CH APTER 6

The developing brain

CO NT EN T S

Structural development of the brain 118

Functional development of the brain 122
Nature and nurture of individual differences 129
Summary and key points of the chapter 141
Example essay questions 142
Recommended further reading 142

Many people are drawn into a subject like psychology because of one nagging
question: What makes us who we are? Were we predestined to become the
person we are today as a result of our genetic endowment? If so, could a parent
of the late twenty-!rst century be reading their child’s genetic “blueprint”
to !gure out their cognitive strengths and weaknesses, their personality and
disposition? Or are we shaped by our experiences and circumstances during
life and during our formative years of development? These questions are
central to the nature–nurture debate, i.e., the extent to which cognition and
behavior can be attributed to genes or environment. While the nature–nurture
debate still has contemporary relevance and continues to excite both scientists
and lay people, this chapter will consider how many of the commonly held
assumptions surrounding this debate are misguided. For example, genes do
not provide a predetermined blueprint, but are themselves switched on and
off by the environment; and the contemporary notion of “environment” is far
broader than is commonly understood. It includes biological circumstances
(e.g., diet, exposure to toxins), as well as personal and social circumstances.
Historically, the pendulum has swung between opposing extremes of
this debate. For example, in 1874, Francis Galton published English Men
of Science: Their Nature and Nurture, arguing that geniuses are born and
not made. As well as coining the phrase “nature or nurture,” he was the !rst
person to realize that heredity could be estimated by comparing identical and
 116 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

K EY T ERMS nonidentical twins. Galton’s advocacy of nature over nurture would become
associated with the discredited eugenics movement, which promoted selective
Nature–nurture debate breeding of the more cognitively able (although in practice this was often
The extent to which cog-
implemented by sterilization of the “feeble-minded”).
nition and behavior can
be attributed to genes or In the early twentieth century, the pendulum had swung to the other
environment. extreme. Freudian theory, for example, emphasized the importance of early
experiences and parenting style in development. The Russian psychologist Lev
Neuroconstructivism
A process of interaction
Vygotsky (1896–1934) also emphasized the role of culture and interpersonal
between environment and communication in development. Behaviorist theories, such as those put
brain-based constraints forward by B. F. Skinner (1904–1990), argued that all behavior was a product
that leads to the mature of learning as a result of rewards and punishments.
cognitive system emerg- Jean Piaget (1896–1980) is regarded as the founding father of modern
ing out of transforma- western developmental psychology. Piaget took a middle ground with regards
tions of earlier ones
to the nature–nurture debate. He regarded development as a cyclical process of
(but does not assume
discrete stages).
interactions between the child and his or her environment (Figure 6.1). In his view,
the genetic contribution consists of developing a brain that is readied to learn in
certain ways, but progression through the stages involves assimilating evidence from
the environment and then developing new mechanisms in light of the feedback
obtained. While many of Piaget’s experimental studies have not stood the test of
time (e.g., children show evidence of reasoning long before Piaget suggested they
should), his basic approach to development has been more enduring.
Following on from the developmental psychology tradition,
developmental cognitive neuroscience has focused on brain-based
explanations of developmental change (Johnson,
2005). One particular current approach is
termed neuroconstructivism (Westermann et al.,
2007). Like Piaget’s approach, this assumes
constant interaction between environment and
genetic factors, with a mature cognitive system
emerging out of transformations of earlier ones.
Unlike Piaget’s approach, the predetermined
aspect of development is construed in terms of
multiple, brain-based constraints (developmental
changes in synapse formation, myelination,
etc.), rather than the less well-de!ned notion of
predetermined “stages.”
This chapter will !rst consider the structural
development of the brain, both prenatally and
postnatally. It will then go on to consider the nature
of developmental change, including evidence for
critical/sensitive periods and innate knowledge. An
FIGURE 6.1: In Piaget’s sensorimotor stage (0–2 years), overview of the origin of genetic differences and
a child learns about the nature of objects (e.g., that they
behavioral genetics will then explore some speci!c
still exist when hidden) and about the nature of cause and
effect (e.g., that actions have consequences on the objects
examples of genetic influences in developmental
around). The child then passes through other stages cognitive neuroscience. Together with the advances
(preoperational, concrete and formal operational) with made in molecular genetics, it is now becoming
greater degrees of abstraction. Although the stages can possible to understand how genetic influences and
be regarded as !xed and predetermined, Piaget stressed experience create changes in the structure and
the role of the environment to successfully develop the function of the brain. This is leading to an exciting
cognitive processes required for the next stage.
rethink of the nature–nurture debate.
© Brooke Fasani/Corbis.
 THE DEVELOPING BRAIN 117

ADAPTING THE METHODS OF COGNITIVE NEUROSCIENCE FOR INFANTS

AND CHILDREN

Methods such as fMRI and EEG are generally considered suitable for infants and children. One
advantage of using these methods in younger people is that they do not necessarily require a
verbal or motor response to be made.
Functional MRI
Gaillard et al. (2001) provide an overview of some of the considerations needed. If one wants to
compare across different ages, then the most signi!cant problem is that the structural properties
of the brain change during development. Although the volume of the brain is stable by about 5
years of age, there are differences in white and gray matter volumes until adulthood (Reiss et al.,
1996). The hemodynamic response function is relatively stable after 7 years of age but differs
below this age (Marcar et al., 2004). The differences in both brain structure and blood flow make
it harder to compare activity in the same region across different ages. Younger children also !nd it
harder to keep still in the scanner, and this motion can disrupt the reliability of the MR signal.
Functional near-infrared spectroscopy (fNIRS)
One relatively new method that is now being used in developmental cognitive neuroscience is func-
tional near-infrared spectroscopy (fNIRS) (e.g., Lloyd-Fox et al., 2010). This measures the amount
of oxygenated blood and is—like fMRI—a hemodynamic method. Unlike fMRI, it accommodates a
good degree of movement and is more portable. The infant can sit upright on their parent’s lap.
However, it has poorer spatial resolution and does not normally permit whole-head coverage.
ERP/EEG
When working with young participants using ERP/EEG, a limiting factor is the child’s willingness to
tolerate the electrodes, the task and the time commitment required (Thomas & Casey, 2003). Chil-
dren and adults can show quite different patterns of ERP (e.g., in terms of latency, amplitude or
scalp distribution), even for tasks that both groups !nd easy (Thomas & Nelson, 1996), as shown
in Figure 6.2. These could either reflect age-related cognitive differences (i.e., the same task can
be performed in different ways at different ages) or non-cognitive differences (e.g., the effects of
skull thickness, cell packing density or myelination).

Adults Eight-year-olds
20 40
15 Target 30 Target
Microvolts

Microvolts

10 Novel 20 Novel
5 10
0 0
–5 –10
–10 –20
10

10
210

410

610

810

210

410

610

810
1010

1210

1410

1010

1210

1410

Time (ms) Time (ms)

FIGURE 6.2: Adults and children show very different visual ERP waveforms, despite having equivalent
behavioral performance.
Adapted from Thomas and Nelson, 1996.
118 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

Brain stimulation: TMS and tES

Single and paired pulse TMS is considered to pose minimal risk in children (Gilbert et al., 2004),
although repetitive TMS isn’t recommended except for compelling therapeutic purposes (e.g., treatment
of depression; Gaynes et al., 2014). Transcranial electrical stimulation (tES) is not believed to pose any
greater risks or side-effects in children relative to adults (Krishnan et al., 2015), and has been used in
conjunction with cognitive training to treat certain learning dif!culties (Krause & Kadosh, 2013).

S T R U CT U R AL DE VE L O PM E N T O F T HE B R AI N
One common idea is that the genetic code provides a blueprint for the structure
of the brain. At some gross level, this must be true: all human brains are similar
to each other, but differ from the brains of other species. This reflects our different
genetic makeup. However, the term “blueprint” is misleading as it suggests that
structural details of individual brains are speci!ed at a very !ne level of detail.
However, it is inconceivable that the genetic code contains a detailed wiring
diagram of the brain (recall that the human brain has 86 billion neurons each
with 10,000 connections). Instead of the blueprint analogy, one could instead
imagine the genetic code more like a recipe for making a brain. Gottlieb (1992)
makes a distinction between two key ideas in development which he termed
predetermined development versus probabilistic development. In predetermined
development, genes dictate the structure of the brain, which enables the particular
functions of the brain, which determines the kinds of experiences we have. This
is a traditional view of how genes affect cognition. Gottlieb contrasts this with
probabilistic development in which brain structure, and even the expression of
genes, can be influenced by experience as well as vice versa. Also the effects of genes
on brain structure are themselves probabilistic (irrespective of any environmental
influences) such that they specify approximately how many neurons to grow, and
where to grow them, but not exactly how/where neurons will grow (Hassan &
Hiesinger, 2015). Thus, the same genetic program run in the same environment
would not produce an exact carbon copy but, instead, some kind of variation on
a theme. This is perhaps a more realistic way of understanding “identical” twins
(monozygotic, MZ) who, we shall see, have very similar but not identical brains.
Probabilistic development represents the dominant view in modern
developmental cognitive neuroscience and the sections below will unpack this
concept in more detail.

