic processing (Friederici,
2011; Zaccarella et al., 2015) and is systematically more activated
by semantic than syntactic tasks (Hagoort and Indefrey, 2014).
Cytoarchitectonically, we can also distinguish two regions within
Broca’s area with an approximate correspondence to the anatom-
ically defined pars opercularis and pars triangularis. BA 44 shows
a significantly less developed layer IV compared to the anteriorly
located BA 45, being the formed classified as dysgranular and the
latter being properly granular. BA 44 also borders posteriorly with
the premotor BA 6, which is in turn agranular. (Brodmann, 1909;
Amunts et al., 2010).
Additionally, it is important to distinguish the dysgranular BA
44 in the lateral cortical convexity from the adjacent FOP and the
anterior dorsal insula (Sanides, 1962), which are agranular and
evolutionarily older, lying ventrally and medially to Broca’s area.
These areas are not only cytoarchitectonically different, but also
play different roles in hierarchical processing. Functional stud-
ies on artificial grammar processing showed that BA 44 is rather
engaged by more complex grammars with a hierarchical structure
(e.g. Friederici et al., 2006; Opitz and Friederici, 2007), whereas
more simple linear grammars only recruited the more ventral FOP
(e.g. Friederici et al., 2006). This difference between two kinds of
grammars was explored in the literature especially through the
distinction between so-called phrase structure grammars (PSG)
and finite state grammars (FSG). PSG allow for center embedded
structures with long distance dependencies, FSG are based on local
transitions. By distilling down the combinatorial rule to a very sim-
ple pairing of items of two classes A and B, PSG issues structures
of the generalized form An Bn , whilst FSG generate sequences of
the type (AB)n . Studies with these very simplified grammars were
very useful to set the frame in terms of evolutionary differences
in sequence learning and also regarding its brain implementation.
Whereas humans can process both grammars effortlessly, NHPs,
in particular cotton-top tamarins, can only capture the rule of the
simpler FSG (Fitch and Hauser, 2004). Furthermore, the two differ-
ent kinds of grammars have differential patterns of brain activation
in humans (Friederici et al., 2006). While violations of the FSG only
activates the evolutionarily older FOP, with a main ventral white
matter connection, the PSG additionally engages the lateral BA 44,
with a main dorsal white matter connection through the arcuate
fascicle. Not only violation, but in general the presence of the hier-
archical rule engaged BA 44 (Bahlmann et al., 2008), even in the
visual domain (Bahlmann et al., 2009). Other grammars based on
several combinatorial rules, e.g. Reber grammars (Reber, 1967),
were also investigated in both neuroimaging and brain stimula-
tion studies (for a review see Petersson and Hagoort, 2012), whose
main neural correlate was the activation of the left IFG. This was
true for studies involving sequence processing with an underly-
ing generative rule, in particular for Reber grammars (Forkstam
et al., 2006), also showing a good overlap with natural language
processing in this brain area (Petersson et al., 2012). Studies using
brain stimulation also highlighted the importance of the IFG for the
processing of these artificial grammars, for instance in the detec-
tion of violations (Uddén et al., 2008), but also in underlining its
importance in the actual acquisition of the rules rather than the
superficial familiarity (De Vries et al., 2010), also showing corre-
lates in structural connectivity in the adjacent white matter (Flöel
et al., 2009). In this sense, it seems that evolutionarily more recent
areas (i.e. the IFG) are involved in complex structure building, both
in artificial grammars (Friederici et al., 2006) and natural language
(Makuuchi et al., 2009), whereas areas shared with other mam-
mals are rather involved in concatenation processes, which precede
structure building (Zaccarella and Friederici, 2015a). In general, BA
44 seems to be an area that is able to combine different levels of
information for hierarchical building (Wang et al., 2015), instead
of merely being responsible for the blind chunking of elements
regardless of their internal structure (Clerget et al., 2009).
Even though it is possible to establish a certain functional seg-
regation in terms of cortical areas, the specificity of the processes
performed in a certain brain area strongly depend on its con-
nectivity. We must therefore distinguish general computations
performed by certain area from the specific function it performs
within a certain network. Long range white matter fiber bundles
connect Broca’s area with temporal regions in the left hemisphere.
First, we must distinguish the dorsal fiber tracts that connect the
posterior part of the IFG – pars opercularis (BA 44) and the ven-
tral precentral gyrus (BA 6v) – from the ventral fiber tracts that
connect both the more anterior IFG – pars triangularis (BA 45) and
 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
orbitalis (BA 47) – and also the more ventral FOP (Friederici et al.,
2006; Anwander et al., 2006; Friederici, 2012). Recent models (e.g.
Friederici and Gierhan, 2013) propose the existence of two dorsal
and two ventral fiber bundles displayed schematically on Fig. 1.
Dorsally, part of the superior longitudinal fascicle (SLF) connects
the superior (STG) and middle temporal gyri (MTG) to the premo-
tor cortex via the parietal cortex (Catani and Jones, 2005; Makris
et al., 2005), associated with auditory-motor integration (Frey et al.,
2008; Saur et al., 2008). Also belonging to the dorsal pathway, the
arcuate fascicle (AF) directly links BA 44 to the posterior STG/MTG.
This structure was singled out to play a central role in process-
ing syntactically complex sentences (Friederici, 2009; Wilson et al.,
2011; Brauer et al., 2011), but can also be more generally seen as a
crucial structure for predicting the incoming input and its under-
lying structure (Friederici, 2012), and to play an important role in
the integration of syntactic and semantic information (den Ouden
et al., 2012), with the posterior STG/MTG being a pivotal area for the
integration of different sources of information (Friederici, 2011).
Ventrally, the extreme fiber capsule system (EFCS) includes the
inferior fronto-occipital fascicle (IFOF) connecting BA 45/47 to the
temporal cortex, involved in sound-meaning mapping (Saur et al.,
2008; Rolheiser et al., 2011; Griffiths et al., 2013). Additionally,
the uncinate fascicle (UF) connects the most ventral part of the
prefrontal cortex, including the FOP, with the anterior temporal
lobe (ATL), and has been seen as important for local concatena-
tions (Zaccarella and Friederici, 2015a) and adjacent dependencies
(Friederici et al., 2006). Also for artificial grammar learning, the
functional segregation discussed earlier corresponded to a differ-
ence in anatomical connectivity: The activation peak of grammars
with local transitions – FSG – had a dominant ventral white matter
connection whereas the activation peak of hierarchical grammars –
PSG – had a clear dorsal connection. Crucially, the functions of these
white matter pathways show an overall correspondence with the
functional segregation of the brain areas they connect, especially
in the IFG.
3.2. The neural implementation of labeling in the language
network
As already mentioned, an attempt to fully understand cognition
from a neural perspective requires us to depart from localizationist
approaches. We must instead appreciate how brain areas are inte-
grated into circuits, to which we can assign more circumscribed
computations, by combining the specificities of the areas involved.
In the particular case of labeling, its implementation presupposes
that the operation of hierarchical structure building can access the
categories of the elements that are being combined. For this pur-
pose, the cross-talk between hierarchical structure building and
categorization is mandatory. In the language network, the PFC, in
particular the posterior IFG, is largely involved in hierarchical pro-
cessing of linguistic structures (Fitch and Martins, 2014; Zaccarella
et al., 2015), whereas the temporal lobe is engaged in categoriza-
tion processes ranging from phonetics (e.g. Overath et al., 2015)
to semantics (Clarke and Tyler, 2015) and long-term conceptual
storage (Ullman, 2001; Price, 2012).
In general, the cross-talk between frontal and temporal cor-
tices in syntactic and semantic processing is central in language
processing. On the one hand, we can observe a certain functional
segregation of the aforementioned processing streams accompa-
nied by a division of labor between different brain areas. On the
other hand, the information conveyed by these two processing
streams has to be integrated, which especially takes place at level
of the posterior temporal cortex (Friederici, 2011). The interaction
between the frontal cortex and ventral brain areas both in the lat-
eral and in the medial temporal cortex plays an important role
in categorization processes. Already in primates for non-linguistic
217
information, there is a coordination of the frontal and temporal cor-
tex for categorization (Freedman et al., 2003), although the storage
aspect is a function of the temporal cortex (Szczepanski and Knight,
2014). In the linguistic domain, the storage of lexical items seems
to be rather of function of the temporal lobe, which are then inte-
grated along the ventral stream (DeWitt and Rauschecker, 2012).
