Neuroconstructivism: a developmental

turn in cognitive neuroscience?

Introduction

Since its birth, brain science has been for the most part the study of the structure and
functioning of a brain already formed, the study of the endpoint of a process. Brodmann areas, for
instance, are cortical areas of the adult brain (Brodmann, 1909). In his authoritative Neurobiology,
Shepherd devotes only one chapter (out of thirty) to developmental neurobiology (Shepherd, 1994).
From early attempts at functional localization by Gall or Broca to recent neurocognitive models like
the model of visual cognition proposed by Milner and Goodale (Milner & Goodale, 2006), functional
decomposition of the brain essentially remained the decomposition of the brain of the adult.
Neuroconstructivism, then, as it has been recently vindicated (Mareschal et alii, 2007; Sirois et alii,
2008) could be understood, first, as the idea that we should take brain development more seriously.
This suggestion comes at a time when in many fields of biology, ontogenetic development has become
the object of both fascinating discoveries and intense speculation. But there is more to
neuroconstructivism than a developmental perspective on the brain, as it can be understood as a view
of cognition: it is this view of cognition that motivates a specific, renewed approach to the human
brain. What neuroconstructivism is challenging, in fact, is a view of cognitive explanation, and of
cognitive development.

The view of cognitive explanation it rejects is in part derived from the idea that Marr expressed when
he said that we should not begin by the study of feathers if we want to understand bird flight (Marr,
1982). It is the idea of independent levels of investigation, the idea that psychological explanation is,
in principle, fruitfully divorced from the study of low-level implementation. Neuroconstructivism
suggests a view of cognitive explanation where there is no point to the separation of levels. There is no
such separation in practice, because neural events are causally relevant to the understanding of
cognitive development in its many forms, both typical and atypical. And there is no such separation in
another sense, because investigation at different levels (levels being this time: the cell, the brain, the
whole body) may rely on the same kind of explanatory factors. As mechanisms of the same kind
operate at different levels, explanations in different fields are fundamentally of the same type. The
unity of cognitive research, then, is not obtained via the existence of a discipline that would shape the
whole field (e. g., evolutionary psychology, seen as a guide to neuroscientific research) but through
the use of recurring patterns of explanation. And as neuroconstructivism adopts as its highest core
principle the principle of context-dependence1, brain is not the largest unit, the largest containing
system that it considers. Even the brain in development and its changing abilities have to be
contextualized. This is why neuroconstructivism perceives itself as the convergence of the work in
different disciplines and trends of research, namely: developmental neurobiology (the study of
encellment see Shepherd, 1994, chapter 9); developmental cognitive neuroscience (the study of
enbrainment see Johnson, 2005a); developmental embodied cognition (the study of embodiment
see Thelen and Smith, 1994).

1
Mareschal et alii, 2007, chapter 5; Sirois et alii, 2008, p. 325, and figure 1 below, section III.

1
 The view of cognitive development it rejects is the view that cognitive abilities are highly canalized
biological features2, i.e., features that develop in a similar fashion in widely different environments,
and are essentially insensitive to environmental variation. In contrast, neuroconstructivism also
suggests a view of cognitive development that is highly sensitive to events in both internal and
external environments. It rejects a deterministic view of epigenesis and rejects a view of cognition
where inborn abilities would simply unfold in time. Brain development matters, in this sense, not just
because there is no preexisting, detailed blueprint of its organization - because, for instance, the
protomap view of the cortex (Rakic, 1988), where the identity of any cortical neuron can be traced
back to the specific spatio-temporal circumstances of its formation in the proliferative zone, has now
been abandoned (Sur and Rubenstein, 2005). Brain development matters because we may learn from
it, not only about the adult brain, but about how we acquire our own mental powers, and what
constrains a childs typical or atypical cognitive development.

I do not intend to comment here on every single aspect of the neuroconstructivist program, such as the
importance it gives to computational modeling, or its endorsement of the embodied cognition view.
My first goal is to provide an understanding of the reasons why neuroscience may need such a
program which gaps it is supposed to fill, which inferences it challenges. I shall therefore focus on
early critiques of inference from functional commitment specialization of cortical areas to their inborn
specialization (II), and on the link between the controversy over plasticity that arose in the 1990s and
the current neuroconstructivist model of brain development (III). My second goal is to examine the
core of the program itself, its guiding principles that is, the principle of context-dependence and
the idea of level-independent mechanisms (IV). I suggest that such a program, as it stands, suffers
from both under-determination and over-generalization (V). Because of these defects, and because
developmental neuroscience has not been the main concern of philosophers who analyse neuroscience
in terms of mechanisms, a dialogue between neuroconstructivism and the philosophy of neuroscience
may yield mutual benefits (VI).