GOTTLIEB’S (1992) DIFFERENT VIEWS OF

DEVELOPMENT

Predetermined development:
Genes → Brain structure → Brain function → Experience
Probabilistic development:
Genes Brain structure Brain function Experience
 THE DEVELOPING BRAIN 119

Prenatal development KE Y T ER MS
The human gestation period is around 38 weeks from conception. The Neural tube
newly formed embryo undergoes a rapid process of cell division, followed The embryo’s precursor
by a process of differentiation during which the different cells become to the central nervous
increasingly specialized (Figure 6.3). The nervous system derives from a set system, consisting of a
set of cells arranged in a
of cells arranged in a hollow cylinder, the neural tube. By around 5 weeks,
hollow cylinder.
the neural tube has organized into a set of bulges and convolutions that will
go on to form various parts of the brain (e.g., the cortex, the thalamus and Neuroblasts
Stem cells for neurons.
hypothalamus, the midbrain). Closer to the hollow of the neural tube are
several proliferative zones in which neurons and glial cells are produced by Radial glial cells
division of proliferating cells (neuroblasts and glioblasts). Purves (1994) Support cells that
estimates that the fetal brain must add 250,000 neurons per minute at certain guide neurons from the
neural tube to their !nal
periods in early development.
destination.
The newly formed neurons must then migrate outwards toward the
region where they will be employed in the mature brain. This occurs in two
ways. Passively, older cells tend to be pushed to the surface of the brain.
Structures such as the hippocampus are formed this way. There is also an
active mechanism by which newer cells are guided to particular destinations,
pushing past the older cells. Rakic (1988) identi!ed radial glial cells that
act like climbing ropes, ensuring that newly formed neurons are guided to
their !nal destination. The convoluted surface of the brain, the neocortex,
is formed in this way.
Regional differences in various molecular signals affect the neurons’
structure, migration and survival (see Sur & Rubenstein, 2005). Different ONLINE RESOURCES
doses of these signals determine the dimensions of the various lobes of the
brain, such that, for example, a dose above a certain threshold may instruct Visit the companion
website for an
a new neuron to develop features characteristic of a frontal lobe neuron animated video of
(e.g., in terms of its connectivity) but below that dose it may resemble a a migrating cortical
parietal neuron (Fukuchi-Shimogori & Grove, 2001). This suggests a simple neuron and animated
mechanism for creating individual differences in brain structure and also for video of radial
evolutionary development (e.g., a shifting dose could enable an evolutionary migration from Rakic
laboratory.
jump in frontal lobe enlargement).
The highly folded cortex is a clearly distinguishable feature of the human
brain. It is visible in the last few months before birth, and shows further small
changes over the !rst two years of life (Li et al., 2014). This folding is likely
to be an outcome of packing more neurons within a restricted space together
with stretching the cortical surface by axonal tension (as opposed to the
cortical shape being speci!ed directly by the genome). Genetic mutations in
mice that lead to over-production of neurons also lead to the development of
a more convoluted cortical surface (Haydar et al., 1999). Van Essen (1997)
has proposed that the overall pattern of cortical gyri and sulci is linked to the
development of axon bundles (e.g., such as those shown previously in Figure
2.4) which places the cortical surface under tension. In effect, the axons are
like elastic bands that pull the cortical sheet in particular directions, giving it a
characteristic shape but also allowing for variability across individuals. Axon
guidance, like neural migration, is also affected by the regional concentration
of different molecular signals that attract/repel different axons, ultimately
biasing how/where they form (McLaughlin & O’Leary, 2005). DTI of infants
born prematurely (at 30 weeks after conception, instead of 38 weeks) show
 120 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

K EY T ERM that the major white matter tracts are in place at 30 weeks but continue to
develop a more !ne-grained pattern over the subsequent 10 weeks (Ball et al.,
Hebbian learning 2014). Although the gyral pattern is broadly similar across individuals, there
Strengthening of a
is variability and twin studies suggest that gyri!cation is far less heritable
synapse that occurs
when the presynaptic and than, for instance, brain volume (Bartley et al., 1997).
postsynaptic neurons are Finally, although prenatal neurons have very limited functional inputs
active at the same time from the environment, they can still show spontaneous electrical activity
(“what wires together that enables networks to form in the brain on the basis of Hebbian learning
!res together”). (“what wires together !res together”). For instance, spontaneous waves of
electrical activity emanating from the retina (but not triggered by light)
are important for setting up synaptic pathways from the eyes to the lateral
geniculate nucleus and the visual cortex in readiness for the processing of
visual stimuli (Katz & Shatz, 1996).

Postnatal development
At birth, the head makes up approximately a
quarter of the length of the infant. Although
the brain itself is small (450 g) relative to adult
human size (1,400 g), it is large in comparison
to remote human ancestors and living primates
(a newborn human brain is about 75 percent
of that of an adult chimpanzee). The vast
majority of neurons are formed prior to birth,
so the expansion in brain volume during
postnatal development is due to factors such
as the growth of synapses, dendrites and axon
bundles; the proliferation of glial cells; and the
myelination of nerve !bers.
Huttenlocher and Dabholkar (1997)
measured the synaptic density in various regions
of human cortex (Figure 6.4). This is a measure
of the degree to which neurons are connected
to each other and is unrelated to the number
of neurons per se or how active the synapses
are. In all cortical areas studied to date, there
FIGURE 6.3: The embryonic and fetal development of the is a characteristic rise and then fall in synapse
human brain. Cortical asymmetries between the left and right formation (synaptogenesis). In primary visual
hemispheres, implicated in language acquisition, are present at and primary auditory cortex, the peak density
24 weeks.
is between 4 and 12 months, at which point it is
From Cowan, 1979. © 1979 by Scienti!c American, Inc. All rights
reserved.
150 percent above adult levels, but falls to adult
levels between 2 and 4 years. In the prefrontal
cortex, the peak is reached after 12 months, but does not return to adult levels
until 10–20 years old. PET studies of glucose metabolism in developing brains
show a similar rise and fall to studies of synaptogenesis, although peaking time
tends to be somewhat later (Chugani et al., 1987). Glucose metabolism may
be a measure of actual neural activity rather than neural structural changes.
Why does the number of synapses fall during the course of development? It is
not necessarily the case that more synapses reflect more ef!cient functioning.
 THE DEVELOPING BRAIN 121

70
Birth
Auditory cortex
60
Visual cortex
Prefrontal cortex
Synapses per 100 µm3

50

40

30

20
Adolescence
10

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10,000
Age (days from conception)

FIGURE 6.4: Synapse formation has a rise-and-fall pattern. It peaks soon after birth, although different
cortical regions differ greatly in the time taken to fall again to adult synaptic levels.
From Huttenlocher and Dabholkar, 1997. Reprinted with permission of John Wiley & Sons Inc.

During development a process of !ne-tuning the brain to the needs of the KE Y T ER MS

environment renders some connections redundant.
Myelination refers to the increase in the fatty sheath that surrounds axons Myelination
An increase in the fatty
and increases the speed of information transmission. In structural MRI, the
sheath that surrounds
increase in white matter volume over the !rst two decades of life may reflect axons and increases the
the time course of myelination (Giedd et al., 1999). Again, the prefrontal speed of information
cortex is one of the last areas to achieve adult levels of myelination, and transmission.
this, together with the late !ne-tuning and elimination of synapses in this Plasticity
region, may contribute to the development of mature social behavior during The brain’s ability to
adolescence and the control of behavior in general. change as a result of
Following birth, all of our everyday experiences result in tiny changes experience.
to the structure of our brain, in the form of altering the pattern of synaptic
connections. Sometimes these changes are even visible at the macroscopic
level. Adults who learn to juggle with three balls over a 3-month period
show increased gray matter density, assessed with MRI, in a region, V5/MT,
specialized for detecting visual motion as well as the occipitoparietal region
implicated in hand–eye coordination (Draganski et al., 2004) (Figure 6.5).
This example illustrates a central concept of this chapter—namely, plasticity.
Plasticity refers to experience-dependent changes in neural functioning.
Similar !ndings are noted elsewhere in the book due to the spatial memory
demands of driving a taxi (Maguire et al., 2000), or acquiring musical
expertise (Bermudez et al., 2009). However, whilst one might intuitively
expect a positive relationship between individual differences in ability and
local increases in gray matter, this is not always the case. Thus, congenitally
blind people have more gray matter than sighted people in their visual cortex
(Jiang et al., 2009) and people with congenital amusia, linked to problems in
pitch perception, have more gray matter in their auditory cortex (Hyde et al.,
2007). Thus, one can’t take gray matter density/thickness as a simple proxy of
cognitive ability as it depends on the underlying mechanisms: developmental
pruning of synapses (thinner is better) or experience-dependent changes
(thicker is better).
122 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

% change in gray matter

4
3
2
1
0
–1
–2
Scan Scan Scan
1 2 3

FIGURE 6.5: MRI scans were obtained at three time intervals: before learning to juggle; after 3 months
of training; and after a further 3 months of no practice. The graph shows increases in gray matter
density in an area associated with visual motion perception, area V5/MT.
Reprinted by permission from Macmillan Publishers Ltd: Draganski et al., 2004. © 2004.