Specific manipulations involving the retrieval of lexical items from
the mental lexicon engage the posterior ventral temporal cortex
(Lau et al., 2008). Besides, the posterior middle and inferior tem-
poral cortices are significantly more recruited by lexical semantics
than syntax processing in a meta-analysis (Hagoort and Indefrey,
2014). When linguistic information is confined to word semantics,
the brain activations only engage the ventral stream (Graves et al.,
2007). Indeed, disorders affecting temporal regions anterior to the
primary auditory cortex, i.e. along the ventral stream, spare the pro-
duction of abstract sentence structure, whereas the word retrieval
and also the use of words in context are severely altered (Mesulam
et al., 2015). In turn, if we are tapping into merely structural pro-
cessing, independently of content, the brain activations are rather
circumscribed to BA 44, which is the case of studies using syntac-
tic manipulations controlling for working memory like scrambling
(Meyer et al., 2012) or embedding (Makuuchi et al., 2009), or stud-
ies using pseudowords (Röder et al., 2002; Goucha and Friederici,
2015). Its neighboring BA 45, mostly involved in sentence level
semantic processing, rather comes into play when the syntactic
manipulation has additional semantic demands, e.g. in the case
of anaphora in movement (Santi and Grodzinsky, 2010). Regard-
ing interactions between syntax and semantics, which is especially
relevant in thematic role assignment, the posterior superior tem-
poral cortex comes into play (Friederici et al., 2003; Bornkessel
et al., 2005), highlighting its important role in the integration of
the two processing streams. In terms of white matter connectivity,
we find a dominance of the dorsal stream in the processing of com-
plex hierarchical structures, which is reflected in brain anatomy
(Skeide et al., 2015) and in effective connectivity (den Ouden et al.,
2012), whereas the ventral connections rather account for seman-
tic aspects (Rolheiser et al., 2011). In fact, the fundamental role
of the arcuate fascicle has already been stressed in the context
of language evolution (Friederici, 2012; Berwick et al., 2013), lan-
guage acquisition (Perani et al., 2011; Skeide et al., 2015) and its
anatomical uniqueness in humans (Rilling et al., 2008). However,
here we extend the interpretation of the functional role of the arcu-
ate fascicle to the crucial connection between two brain areas with
two essentially different kinds of computation, whose crosstalk is
fundamental for building language structures.
In the same way that the neural implementation of language
can be seen as dichotomy between two processing streams, we can
also discern two sides in the language’s building blocks − words.
On the one hand, words possess a categorical dimension that is
related to the syntactic rules used in hierarchal structure building
and, on the other hand, words contain a conceptual dimension that
is related to their content. The rules that guide the combination
of words and allow us to label the resulting hierarchical structure
generally need two different sources of input. First, we need to
retrieve the word category, which determines the general selection
rules (e.g., headedness, constituent structure), and seems to be pro-
cessed in the inferior frontal cortex (Friederici et al., 2000a). Second,
the brain also has to access the grammatical relations established
by each specific lexical item, which bring together syntactic and
semantic information. This is, for example, the case of the argu-
ment structure of a verb, where a certain hierarchical structure
with specific constituents marked by case, such as nominative or
accusative, correspond to the different thematic roles of the verb,
such as agent or patient. This information, in turn, seems to rather
be stored in the in the posterior temporal lobe (Bornkessel et al.,
2005). The dorsal stream, in particular the arcuate fascicle, would
218 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
Fig. 2. Types of operations performed in the language network. Our framework aims to disentangle different kinds of computations performed within the perisylvian
cortex and identify which circuits are responsible for their neural implementation. We also distinguish artificial grammar processing from natural language processing.
In language comprehension and production, semantic combinatorial processing mainly takes place along the ventral stream, whereas syntactic hierarchical processing is
located in the dorsal stream.
then enable the establishment of these syntactic-semantic relations
that require the interaction of these two sources of information.
This way, labeling would be anatomically be enabled by the arcu-
ate fascicle, which allows the connection with the interfaces. The
dorsal pathway, especially the arcuate fascicle, is hence crucial part
of the rule-based system engaged in the processing of hierarchi-
cal structures. The uniquely semantic dimension of words, in turn,
seems to be processed along the ventral pathway, with associative
principles of word combination. We must thus distinguish between
the rule-based mechanism of the dorsal stream connecting BA 44
and the posterior temporal cortex, where sequences are assigned
a hierarchical structure, and the combinatorial mechanisms of the
ventral stream connecting BA 45 and the temporal cortex. The kind
of concatenation that we find in the ventral stream and in the evo-
lutionarily older regions of the frontal cortex is essentially different
from the hierarchical labeled structures produced by the dorsal
stream (Friederici et al., 2006; Zaccarella and Friederici, 2015b). The
combinations performed along the ventral stream are mainly based
on associative operations, possibly resorting to template match-
ing, which can give rise to linear sequences. The ventral stream
was shown to be fundamental in concept combination (DeWitt
and Rauschecker, 2012), in particular in the anterior temporal lobe
(Bemis and Pylkkänen, 2013a), an area that has also been associated
with retrieval of word meaning (Mesulam et al., 2015). The com-
binatorial processes in the ventral stream thus seem to depend on
local transition rules, as we can also see in artificial grammar for
adjacent non-hierarchical dependencies studies (Friederici et al.,
2006). Language processing along the ventral stream appears to
be dependent on declarative memory, operating on the associative
strengths between the elements of a sequence. These mechanisms
of concept association have been put forward as spreading activa-
tion (Collins and Loftus, 1975), where representations engage other
meanings that neighbor their conceptual network (Pulvermüller,
2013), with a crucial role played by the temporal lobe. The con-
ceptual networks established by association rather give rise to
mental scenes, without any hierarchical projection of its elements.
In Fig. 2 we schematically summarize the possible different kinds of
computation taking place in the language network and their neu-
ral implementation, spanning from artificial grammar learning to
combinatorial semantics and hierarchical structure building.
The possibility of different combinatorial mechanisms in lan-
guage was criticized by Bornkessel-Schlesewsky et al. (2015), who
claim that they should be narrowed down to a single computa-
tion, thus making any distinction between simple, linear structures
and complex, hierarchical structures rather futile. Although it is
clear from a linguistic theoretical point of view that the complex
syntactic structures result from the reapplication of the same ele-
mentary combinatorial algorithm (cf. 2. The Merge-only hypothesis
and why we need labels), we put forward that not all computations
performed in the language network are necessarily hierarchical
(Frank et al., 2012; Christiansen and Chater, 2015), in particular the
combinatorial processes that happen in the semantic domain. Here,
the distinction between competence and performance (Chomsky,
1965) might prove very helpful. Although our language competence
implies the aforementioned rule-based mechanism of processing
in the dorsal stream, where sequences are assigned a hierarchi-
cal structure, the performance that is implemented in the ventral
stream can be based in more simple heuristics of template match-
ing and resort to associative thinking, producing linear sequences
instead. Such associative mechanisms do not rely on hierarchical
structures, thus not being able to fully explain syntactic phe-
nomena. These strategies for language comprehension have lower
processing demands, since they are based on a superficial under-
standing that does not require the full projection of the syntactic
structure. Participants have been shown to follow such strategies
when not specifically instructed to pay attention to the precise
structure (Ferreira and Patson, 2007). In general, this kind of
associative strategies was recently highlighted as an important
component of complex cognition (Heyes, 2012), despite the fact
that they are more error-prone, based on fast and frugal heuristics
(Gigerenzer and Goldstein, 1996).
Cognitive models presenting a dissociation of rule-based learn-
ing and associative learning are often criticized (Fitch, 2014) since
both types of learning are still based on common mechanisms, in
particular through predictive coding (Clark, 2013). Human cog-
nition, including language, mainly depends on predictive models
formed in higher cognitive areas about the incoming stimulus and
its continuous adaptation as a function prediction errors coming
from primary cortices (Giraud and Poeppel, 2012). Here, we have
to take into account that the predictions made in online process-
ing are based upon mental models. This is precisely where the
difference between the two kinds of processing in the dorsal and
ventral stream resides. In the dorsal stream, we find a predictive
model based on an internal meta-representational level of gram-
 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
mar, allowing for non-adjacent dependencies (Friederici, 2012).
Such a model generates predictions that will be radically different
from those generated by another model exclusively based on local
transition probabilities, for example depending on the frequency of
item co-occurrence, generally implemented in the ventral stream.