II

The broader context: the Wundt-Munk controversy about the origins of brain function

The anti-nativist stance of neuroconstructivism can be situated in its opposition to recent Chomskyan,
Fodorian, or massive modularity views about neurocognitive explanation. But it has roots in more
ancient debates that are worth mentioning. One reason is that there are good reasons to prefer a
continuist view about the history of cognitive neuroscience over a discontinuist one. Even if the
expression cognitive neuroscience itself has not been widely used until recently, what this
expression refers to may be seen a research program that finds its origins in the 19 th century, and
whose development can be understood as the application to the brain of these heuristic strategies that
Bechtel and Richardson have called decomposition and localization (Bechtel and Richardson, 1993).
The central idea in this program is that the discovery of brain mechanisms is key to the understanding
of the corresponding mental powers. Notable differences in investigation methods, scientific tools and
intellectual background notwithstanding, privileging continuity should not be controversial, as many
topics that belong today to cognitive science were treated in the 1880s by scientists like Meynert,
Munk, Ferrier, and Jackson. One of the first philosophers to have taken the measure of such a program
was Wundt in his Elements of Psychophysiology (Wundt, 1880). In this monumental work as in related
papers, his aim was to make explicit the founding principles of such a science, to sum up its main
discoveries, and to offer critical views about some of its central claims. The critical views were meant

2
Ariew, 1996.

2
as a positive contribution to a young domain of investigation. Researchers like Munk, according to
Wundt, offer only a mixture of important scientific discoveries and old prejudices. Making these
prejudices explicit is a means to an end: clearing the ground for further questions and investigations.
In order to achieve this, Wundt sketches what we may call a constructivist answer to nativist views.

Wundt does not reject the principle of localization of function within the brain, a principle that is
central to the mechanistic methodology of brain science. But he thinks that the kinds of methods used
by neuroscientists may lead them to adopt views that are not, in fact, supported by available evidence.
First, the localization of lesions should not be conflated with the localization of functions. Focal
lesions in pathological cases are signs of the involvement of brain regions in cognitive tasks, but they
do not, by themselves, indicate what the exact nature of this involvement may be. Second, and more
importantly, evidence of localized functions is not evidence of inborn commitment of brain areas to
perform definite functions. To the old holistic error (no functional decomposition of the brain is
possible, or fruitful) we should not substitute what Wundt calls the phrenological error (Wundt, 1891):
the claim that each part of the brain has its own, immutable, pre-specified function. In order to reject
phrenology, according to Wundt, it is not enough to give up the entire list of mental faculties as they
were defined by Gall. Quite strikingly, Wundt sees (much before Fodor, but for opposite reasons) that
there is more to Gall than a set of old prejudices that everyone has already overcome. There is a
mixture of nativism and what we would call modularity (that is, functional specificity of largely
autonomous units or powers, units that may be localized in the brain) and this mixture, in 1880,
according to Wundt, still provides the framework of much scientific work done in brain science. This
point: no sound inference from functional specialization in the adult brain to inborn commitment of
the corresponding brain parts) is central to Wundts proposal. The task of physiological psychology
(roughly, our cognitive neuroscience), as it is defined by Wundt, is not only to establish what
functional localizations are, but more importantly, where they come from. Nativist views, according to
Wundt, are just the product of our ignorance in this matter. The first part of Wundts work, then, is to
explain why the nativist view of inborn commitment of brain areas should be rejected. To this end,
Wundt relies on recent work on neural plasticity. Brain function, when abolished by the destruction of
some quantity of brain tissue, may be spontaneously restored during recovery. This is explained,
according to Wundt, by other neural elements taking over the defective ones. This shows that with
different connections, different input/ output conditions, functional specificity of brain components
may be altered. The possibility of a substitution of one nervous element to another is evidence against
strict preexisting definition of specific neural function. This principle of Ersatz function itself falls
under another, more general principle that Wundt calls the principle of adaptation: any central
element is adapted to its function as it has to perform it more frequently under the pressure of external
conditions3. Plasticity, then, when it is linked to physiological recovery after brain damage, is not a
sui generis phenomenon: it is just an extreme case of adaptation as defined in this broad sense4: in this
case, being progressively adapted does not mean doing something better with time (improvement of
performance, as in a Jamesian definition of plasticity5) but being able to alter ones pre-existing
function under external (contextual) influences. But Wundt does not stop there. He asks: why is this
principle of adaptation valid? Where does the power of neural elements to adjust to external
circumstances come from? And his answer is: it comes from the lack of specific functional
commitment of central, neural elements at birth. If the definition of function is the result of the process
of brain organization, then it is not that remarkable that neural elements may change their functional
role in special circumstances during adult life. It is just that no restriction of fate during life, no

3
Wundt, 1880, Part I, chapter V, 7: General principles of central functions.
4
A strikingly similar view is expressed in Elman and alii, 1996, p. 247-248.
5
James, 1890.

3
special adaptation leads to a unique, non-reversible functional commitment. Evidence of plasticity
brings back to the necessity of a scientific explanation of the outcome of epigenetic processes.
Nativism, then, is not a solution to the problem of the origin of functional specialization within the
brain, it is just a way to ignore it6. Plasticity depends on the adaptive power of neural elements, which
itself derives from the fact that their standard, domain-specific activity during adult life should be seen
as the temporary, reversible product of the changes imposed by brain organization (for instance, input
and output conditions) on initial, non-specific response properties of neural elements. The fact that
Wundt has little to say about the detail of these changes makes his clear, synthetic view of this cluster
of important questions even more remarkable.

III

The debate on the significance of neural plasticity for the controversy about representational nativism.