Zatorre et al. (2012) discuss the potential cellular and molecular

mechanisms that could underpin experience-related structural changes that
are visible with MRI in humans (VBM approaches for gray matter, and DTI
for white matter). First of all, they suggest that neurogenesis is an unlikely
candidate outside of the hippocampus. However, growing more dendritic
branches, synapses or axon collaterals could increase gray matter density. Glia
cells can also divide or change in size or structure which could contribute to
both gray matter and white matter volume, as could changes to the capillary
vasculature. With regards to fractional anisotropy (in DTI), this could
increase as a result of axons becoming more aligned or more myelinated.

Evaluation
Although the gross anatomical features of our brain are driven by our genes
(e.g., making a human brain versus making the brain of another species),
there are limits to what our genes can do. This reflects both a positive influence
of the environment (including our own behaviors), but also the fact that the
genetic information for constructing a brain is imprecise and, hence, variable.
This is consistent with Gottlieb’s (1992) idea of probabilistic development.

F U N CT IO N AL DE VE LO PME N T O F TH E B R A I N
Having considered how brain structure is changed during development, the
present section is primarily concerned with how brain function (i.e., different
types of ability and knowledge) changes developmentally. In particular,
several broad issues will be considered: !rst, the extent to which brain function
is altered by major structural changes; the role of critical/sensitive periods
in development; and, !nally, the extent to which any kind of knowledge or
ability can be said to be innate.
 THE DEVELOPING BRAIN 123

Functional brain plasticity in

rewired brains
The previous section introduced the idea that
whilst early development of the brain is influenced
by genes, there isn’t a genetic blueprint for how,
exactly, each individual brain will turn out.
One consequence of this is that, in the face of
some major insult, the brain may be capable
of reorganizing itself in some fundamentally
different ways. Consider the case of AH, a 10-year-
old girl, who failed to develop a right hemisphere
and right eye prenatally but had only very minor
visual impairments (Muckli et al., 2009). Detailed
fMRI scanning revealed that visual information
that would normally cross the optic chiasm into
the “missing” hemisphere could be rerouted
into the intact (ipsilateral) hemisphere. Neurons
coding left and right sides of space were inter-
mingled within the same cortical map where
they are normally segregated into the different
hemispheres (Figure 6.6).
In non-human animals, it is possible to FIGURE 6.6: The structural MRI scan of a 10-year-old girl
surgically transplant a region of cortex or sever who failed to develop a right hemisphere in the womb,
pathways such that novel ones emerge. In these showing the intact LGN in the left hemisphere (yellow).
extreme instances, it is possible to explore how Front: fMRI brain activity in her intact left visual cortex to
stimuli presented in both the right visual !eld (green-blue)
these major structural changes impact on neural and – unusually – the left visual !eld (yellow-red).”
function and behavior. Prenatal visual cortex From Muckli et al. (2009).
transplanted into somatosensory cortex responds
to touch on a mouse’s whiskers and reconnects
to the somatosensory region of the thalamus (Schlagger & O’Leary, 1991).
If the pathway from the cochlear to the medial geniculate nucleus (MGN)
is severed in a ferret, then visual inputs (from the retina) spontaneously
reroute themselves into the MGN (Sur et al., 1988). The consequence of this
is that visual input into the eyes ultimately leads to activity in the auditory
cortex (as well as visual cortex). Interestingly, these “auditory” neurons
take on functional characteristics of visual neurons such as preferential
responding to different orientations and movement directions (Sharma et al.,
2000; Sur et al., 1988).
Studies such as these show that there is a high degree of structural and
functional plasticity in the early brain. However, it doesn’t mean that all
neurons are fully interchangeable. Spontaneous patterns of activity prior to
birth are already shaping neural activity and parcellating them into different
networks (Katz & Shatz, 1996). In the example above, visually rewired
“auditory” cortex still retains vestiges of normal auditory cortex connections
and the visual representations are poorer than those found in true visual
cortex (Sharma et al., 2000). Moreover, these kinds of opportunities for
major reorganization appear to be strictly time-limited. The next section
considers in more detail the notion of critical/sensitive time periods for
functional brain development.
 124 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

RECOVERY OF FUNCTION AFTER EARLY BRAIN DAMAGE

Although strokes are rare in infancy and childhood, they do occur. Typically, however, the long-term
effects on cognition are neither as severe nor as speci!c as those arising from strokes in adulthood.
This is consistent with the view that plasticity and recovery is greatest earlier in life, often referred
to as the Kennard Principle after the neurologist Margaret Kennard (Dennis, 2010). Several studies
have found that children who had strokes around the time of birth go on to develop intellectual
and language skills in the normal range (Aram & Ekelman, 1986; Ballantyne et al., 2008). With
regards to language, it is often found that early lesions to the left hemisphere can result in later
right hemisphere language as assessed using fMRI (Liegeois et al., 2004). In this study, even
lesions outside of “classical” language areas (e.g., Broca’s area) were just as likely to result
in right hemispheric language consistent with the view that functional specialization of regions
emerges gradually and in a way that is not completely predetermined. Given that the brain has very
limited scope to grow new neurons, one may wonder whether accommodating language in the right
hemisphere would have a detrimental outcome on traditional right hemispheric functions (e.g.,
visuo-spatial skills). There is some evidence for this. Lidzba et al. (2006) report that the extent of
right hemispheric language (assessed by fMRI) resulting from early stroke correlated negatively
with performance on visuo-spatial tasks (i.e., greater right hemisphere language is associated with
poorer visuo-spatial skills). This suggests that, while early plasticity can aid recovery, this may not be
completely without a cost.

K EY T ERMS Critical and sensitive periods in development

Kennard Principle In 1909, a young Austrian boy named Konrad Lorenz and his friend (and
The idea that the later wife), Gretl, were given two newly hatched ducklings by a neighbor. The
earlier brain damage is ducklings followed them everywhere, apparently mistaking them for their
sustained, the better the parents (see Figure 6.7). This process, now termed filial imprinting, was studied
functional outcome.
intensively by the adult Lorenz using goslings and earned him a Nobel Prize
Filial imprinting (see Tinbergen, 1951). Lorenz observed that there was a narrow window of
The process by which a opportunity, between 15 hours and 3 days, for a gosling to imprint. Once
young animal comes to
imprinted, the gosling is unable to learn to follow a new foster parent. The
recognize the parent.
movement of a stimulus was deemed to be crucial for determining what object
Critical period the gosling will imprint to. A region of the chick forebrain known as intermediate
A time window in which and medial of the hyperstriatum ventrale (IMHV), which may correspond to
appropriate environmen-
mammalian cortex, is critical for enabling imprinting (Horn & McCabe, 1984).
tal input is essential for
learning to take place. The studies above suggest that there is a critical period for imprinting.
A critical period has two de!ning features: !rst, learning can only take place
Sensitive period
within a limited time window; and, second, the learning is hard to reverse in
A time window in which
appropriate environ-
the face of later experience. Subsequent evidence suggests that the window
mental input is particu- of opportunity can be extended by lack of suitable early experience (e.g.,
larly important (but not initial absence of a moving object), and that learning can be reversed in
necessarily essential) for certain circumstances. As such, many researchers prefer the more moderate
learning to take place. terminology of a sensitive period. For instance, a chick imprinted to one
object will often generalize to other objects of similar appearance (e.g., color
and shape). By gradually changing the features of the objects to which it is
exposed, the chick’s !nal preference can be different from its initial preference,
even after the end of the “critical” period (Bolhuis, 1990).
 THE DEVELOPING BRAIN 125

FIGURE 6.7: These goslings

follow the Austrian professor,
Konrad Lorenz, as if he is
their mother! This process is
called !lial imprinting.
© Science Photo Library.

The development of visual abilities also shows evidence of a sensitive

period. For example, Hubel and Wiesel (1970b) took single-cell recordings
from the primary visual cortex of cats in whom one eye had been deprived
of visual input in early life (by sewing it shut). They found that the cells
responded to input from the sighted eye only, whereas normally reared cats
possess cells that respond to inputs from both eyes. During a sensitive period
between 4 and 5 weeks after birth, eye closure for 3–4 days leads to a sharp
decline in the number of cells that will respond to input from both eyes.
What of “higher” cognitive abilities, such as language? Lenneberg
(1967) initially argued that language acquisition has a critical period that
ends abruptly at puberty. However, the ability to comprehend and produce
language is likely to depend on other skills such as hearing, motor ability,
working memory capacity and so on. Each of these basic skills may have its
own sensitive period, which means that different components of language may
have their own sensitive period rather than a !xed cut-off point at puberty.
For example, the sensitive period for making phonemic discriminations such
as the distinction between r and l occurs during infancy and is resistant
to subsequent exposure (McCandliss et al., 2002). In contrast, accents are
more fluid during childhood but become notoriously hard to change from
the onset of adulthood.
Studies of feral children offer some support to Lenneberg’s idea. Genie
had been locked away by her mentally unstable family from the age of 20
months to 13 years when she was discovered in Los Angeles in 1970 (Curtiss,
1977). During this period she was severely maltreated and was not allowed
to speak or be spoken to. On being rescued she was almost entirely mute,
with a vocabulary of around 20 words. Within the !rst 18 months of being
placed with a foster parent, her language was reported to have developed
well on all fronts, including both vocabulary and grammar, and this was
cited as evidence against a sensitive period (Fromkin et al., 1974). However,
subsequent studies are more consistent with a sensitive period and have
 126 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