In fact, it was possible to see in recent fMRI studies that differ-
ent cues used to establish predictions about the incoming input
engaged segregated processing streams. If cues about content were
used, i.e. supported by combinatorial processes based on rela-
tional semantics, the areas connected by the ventral stream were
engaged, that is the anterior temporal and anterior frontal cortices
(Tuennerhoff and Noppeney, 2016; Scharinger et al., 2015) and also
the hippocampus (Piai et al., 2016). In turn, if the cues used for sen-
tence processing and to make online predictions were limited to
syntactic cues, i.e., cues about sentence structure, we rather see
the involvement of areas that are connected by the dorsal stream
(Goucha and Friederici, 2015). Furthermore, brain oscillations at
different timescales were shown to reflect combinatorial process-
ing in language in phrases and sentences in a way that could not
be totally accounted for by mere local transitional probabilities
(Ding et al., 2016). In fact, even in the case of adjacent dependen-
cies, we do not find exclusively the involvement of the ventral
stream as a function of the violation of local structure building
(Herrmann et al., 2009). When this violation is a product of hier-
archical rules, including agreement, there is a clear involvement of
the IFG (Friederici et al., 2000b), and in particular BA 44 (Herrmann
et al., 2012; Hanna et al., 2014). These studies make a point for
the neural implementation of a predictive model based on a gen-
erative structural combinatorial algorithm. From a neural point
of view, we should distinguish two main combinatorial systems
in language: a dorsal and a ventral one. Both systems necessarily
encompass bidirectional connections with top-down predictions
being sent downstream and prediction errors being sent upstream
in online language comprehension. The main difference neverthe-
less is which predictive models are used by these two processing
streams. Whereas the dorsal stream acts according to a hierarchical
model based on categorical knowledge, permitting long-distance
dependencies, the ventral stream is rather dependent on associa-
tive learning based on transition probabilities in local dependencies
and usually at fixed distances, allowing for the semantic combina-
tion of the sequence elements.
However, the engagement of BA 44, the IFG and the dorsal
stream in general in hierarchical processing is still often attributed
to mere linearization processes, i.e. the IFG is rather responsible
to establish the serial order of the constituents (Bornkessel-
Schlesewsky et al., 2015). According to this view, this region should
be rather engaged in converting a hierarchical structure still with-
out a fixed order to a linear sequence that can be sent to the
interfaces (especially sensorimotor), which is the point at which
labeling would intervene. However, the capacity of a brain structure
to linearize hierarchical structures entails the capacity of build-
ing hierarchical structures itself. If a certain brain area hosts the
computations that allow it to infer the labeling of a hierarchi-
cal structure in order to produce a linear sequence, this area is
necessarily capable of building that same hierarchical structure.
Furthermore, as discussed in the previous section, the labeling of
a hierarchical structure influences its word order, but they are
still two separate dimensions. Boeckx (2014) also voiced criticism
regarding the lexicocentricity of theoretical accounts of label-
ing portrayed in recent work of Hornstein (2009) and Bickerton
(2014a), more recently discussed by Murphy (2015a). We would
like to stress that labeling can still take place in the absence of con-
crete lexical items and is rather based on an abstract dimension,
which in turn allows labels to be marked by inflectional morphol-
ogy. These labels are rather bound to the categories that make the
interface between the mental lexicon and the rule-based system
219
of syntax. In fact, Goucha and Friederici (2015) and Zaccarella and
Friederici (2015c) show how individuals are able to successfully
retrieve the combinatory rules that operate on labels, even though
the access to the actual lexical items was blocked by the use of
pseudowords in both experiments. The two studies were further-
more able to trigger the same pattern of activation seen in syntactic
manipulations with real words.
By distinguishing two combinatorial systems in human cogni-
tion with two segregated implementations, we put into question
the thesis that the multimodal combination of concepts allowed
by Merge is the actual novelty of language (Boeckx and Benítez-
Burraco, 2014). If applied to the linguistic domain, there is a corpus
of evidence that conceptual/semantic combination takes place in
the anterior temporal lobe, a central semantic hub (Bemis and
Pylkkänen, 2013b). Multimodal association of other kinds of men-
tal representations, however, does not involve the same neural
mechanisms as conceptual combination in the semantic domain
let alone syntactic combination. This kind of combination, gener-
ally coordinated by the visual and auditory ventral streams in the
temporal lobe, is not necessarily specific of the language system
nor it is of humans. The medial temporal lobe, part of the declar-
ative system (Ullman, 2001), is highly conserved in mammals and
its mechanisms have similar functioning principles that in gen-
eral allow for the association and combination of mental objects
in memory. The connections of the hippocampus with the cere-
bral neocortex allow for this multimodal integration of mental
objects in the process of memory formation and can be seen even
at very basic domains of cognition present in other species (Suzuki,
2007; Langston and Wood, 2010). Indeed, it can account for the
representation of time-oriented sequences (McKenzie et al., 2014;
Howard et al., 2014), even in artificial grammars, but only when
item-based and not for rule-based learning (Opitz and Friederici,
2004).3 Further, this combinatory mechanism has no constraints
on the representations it combines, but it clearly is not enough to
bring language about in other species. That’s why the mechanis-
tic difference between the combinatorial faculty of language and
other domains proposed here should account for the emergence of
language in humans.
In the investigation of the language faculty, some particu-
lar theoretical proposals received especial attention, in particular
recursion (Hauser et al., 2002). The empirical investigation of
recursion showed that the cognitive resources used for the rep-
resentation of self-similarity in non-linguistic domains does not
depend on language mechanisms (Martins et al., 2015), and also
possesses a very distinct neural representation in the visual domain
(Martins, Fischmeister et al., 2014), with no actual overlap with
linguistic recursion, as seen in Makuuchi et al. (2009). This is an
interesting observation and suggests that recursion of asymmet-
ric structures may be specific. Within language, however, minimal
phrasal structures in language (Zaccarella et al., 2015) largely share
the brain activation patterns of more complex embedded recursive
structures (Friederici et al., 2009). This pinpoints how the reap-
plication of the same computation of labeling not only accounts
for minimal structures but also for complex structures, keeping a
similar neural representation.
3.3. Integrating the hypothesis with evolutionary evidence in
neuroanatomy
In order to consolidate our anatomical hypothesis regarding the
arcuate fascicle, we should integrate it within the recent devel-
3
Rule-based learning of an artificial grammar due to grammatical categories
allowing labeling and hiearchy building showed activation in Broca’s area once the
hierarchical rules were learned (Opitz and Friederici, 2004).
220 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224

opments in comparative genetic studies and more specifically in cognition through the hierarchical embedding of oscillations are
the genetics of language. Geschwind and Rakic (2013) highlight transversal across animals.4
how the big innovation in the human central nervous system In summary, our hypothesis about the neural implementation
have to do with a reshaping of the cortex, especially the pre- of the faculty of language concerns the plasticity in the brain net-
frontal and parietal cortices. More specifically, most of the genes works in both developmental and evolutionary scales. In particular,
associated with vocal behavior and language play a major role we consider the possibility of wiring of brain areas with different
in neural plasticity (Szalontai and Csiszar, 2013). Several stud- contents and functioning modes, instead of assuming a closed hard-
ies show that these genes are related to microscopic connectivity wired circuit at birth. This is the kind of non-disruptive evolution
(Charrier et al., 2012; Enard, 2011; French et al., 2012; Groszer with a plausible genetic background that can still generate great
et al., 2008). Crucially, this microstructural plasticity could be novelty. This point of view allows for coexistence of continuity in
correlated with behavioral changes, leading to a selective improve- the brain mechanisms, which can be seen in brain oscillations com-
ment in rule-based procedural memory (Schreiweis et al., 2014). mon to humans and NHPs, and discontinuity in behavior brought
Overall, the recent phylogenetic changes in the network of genes about by novelty in brain structure.
related to neural plasticity should also influence in the macro-
scopic long-range wiring of brain regions (Geschwind and Rakic,
3.4. Implications for future research: Which predictions can we
2013).
make?