Just as Wundt once challenged a phrenological view of inborn functional determination,

constructivists have challenged a deterministic view of brain development one century later. The
intellectual debt of the co-authors of Neuroconstructivism towards Jeffrey Elman, Annette Karmiloff-
Smith and the other co-authors of the manifesto, Rethinking innateness7 is obvious enough. What is
needed in order to understand neuroconstructivism is rather to clarify the meaning of the shift of
emphasis from the analysis of neural plasticity (considered in 1996 as evidence against
representational nativism) to developmental cognitive neuroscience as it flourished in the following
years. This can be done through the debate between the philosopher Richard Samuels and the co-
authors of Rethinking innateness8. Samuels holds that Elman and his colleagues have misinterpreted
the neurobiological facts: neural plasticity does not, contrary to their claims, falsify representational
nativism. OLeary and Sur, in a striking series of experiments, offered evidence that in the brain of
mammals, the sensory cortex of a given region may take response properties of another region under
exceptional circumstances. Studies have shown that fetal neurons taken from the visual area may
come to exhibit the organizational and functional properties of the neurons of the somatosensory
region where they have been transplanted (O Leary and Stanfield, 1989). Rewiring experiments have
also shown that the somatosensory cortex of ferrets, when it is deprived of its normal input, and when
it receives visual inputs early in life, may develop response properties that are typical of the visual
cortex (Sur, Pallas & Roe, 1990). Are we to think, then, that parietal regions that are usually
described as somatosensory areas are not intrinsically dedicated to the representation of somatic
states? According to Samuels, neural plasticity has no such implications. First, innate properties are
not necessarily intrinsic properties (that is, properties that are non-relational). If representational
properties are extrinsic, rather than intrinsic properties, representational properties of a given set of
neurons may both be innately specified and depend on the relation of these neurons to the rest of the
cortex and to the environment. In this case, when we are modifying the input conditions of these
neurons, it is no surprise that their representational properties are altered. Samuels thinks that
constructivists suggest mistakenly that nativists have to adopt an Invariance principle such as

6
For instance, Wundt sees very clearly the implications of a constructivist view for a question like the specific
energies of nerves inherited from Mller, a problem that had become in his days the problem of the relation
between sensory qualia and brain regional activation. For him, there is no individualistic, internalist explanation
of the differences between sensory modalities. The supervenience base of visual experience cannot be the
activation of the visual region of the brain alone, independently from the nature of sensory input.
7
Elman & alii, 1996.
8
Samuels, 1998.

4
The innately specified (representational) properties of a piece of cortical tissue T are invariant under
alterations in Ts location within the brain and alterations in the afferent inputs to T.

But Samuels thinks that because of the distinction between intrinsic and innate properties, nativists do
not have to accept this Principle. Moreover, they are, in fact, committed to what he calls Organism
nativism rather than Tissue Nativism. Tissue nativism is a claim about representational properties of
specific brain parts. Organism nativism is a claim about inborn cognitive abilities of whole organisms
or people and it is entirely independent from claims about innate commitment of brain parts or specific
localization of function. Even a refutation of tissue nativism (a refutation that, according to Samuels,
experiments on plasticity do not provide) would not be a refutation of representational nativism in
general.

Lets focus on tissue nativism, as it is linked to the definition of the functions of brain parts, and as
Samuels holds that it can be vindicated against Elmans views. First, innate specification of
representational properties is difficult to reconcile with some existing neurophysiological data. The
crucial role of the activity of visual areas in blind subjects during Braille reading seems to support a
view of sensory areas (the meta-modal organization of the brain) where they possess a purpose
general ability to treat incoming sensory signals rather than an inborn commitment to treat one or
another kind of such signals9. In the case of Braille reading by blind subjects, response properties of
the so-called visual area are different from what they are in standard cases, without any transplant or
re-wiring. In congenitally deaf mice, it has been shown that some neurons of the auditory cortex
develop responses to visual and somatosensory stimulation, and that the response properties of other
regions, like the visual cortex, are themselves altered (Hunt, Yamoah and Krubitzer, 2006). What
evidence do we have, then, of an innate specification of the representational properties of visual or
auditory cortex, and why would we prefer this nativist view to the parsimonious alternative of a lack
of the inborn functional commitment of sensory areas? Second, we can ask ourselves what it means
for the representational property of a given set of neurons to be innately specified. Extrinsic
properties, says Samuels, are constitutive of representational properties. But extrinsic properties are
defined during epigenesis and as the result of the specification of neural paths. In this case, how could
representational properties be innately specified before the outcome of this epigenetic process or
independently of it? How could they have a pre-existing, definite content? And if they dont, how
could they exist at all? What is called the innate specification of the representational property of
visual neurons seems to be a convoluted way of saying that neurons of the striate cortex have early in
life a higher probability to receive visual inputs than, say, somatosensory inputs. Calling this an
innate representational property conflates different levels: it re-describes a frequent but non-
necessary correlate of the outcome of a neurobiological process as what is (at the representational or
psychological level) meant to be. However, defeating a nativist critique of conclusions drawn from
plasticity experiments is one thing; explaining where the typical organization of the brain of mammals
comes from (especially, the existence of discrete units like cortical areas, and their functional
specialization) is another. Neuroconstructivism could be understood as the empiricist answer to the
objections made by Samuels to the argument from plasticity against innate specification of
representational properties of given brain parts. What is needed is not only manipulations that
experimentally alter the extrinsic properties of cortical areas, but an understanding of how functional

9
Pascual-Leone and Hamilton, 2001. The meta-modal organization hypothesis takes a middle ground in the
debate on domain-specific or domain-general abilities of neuro-cognitive systems. It does not involve any claim
of equipotentiality of brain regions, but rejects an inference from specialization of a sensory region to a specific
domain to an inborn commitment to this specific domain.