FIGURE 6.8: When brain

activity linked to grammatical
processing is contrasted
for !rst (L1) and second
(L2) languages, there are
no differences for early
bilinguals (L2 from birth).
However, L2 (relative to L1)
is linked to more activity in
language-related regions
when late bilinguals (L2 after
6 years) make grammatical
judgments, irrespective
revealed that her language acquisition remained very poor compared with
of their pro!ciency in that
language. This suggests a
young children; although it remains debated as to the extent to which her
sensitive period for ef!cient grammar was speci!cally affected or whether all aspects of language were
grammatical acquisition. affected (Jones, 1995).
From Perani & Abutalebi, 2005. Thankfully, research in which exposure to a !rst language is withheld
from a child is limited to a tiny number of cases. However, second language
acquisition offers a richer source of evidence to test for the existence of a
sensitive period. Brain imaging studies reveal that both age of acquisition
and level of pro!ciency determine the neural substrates of second language
processing in adults. One study compared syntactic and semantic language
tasks in Italian-German bilinguals using fMRI (Wartenburger et al., 2003).
For syntactic judgments, the age of acquisition was critical: those who
learned the second language later in life showed more activity in language-
related brain regions when processing syntax irrespective of their level of
pro!ciency (Figure 6.8). This suggests a sensitive period for grammar in
terms of neural ef!ciency (more activity is interpreted here as less ef!ciency).
For semantic judgments, by contrast, the pattern of activity was related to
pro!ciency level in the second language rather than age of acquisition (i.e.,
little influence of sensitive periods for language semantics). Hartshorne et al.
(2018) examined syntactic pro!ciency (judging whether sentences are
grammatical) in second language learners in a very large sample (two-thirds
of a million) and separately modeled age of acquisition and pro!ciency. They
estimated a sensitive period as late as 17.4 years beyond which syntax learning
ability declines steadily.
What general properties of the nervous system give rise to sensitive
periods in development? Thomas and Johnson (2008) provide an overview.
One possibility is that there is a strict maturational timetable in which a set
of neurons are readied for learning (e.g., by synaptogenesis) and are then
later “fossilized” (e.g., reducing plasticity, removing weaker connections)
according to a strict timetable. A second possibility is that a set of neurons
are readied for learning but that the process is self-terminating to some extent,
i.e., the sensitive period will “wait” for suitable exposure. For example, in !lial
imprinting there is evidence that a particular gene is switched on at the start
of the sensitive period but is switched off again 10 hours after imprinting has
occurred (Harvey et al., 1998). In human infants born with dense cataracts
over both eyes, there is a rapid increase in visual acuity when the cataracts are
surgically removed, even as late as 9 months after birth (Maurer et al., 1999).
This suggests that the development of visual acuity will, to some extent,
“wait” for an appropriate environment. However, this is only partly true, as
 THE DEVELOPING BRAIN 127

9-year-old children who had cataracts removed in the !rst 6 months of life KE Y T ER MS
had some dif!culties in visual processing of faces (Le Grand et al., 2001).
Empiricism
In philosophy, the view
Innate knowledge? that the newborn mind is
a blank slate.
Perhaps the most controversial topic in developmental cognitive neuroscience
Nativism
is the extent to which any form of knowledge or ability can be said to be innate
In philosophy, the view
(Karmiloff-Smith, 2006; Spelke, 1998). This division has a long historical and that at least some forms
philosophical tradition between so-called empiricists (who believed that the of knowledge are innate.
mind is a blank slate) and nativists (who believed that at least some forms of
Instinct
knowledge are innate). A behavior that is a prod-
The word innate itself conjures up somewhat different connotations to uct of natural selection.
different researchers. For some, the word is synonymous with the idea that
behavior is a product of natural selection (Ridley, 2003). The word instinct
is often used in this context and suitable examples would be !lial imprinting
in birds (Tinbergen, 1951) or even language in humans (Pinker, 1994). In this
usage of the word “innate,” there is still a role for experience to play, perhaps
within a sensitive period of development. A chick will only imprint if it is
exposed to a suitable stimulus in the environment, and a child will only learn
sophisticated language given suitable inputs. However, in both examples the
particular content of the behavior cannot be said to be innate. The chick will
as happily imprint to an Austrian professor as to its mother, and a child is
capable of learning a diverse range of vocabulary and syntax, and not even
the manner of production (e.g., speaking versus sign language) is strongly
predetermined. In this sense of the word “innate,” there is a readiness for
certain knowledge to be acquired, but the knowledge itself is not strictly innate.
This leads to a consideration of the second way in which the word “innate”
is applied: namely, that knowledge or behavior can be said to be innate if it
comes about in the absence of appropriate experience. It is this particular usage
of the term that has attracted much controversy (Spelke, 1998). The very early
development of the primary visual cortex of the cat can, in this sense, be said to
be innate, because it makes no difference whether the cat has visual experience
or not (Blakemore & Vansluyters, 1975), as shown in Figure 6.9. Both normally
developing cats and cats that have been visually deprived in both eyes have cells
that respond to lines of particular orientations up to around 3 weeks after birth

FIGURE 6.9: Orientation

selectivity at 14, 21 and 45
days in the primary visual
cortex of cats reared in a
normal visual environment
(top) and a dark-reared
environment (bottom). The
dark-reared cats show
normal development up
to 21 days but then show
a decrease. The different
colors represent the extent
to which neurons respond to
particular orientations.
Adapted from Crair et al., 1998.
 128 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

(Blakemore & Vansluyters, 1975). However, experience is

needed for a mature system to form. In the presence of
complex visual experience, these cells become even more
!nely tuned and resemble those of an adult by 4 weeks, but
in the absence of appropriate visual experience the blind
cats lose this speci!city.
Similar conclusions arise when one considers the
development of phobias. Humans can easily learn to
become fearful of certain stimuli such as snakes (e.g., by
pairing with an electric shock), but it is hard to become
fearful of stimuli such as flowers—a phenomenon that
has been termed prepared learning (Seligman, 1971). In
a series of studies, Mineka and colleagues studied fear
conditioning in monkeys (for a review, see Ohman &
Mineka, 2001). Whereas monkeys born in captivity to
wild-born monkeys show fear of snakes, monkeys who
were born from mothers raised in captivity do not.
The fearless monkeys could acquire fear of snakes by
watching videos of other monkeys reacting with fear to
snakes, but they could not acquire fear of flowers using
the same method. This suggests that fear of snakes does
FIGURE 6.10: Harlow, in his famous article “The require suitable experience, even if that fear is transmitted
nature of love,” argues that monkeys have an innate
vicariously via other monkeys rather than through contact
preference for a soft versus wire “mother” even if
the wire “mother” provides the infant with milk.
with snakes. That is, this behavior can be said to be innate
From Harlow, 1958. Reproduced with kind permission of
in the sense of being a product of natural selection, but
Harlow Primate Laboratory, University of Wisconsin. not in the sense of developing without experience.
Some preferences could, arguably, be said to be innate
K EY T ERM in the sense that they do not appear to depend on experience. Newborn infants
prefer sweet tastes over neutral and sour ones (Desor et al., 1975), and they
Prepared learning
prefer some visual patterns over others (Damon et al., 2017). Harlow (1958)
The theory that common
phobias are biologically
reported a series of ethically dubious experiments in which newborn monkeys
determined from evolu- were isolated from their natural mothers but “reared” by arti!cially created
tionary pressures. mothers such as a stuffed toy monkey or a metal wire monkey (Figure 6.10).
The monkeys preferred to cling to the furry stuffed toy rather than the metal
one, even if the metal one provided the monkey with milk. This went against
the standard behaviorist doctrine at the time that maternal love was merely
a learned reward for satisfying basic needs such as hunger (in which case the
monkey should show affection to the wire mother).
Some abilities could also, arguably, be said to be innate in the sense that
they do not appear to depend on experience. Newborn infants will imitate
tongue protrusion (Meltzoff & Moore, 1977), as shown in Figure 6.11. That
is, they demonstrate an understanding that a seen tongue being protruded
corresponds to their own, unseen, motor ability to do the same. Meltzoff and
Moore concluded that “the ability to use intermodal equivalences is an innate
ability of humans” (1977, p. 78).
The studies above suggest that certain dispositions (e.g., to fear certain types
of thing), preferences (e.g., sweet) and abilities (e.g., to detect edges, intermodal
matching) can—in some sense of the word—be said to be innate. However, the
issue of whether the speci!c content of knowledge (or so-called representations)
is innate is much harder to substantiate. For example, newborn infants prefer to
 THE DEVELOPING BRAIN 129

look at real faces relative to faces with the parts

rearranged, but this could reflect a tendency to
prefer certain symmetrical patterns (Johnson
et al., 1991). However, they will also prefer to
look at a jumbled-up face provided it is top-
heavy (Macchi Cassia et al., 2004). This makes it
hard to argue that the speci!c knowledge of what
a face looks like is innate, although one could still
reasonably claim that a preference for particular
kinds of pattern is an evolutionary adaptation.