This novel brain rewiring in humans could be the origin of the
connectivity differences between the prefrontal cortex and the
If we intend to investigate the language faculty and its combi-
temporal cortex that we observe in the case of the arcuate fasci-
natorial algorithm isolating its core computation, the easiest way
cle (Rilling, 2014). The cortical terminations of the arcuate fascicle
to approach it without having the strong confound of lexicality is
also seem to be particularly different in comparison with NHPs.
to test it through an artificial grammar. In order to address the role
Especially, BA 44 suffered a big expansion and acquired a left-
of the IFG and the arcuate fascicle in the neural implementation
right asymmetry (Schenker et al., 2010; Spocter et al., 2010), which
of labeling and ultimately in language emerge, it is important to
is accompanied by the strong asymmetry of the arcuate fascicle
design experimental paradigms that directly compare the acquisi-
itself (Nucifora et al., 2005; Vernooij et al., 2007; Takaya et al.,
tion of grammars based on labeled structures with other possible
2015). Rilling et al. (2008, 2011) show a striking difference between
strategies. We should therefore compare rule-based learning, to
the ventral and dorsal anatomical connectivity between NHPs and
grammars that are exclusively based on item similarity, controlling
humans. Even in our closest relatives, the chimpanzees, there is
for other variables like dependency length, that we do not think to
only a faint connection of Broca’s homologues to the posterior
be central to the computational aspect.
planum temporale, whereas humans display the robust termina-
Although most studies in AGL did not include labeling rules in
tions of the arcuate fascicle in the posterior superior and middle
their paradigms, some studies did implement the necessarily labels
temporal gyri (Fig. 3, upper panel). The ventral stream, however, is
for language-like phrase structure rules, in an attempt to attain an
already well developed in NHPs, as shown anatomically with diffu-
abstract structure that would be more comparable to language than
sion tractography and autoradiography, especially from the ventral
other artificial grammars in use. Rohrmeier et al. (2012) showed
PFC to the anterior and middle temporal gyrus (Schmahmann et al.,
behaviorally that participants acquired knowledge of the grammat-
2007), also corroborated by functional connectivity (Neubert et al.,
ical classes and their relations, i.e. the selection rules in use, which
2014; Petkov et al., 2015). In general, these crucial white matter
also applied for long distance dependencies, and were not based on
differences we have been discussing might be related to more gen-
n-gram knowledge. Studies using BROCANTO, an artificial language
eral principles of brain organization distinguishing humans from
also including language-like labels without content, could provide
NHPs (Buckner and Krienen, 2013). Already prenatally, there is
interesting insights in the processing of such structures, more akin
an accentuated slowing down in brain volume growth present in
to language hierarchies. In particular, these studies could show
chimpanzees not observed in humans (Sakai et al., 2012), to be
dissociations between more rule-based learning, based on the arti-
followed by a faster post-birth prefrontal growth in humans in com-
ficial hierarchical structure, and more associative learning, based
parison to both chimpanzees and Rhesus macaques (Sakai et al.,
on sequence similarity and local transitions (Opitz and Friederici,
2011), especially in white matter compared to gray matter (Sakai
2004). The hippocampus was initially more engaged in the acqui-
et al., 2013). Interestingly, these developmental differences across
sition of the grammar, being rather responsive to similarity-based
species find a correspondent in the particular ontogeny of the dor-
learning, but its activation dropped whereas the posterior IFG was
sal pathway in human infants (Dubois et al., 2015). In particular,
progressively more engaged (Opitz and Friederici, 2003). Crucially,
whereas the part of the dorsal stream responsible for articulation
the posterior IFG, especially BA 44, was primarily engaged in rule-
is already fully developed in newborns, the arcuate fascicle has yet
learning, in comparison with the ventral premotor cortex or the
to undergo myelination (Perani et al., 2011; Fig. 3, lower panel).
hippocampus, rather related to the local character of rule change
These insights from developmental and comparative studies sup-
(Opitz and Friederici, 2007). Furthermore, the use of analogous
port our hypothesis regarding the arcuate fascicle and its crucial
rules in cotton-top tamarins and infants showed again a differ-
role in the emergence of language.
ential between humans and NHPs (Saffran et al., 2008). Whereas
Alternative mechanisms based on brain oscillations have been
both infants and cotton-top tamarins were able to detect regular-
proposed as a crucial element for the emergence of language
ities in simple combinatorial patterns, only infants were able to
(Murphy, 2015b). The oscillatory processing of speech and the
acquire complex grammatical structures with a phrase structure
general issuing computational mechanisms are in fact crucial to
by the predictive patterns established by a grammar with selection
understand how the brain encodes different linguistic aspects
rules.
(Giraud and Poeppel, 2012). However, those mechanisms seem to
already be in place in other species. For example, despite the crucial
brain expansion that took place in primates and especially humans
4
compared to other mammals, the rhythmical hierarchy of oscilla- However, there is first evidence that brain oscillations in the theta band related
to verbal working memory can be segregated both functional and topographi-
tions is mainly kept unchanged (Buzsáki et al., 2013). As Friederici cally into a domain-general and a syntax-specific component, with the latter being
and Singer (2015) pinpoint, the basic neural mechanisms behind located in Broca’s area (Beese, Meyer, Vassileiou & Friederici, in preparation). This
once again speaks rather for an anatomical distinction.
 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224 221

Fig. 3. Phylogeny and ontogeny of the dorsal stream. The arcuate fascicle has a particular profile in both comparative and developmental research. The upper panel shows
how the dorsal connections from the IFG to the posterior temporal cortex are highly strengthened in humans in comparison with chimpanzees. The lower panel displays the
fronto-temporal structural connectivity of newborns and adults, where we can observe that the arcuate fascicle is not yet fully myelinated at birth, in contrast to the other
fiber tracts involved in language processing, which are trackable from birth.

Fig. 4. Recent developments in the neural implementation of syntactic processing. A. In the absence of all semantic cues, including derivational morphology, only the
posterior part of the IFG, BA 44, is activated. These data indicate that BA 44 is the core area for abstract syntax processing. B. Even in the processing of local structures (3-word
phrases or sentences) involving lexical items, there is a co-activation of the IFG and the posterior temporal cortex. C. The current framework focuses on the arcuate fascicle
as the structure that enables the cross-talk between abstract structural building in the IFG and the lexically-based argument structure in the posterior temporal cortex.

One important question that remains to be addressed empiri-
cally is the precise function of the arcuate fascicle in hierarchical
structure building. The literature points to its crucial role in linguis-
tic syntax because its degree of myelination determines syntactic
performance in processing syntactically complex sentences in
patients (Wilson et al., 2011) and in children (Skeide et al., 2015).
The arcuate fascicle connects the IFG and the posterior temporal
cortex, which in adults are often simultaneously involved in the
processing of hierarchically complex sentences (Friederici et al.,
2009; Meyer et al., 2012; Pattamadilok et al., 2015; Skeide et al.,
2015), and also in structure building of phrases (Zaccarella et al.,
2015; Fig. 4b). However, the posterior temporal cortex is not
involved in most artificial grammar studies and also in language
studies that isolated the abstract language structure regardless of
meaning (Goucha and Friederici, 2015; Fig. 4a) or in the local syn-
tactic combination of a determiner and a pseudoword resulting
from a single Merge (Zaccarella and Friederici, 2015c), discussed in
another paper in this issue (Zaccarella and Friederici, 2016). Note
that the combination of two words in language can be seen as a
hierarchical structure since this paradigm is productive, as dis-
cussed by Yang (2013), that is, a determiner phrase (DP) allows
for all combinations of determiners and nouns. Do to the genera-
tive properties of this syntactic element, we can already talk about
hierarchy—a structure that is impossible in adjacent relationships
in an artificial grammar. Therefore, according to our interpreta-
tion, this single combination of two words already encompasses
222 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
the labeling of two linguistic elements that have previously been
concatenated. It appears that in order to recruit the temporal cor-
tex the second element needs to be a real noun—a lexical element
with a clear label triggering lexical access. The word categories that
constitute the labels are delivered by the lexicon, which is located
in the middle temporal cortex (Lau et al., 2008; Price, 2012) in the
ventral stream. This raises the question at which step the arcuate
fascicle comes into play. The framework put forward in this paper
highlights the role of white matter pathways in the establishment
of brain computations across distant brain regions. One hypothe-
sis for the function of the arcuate fascicle would be to allow that
the IFG operates hierarchically on the labels and matches them to
lexically-based argument structure frames stored in the posterior
temporal cortex. The integration of these information types is made
possible through the cross-talk established via the arcuate fascicle
(as portrayed in Fig. 4c).
4. Conclusions: Is there really something special about
language?