5
specialization and extrinsic properties are specified during brains development. This is what
neuroconstructivism hopes to provide.

IV

Neuroconstructivism at work

One way of presenting the neuroconstructivist view may be to begin with the contrast between two
theories of biological functions in the philosophical literature: one is the etiological view functions
are nothing but effects selected during biological evolution (Neander, 1991), the other is the systemic
view of Cummins: the function of a component x in a system S is its contribution to the explanation
of the ability of S to (Cummins, 1975). Although it is fairly uncontroversial that in brain science,
research is aimed at discovering Cummins functions, and that neuroscience textbooks provide
information about the contribution of activities and/or components to larger systems, some still
maintain that where there is no history, there is no function (Jacob and Jeannerod, 2003), and that we
should consider the functions of brain components as products of evolution by natural selection. I
dont want to discuss here the merits of these philosophical views, but rather to point out that we have
to take in consideration the finesse of grain of our analysis. For instance, the ability of place cells in
the hippocampus to contribute to the formation of maps of the environment (O Keefe and Nadel,
1978) can be seen as (one of ) their function (s), and what we mean by that may be that it is a product
of a certain evolutionary history where natural selection has played a role. However, the ability of
these cells to contribute to the individuals knowledge of his environment requires more than the
existence of these cells and the corresponding history of the species; it requires a certain kind of
individual history where tokens of place cells end up coding for specific places. An explanation of
actual orientation of a given individual will require a causal analysis where specific activations of
place cells contribute to the ability of the hippocampus to form a map of his familiar environment: no
actual orientation is provided by past selection for places cells as a type. This means that different
interesting stories may be told about the origin of functions and that one of them may be about their
ontogeny. In this sense, developmental cognitive neuroscience and neuroconstructivism fill a gap.
They require that we address additional questions that are left unanswered by the evolutionary
perspective, and that Cummins style functional analysis is not meant to solve either:

Question 1: in a given system S, in virtue of what does a component x receive its own power to ?

Question 2: how did S become able to produce its own, characteristic output? Why (rather than
some other activity) in S ?

Developmental Cognitive Neuroscience has to provide answers to these related questions, to give a
mechanistic explanation of how components receive the distinctive powers they have. And the idea of
neuroconstructivists is that, if we cannot be satisfied with a view where brain development is just brain
maturation, we should adopt what Johnson calls the Interactive specialization view. To know why
component C has received its distinctive role (question 1), we have to look at the developmental
history of the System in which C is embedded and at the corresponding external environment. This
developmental history causally explains, in particular, the current pattern of connectivity of C and its
response properties. To the second question: how does S become able to ?, the neuroconstructivist
answer is: we can explain the emergence of Ss characteristic output through its own interactive
specialization and the interactive specialization of its parts. Now to explain of how specialization
occurs, constructivism offers a set of domain-general and level-independent mechanisms:

6
Competition: for any given function, neural systems evolve from widespread, aspecific activity to
specialized correlates

Cooperation : which is another word for functional integration

Chronotopy: key aspects of development rely on sequences of events that are closely related: in
particular early specialization of a component A constrains the posterior developmental trajectory of a
related component B; explaining the latter is impossible without explicit reference to the former.

Figure 1. The core principle of the Neuroconstructivist Program

(from Sirois S. & alii, 2008)

Lets take, for instance, face recognition as the explanandum. Instead of considering the fusiform
area as a cognitive module with an inborn, domain-specific commitment to process face
representations, the neuroconstructivist framework invites us, on the cognitive level, to pay attention
to the difference between early sensitivity to face-like visual patterns and later development of face
recognition itself, a development that may take advantage of an early emerging ability for aspecific
(domain-general) visual expertise. It invites us to consider the difference between a sub-cortical route
responsible for face detection and a cortical network of which the fusiform gyrus is a part, involved in
face identification (Johnson, 2005b): in this case, the specialization for faces of the fusiform area is the
product of its interactions with the sub-cortical route and the constant exposure to faces in the social
environment. Competition, cooperation of brain components and chronotopy are jointly responsible

7
for the emergence of face recognition during development as a deeply entrenched cognitive feature.
According to its proponents, this view is able to account for several phenomena: a) widespread brain
activation in response to faces in young subjects (when specialization through competition has not yet
occurred) that contrasts with specific local activation in elder subjects (Scherf and alii, 2007); b) the
recruitment of the fusiform gyrus in tasks of visual recognition of non-face stimuli by experts in a
given domain (Gauthier, Skudlarski, Gore, and Anderson, 2000); c) atypical developmental
trajectories where defects of the sub-cortical processing of faces have cascading effects on other
cognitive abilities and later phases of development (Johnson, 2005b). Neuroconstructivist explanations
of this kind may be considered as a sub-type of what has been called by philosophers of science like
Carl Hempel and Ernst Nagel genetic explanations. According to Hempel, a genetic explanation
presents the phenomenon under study as the final stage of a developmental sequence, and accordingly
accounts for the phenomenon by describing the successive stages of that sequence (Hempel, 1965, p.
447). This is still an appropriate description of the neuroconstructivist proposal, even if
neuroconstructivist explanations do not fit Hempels covering-law model of scientific explanation.
Mechanisms like cooperation, competition and chronotopy are proposed to identify causally relevant
factors (Craver, 2007) through their abstract, generic description, not merely regular sequences of
events.