Evaluation
The functions of different regions of the brain
depend on the underlying structure (i.e., what
connects with what). Signi!cant changes to
brain structure during development can result
in atypical functional specializations, although
there may be a time-limited window for this to
FIGURE 6.11: This 23-day-old infant imitates the tongue protrusion
occur. Sensitive periods are important for the of the experimenter, suggesting an understanding of the link
development of most, if not all, cognitive abilities between seen actions of another and their own, unseen actions.
(from vision to grammar) and reflect a mixture Photo by Andrew N. Meltzoff and E. Ferorelli, with permission from Andrew
of innate influences and environmental exposure. N. Meltzoff.
Innate influences can be construed both as
adaptive dispositions (i.e., instincts) or, in some cases, as a tendency for certain
behaviors to develop (at least up to some point) without a suitable environment.

NATURE AND NURTURE OF INDIVIDUAL DIFFERENCES

The previous sections have focused primarily on how the brain and cognition tends
KE Y T ER M
to develop structurally and functionally across the population, but the remaining
sections of this chapter focus on how and why there is variability from one person Behavioral genetics
to another. Behavioral genetics is concerned with studying the inheritance of A !eld concerned with
behaviors and cognitive skills. The classic methods of behavioral genetics are studying the inheritance
of behavior and cogni-
twin studies and adoption studies which provide ways of disentangling nature
tion.
and nurture. However, genetic differences themselves (i.e., variations in the DNA
sequence) can now be studied relatively easily. This holds out the promise of
being able to link behavioral differences back to molecular signals involved in
building the brain (e.g., axon guidance, neural migration to the cortex).

Twin studies and adoption studies

Most behaviors run in families, but it is hard to know to what extent this reflects
shared environment or shared genes. When a child is placed into an adopted
home, he or she will effectively have two sets of relatives: biological relatives
with whom the child no longer shares any environment, and adopted relatives
with whom the child shares an environment, but not genes. Will the child more
closely resemble the biological or adoptive family, thus emphasizing a role of
nature or nurture, respectively? In many cases, it is not possible to contact or
test the biological relatives, but the genetic contribution can still be estimated
 130 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

K EY T ERM by comparing the adopted child with non-adopted siblings in the household
Chromosome (i.e., both the adopted and non-adopted siblings share family environment,
An organized package of but not genes). Assisted fertility with donor eggs and/or sperm also represents
DNA bound up with pro- a modern twist on the classic adoption design (Rice et al., 2009).
teins; each chromosome
contains many genes.

THE ORIGINS OF GENETIC DIFFERENCES

The human genetic code is organized onto 23 pairs of chromosomes, making a total of 46 chro-
mosomes. One of the chromosomes of each pair comes from the maternal line and one from the
paternal line. In each individual there are two copies of each gene normally present, one on each
chromosome. However, genes may exist in different forms, termed alleles. The different alleles
represent changes (or mutations) in the sequence of the gene that is propagated over many
generations, unless natural selection intervenes. Many different allelic forms are common and
benign, but they account for the individual differences that are found between humans as well as
differences between species. For example, two different alleles of a single gene determine whether
the earlobes will be hanging or attached. In other instances single gene mutations are not benign,
as in the case of Huntington’s disease (see Chapter 10). A different allele may mean that the
end-product encoded by the gene (such as enzymes) works less ef!ciently, more ef!ciently or not at
all. Alternatively, it may mean that the gene works in an entirely novel way by, for example, altering
the expression of other genes. Most behavioral traits will be an outcome of the concerted action of
many genes. Even though a given gene may exist in only a small number of discrete allelic types,
when many such genetic variants are combined together they may produce an outcome that is con-
tinuously distributed—such as the normal distribution found for height or IQ. Disorders such as au-
tism, dyslexia and schizophrenia also appear to be polygenic in nature (see Tager-Flusberg, 2003).
As well as differences in alleles, individuals differ in the spacing of genes on the chromosomes
(most of the genome contains nongene segments). While it is unclear whether this contributes to
observable individual differences, an analysis of the spacing of various genomic markers is central
to techniques such as genetic “!nger-printing” and attempts to locate candidate genes on the
basis of behavioral markers (e.g., presence of schizophrenia).
During production of eggs and sperm the genes from the maternal and paternal chromosomes
are “shuffled” so that a single new chromosome is created that is a combination of the original two.
This mechanism prevents the number of chromosomes doubling in each generation. This provides
one mechanism leading to genetic variation through producing different combinations of a !nite set
of alleles. This process can also go wrong if segments of DNA get deleted or duplicated. Some rela-
tively common genetic disorders formed in this way are summarized below.

Genetic disorder Origins Cognitive developmental characteristics

Down’s syndrome A duplicated copy General learning dif!culties (IQ < 70), poor !ne motor control, de-
of chromosome 21 layed and impaired expressive language
Turner syndrome A missing copy of Not associated with mental retardation, but verbal IQ tends to be
the X-chromosome (or higher than nonverbal; some dif!culties in executive functions and
deletion of part of it) social skills (Ross et al., 2000)
William’s syndrome A deleted segment General intellectual impairment but with some tendency for language
of chromosome 7 abilities to be better than spatial abilities; high sociability, but not
necessarily high social intelligence (Karmiloff-Smith, 2007)
 THE DEVELOPING BRAIN 131

Twin studies follow a similar logic (Figure 6.12). Twins are

formed either when a single fertilized egg splits in two (monozygotic
or MZ twins) or when two eggs are released at the same time
and separately fertilized (dizygotic or DZ twins). MZ twins are
genetically identical; they share 100 percent of their genes. DZ
twins are nonidentical and share only 50 percent of their genes
(i.e., the same as non-twin siblings). Given that both are assumed
to share the same family environment, any difference between MZ
and DZ twins is assumed to reveal genetic influences. Studies of
twins reared apart combine the advantages of the standard twin
study and adoption study.
There are a number of ifs, ands or buts to the usefulness of
these study designs. With regards to twin studies, it is assumed
that MZ and DZ twins experience similar environments. However,
MZ twins could be treated more similarly by others. Also, MZ
FIGURE 6.12: Identical twins look the
twins often have more similar prenatal environments: many MZ
same, but do they think the same?
twins share the same sac (called the chorion) within the placenta,
uniball/iStock
but DZ twins never do. As such, MZ twins may be more likely to
be exposed to the same viruses prenatally. With regards to adoption studies, KEY T E R M S
selective placement could mean that children tend to get adopted into similar Allele
environments (e.g., with regard to race or socioeconomic status). Another Different versions of the
issue is whether families who adopt or who give up their children for adoption same gene.
are representative of the general population. Plomin et al. (2001) provide MZ twins (monozygotic)
an assessment of this debate and argue that the main !ndings are relatively Genetically identical
robust to these potential drawbacks. twins caused when a fer-
tilized egg splits in two.

Heritability estimates of brain and behavior DZ twins (dizygotic)

Twins who share half of
Twin studies and adoption studies are ways of establishing whether there is their genes, caused when
genetic influence. Heritability is an estimate of how much genetics contributes two eggs are fertilized by
to a trait. In particular, heritability is the proportion of variance in a trait, in two different sperm.
a given population, that can be accounted for by genetic differences among Heritability
individuals. In twin studies, if MZ twins correlate with each other by 1.00 and The proportion of variance
if DZ twins correlate with each other by 0.50, then heritability is 100 percent. in a trait, in a given popula-
tion, that can be accounted
A rough estimate of heritability in a twin study can be made by doubling the
for by genetic differences
difference between the MZ and DZ correlations (Plomin et al., 2001). The among individuals.
degree to which an MZ correlation is less than the perfect 1.0 is assumed
to reflect unshared environment—i.e., those differences that distinguish Unshared environment
The proportion of variance
between twins (e.g., different peers, different illnesses). The remaining portion
in a trait, in a given popula-
of variance is attributed to shared environment (e.g., family socioeconomic tion, that can be accounted
status, common parenting). for by events that happen
Figure 6.13 shows heritability estimates of various psychological abilities to one twin but not the
and conditions. It is to be noted that these are comparable to many heritability other, or events that affect
estimates from non-psychological traits and functions that we might naively them in different ways.
expect to be “more biological.” In a meta-analysis of 50 years of twin studies, Shared environment
Polderman et al. (2015) calculated the heritability of cognitive abilities to be The proportion of variance
.468 and for psychiatric conditions to be .463. This contrasts with .436 and in a trait, in a given popula-
.545 for cardiovascular and respiratory functioning, respectively (of course, tion, that can be accounted
for by events that happen
the brain is as much a part of our biology as our heart and lungs).
to both twins, affecting
them in the same way.
132 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

FIGURE 6.13: The approximate heritability of various psychological abilities and conditions:
attention de!cit and hyperactivity disorder, ADHD (Eaves et al., 1997); schizophrenia
(Gottesman, 1991); dyslexia (Hawke et al., 2006); autistic traits (Hoekstra et al., 2007); IQ
(Bouchard & McGue, 1981); spatial visualization, memory, verbal fluency (Nichols, 1978);
reading ability in elementary school (Thompson et al., 1991); creativity (Nichols, 1978); and
social phobia (Kendler et al., 1992).