It is undeniable that language is a uniquely human property,
but it is less clear whether there is a language-specific faculty
behind or if language is simply enabled by the general cognitive
capacity of humans, including enhanced memory and more sophis-
ticated sociality. Despite claims about the linearity of language
and the possibility of processing linguistic strings in a parallel
non-hierarchical fashion (Frank et al., 2012), there are a num-
ber of linguistic phenomena that can only fully be accounted for
by representing utterances as structures with relations of domi-
nance between their elements (Everaert et al., 2015). Following the
rationale of the Minimalist Program, we aimed to identify the fun-
damental computational basis of hierarchical structure building in
language. By integrating linguistic theory in an evolutionary per-
spective, we isolated the need for a labeling algorithm operating on
the output of the elementary combinatorial computation. While
grounding this labeling algorithm and linguistic structure build-
ing in neural circuits, we reinstated a functional divide between
dorsal and ventral white matter pathways in language processing.
The ventral pathway, already developed in NHPs, allows for serial
concatenation and is able to combine meanings based on their asso-
ciative strength, accounting for a non-hierarchical dimension of
language (Jackendoff and Wittenberg, 2014) and non-human pri-
mate communication. In turn, the dorsal pathway is responsible
for the full-fledged hierarchical structures that characterize lan-
guage, allowing for local combinations of words with a labeling
algorithm that can be recursively applied and generate complex
hierarchical structures. Anatomically, we highlight the crucial role
of the arcuate fascicle for this rule-based processing mechanism,
with its important evolutionary path (Rilling, 2014), integrated in
general principles of brain evolution (Geschwind and Rakic, 2013).
This evolution from transitional probabilities to the rule-based con-
struction of linguistic trees (Fitch, 2014; Dehaene et al., 2015)
should constitute the focus of further research. At the interface
between linguistic theory and neurolinguistics, we finally believe
that future neurobiological investigations aiming at describing the
specific nature of linguistic computations in the human brain will
have to consider the fundamental dichotomy between competence
and performance (Chomsky, 1965; Chesi and Moro 2015). Any true
endeavor in biolinguistics must be able to reconcile empirical neu-
robiological research on the basis of performative evidence with the
formal constraints imposed by the internal linguistic competence.
We therefore hope to have provided a constructive hypothesis
concerning a human specific language faculty that is both com-
putationally sound and anatomically grounded.
References
Amunts, K., Schleicher, A., Bürgel, U., Mohlberg, H., Uylings, H., Zilles, K., 1999.
Broca’s region revisited: cytoarchitecture and intersubject variability. J. Comp.
Neurol. 412 (2), 319–341.
Amunts, K., Lenzen, M., Friederici, A.D., Schleicher, A., Morosan, P.,
Palomero-Gallagher, N., Zilles, K., 2010. Broca’s region: novel organizational
principles and multiple receptor mapping. Plos Biol. 8 (9), e1000489.
Badre, D., D’Esposito, M., 2009. Is the rostro-caudal axis of the frontal lobe
hierarchical? Nat. Rev. Neurosci. 10 (9), 659–669.
Badre, D., 2008. Cognitive control, hierarchy, and the rostro-caudal organization of
the frontal lobes. Trends Cogn. Sci. 12 (5), 193–200.
Bahlmann, J., Schubotz, R.I., Friederici, A.D., 2008. Hierarchical artificial grammar
processing engages Broca’s area. Neuroimage 42 (2), 525–534.
Bahlmann, J., Schubotz, R.I., Mueller, J.L., Koester, D., Friederici, A.D., 2009. Neural
circuits of hierarchical visuo-spatial sequence processing. Brain Res. 1298,
161–170.
Beese, C., Meyer, L., Vassileiou, B., Friederici, A.D., (in preparation). Resting state
theta power as lifespan marker of sentence comprehension.
Bemis, D.K., Pylkkänen, L., 2013a. Basic linguistic composition recruits the left
anterior temporal lobe and left angular gyrus during both listening and
reading. Cereb. Cortex, bhs170.
Bemis, D.K., Pylkkänen, L., 2013b. Combination across domains: an MEG
investigation into the relationship between mathematical, pictorial, and
linguistic processing. Front. Psychol. 3, 583.
Berwick, R., Chomsky, N., 2011. The biolinguistic program: the current stage of its
development. In: di Sciullo, Anna Maria, Boeckx, Cedric (Eds.), The Biolinguistic
Enterprise. Oxford University Press, Oxford, pp. 19–41.
Berwick, R.C., Chomsky, N., 2016. Why Only Us: Language and Evolution. MIT
Press, Cambridge, MA.
Berwick, R.C., Friederici, A.D., Chomsky, N., Bolhuis, J.J., 2013. Evolution, brain, and
the nature of language. Trends Cognit. Sci. 17 (2), 89–98.
Bickerton, D., 2014a. More than nature needs. In: Language, Mind, and Evolution.
Harvard University Press, Cambridge, MA.
Bickerton, D., 2014b. Some problems for biolinguistics. Biolinguistics 8, 73–96.
Boeckx, C., Benítez-Burraco, A., 2014. The shape of the human language-ready
brain. Front. Psychol. 5.
Boeckx, C., 2013. Biolinguistics: forays into human cognitive biology. J. Anthropol.
Sci. 91, 1–28.
Boeckx, C., 2014. Elementary Syntactic Structures: Prospects of a Feature-Free
Syntax. Cambridge University Press, Cambridge, http://dx.doi.org/10.1017/
CBO9781139524391.
Bolhuis, J.J., Tattersall, I., Chomsky, N., Berwick, R.C., 2014. How could language
have evolved? Plos Biol. 12 (8), e1001934.
Bornkessel, I., Zysset, S., Friederici, A.D., von Cramon, D.Y., Schlesewsky, M., 2005.
Who did what to whom? The neural basis of argument hierarchies during
language comprehension. Neuroimage 26 (1), 221–233.
Bornkessel-Schlesewsky, I., Schlesewsky, M., Small, S.L., Rauschecker, J.P., 2015.
Neurobiological roots of language in primate audition: common computational
properties. Trends Cogn. Sci. 19 (3), 142–150.
Brauer, J., Anwander, A., Friederici, A.D., 2011. Neuroanatomical prerequisites for
language functions in the maturing brain. Cereb. Cortex 21 (2), 459–466.
Brodmann, K., 1909. Vergleichende Lokalisationslehre der Grosshirnrinde in ihren
Prinzipien dargestellt auf Grund des Zellenbaues. Barth.
Buckner, R.L., Krienen, F.M., 2013. The evolution of distributed association
networks in the human brain. Trends Cogn. Sci. 17 (12), 648–665.
Buzsáki, G., Logothetis, N., Singer, W., 2013. Scaling brain size: keeping timing:
evolutionary preservation of brain rhythms. Neuron 80 (3), 751–764.
Catani, M., Jones, D.K., 2005. Perisylvian language networks of the human brain.
Ann. Neurol. 57 (1), 8–16.
Charrier, C., Joshi, K., Coutinho-Budd, J., Kim, J.E., Lambert, N., De Marchena, J.,
et al., 2012. Inhibition of SRGAP2 function by its human-specific paralogs
induces neoteny during spine maturation. Cell 149 (4), 923–935.
Chesi, C., Moro, A., 2015. The subtle dependency between competence and
performance. In: 50 Years Later: Reflections on Chomsky’s Aspects. MIT
Working Papers in Linguistics, Cambridge, M.
Chomsky, N., 1965. Aspects of the Theory of Syntax. MIT Press, Cambridge, MA.
Chomsky, N., 1995. The Minimalist Program, Vol. 1765. MIT Press, Cambridge, MA.
Chomsky, N., 2007. Approaching UG from Below. Interfaces+
Recursion = Language?, Vol. 89. Mouton de Gruyter, Berlin, pp. 1–30.
Chomsky, N., 2013. Problems of projection. Lingua 130, 33–49.
Christiansen, M.H., Chater, N., 2015. The language faculty that wasn’t: a
usage-based account of natural language recursion. Front. Psychol. 6.
Clark, A., 2013. Whatever next? Predictive brains, situated agents, and the future of
cognitive science. Behav. Brain Sci. 36 (03), 181–204.
Clarke, A., Tyler, L.K., 2015. Understanding what we see: how we derive meaning
from vision. Trends Cogn. Sci. 19 (11), 677–687.
Clerget, E., Winderickx, A., Fadiga, L., Olivier, E., 2009. Role of Broca’s area in
encoding sequential human actions: a virtual lesion study. Neuroreport 20
(16), 1496–1499.
Collins, A.M., Loftus, E.F., 1975. A spreading-activation theory of semantic
processing. Psychol. Rev. 82 (6), 407.