A view of brain development that is context-dependent in this sense may not only be accurate, but
prove crucial for neuroethical issues. Some studies have shown a marked disadvantage for children of
low economic status in tasks involving the prefrontal executive system, the left peri Sylvian language
system, and the medial temporal memory system (Farah, Noble and Hurt, 2006). Potential causes
range from prenatal substance exposure to nutritional factors (resulting in iron deficiency anemia),
effects of environmental stress (the release of hormones that have a negative impact on hippocampal
development, for instance) and lack of cognitive stimulation. Maturational views of neurocognitive
development that suggest only a triggering role for environmental factors may seriously underestimate
the impact of these factors not only on neurobiological, but also on cognitive development.
Accordingly, neuroconstructivism may be important at two levels. First, from a theoretical point of
view, it sides with an interactive view of individuation, where abilities supervene on the interactions of
individuals and their environment during development. Neuroconstructivism may stimulate research
on the nature and extent of such interactions. On a practical level, developmental neurobiology may
give precious information on how, in matters of public health, we may become able to implement our
norms of justice and fairness when it comes to child development.

Universal context-dependence? A critique

There is a close link, in the neuroconstructivist framework, between an empirical claim about
neurobiological development and its central theoretical claim. The empirical claim concerns the
relevant factors of an explanation of cortical development; it says that we should downplay the
importance of genetic factors in such an explanation. In favor of that claim, neuroconstructivists offer
two main reasons. The first is the role during development of epigenetic factors, in particular, activity-
dependent mechanisms and adaptations like the ones that are crucial to the definition of ocular
dominance columns (Mareschal and al., 2007, p. 21). The second is that true instances of region-
specific gene expression in the cortex are not common: one notable case is the H-2Z1 transgene that is

8
expressed in only one region in mice, the layer IV of its somatosensory cortex (OLeary and
Nakagawa, 2002, p. 22). Accordingly, neuroconstructivists hold that although genes are involved in an
early definition of broad regional differences, later specification of well-defined areas is mostly the
product of activity-dependent processes (Mareschal and alii, 2007, p. 22). This empirical claim is
offered as evidence by neuroconstructivists for the validity and heuristic value of the core principles of
their model: universal context-dependence, interactive specialization of brain components governed by
competition and cooperation.

Context-dependence is often understood in terms of dependence on interactions with the

environment, internal but also external. One risk here is overgeneralization: from an evolutionary
perspective, while it is obvious that both the level and the type of related activity may be crucial when
it comes to the size or the function of a given area (Hunt, Yamoah and Krubitzer, 2006), it is hardly
obvious that the global cortical architecture can be understood in terms of context-dependence.
Comparison between species shows that the global organization of the mammalian cortex (its
Bauplan) is fairly conservative, differences in behavior notwithstanding (Krubitzer, 2007). For
instance, sensory cortical fields are not context-dependent to the point that blind species like mole rats
would be entirely deprived of visual structures: the architectural pattern remains strictly constrained
and, to a large extent, context-independent. To take another example, recent work devoted to ocular
dominance columns in primary visual cortex (Crowley and Katz, 2002) suggests that their emergence
predates the first months of life, and as a consequence, cannot be the result of activity-dependent
competition between thalamic inputs resulting from retinal stimulation. Although the formation of
such columns may not be entirely independent from activity, as recent work on the role of retinal
waves has shown (Torborg and Feller, 2005), what is crucial in this case is endogenous activity
triggered by internal factors rather than actual visual experience.

Concerning epigenetic factors in general and activity-dependent change, some similar lessons could
be drawn from the phenomenon of axon guidance. Constructivism since the days of Elman and his co-
workers10 have argued that we have to make a distinction between additive and substractive events,
initial proliferation of synaptic connections during development and a later phase of pruning that
corresponds to the degeneration of non-functional paths. This is conform to what is predicted by the
epigenetic, specialization through competition model. But neural pathways do not develop in a
purely anarchic manner before a negative phase of selective apoptosis and degeneration driven by
competition. In particular, projection from thalamic regions to the isocortex may largely depend on
patterns of regional expression of molecules that function as guidance cues (both positive and
negative) for neurite outgrowth. For instance, expression of ligand Ephrin-2A5 in the somatosensory
cortex inhibits projection from limbic thalamic afferents (Gao and alii, 1998). As expression of these
molecules happens early in development, and precedes the invasion of the cortex by thalamic axons, it
may be viewed as a context-independent factor of regional differentiation. This involves something
that is very different from the predictions of the proliferation-and-pruning-model. The risk, then, is to
neglect explanatory factors that do not fit the model, such as the ones suggested by the pioneering
work of Sperry (Sperry, 1943) and his idea of chemo-affinity as a factor of organization during
development. To sum up, from genuine instances of activity-dependence and thalamic influence we
cannot conclude to their explanatory relevance in any given context. And although neuro-
constructivists are right to distinguish between activity-dependence and (external) context-dependence
(Mareschal et alii, 2007, p. 32), not only does spontaneous endogenous activity have a role where
sensory experience has none for the establishment of visual circuity, but it seems that we also have to
take into account activity independent factors (Huberman, Feller and Chapman, 2008).