The concept of heritability, although useful, is easily misunderstood.

It measures how much variability is due to genetic factors within a given
population, not the contribution it makes in a given individual. If the
heritability of height is about 0.69 (Hemani et al., 2013), it doesn’t mean that
69 percent of a person’s height has come from their genes and 31 percent
from their environment. It means that 69 percent of the differences in height
between different people, within that population, are due to their genes. To give
another example, most people have 10 !ngers and this is genetically speci!ed.
However, the heritability measure for number of !ngers is low, because
the variability in the number of !ngers is due primarily to environmental
reasons—industrial accidents and so on (Ridley, 2003).
To consider a cognitive example, the fact that heritability for reading ability
in elementary school pupils is 0.30 (Thompson et al., 1991) does not mean that
30 percent of a child’s ability is due to genes and 70 percent due to environment.
Reading requires an appropriate environment (i.e., living in a literate culture),
otherwise literacy will not exist at all. It also requires an appropriate brain
architecture that will support reading. Both are equally essential. The measure
of heritability may also vary according to the population studied. If one were to
measure reading ability in a country in which education was not universal, then
heritability would almost certainly be lower because reading ability would be an
outcome of opportunity, i.e., an environmental factor. It is curious, but true, that
the more that our society is based upon equal opportunities, the more that genetic
differences will matter (proportionally speaking). To give one !nal example, the
heritability for reading disability, or dyslexia, in western societies is higher, at 0.60
 THE DEVELOPING BRAIN 133

FIGURE 6.14: Correlations in

gray matter density between
pairs of MZ twins (left
column) and DZ twins (right
column) for three different
views of the brain (F = frontal
pole, W = Wernicke’s area,
S/M = somatosensory/motor
cortex).
Thompson et al., 2001.

(Hawke et al., 2006), than for reading ability per se, because the diagnostic criteria
for dyslexia typically assume adequate opportunity and intellect; i.e., variability
in environmental factors is minimized by the selection criteria.
Heritability estimates can also be applied to structural and functional brain
differences as well as to cognitive and behavioral traits (Jansen et al., 2015). For
instance, Figure 6.14 shows the correlation in localized gray matter volumes
between MZ and DZ twins (heritability is, of course, related to the difference of
the two sets of correlations). The correlations tend to be far greater in MZ twins.
Heritability estimates for total brain volume are 0.94 (Bartley et al., 1997). For
cortical surface area they range from .30 to .43 depending on the region (Chen et
al., 2012), and for gray matter and white matter density using VBM they range
from .69 to .85 (Hulshoff Pol et al., 2006). The Human Connectome Project
aims to have detailed MRI and MEG scans from 1,200 people comprising sets
of twins (Van Essen et al., 2013) and initial research has shown that whole
brain measures of functional connectivity have a far stronger heritable (genetic)
component than do effects of shared environment (Colclough et al., 2017).
There was still a sizeable influence of non-shared environment, which could
include things such as measurement error and variability in implementing the
genetic program which, as noted before, does not specify a precise connectome
but provides guiding constraints for its construction.

Linking genetic differences to brain and behavior

Heritability is a statistical measure that doesn’t say anything directly about
particular genes or their function. In order to do that one needs to link relevant
data from cognitive neuroscience with individual differences in the genetic code
itself (i.e., variations in the DNA sequence). Taking cheek swabs remains the
 134 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

most common and most simple way of extracting

cells for human genetic analysis (Figure 6.15).
Cells on the inside of the cheek are loose and can
be removed by light abrasion with a swab. Recall
that the genetic code is the same across all the
cells in the body so it makes no difference whether
it is extracted from a cheek cell or a neuron
itself. Sterile kits are cheaply available and swabs
can be obtained remotely with participants
returning their DNA sample via post. The two
approaches that one could adopt for analysis
of genetic differences are called genotype-first
and phenotype-first. In conducting this kind
of research it is important to obtain accurate
information about the ethnicity of your genetic
FIGURE 6.15: DNA can be obtained from cheek cells using sample (or have a homogeneous sample) as the
commercially available swab kits. The sequence of DNA in an
prevalence of many polymorphisms can vary
individual is the same in all the cells of their body—so cheek
cells are as useful as neurons for this purpose.
considerably depending on race.
Science History Images / Alamy Stock Photo
An example of a genotype-!rst approach
would be to take a single gene that is known to
exist in multiple variants (polymorphisms) and that may be relevant to a given
research question (i.e., based on previous research). As an example from the
literature, oxytocin is a hormone involved in prosocial interactions (including
falling in love). Oxytocin binds to receptors in the brain which are coded by a
gene with a relatively common mutation (termed “A” and “G” variants that reflect
a single difference in the DNA sequence). It is possible to group participants
according to their genotype (AA, GG, AG) and then show that they differ on
behavioral measures such as parenting style (Bakermans-Kranenburg & van
Ijzendoorn, 2008) or neural responses, in fMRI, to rewards (Damiano et al.,
K EY T ERMS
2014). The advantage of this genotype-!rst approach is that the genetic analysis
Genotype-first is limited to one speci!c gene and so is relatively straightforward to conduct
An analysis approach in (commercial companies can provide this service to the neuroscience community).
which different genotypes
It also avoids the problem of multiple comparisons when testing multiple genes
(e.g., different alleles)
are used to explore for (the problem of Type I errors).
phenotypic variation. An example of a phenotype-!rst approach would be to take a given
trait that is known to vary in the population (e.g., empathy) or to take
Phenotype-first
An analysis approach in a clinically de!ned condition (e.g., autism spectrum disorder) and to
which different pheno- determine which portions of the genome contribute most to variations in
types are used to explore that trait (as a continuous measure) or the presence/absence of a condition
genetic differences. (as a binary measure). One method that uses this approach is termed a
Genome-wide genome-wide association study (GWAS) and tends to involve many
association study thousands of participants. This method is based on the fact that there
(GWAS) are many small variations in the genome across individuals termed single
A phenotype-!rst nucleotide polymorphisms (SNPs, pronounced “snips”). These SNPs are
approach in which the not of interest in their own right but provide useful clues as to which parts
presence/absence, or
of the genome contain a “hot spot” (i.e., regions where individuals with
continuous variation, in
a trait is linked to varia-
the same phenotype have genetic similarities that deviate from chance). To
tions at many different give an example of this approach, Ma et al. (2009) studied 487 Caucasian
sites in the genetic code. families (1,537 individuals) with autism and examined their genomes using
over one million SNPs. Figure 6.16 shows the way that these data can be
 THE DEVELOPING BRAIN 135

FIGURE 6.16: This genome-

wide association study
(GWAS) of autism divides
the genome into different
regions (Chr 1 to 22 refer to
the different chromosomes)
and measures the statistical
probability (plotted on the
y-axis) that variability in
the presence of the trait is
linked to variability in that
region of the genome. The
values of 4.00 and 5.00
refer to P < 10$4 (.0001)
and P < 10$5 (.00001),
respectively.
Reproduced with permission
from John Wiley and Sons

visualized: the genome is plotted on the x-axis and (log) probability values on
the y-axis. They found 96 genetic locations (SNPs) that have P < 10$4 (.0001)
and 6 that have P < 10$5 (.00001). This suggests that genes close to these
locations are commonly implicated in the development of autism. The same
approach can be applied to brain-based data itself, although this approach is ONLINE RESOURCES
still in its infancy because suf!ciently large samples (thousands) of scanned For a link to the UK
people have not been readily available. To give one example, the UK Biobank Biobank Imaging
has health, behavioral and genetic data on half a million participants and Study, visit the
companion website
is collecting brain scans (MRI) on up to 100,000 of them. A preliminary
([Link]/
GWAS (albeit with N=8,428) revealed 148 areas of the human genome that cw/ward).
were linked to brain structure and function including those involved in axon
guidance (e.g., the ROBO3 gene affects development of midbrain tracts),
white matter repair (e.g., relevant to disorders such as multiple sclerosis)
and iron transport relevant to neuro-vascular coupling (Elliott et al., 2018).
However, identifying these genes is only the starting point. The ultimate aim
is to be able to link different levels of explanation together coherently, along
the lines of Gottlieb’s (1992) simple framework described earlier: Genes ↔
Brain structure ↔ Brain function ↔ Experience.