Comrie, B., 1989. Language Universals and Linguistic Typology: Syntax and
Morphology. University of Chicago Press, Chicago, IL.
Dapretto, M., Bookheimer, S.Y., 1999. Form and content: dissociating syntax and
semantics in sentence comprehension. Neuron 24 (2), 427–432.
 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
den Ouden, D.B., Saur, D., Mader, W., Schelter, B., Lukic, S., Wali, E., Thompson, C.K.,
2012. Network modulation during complex syntactic processing. Neuroimage
59 (1), 815–823.
De Vries, M.H., Barth, A.C., Maiworm, S., Knecht, S., Zwitserlood, P., Flöel, A., 2010.
Electrical stimulation of Broca’s area enhances implicit learning of an artificial
grammar. J. Cogn. Neurosci. 22 (11), 2427–2436.
DeWitt, I., Rauschecker, J.P., 2012. Phoneme and word recognition in the auditory
ventral stream. Proc. Natl. Acad. Sci. U. S. A. 109 (8), E505–E514.
Dehaene, S., Meyniel, F., Wacongne, C., Wang, L., Pallier, C., 2015. The neural
representation of sequences: from transition probabilities to algebraic patterns
and linguistic trees. Neuron 88 (1), 2–19.
Ding, N., Melloni, L., Zhang, H., Tian, X., Poeppel, D., 2016. Cortical tracking of
hierarchical linguistic structures in connected speech. Nat. Neurosci. 19 (1),
158–164.
Dubois, J., Poupon, C., Thirion, B., Simonnet, H., Kulikova, S., Leroy, F., et al., 2015.
Exploring the early organization and maturation of linguistic pathways in the
human infant brain. Cereb. Cortex, bhv082.
Enard, W., 2011. FOXP2 and the role of cortico-basal ganglia circuits in speech and
language evolution. Curr. Opin. Neurobiol. 21 (3), 415–424.
Everaert, M.B., Huybregts, M.A., Chomsky, N., Berwick, R.C., Bolhuis, J.J., 2015.
Structures, not strings: linguistics as part of the cognitive sciences. Trends
Cogn. Sci. 19 (12), 729–743.
Ferreira, F., Patson, N.D., 2007. The ‘good enough’approach to language
comprehension. Lang. Linguist. Compass. 1 (1–2), 71–83.
Fitch, W.T., Hauser, M.D., 2004. Computational constraints on syntactic processing
in a nonhuman primate. Science 303 (5656), 377–380.
Fitch, W., Martins, M.D., 2014. Hierarchical processing in music, language, and
action: lashley revisited. Ann. N.Y. Acad. Sci. 1316 (1), 87–104.
Fitch, W.T., 2014. Toward a computational framework for cognitive biology:
unifying approaches from cognitive neuroscience and comparative cognition.
Phys. Life Rev. 11 (3), 329–364.
Flöel, A., de Vries, M.H., Scholz, J., Breitenstein, C., Johansen-Berg, H., 2009. White
matter integrity in the vicinity of Broca’s area predicts grammar learning
success. Neuroimage 47 (4), 1974–1981.
Forkstam, C., Hagoort, P., Fernández, G., Ingvar, M., Petersson, K.M., 2006. Neural
correlates of artificial syntactic structure classification. Neuroimage 32 (2),
956–967.
Frank, S.L., Bod, R., Christiansen, M.H., 2012. How hierarchical is language use?
Proc. R. Soc. Lond. B Biol., rspb20121741.
Freedman, D.J., Riesenhuber, M., Poggio, T., Miller, E.K., 2003. A comparison of
primate prefrontal and inferior temporal cortices during visual categorization.
J. Neurosci. 23 (12), 5235–5246.
French, C.A., Jin, X., Campbell, T.G., Gerfen, E., Groszer, M., Fisher, S.E., Costa, R.M.,
2012. An aetiological Foxp2 mutation causes aberrant striatal activity and
alters plasticity during skill learning. Mol. Psychiatry 17 (11), 1077–1085.
Frey, S., Campbell, J.S., Pike, G.B., Petrides, M., 2008. Dissociating the human
language pathways with high angular resolution diffusion fiber tractography. J.
Neurosci. 28 (45), 11435–11444.
Friederici, A.D., Gierhan, S.M.E., 2013. The language network. Curr. Opin. Neurobiol.
23 (2), 250–254.
Friederici, A.D., Singer, W., 2015. Grounding language processing on basic
neurophysiological principles. Trends Cogn. Sci. 19 (6), 329–338.
Friederici, A.D., Opitz, B., von Cramon, D.Y., 2000a. Segregating semantic and
syntactic aspects of processing in the human brain: an fMRI investigation of
different word types. Cereb. Cortex 10 (7), 698–705.
Friederici, A.D., Wang, Y., Herrmann, C.S., Maess, B., Oertel, U., 2000b. Localization
of early syntactic processes in frontal and temporal cortical areas: a
magnetoencephalographic study. Hum. Brain Mapp. 11 (1), 1–11.
Friederici, A.D., Rüschemeyer, S.A., Hahne, A., Fiebach, C.J., 2003. The role of left
inferior frontal and superior temporal cortex in sentence comprehension:
localizing syntactic and semantic processes. Cereb. Cortex 13 (2), 170–177.
Friederici, A.D., Bahlmann, J., Heim, S., Schubotz, R.I., Anwander, A., 2006. The brain
differentiates human and non-human grammars: functional localization and
structural connectivity. Proc. Natl. Acad. Sci. U. S. A. 103 (7), 2458–2463.
Friederici, A.D., Makuuchi, M., Bahlmann, J., 2009. The role of the posterior superior
temporal cortex in sentence comprehension. Neuroreport 20 (6), 563–568.
Friederici, A.D., 2009. Allocating functions to fiber tracts: facing its indirectness.
Trends Cogn. Sci. 13 (9), 370–371.
Friederici, A.D., 2011. The brain basis of language processing: from structure to
function. Physiol. Rev. 91 (4), 1357–1392.
Friederici, A.D., 2012. The cortical language circuit: from auditory perception to
sentence comprehension. Trends Cogn. Sci. 16 (5), 262–268.
Fujita, K., 2014. Recursive merge and human language evolution. In: Roeper, T.,
Speas, M. (Eds.), Recursion: Complexity in Cognition. Springer International
Publishing, Switzerland, pp. 243–264.
Geschwind, D.H., Rakic, P., 2013. Cortical evolution: judge the brain by its cover.
Neuron 80 (3), 633–647.
Gigerenzer, G., Goldstein, D.G., 1996. Reasoning the fast and frugal way: models of
bounded rationality. Psychol. Rev. 103 (4), 650.
Giraud, A.L., Poeppel, D., 2012. Cortical oscillations and speech processing:
emerging computational principles and operations. Nat. Neurosci. 15 (4),
511–517.
Goucha, T.B., Friederici, A.D., 2015. The language skeleton after dissecting meaning:
a functional segregation within Broca’s area. Neuroimage 114 (6), 294–302.
Graves, W.W., Grabowski, T.J., Mehta, S., Gordon, J.K., 2007. A neural signature of
phonological access: distinguishing the effects of word frequency from
223
familiarity and length in overt picture naming. J. Cogn. Neurosci. 19 (4),
617–631.
Griffiths, J.D., Marslen-Wilson, W.D., Stamatakis, E.A., Tyler, L.K., 2013. Functional
organization of the neural language system: dorsal and ventral pathways are
critical for syntax. Cereb. Cortex, bhr386.
Groszer, M., Keays, D.A., Deacon, R.M., De Bono, J.P., Prasad-Mulcare, S., Gaub, S.,
et al., 2008. Impaired synaptic plasticity and motor learning in mice with a
point mutation implicated in human speech deficits. Curr. Biol. 18 (5),
354–362.
Hagoort, P., Indefrey, P., 2014. The neurobiology of language beyond single words.
Ann. Rev. Neurosci. 37 (1), 347–362.
Hagoort, P., 2005. On Broca, brain, and binding: a new framework. Trends Cogn.
Sci. 9 (9), 416–423.
Hanna, J., Mejias, S., Schelstraete, M.A., Pulvermüller, F., Shtyrov, Y., Van der Lely,
H.K., 2014. Early activation of Broca’s area in grammar processing as revealed
by the syntactic mismatch negativity and distributed source analysis. J. Cogn.
Neurosci. 5 (2), 66–76.