10
Elman, 1996, p. 245.

9
 Concerning the second empirical claim, the one concerned directly with genes, even if it remains
true that there is no one-to-one correspondence between genes and cortical areas, and even if
knowledge in these matters is still fragmentary, and based mainly on studies that focus on a single
species (mice), not only has evidence of genetic control of arealization been growing in the last twelve
years, but this may be considered as the main recent event in the field of the neurobiology of cortical
development (O Leary and Sahara, 2008). For instance, gene Emx2 is normally expressed in low
rostral to high caudal and low lateral to high medial gradients, and cadherin Cad 8 is a special attribute
of motor cortex situated in the rostrally located motor cotex. In Emx2 homozygous mutant mice,
however, it has been discovered that the pattern of cadherin expression is markedly altered, expression
of Cad8 being expanded both caudally and medially while caudal areas contract. Moreover, in Emx2
mutant mice, connections between cortical areas and thalamic nuclei are significantly altered, arguably
because of the involvement of Emx2 in the differential production of molecules controlling axon
guidance: while in wild-type mice, the anterior occipital cortex receives projections from the dorsal
lateral geniculate nucleus, which conveys visual inputs, in mutant mice, the same region receives
projections from the ventroposterior nucleus, which are normally characteristic of the somatosensory
cortex, a clear sign of the contraction of the visual area (Bishop, Goudreau and O'Leary, 2000).
Accordingly, as molecular expression and patterns of connectivity are two of the most important
attributes of cortical areas, we may conclude that genetic control, in some species at least, goes much
further than the rough preliminary definition of whole regions whose internal architecture would be
fine-tuned under the influence of epigenetic factors. Moreover, the choice is not between a one to one
correspondence between genes and cortical areas and no genetic mediation of cortical development
whatsoever: genes like Emx2, Pax6 and COUP-TFI are expressed in the cortex according to gradients
that may overlap in such a manner that taken together, they play a crucial role in the definition of the
emergence of the combination of features that is unique for each area (Kingsbury and Finlay, 2001).

However suspicious we may be, then, of the metaphor of traits being directly coded or represented in
sequences of DNA basis, we do not have to conclude from the fallacy of outdated genetic determinism
to the validity of rival constructivist proposals when it comes to the explanation of cortical
development. Explanations of arealization now begin with patterning centers contained in the dorsal
telencephalon of the developing brain (O Leary and Sahara, 2008). These patterning centers secrete a
series of molecules (like the fibroblast growth factor FGF8) which are in turn, responsible for the
differential expression of genes like Emx2, Pax6 and COUP-TFI in progenitor cells and their progeny
in cortical regions. This proposal deserves several comments. First, it could be said with reason that
expression of genes in this model is context-dependent; but this kind of dependence has to be
understood in the perspective of a regulatory hierarchy (O Leary and Sahara, 2008) that secures the
emergence of a quite uniform and highly adaptive cortical structure. It is not horizontal interaction
between equals (genes, cells), but hierarchical control that seems to matter the most. Second, even if
constructivists are right to insist on the importance of the timing of events during development
(chronotopy) it seems difficult to understand what this timing depends on without reference to the
above-mentioned regulatory hierarchy, and for instance, to early secretion of signaling molecules on
which gene expression is dependent. Third, in agreement with an influential view of mechanistic
explanation (Machamer, Darden and Craver, 2000), explanation of cortical development is not
achieved by pointing exclusively to low-level, bottoming-out components of cortical structures, but
through the careful description of the integration of entities and activities located at different levels of
the mechanism. We have to adapt this model in a developmental context: mechanism does not refer
here to a static set of components, but to a self-modifying structure where interactions are responsible
for the addition of new features and operations. But the lesson remains: explanation of phenomena at
higher -levels of mechanism is neither reducible to bottoming-out entities and activities, nor divorced

10
from them. To sum up, even if the shift from gradients of gene expression to the abrupt contrast
between discrete cortical areas is not fully understood, it seems that there is no mechanistic
explanation of cortical development without a reference to this regulatory hierarchy of which gene
expression is a key part, a proposition that is not easy to reconcile with the spirit of the
neuroconstructivist program which favors horizontal interactions and epigenetic factors.

The neuroconstructivist program aims to make explicit lessons from past studies that should help
us to identify the relevant questions, factors and variables that will lead us to a deeper understanding
of development (2007, p. 91). It does not seem that its principles are to be understood as statements
of universal laws of nature; it is nowhere said, for instance, that comparative developmental
neurobiology would support claims of necessity or universality in these matters. What is offered by
neuroconstructivists seems rather to be heuristic principles that may function as guidelines for future
research. However, neuroconstructivist principles may suffer, first, from overgeneralization : it may be
that what is valid and heuristically useful for late stages of development and mid-level organization
may not have the same value for earlier phases and/or lower levels of organization (see Kingsbury and
Finlay, 2001, and their distinction between early cortical regionalization and late cortical
regionalization). Second, these same principles may suffer from under-specification, as was already
noticed, not only because specialization may be understood in more ways than one (Anderson, 2008)
but because of the wide differences between types of context-sensitivity. Unless we define
unambiguously a) which degrees of change are significant enough to be counted as evidence of
sensitivity (that is, as a mark of dependence), b) what is evidence of causal dependence to context, and
above all c) what is exactly the context (with its specific boundaries and properties) a given event is
supposed to depend on in a given case; it will be very difficult to establish what exactly confirms or
disconfirms the principle of context-dependence. Moreover, the kind of interactions between
components we have to understand is not always the one that is predicted by neuroconstructivism:
often, context-dependence involves hierarchical control within a multi-level developmental
mechanism, rather than cross-talk among equals at a given level.