Mechanisms for gene–environment interplay

The new frontier of the nature–nurture debate is concerned with how genes and
environments influence each other mechanistically. In their book Rethinking
Innateness, Elman and colleagues (1996) put it this way: “The answer is not Nature
or Nurture; it’s Nature and Nurture. But to say that is to trade one platitude for
another; what is necessary is to understand the nature of the interaction” (p. 357).
At least three broad scenarios have been identi!ed (Rutter et al., 2006), namely:
epigenetics, gene–environment correlations and gene–environment interactions.
136 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

LINKING GENES TO COGNITION: THE CASE OF FOXP2, SPEECH AND GRAMMAR

In 1990, a remarkable family came to the atten-

tion of the scienti!c community. Around half of the
members of the so-called KE family had problems
in producing speech and language and, moreover,
the pattern of inheritance was consistent with a
single gene mutation (Figure 6.17). Affected family
members would come out with sentences like
“The boys eat four cookie” and “Carol is cry in the
church.” Indeed, early reports of the family sug-
gested that they may have problems in speci!c as-
pects of grammar (Gopnik, 1990; Gopnik & Crago,
1991), i.e., a potential “gene for grammar.” Since
then, the affected mutation in the FOXP2 gene has FIGURE 6.17: The family tree of three generations of
been identi!ed, the nature of the speech problems the KE family shows that around half of the members
have been described in more detail and the neural have signi!cant problems in speech and language. This
problem has now been linked to a mutation in a single
substrates have been explored in both humans gene called FOXP2. Does this gene have a role to play in
and other species (for a review, see Fisher, 2017). the evolution of human language?
While the core de!cit in the family remains Reprinted by permission from Macmillan Publishers Ltd: Watkins
et al., 2002. © 2004.
debated, the de!cits are certainly not limited to
grammar. Affected KE family members, relative to unaffected ones, score poorly on tests of pronunci-
ation, grammar, semantics, verbal IQ and even nonverbal IQ, although the scores from the two groups
overlap (Vargha-Khadem et al., 1995). Tests of oral praxis and orofacial praxis (e.g., copying tongue
protrusions or lip pouting) do produce nonoverlapping test scores, suggesting that orofacial dyspraxia
is a core de!cit. There is reduced volume in the basal ganglia (caudate nucleus) that correlates with
their level of orofacial dyspraxia (Watkins et al., 2002). The basal ganglia have a key role in the control
of voluntary movement. The basal ganglia, and the caudate in particular, have also been linked to
implicit rule learning in arti!cial grammars (Lieberman et al., 2004), suggesting a possible link to the
grammatical de!cits. Inserting different versions of the FOXP2 gene into the mouse affects the plastici-
ty of basal ganglia circuits, with human versions increasing plasticity (see Fisher, 2017). Other families
with Speci!c Language Impairment (SLI) of developmental origin do not appear to have the FOXP2 gene
affected (Newbury et al., 2002), although some of them do perform poorly on grammar in the absence
of orofacial dyspraxia (Falcaro et al., 2008). As such, there are likely to be multiple genes that affect
grammar and, at present, there are no known genes that speci!cally affect only grammar.
What do studies of the normal version of the FOXP2 gene reveal about its possible function?
The product of the FOXP2 gene is what is called a transcription factor, i.e., its molecular func-
tion is to affect the expression of other genes. As such, its effects may be wide-ranging and it is
expressed in various tissues in the body, not just the brain. Haesler et al. (2004) found that FOXP2
expression in birds who need to learn their vocalization (e.g., canaries) had greater expression in
the avian equivalent of the basal ganglia during song learning than song production. Intriguingly, the
FOXP2 proteins of chimpanzees, gorillas and rhesus macaques are identical to each other but differ
from humans in terms of two small sequence changes, one of which is likely to be functional and
has been dated to 200,000 years ago, about the time that anatomically modern humans emerged
(Enard et al., 2002). The !nal word concerning the function of this gene has yet to be written.
 THE DEVELOPING BRAIN 137

Epigenetics KE Y T ER MS
Although the structure of the genetic code for each person is !xed at Orofacial dyspraxia
conception, the functioning (or “expression”) of the genetic code is highly An impaired ability to
dynamic. Different genes become active or inactive at different times of life perform the coordinated
(e.g., causing us to go through puberty, or our hair to turn gray). The expression movements that are
required for speech.
of the genetic code is also influenced by the environment—a phenomenon
termed epigenetics. In epigenetic marking the genes are not changed but Transcription factor
get tagged with a chemical marker that dampens (e.g., a methyl group) or A gene product that
affects the function of
accentuates (e.g., an acetyl group) their expression. At present, epigenetic
other genes.
markers have to be explored in vitro. A lack of availability of human brain
tissue, and the dif!culty in linking it to cognitive or behavioral measures, has Gene–environment
resulted in limited progress in humans (Miller, 2010). It is also unclear how correlations
Genetic influences in
epigenetic differences that can be more easily measured in the DNA of other
people’s exposure to
tissues (e.g., blood cells) would relate to those of neurons. Thus, epigenetic different environments.
differences tend to be studied in animal models (e.g., Meaney, 2001).
One well-known example of epigenetic effects concerns
the influence of abuse or poor caregiving (neglect) in early *
development. In rats, mothers vary in the amount of care
n.s.
(licking and grooming) given to their pups. Low care is
1.8
linked to an increased, and lasting, stress response in the
pups to both neutral and stressful events (e.g., Meaney, 1.6
2001). The effect has been related to an epigenetic reduced
GRtotal/GAPDH (log(conc))
1.4
expression of a gene coding for a glucocorticoid receptor,
1.2
leading to an increased stress response (Weaver et al.,
2004). In humans, McGowan et al. (2009) examined the 1.0
postmortem brains of people who had committed suicide 0.8
versus control brains. They found epigenetic influences on
the gene coding for the glucocorticoid receptor in suicide 0.6
victims who had experienced early neglect/abuse, but this 0.4
was not found on the other suicide victims or the controls— 0.2
see Figure 6.18. This con!rms that the epigenetic effects
were linked to early abuse/neglect rather than other factors 0
Control Suicide Suicide
contributing to suicide. nonabused abused

FIGURE 6.18: Abuse in early childhood, rather than

Gene–environment correlations (rGE) suicide-related mental illness, is linked to lower
Gene–environment correlations (rGE) are genetic expression of the stress-related glucocorticoid
influences on people’s exposure to different environments receptor gene in the human hippocampus. This is
evidence of an influence of the social environment
(Plomin et al., 1977). For example, people will seek out
on the functioning of the genome, i.e., epigenetics.
different environments (e.g., drug taking and novelty
From McGowan et al. (2009).
seeking) depending on their genotype (Benjamin et al.,
1996; Kotler et al., 1997). Benjamin et al. (1996) reported a link between long
versions of the D4 dopamine receptor gene and personality questionnaires
that measure novelty seeking (Figure 6.19) and extraversion (but linked
negatively to conscientiousness). However, in order to properly understand
these results that link genes and behavior it will be important to situate them
within models of brain function which, perhaps surprisingly, remains to be
done (Robbins, 2018). There is unlikely to be a simple mapping between
neurotransmitter function and behavior.
 138 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

40
Gene–environment correlations have often been
studied in the context of parenting styles and this has
led to the notions of evocative and passive processes (see
Percent of Subjects

30
Figure 6.20). An evocative process, in this context, is one
20 in which a child’s negative behavior leads to a harsher
parenting style. A passive correlation is that both a risk
D4DR Genotype
10 for negative child behaviors and negative parenting
S
behaviors are transmitted genetically. Although these
L
0
-36- -42- -48- -54- -60- -66- -72- -78- differences are hard to pull apart in most families, there
are some “natural experiments” that lend themselves to
Estimated TPQ-Novelty Seeking Score
this approach. For instance, adopting a child or assisted
FIGURE 6.19: The D4 dopamine receptor gene (D4DR) fertility with donor embryos eliminates the possibility
exists in short and long (S, L) polymorphisms and the long of passive correlations but provides positive evidence
variant is linked to higher questionnaire scores of novelty of evocative rGE effects where the child’s behavior
seeking (TPQ = tridimensional personality questionnaire). influences parenting style (Harold et al., 2013).
From Benjamin et al. (1996)

(a) Environmental causality model Gene–environment

interactions (G x E)
Environment Psychopathology
Gene X environment interactions (G
x E) occur when susceptibility to a trait
(b) Passive rGE depends on a particular combination
Inheritance
of a gene and environment. The
Parent’s genotype Child’s genotype effects of the gene and environment
together exceed what would be
expected from the sum of the parts.
Heritability:

Heritability:
behavior
parental

behavior
child’s

Three classic demonstrations of G

x E were reported in the early 2000s
that attracted much attention at the
Parenting or parent Child’s time, but have subsequently led to
psychopathology psychopathology
controversy. These were interactions
between childhood abuse and
Correlation reflecting passive rGE
variants of monoamine oxidase A
gene in eliciting antisocial behavior
(c) Evocative rGE
(Caspi et al., 2002); interactions
Child’s genotype between short and long versions
of a serotonin transporter gene
and adverse life events in eliciting
Heritability:
behavior
child’s

depression (Caspi et al., 2003); and

Evoking response interactions between the COMT
gene and cannabis use in eliciting
Environment Child’s symptoms of schizophrenia (Caspi
reacting to child psychopathology et al., 2005). All studies involved a
Reaction of environment sample of over 1,000 consecutive
births from Dunedin, New Zealand,
FIGURE 6.20: Different ways in which child psychopathology and environment
that have been systematically studied
might be related. (a) the environment directly causes psychopathology
(irrespective of genetics); (b) both the parent and child have similar genotypes
at subsequent time points (between
that predispose toward certain behaviors; (c) the child’s genotype predisposes 3 and 45 years, so far). Hence, this
toward both psychopathology and predisposes toward particular parenting sample have had a detailed assessment
responses. of their health and environment
From Knafo and Jaffee (2013).
 THE DEVELOPING BRAIN 139

throughout life (Poulton et al., 2015). These studies are described briefly KE Y T ER M
below and then reevaluated in terms of subsequent evidence and debates.
Caspi et al. (2002) reported that children with the low-activity (L) variant Gene X environment
interactions
of the monoamine oxidase A (MAOA) gene who were maltreated are more
Susceptibility to a trait
likely to show antisocial behavior as an adult. That is, the effect of having depends on a particular
this combination of gene and environment exceeds what one would expect combination of a gene
from the simple sum of the effect of the gene alone and the effect of the and environment.
environment alone. The gene codes for an enzyme involved in the metabolism
of dopamine, norepinephrine and serotonin. A different mutation in this
gene had previously been found in a large Dutch family with a strong history
of violence in its male members (Brunner et al., 1993). The effect is more
pronounced in men because the gene is coded on the X-chromosome. Men
(XY) have only a single copy of the gene whereas women (XX) have two
copies, one of which can compensate for the other.
Caspi et al. (2003) examined the serotonin transporter gene (5-HTT) that
occurs in two variants termed short and long. Carriers of the short allele
show a higher prevalence of depression and suicidal thoughts following a
negative life event (such as divorce).
In their third study, Caspi et al. (2005) reported a gene X environment
interaction between variants of the COMT gene and cannabis smoking in
triggering symptoms linked to schizophrenia (e.g., hearing voices that others
couldn’t hear). A common mutation at one place in the COMT gene leads
to the substitution of an amino acid (from valine, Val, to methionine, Met),
such that the presence of a Val allele is linked to more ef!cient dopamine
breakdown. The presence of the Val/Val genotype together with the smoking
of cannabis during adolescence was linked to greater report of symptoms
at age 26 (see Figure 6.21). The effect was not
found for cannabis smoking at a later age,
suggesting a sensitive period.
It is not hard to see why these studies
attracted so much attention. In all three examples
these were essentially “normal” genes (i.e.,
present in a large proportion of the population)
that, given a particular environment, leads to an
increased vulnerability of a psychiatric problem
but without a need for simple reductionism
(e.g., a “gene for” depression) or determinism
(i.e., that an outcome is guaranteed). So why
the subsequent controversy? The original
studies relied on the identi!cation of candidate
genes (a genotype-!rst approach) but this
approach has largely been overtaken by GWAS
(a phenotype-!rst approach). Ideally the two
approaches should converge to give the same FIGURE 6.21: The COMT gene is involved in the metabolism
answers but the results of GWAS have tended of the neurotransmitter dopamine and the gene exists in two
not to con!rm the involvement of the genes that main forms (termed Val and Met). Each person has two copies
were expected from this earlier research, despite of the gene. If you have a Val copy of the gene and you smoke
cannabis during adolescence, then there is an increased risk
having far larger sample sizes (for a discussion
of displaying symptoms of schizophrenia at age 26—a gene X
see Dick et al., 2015). Others have taken a environment interaction.
similar phenotype-!rst approach as the original Reprinted from Caspi et al., 2005. © 2005, with permission from Elsevier.
140 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

authors (e.g., regarding schizophrenia and COMT) and whereas some groups
replicate the original !ndings (Nieman et al., 2016), others do not (Zammit
et al., 2007). However, in support of the original claims, there has also been
converging evidence from experimental paradigms exploring the effects of
these genetic differences. For instance, carriers of the different versions of
the MAOA gene (high or low activity) but from normal environments, i.e.,
without a history of maltreatment, show structural differences in regions
such as orbitofrontal cortex as well as functional differences when processing
fear and anger (Meyer-Lindenberg et al., 2006).
Although the status of these particular !ndings remains for future research
to corroborate (or de!nitively discard), there is a general consensus in the
literature that gene X environment interactions do exist. For instance, Dick et al.
(2015) argue that G X E for single genes are likely to be linked to small effect sizes
(because the genetic influences from GWAS are of this order of magnitude),
but that the overall effects of multiple genes could be large. This is effectively
what the earlier research on heritability estimates showed (i.e., the degree of
heritability can vary according to environment) but where heritability was a
summed estimate of all genetic effects. Others have argued that inconsistency
in the G X E results reflects inconsistencies and uncertainties in how to measure
relevant environmental influences (Rutter, 2012). It may depend on the severity
of the environmental trigger or the timing of it (e.g., during a sensitive period)
in ways that are not fully understood or that were not always taken into account
by subsequent replication attempts. Belsky and Beaver (2011) have argued that
previous research has incorrectly over-emphasized the importance of adverse
environments but that, instead, some genetic influences may make people more
susceptible to environmental influences per se (both good and bad ones). They
conceptualize this in terms of genetic differences in environmental plasticity
(how this concept relates to neural plasticity is unknown). For example, Figure
6.22 shows the relationship between self-regulation in adolescents and quality
of parenting as a function of how many “plastic” genes are present (including
the 5-HTT and MAOA polymorphisms discussed already). Note that the G X
E interaction extends to both sides of the graph, i.e., people with more of these
genes are more sensitive to parenting style per se, whether the parenting style is
unusually good or unusually bad.

Evaluation
Twin and adoption studies have provided an important means for estimating
how much variability in a trait is due to genetic influences (heritability) or
environmental influences. These approaches can be applied to data from
cognitive neuroscience (e.g., inter-individual variability in structural and
functional MRI scans) as well as traditional behavioral genetic approaches.
Heritability estimates, however, can’t be used to infer causal mechanisms
(i.e., to determine whether a trait is caused by genes or environment). There
are several mechanisms by which genes and environments interact including
epigenetics (changes in gene expression), gene–environment correlations
(genetic influences on people’s exposure to different environments) and gene–
environment interaction (in which a combination of a gene and environment
together are crucial). These can now be studied at the whole genome level
using large participant databases shared amongst the research community.
 THE DEVELOPING BRAIN 141

4 or 5 alleles
45
Level of Self-Control/Regulation

3 alleles
40
2 alleles

35

0 or 1 alleles

30

25

20
−3 SD −2 SD −1 SD Mean +1 SD +2 SD +3 SD
Parenting Quality

FIGURE 6.22: There is a relationship between parenting quality and level of self-control/
emotion regulation in adolescents but the strength of that relationship depends on the
child’s genetic predisposition (number of genes that are found to have a more “plastic”
allele: 0/1, 2, 3 or 4/5).
From Belsky and Beaver (2011).

SUMMARY AND KEY POINTS OF THE CHAPTER

• Structural changes in the brain occur throughout life. Genes not

only influence prenatal development (e.g., the formation of large
numbers of synapses around birth) but have a lifelong influence.
Genes act as a guiding influence for the development of the brain
but do not provide a detailed “blueprint.”
• There is good evidence from both animal studies and human
research for sensitive periods in development in which the brain is
optimally readied to acquire certain skills or knowledge. However,
these periods may be more flexible than were !rst thought.
• The word “innate” can mean at least two different things: either
referring to instincts that have been shaped by natural selection
(language being one candidate); or referring to knowledge/skills/
resources acquired in the absence of experience (candidates
include certain preferences and dislikes, and the existence of
orientation-selective cells in the visual cortex).
• Twin and adoption studies provide one way of showing that there
is a genetic contribution for a given trait. However, the concept of
“heritability” is not a pure measure of genetics because the amount
of variance in a trait that is due to genetic factors will also depend
on the amount of variance that is due to non-genetic factors.
 142 THE STUDENT’S GUIDE TO COGNITIVE NEUROSCIENCE

• Genetic disorders can often affect some cognitive abilities

more than others. However, genes are rarely highly speci!c for
psychological/cognitive traits. This is because genes interact
ONLINE RESOURCES with other genes, and there is a complex interplay between
Visit the companion genes and environment. Sometimes a gene makes certain
website at www. environments more likely to be sought out (a gene–environment
[Link]/cw/ward correlation), and sometimes a gene leads to a susceptibility that
for:
also depends crucially on environmental circumstance (a gene–
• References to key environment interaction).
papers and readings
• Video lectures and
interviews on key
topics with leading EXAMPLE ESSAY QUESTIONS
psychologists
Charles Nelson,
• What is the evidence for sensitive periods in development and
Elizabeth Spelke and
David Hubel, and what kinds of neural mechanism could give rise to them?
author Jamie Ward • To what extent can any kind of behavior be said to be innate?
• Multiple-choice • What is meant by the term “heritability” and how can it be
questions and measured?
interactive flashcards • How might gene–environment interplay contribute to
to test your
developmental and psychiatric disorders?
knowledge
• Downloadable
glossary
RECOMMENDED FURTHER READING

• Goswami, U. (2008). Cognitive development: The learning brain.

Hove, UK: Psychology Press. An introductory textbook on all
aspects of development, including evidence from developmental
cognitive neuroscience.
• Johnson, M. H., & De Haan, M. (2015). Developmental cognitive
neuroscience: An introduction (4th edition). New York: John Wiley
& Sons. A good introductory overview.
• Nelson, C. A., & Luciana, M. (2008). Handbook of developmental
cognitive neuroscience (3rd edition). Boston, MA: MIT Press. A
comprehensive selection of advanced topics.