Hauser, M.D., Chomsky, N., Fitch, W.T., 2002. The faculty of language: what is it,
who has it, and how did it evolve? Science 298 (5598), 1569–1579.
Herrmann, B., Maess, B., Hasting, A.S., Friederici, A.D., 2009. Localization of the
syntactic mismatch negativity in the temporal cortex: an MEG study.
Neuroimage 48 (3), 590–600.
Herrmann, B., Obleser, J., Kalberlah, C., Haynes, J.D., Friederici, A.D., 2012.
Dissociable neural imprints of perception and grammar in auditory functional
imaging. Hum. Brain Mapp. 33 (3), 584–595.
Heyes, C., 2012. Simple minds: a qualified defence of associative learning. Philos.
Trans. R. Soc. B 367 (1603), 2695–2703.
Hornstein, N., 2009. A Theory of Syntax: Minimal Operations and Universal
Grammar. University Press Cambridge, Cambridge.
Howard, M.W., MacDonald, C.J., Tiganj, Z., Shankar, K.H., Du, Q., Hasselmo, M.E.,
Eichenbaum, H., 2014. A unified mathematical framework for coding time,
space, and sequences in the hippocampal region. J. Neurosci. 34 (13),
4692–4707.
Jackendoff, R., Wittenberg, E., 2014. In: Newmeyer, Fritz, Preston, Lauren (Eds.),
What You can Say Without Syntax: a Hierarchy of Grammatical Complexity.
Measuring Grammatical Complexity. Oxford University Press, Oxford.
Jeon, H.-A., Friederici, A.D., 2013. Two principles of organization in the prefrontal
cortex are cognitive hierarchy and degree of automaticity. Nat. Commun. 4,
2041.
Jeon, H.A., Friederici, A.D., 2015. Degree of automaticity and the prefrontal cortex.
Trends Cogn. Sci. 19 (5), 244–250.
Kayne, Richard S., 1994. The Antisymmetry of Syntax. Linguistic Inquiry
Monograph Twenty-Five. MIT Press, Cambridge, MA.
Koechlin, E., Basso, G., Pietrini, P., Panzer, S., Grafman, J., 1999. The role of the
anterior prefrontal cortex in human cognition. Nature 399 (6732), 148–151.
Langston, R.F., Wood, E.R., 2010. Associative recognition and the hippocampus:
differential effects of hippocampal lesions on object-place, object-context and
object-place-context memory. Hippocampus 20 (10), 1139–1153.
Lau, E.F., Phillips, C., Poeppel, D., 2008. A cortical network for semantics:
(de)constructing the N400. Nat. Rev. Neurosci. 9, 920–933.
Makris, N., Kennedy, D.N., McInerney, S., Sorensen, A.G., Wang, R., Caviness, V.S.,
Pandya, D.N., 2005. Segmentation of subcomponents within the superior
longitudinal fascicle in humans: a quantitative, in vivo, DT-MRI study. Cereb.
Cortex 15 (6), 854–869.
Makuuchi, M., Bahlmann, J., Anwander, A., Friederici, A.D., 2009. Segregating the
core computational faculty of human language from working memory. Proc.
Natl. Acad. Sci. U. S. A. 106 (20), 8362–8367.
Martins, M.J., Fischmeister, F.P., Puig-Waldmueller, E., Oh, J., Geißler, A., Robinson,
S., et al., 2014. Fractal image perception provides novel insights into
hierarchical cognition. Neuroimage 96, 300–308.
Martins, M.D.J.D., Muršič, Z., Oh, J., Fitch, W.T., 2015. Representing visual recursion
does not require verbal or motor resources. Cogn. Psychol. 77, 20–41.
McKenzie, S., Frank, A.J., Kinsky, N.R., Porter, B., Rivière, P.D., Eichenbaum, H., 2014.
Hippocampal representation of related and opposing memories develop within
distinct, hierarchically organized neural schemas. Neuron 83 (1), 202–215.
Mesulam, M.M., Thompson, C.K., Weintraub, S., Rogalski, E.J., 2015. The Wernicke
conundrum and the anatomy of language comprehension in primary
progressive aphasia. Brain 138 (8), 2423–2437.
Meyer, L., Obleser, J., Anwander, A., Friederici, A.D., 2012. Linking ordering in
Broca’s area to storage in left temporo-parietal regions: the case of sentence
processing. NeuroImage 62 (3), 1987–1998.
Moro, A., 2000. Dynamic antisymmetry (No. 38). MIT press.
Moro, A., 2014. On the similarity between syntax and actions. Trends Cogn. Sci. 18,
109–110.
Murphy, E., 2015a. Labels, cognomes and cyclic computation: an ethological
perspective. Front. Psychol. 6, 715.
Murphy, E., 2015b. The brain dynamics of linguistic computation. Front. Psychol. 6.
Narita, H., 2014. Endocentric Structuring of Projection-free Syntax, Vol. 218. John
Benjamins Publishing Company.
Neubert, F.X., Mars, R.B., Thomas, A.G., Sallet, J., Rushworth, M.F., 2014. Comparison
of human ventral frontal cortex areas for cognitive control and language with
areas in monkey frontal cortex. Neuron 81 (3), 700–713.
Nucifora, P.G., Verma, R., Melhem, E.R., Gur, R.E., Gur, R.C., 2005. Leftward
asymmetry in relative fiber density of the arcuate fasciculus. Neuroreport 16
(8), 791–794.
224 T. Goucha et al. / Neuroscience and Biobehavioral Reviews 81 (2017) 213–224
Opitz, B., Friederici, A.D., 2003. Interactions of the hippocampal system and the
prefrontal cortex in learning language-like rules? Neuroimage 19 (4),
1730–1737.
Opitz, B., Friederici, A.D., 2004. Brain correlates of language learning: the neuronal
dissociation of rule-based versus similarity-based learning. J. Neurosci. 24 (39),
8436–8440.
Opitz, B., Friederici, A.D., 2007. Neural basis of processing sequential and
hierarchical syntactic structures. Hum. Brain Mapp. 28 (7), 585–592.
Overath, T., McDermott, J.H., Zarate, J.M., Poeppel, D., 2015. The cortical analysis of
speech-specific temporal structure revealed by responses to sound quilts. Nat.
Neurosci. 18 (6), 903–911.
Pattamadilok, C., Dehaene, S., Pallier, C., 2015. A role for left inferior frontal and
posterior superior temporal cortex in extracting a syntactic tree from a
sentence. Cortex.
Perani, D., Saccuman, M.C., Scifo, P., Anwander, A., Spada, D., Baldoli, C., Friederici,
A.D., 2011. Neural language networks at birth. Proc. Natl. Acad. Sci. U. S. A. 108
(38), 16056–16061.
Petersson, K.M., Hagoort, P., 2012. The neurobiology of syntax: beyond string sets.
Philos. Trans. R. Soc B 367 (1598), 1971–1983.
Petersson, K.M., Folia, V., Hagoort, P., 2012. What artificial grammar learning
reveals about the neurobiology of syntax. Brain Lang. 120 (2), 83–95.
Petkov, C.I., Kikuchi, Y., Milne, A.E., Mishkin, M., Rauschecker, J.P., Logothetis, N.K.,
2015. Different forms of effective connectivity in primate frontotemporal
pathways. Nat. Commun. 6.
Piai, V., Anderson, K.L., Lin, J.J., Dewar, C., Parvizi, J., Dronkers, N.F., Knight, R.T.,
2016. Direct brain recordings reveal hippocampal rhythm underpinnings of
language processing. Proc. Natl. Acad. Sci. 201603312.
Price, C.J., 2012. A review and synthesis of the first 20 years of PET and fMRI studies
of heard speech, spoken language and reading. Neuroimage 62 (2), 816–847.
Pulvermüller, F., 2013. How neurons make meaning: brain mechanisms for
embodied and abstract-symbolic semantics. Trends Cogn. Sci. 17 (9), 458–470.
Röder, B., Stock, O., Neville, H., Bien, S., Rösler, F., 2002. Brain activation modulated
by the comprehension of normal and pseudo-word sentences of different
processing demands: a functional magnetic resonance imaging study.
Neuroimage 15 (4), 1003–1014.
Reber, A.S., 1967. Implicit learning of artificial grammars. J. Verb. Learn. Verb.
Behav. 6, 855–863.
Reinhart, T.M., 1976. The Syntactic Domain of Anaphora (Doctoral Dissertation).