VI Conclusion

Very often, constructivism is perceived and debated as an alternative to nativism, as if arguments and
empirical predictions could lead to a final settlement of the dispute. It is reasonable to think, however,
that scientific investigation itself is inherently pluralistic, that developmental neuroscience, as
understood by neuroconstructivists, evolutionary neuroscience, and systemic neuroscience, ask, to
use van Fraasens terminology, different why-questions, the topic of each of them being associated
with a different contrast class (van Fraassen, 1980). One of the main interests of the developmental
perspective is that atypical development is not necessarily synonymous with dysfunction, impairment
and cognitive failure, as was shown by recent work on high-level autism (Happ, 1999). But the
classical framework of neuropsychology, with its contrast between brain or cognitive integrity and
deficits associated with lesions, if not relevant to the field of developmental syndromes, is still valid in
its proper context. Neuroconstructivism may (and does) inspire quite promising research (Rippon,
Brock, Brown, and Boucher, 2007) with valuable theoretical and social implications, but different
perspectives are still needed. Pluralism, as it is advocated here, does not preclude cross talk between
existing disciplines, and the birth of new integrating disciplines at their borders (evolutionary
developmental neuroscience would be an example). But it precludes seeing a proposal like
neuroconstructivism and the new emphasis on development in terms of developmental turn or
paradigm shift.

11
Philosophers interested in neuroconstructivism may face the following alternative. One possibility is
to use the neuroconstructivist framework to build a broad view of development and human nature. The
other possibility is to reflect on the program itself and its current limitations. The second possibility
defines one possible task for the philosophy of neuroscience. On the one hand, if mechanisms have
been the focus of attention in recent years in the field of philosophy of neuroscience, the
developmental perspective is an occasion to consider these mechanisms in a different light:
mechanisms and their characteristic activities are not only the producers of change, they are also the
products of change, something we need to understand if we want to know how they become capable of
doing what they do. On the other hand, the core idea that mechanisms typically span multiple levels
(Machamer, Darden and Craver, 2000; Craver, 2007) is still fruitful in this different context, as it cures
us both from strict fundamentalism (only lower levels matter) and vague emergentism. In particular, it
is only in defining the role of genes in the containing systems where they are embedded, and it is only
in considering the phenomenon of arealization in its relation with its tight but complex genetic control,
that developmental neuroscience will be able to overcome the present limitations of the rhetoric of
construction.

Denis Forest

University of Paris Ouest (Nanterre) and IHPST, Paris

Bibliography

Anderson (M. A.), 2008, Are interactive specialization and massive redeployment compatible?,
Behavioral and brain sciences, 31, p. 331-34.

Ariew (A.), 1996, Innateness and canalization, Philosophy of science, 63, p. 19-27.

Armentano (M.) et al., 2007, COUP-TFI regulates the balance of cortical patterning between frontal/
motor and sensory areas, Nature Neuroscience, 10, p. 1277-1286.

Bechtel (W.) and Richardson (R.), 1993, Discovering complexity, decomposition and localization as
strategies in scientific research, Princeton, Princeton University Press.

Bishop( K. M.), Goudreau (G.), O'Leary (D.), 2000, Emx2 and Pax6 regulate area identity in the
mammalian neocortex, Science , 288, p. 344-349

Brodmann (K.), 1909/1994, Localisation in the cerebral cortex [translation of Vergleichende

Lokalisationslehre der Grosshirnrinde], London, Smith-Gordon.

Craver (C.), 2007, Explaining the brain, Oxford, Oxford University Press.

Crowley (J. C.) and Katz (L. C.), Ocular dominance revisited, Current Opinion in Neurobiology, p.
104-109.

Cummins (R.), 1975, Functional Analysis , Journal of Philosophy, 72, p. 741-764.

12
Dewey (J.), 1925, Experience and nature, in The Collected works of John Dewey, Later Works,
Southern Illinois University Press

Elman (J.), Bates (E. A.), Johnson (M. K.), Karmiloff-Smith (A.), Parisi(D.), Plunkett (K.), 1996,
Rethinking innateness, MIT Press.

Farah (M.), Noble (K. G.), Hurt (H.), 2006, Poverty, privilege, and brain development, in
Neuroethics, J. Illes ed., Oxford, Oxford University Press, p. 277-287.

Gao (P.-P.), Yue (Y.), Zhang (J.-H.), Cerretti (D.), Levitt (P.), Zhou (R.), 1998, Regulation of
thalamic neurite outgrowth by the Eph ligand ephrin-A5, Proceedings of the national Academy of
science, 95, p. 5329-5334.

Gauthier (I.), Skudlarski (P.), Gore (John C.), Anderson (Adam W.), 2000, Expertise for cars and birds
recruits brain areas involved in face recognition , Nature neuroscience, 3/2, p. 191-197.

Happ (Francesca), 1999, Autism: cognitive deficit or cognitive style?, Trends in cognitive sciences,
3/6, p. 216-222.

Hempel (Carl), 1965, Aspects of scientific explanation, Free Press.

Huberman (A.), Feller (M.) and Chapman (B), 2008, Mechanisms Underlying Development of
Visual Maps and Receptive Fields, Annual review of neuroscience, 31, p. 479-509.

Hunt (D. L.), Yamoah (E. N.) and Krubitzer (L.), 2006, Multisensory Plasticity in congenitally deaf
mice: how are cortical areas functionally specified?, Neuroscience, 139, p. 1507-1524.

Jacob (P.) and Jeannerod (M.), 2003, Ways of Seeing, The scope and limits of visual cognition, Oxford
University Press.

James (W.), 1890, The principles of psychology, New York, Holt.

Johnson (Mark), 2005a, Developmental cognitive neuroscience, 2nd edition, Oxford, Blackwell.

2005b, Subcortical face processing , Nature Reviews Neuroscience, 6/10, p. 766-

774.

Kanwisher (N.), Mc Dermott (Josh), Chun (Marvin M.), 1997, The fusiform face area : a module in
human extrastriate cortex specialized for face perception , The journal of neuroscience, 17/11, p.
4302-4311.

Kingsbury (Marcy A.), Finlay (Barbara L.), 2001, The cortex in multidimensional space : where do
cortical areas come from ? , Developmental science, p. 125-142.

Krubitzer (L.), 2007, The magnificent compromise: cortical field evolution in mammals, Neuron, 56,
p. 201-208.

Machamer (P.), Darden (L.), Craver (C.), 2000, Thinking about mechanisms, Philosophy of science,
67/1, p. 1-25.

13
Mareschal (D.), Johnson (M.J.), Sirois (S.), Spratling (M. W.), Thomas (M. S.), Westermann (G.), 2007,
Neurocosntructivism. How the brain constructs cognition, Oxford, Oxford University Press.

Milner (A. D.) and Goodale (M. A.), 2006, The visual brain in action, Oxford University Press.

Marr (D.),1982, Vision, New York, Freeman and co.

Neander (Karen), 1991, 'The Teleological Notion of "Function'", Australasian Journal of Philosophy,
69, 4, p. 454-468.

OKeefe (John) & Nadel (Lynn), 1978, The hippocampus as a cognitive map, Oxford, Oxford University
Press.

OLeary (D. D.) and Standfield (B. B.), 1989, Selective elimination of axons extended by
developing cortical neurons is dependent on regional locale : experiments utilizing fetal cortical
transplants , Journal of neuroscience, 9, p. 2230-2246.

O Leary (D. D.) and Nakagawa (Y.), 2002, Patterning centers, regulatory genes and extrinsic
mechanisms controlling arealization of the neocortex, Current opinion in neurobiology, p. 14-25

O Leary (D. D.) and Sahara (S.), 2008, Genetic regulation of arealization of the neocortex, Current
Opinion in neurobiology, 18/1, p. 90-100.

Pascual-Leone (Alvaro) and Hamilton (Roy), 2001, The metamodal organization of the brain, in
Casanova (C.) et Ptito (M.), Progress in brain research, 134, Elsevier, p. 1-19.

Rakic (P.), 1988, Specification of cerebral cortical areas, Science, 241, p. 170-176.

Rippon (G.), Brock (J.), Brown (C.), Boucher (J.), 2007, Disordered connectivity in the autistic brain:
challenges for the new psychophysiology, International journal of psychophysiology, p. 164-172.

Samuels (Richard), 1998, What brains wont tell us about the mind: a critique of the neurobiological
argument against representational nativism, Mind and language, 13/4, p. 588-597.

Scherf (K. Suzanne), Behrmann (Marlene), Humphreys (Kate), Luna (Beatriz), 2007, Visual category-
selectivity for faces, places and objects emerges along different developmental trajectories,
Developmental Science, 10/4, p. 15-30.

Shepherd (Gordon M.), 1994, Neurobiology, Oxford University Press.

Sirois (S.), Spratling (M.), Thomas (M. S.), Westermann (G.), Mareschal (D.), Johnson (M. H.), 2008,
Prcis of neuroconstructivism, How the brain constructs cognition, Behavioral and brain sciences,
31, p. 321-331.

Sperry (Roger), 1943, Effect of a 180 degree rotation of the retinal field on visuomotor coordination,
Journal of experimental zoology, 92, p. 263-279.

Sur (Mriganka), Pallas (S. L.) et Roe (A. W.), 1990, Cross-modal plasticity in cortical development :
differenciation and specification in sensory neocortex , Trends in neuroscience, 13, p. 227-233.

Sur (M.) and Rubenstein (J. R. L.), 2005, Patterning and plasticity of the cerebral cortex, Science,
310, p. 805-810.

14
Thelen (E.), and Smith (L. B.), 1994, A dynamic approach to the development of cognition and action,
Cambridge, MA, MIT press.

Tomassy (G. S.) and al., 2010, Area-specific control of cortico-spinal motor differentiation by
COUP-TFI, PNAS, 107/8, p. 3576-3581.

Torborg (C. L.) and Feller (M. B.), 2005, Spontaneous patterned retinal activity and the refinement of
retinal projections, Progress in Neurobiology, 76/4, p. 213-35.

van Fraassen (B. C.), 1980, The scientific image , Oxford, Clarendon Press.

Wundt (W.), 1880, Grundzge der physiologischen Psychologie, Second edition, Leipzig,
Engelmann.

1891, Zur Frage der Localisation der Grosshirnfunctionen , 1891, Philosophische

Studien, Leipzig, Wilhelm Engelmann Verlag

15