Massachusetts Institute of Technology.
Rilling, J.K., Glasser, M.F., Preuss, T.M., Ma, X., Zhao, T., Hu, X., Behrens, T.E., 2008.
The evolution of the arcuate fasciculus revealed with comparative DTI. Nat.
Neurosci. 11 (4), 426–428.
Rilling, J.K., Glasser, M.F., Jbabdi, S., Andersson, J., Preuss, T.M., 2011. Continuity,
divergence, and the evolution of brain language pathways. Front. Evol.
Neurosci. 3.
Rilling, J.K., 2014. Comparative primate neuroimaging: insights into human brain
evolution. Trends Cogn. Sci. 18 (1), 46–55.
Rodd, J.M., Vitello, S., Woollams, A.M., Adank, P., 2015. Localising semantic and
syntactic processing in spoken and written language comprehension: an
activation likelihood estimation meta-analysis. Brain Lang. 141, 89–102.
Rohrmeier, M., Fu, Q., Dienes, Z., 2012. Implicit learning of recursive context-free
grammars. PloS One 7 (10), e45885.
Rolheiser, T., Stamatakis, E.A., Tyler, L.K.1, 2011. Dynamic processing in the human
language system: synergy between the arcuate fascicle and extreme capsule. J.
Neurosci. 31 (47), 16949–16957.
Saffran, J., Hauser, M., Seibel, R., Kapfhamer, J., Tsao, F., Cushman, F., 2008.
Grammatical pattern learning by human infants and cotton-top tamarin
monkeys. Cognition 107 (2), 479–500.
Sakai, T., Mikami, A., Tomonaga, M., Matsui, M., Suzuki, J., Hamada, Y., et al., 2011.
Differential prefrontal white matter development in chimpanzees and
humans. Curr. Biol. 21 (16), 1397–1402.
Sakai, T., Hirata, S., Fuwa, K., Sugama, K., Kusunoki, K., Makishima, H., et al., 2012.
Fetal brain development in chimpanzees versus humans. Curr. Biol. 22 (18),
R791–R792.
Sakai, T., Matsui, M., Mikami, A., Malkova, L., Hamada, Y., Tomonaga, M., et al.,
2013. Developmental patterns of chimpanzee cerebral tissues provide
important clues for understanding the remarkable enlargement of the human
brain. Proc. R. Soc. Lond. B Biol. 280 (1753), 20122398.
Sanides, F., 1962. Entwicklungsprinzipien des menschlichen Stirnhirns.
Naturwissenschaften 49 (7), 160–161.
Santi, A., Grodzinsky, Y., 2010. fMRI adaptation dissociates syntactic complexity
dimensions. Neuroimage 51 (4), 1285–1293.
Saur, D., Kreher, B.W., Schnell, S., Kümmerer, D., Kellmeyer, P., Vry, M.S., et al.,
2008. Ventral and dorsal pathways for language. Proc. Natl. Acad. Sci. U. S. A.
105 (46), 18035–18040.
Scharinger, M., Bendixen, A., Herrmann, B., Henry, M.J., Mildner, T., Obleser, J.,
2015. Predictions interact with missing sensory evidence in semantic
processing areas. Hum. Brain Mapp. 37 (2), 704–716.
Schenker, N.M., Hopkins, W.D., Spocter, M.A., Garrison, A.R., Stimpson, C.D., Erwin,
J.M., et al., 2010. Broca’s area homologue in chimpanzees (Pan troglodytes):
probabilistic mapping, asymmetry, and comparison to humans. Cereb. Cortex
20 (3), 730–742.
Schmahmann, J.D., Pandya, D.N., Wang, R., Dai, G., D’Arceuil, H.E., de Crespigny, A.J.,
Wedeen, V.J., 2007. Association fibre pathways of the brain: parallel
observations from diffusion spectrum imaging and autoradiography. Brain 130
(3), 630–653.
Schreiweis, C., Bornschein, U., Burguière, E., Kerimoglu, C., Schreiter, S.,
Dannemann, M., et al., 2014. Humanized Foxp2 accelerates learning by
enhancing transitions from declarative to procedural performance. Proc. Natl.
Acad. Sci. U. S. A. 111 (39), 14253–14258.
Sciullo, A.M.D., Jenkins, L., 2016. Biolinguistics and the human language faculty.
Language 92 (3), e205–e236.
Skeide, M.A., Brauer, J., Friederici, A.D., 2015. Brain functional and structural
predictors of language performance. Cereb. Cortex, bhv042.
Spocter, M.A., Hopkins, W.D., Garrison, A.R., Bauernfeind, A.L., Stimpson, C.D., Hof,
P.R., Sherwood, C.C., 2010. Wernicke’s area homologue in chimpanzees (Pan
troglodytes) and its relation to the appearance of modern human language.
Proc. R. Soc. Lond. B Biol. Sci. 277.
Suzuki, W.A., 2007. Making new memories. Ann. N.Y. Acad. Sci. 1097 (1), 1–11.
Szalontai, A., Csiszar, K., 2013. Genetic insights into the functional elements of
language. Hum. Genet. 132 (9), 959–986.
Szczepanski, S.M., Knight, R.T., 2014. Insights into human behavior from lesions to
the prefrontal cortex. Neuron 83 (5), 1002–1018.
Takaya, S., Kuperberg, G.R., Liu, H., Greve, D.N., Makris, N., Stufflebeam, S.M., 2015.
Asymmetric projections of the arcuate fasciculus to the temporal cortex
underlie lateralized language function in the human brain. Front. Neuroanat. 9.
Tuennerhoff, J., Noppeney, U., 2016. When sentences live up to your expectations.
Neuroimage 124, 641–653.
Uddén, J., Folia, V., Forkstam, C., Ingvar, M., Fernández, G., Overeem, S., Petersson,
K.M., 2008. The inferior frontal cortex in artificial syntax processing: an rTMS
study. Brain Res. 1224, 69–78.
Uddén, J., Bahlmann, J., 2012. A rostro-caudal gradient of structured sequence
processing in the left inferior frontal gyrus. Philos. Trans. R. Soc. Lond. B Biol.
Sci. 367 (1598), 2023–2032.
Ullman, M.T., 2001. A neurocognitive perspective on language: the
declarative/procedural model. Nat. Rev. Neurosci. 2 (10), 717–726.
Vernooij, M.W., Smits, M., Wielopolski, P.A., Houston, G.C., Krestin, G.P., van der
Lugt, A., 2007. Fiber density asymmetry of the arcuate fasciculus in relation to
functional hemispheric language lateralization in both right-and left-handed
healthy subjects: a combined fMRI and DTI study. Neuroimage 35 (3),
1064–1076.
Wang, L., Uhrig, L., Jarraya, B., Dehaene, S., 2015. Representation of numerical and
sequential patterns in macaque and human brains. Curr. Biol. 25 (15),
1966–1974.
Wilson, S.M., Dronkers, N.F., Ogar, J.M., Jang, J., Growdon, M.E., Agosta, F.,
Gorno-Tempini, M.L., 2010. Neural correlates of syntactic processing in the
nonfluent variant of primary progressive aphasia. J. Neurosci. 30 (50),
16845–16854.
Wilson, S.M., Galantucci, S., Tartaglia, M.C., Rising, K., Patterson, D.K., Henry, M.L.,
et al., 2011. Syntactic processing depends on dorsal language tracts. Neuron 72
(2), 397–403.
Yang, C., 2013. Ontogeny and phylogeny of language. Proc. Natl. Acad. Sci. 110 (16),
6324–6327.
Zaccarella, E., Friederici, A.D., 2015a. Syntax in the brain. In: Toga, A. (Ed.), Brain
Mapping: An Encyclopedic Reference. Elsevier, pp. 461–468.
Zaccarella, E., Friederici, A.D., 2015b. Reflections of word processing in the insular
cortex: a sub-regional parcellation based functional assessment. Brain Lang.
142, 1–7.
Zaccarella, E., Friederici, A.D., 2015c. Merge processing in the human brain: a
sub-region based functional investigation in the left pars opercularis. Front.
Psychol. 6, 1818.
Zaccarella, E., Meyer, L., Makuuchi, M., Friederici, A.D., 2015. Building by syntax:
the neural basis of minimal linguistic structures. Cereb. Cortex, Bhv234.
Zaccarella, E., Friederici, A.D., 2016. 2016. The neurobiological nature of syntactic
hierarchies. Neurosci. Biobehav. Rev.