Content uploaded by Marc D Lewis
Author content
All content in this area was uploaded by Marc D Lewis on Apr 09, 2020
Content may be subject to copyright.
0273-2297/$ - see front matter 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.dr.2005.10.006
Developmental Review 25 (2005) 252–277
www.elsevier.com/locate/dr
Self-organizing individual diVerences
in brain development
Marc D. Lewis
University of Toronto, 252 Bloor St. West, Toronto, Ont., Canada M5S 1V6
Received 30 September 2005; revised 20 October 2005
Available online 13 December 2005
Abstract
Brain development is self-organizing in that the unique structure of each brain evolves in unpredict-
able ways through recursive modiWcations of synaptic networks. In this article, I review mechanisms of
neural change in real time and over development, and I argue that change at each of these time scales
embodies principles of self-organizing systems. I demonstrate how corticolimbic conWgurations that
emerge within occasions lay down synaptic structure across occasions, giving rise to individual trajecto-
ries that become entrenched with age. Emotions have a powerful inXuence on this process.This is
because the neural processes mediating emotion consolidate patterns of activation across the brain,
through their enhancement of inter-regional coordination in real time and their contribution to synaptic
shaping over development. The loss of corticolimbic plasticity with age is an unfortunate fact of devel-
opment, but it is compensated in part by transitional phases and individual learning experiences
through which habits are modiWed or replaced. I emphasize variations in inter-systemic coupling as a
key mediator of developing individual diVerences, and I discuss the acquisition of anxious/depressive
appraisals as an example.
2005 Elsevier Inc. All rights reserved.
Keywords: Individual diVerences; Development; Neurobiology; Self-organization; Dynamic systems; Emotion;
Synaptic change
Brain development has come to be conceptualized by many investigators as a process
of self-organization. This view is based on a number of features: feedback between neu-
ral regions promoting long-term coupling and synchronization (Edelman, 1987), func-
E-mail address: mlewis@oise.utoronto.ca.
M.D. Lewis / Developmental Review 25 (2005) 252–277 253
tional selection of cells or synapses through competition and elimination (Changeux &
Dehaene, 1989), interplay between growth and elimination in the sculpting of cortical
tissue (Greenough & Black, 1992), the convergence of structure from initial plasticity
through corticolimbic resonance (Tucker, 1992), and recursive cycles of cortical coher-
ence corresponding with developmental stages (Thatcher, 1998). Some of these processes
and phenomena have been studied through sophisticated modeling techniques. These
include neural network models that demonstrate the development of concepts over
repeated trials (e.g. Elman et al., 1996). Some of these models speciWcally highlight mech-
anisms of feedback, synchronized oscillations, and other neurally inspired concepts (e.g.
Grossberg & Somers, 1991). Other network-type models informed by neural plausibility
are predicated on emergent cognitive phenomena such as conceptual coherence (Tha-
gard & Verbeurgt, 1998). Some theorists conceptualize neural development in terms of
constructivist ideas that cross the line, either implicitly or explicitly, into fundamental
precepts of self-organizing systems. For example, Johnson (2000) identiWes broad struc-
tural constraints in cortical tissue as the basis for progressive speciWcation of structure
through interactive processes. Still others integrate constructs from cognitive science and
philosophy with critical mechanisms of biology and/or evolution to specify a role for
self-organization in the neural basis of cognition (Thompson & Varela, 2001; Varela,
Thompson, & Rosch, 1991). Finally, a few neuroscientists integrate the basic language of
dynamic systems theory with detailed research on brain activity (Bressler & Kelso, 2001;
Freeman, 1995).
These theorists have made immense contributions to our understanding of mechanisms
of neural function and neural development at a variety of scales and across a wide spec-
trum of neural entities from synapses to macroscopic systems. Yet, few of them speak to
issues of direct relevance to developmental psychologists, and none focus primarily on
individual diVerences in development. Neuroscientists and those in related disciplines are
most concerned with modeling basic mechanisms of coordination among multiple units,
the biological or computational properties of the units themselves, and the general princi-
ples by which brains and artiWcial networks learn and evolve. In keeping with this focus,
their interest in development is mostly restricted to normative processes, shared by all
humans and often by nonhuman animals as well. The present article is intended for devel-
opmental psychologists, particularly those interested in social, emotional, and personality
development. Moreover, it attempts to explain the vast proliferation of individual traits,
capacities, styles, and even pathologies that characterize the socioemotional domain.
Developmental trajectories have been modeled as self-organizing processes for nearly 20
years (e.g. Fogel, 1993; Keating, 1990; Lewis, 1995; Magai & Haviland-Jones, 2002; Thelen,
1990). It seems important to integrate this perspective with current knowledge about the
brain, to specify a realistic biological foundation for detailed modeling of developmental
processes.
In the following sections, I look at data on brain function and brain development
acquired from neurobiological research, and integrate and interpret these data in relation
to principles and mechanisms of self-organization, broadly conceived. In doing so, I pay
particular attention to the role of emotion in organizing activity patterns spanning multi-
ple neural systems, both within occasions and over development. Neural substrates of emo-
tion can be seen to inXuence structural changes underlying all domains of development.
However, the study of social and personality development demonstrates their power most
clearly.
254 M.D. Lewis / Developmental Review 25 (2005) 252–277
Principles of developmental self-organization for neuroscience
Self-organizing processes—or, more broadly, the activities of any complex adaptive
system—become easier to conceptualize once we sort them out into “micro” and
“macro” time scales. Thelen and Ulrich (1991) provided an important heuristic for this
kind of analysis by demonstrating how real-time processes such as infant stepping Wt
developmental trajectories such as learning to walk. Thelen and colleagues went on to
show that behavior converged to attractors in real time while development could be
described as the formation and disappearance of those very attractors (Thelen & Smith,
1994). A number of theorists discuss reciprocal inXuences between microscopic and mac-
roscopic time scales. SpeciWcally, real-time processes give rise to developmental trajecto-
ries, and developmental trajectories constrain the activities of a system in real time. How
does this work? In the Wrst case, real-time activities converge to a particular pattern that
lays down traces that facilitate the emergence of the same pattern in the future. To lay
down traces is to permanently alter the structure of the system. For example, streams of
rainwater that self-organize during each thunderstorm create trenches through one’s
Xowerbed over weeks and months. In the second case, developmental structuration—
that is, the consolidation of structure over development—alters the elements in a system
and their relation to each other. These alterations determine the system’s propensities—
what it can and cannot do, and what it tends to do, on each occasion. In dynamic systems
terms, we can say that developmental structure determines the state space of possibilities
in the moment—in real time. The contours of the Xowerbed, sculpted over many thun-
derstorms, constrain the movement of rainwater as it begins to Xow and pool during
each storm.
Brain activities can also be divided into time scales. In real time (milliseconds, seconds)
the Wring patterns of diverse neuronal assemblies become synchronized, cell groups trans-
mit information to each other, and the activities of the brain literally cohere as a particular
cognitive or behavioral act emerges from multiple possibilities (e.g., Lutz, Lachaux, Mar-
tinerie, & Varela, 2002; Thompson & Varela, 2001). In thermodynamic terms, real-time
self-organization couples the metabolic activities of many component neurons, maximizing
the Xowthrough of energy in a far-from-equilibrium system (Prigogine & Stengers, 1984). I
have modeled real-time neural self-organization in detail elsewhere (Lewis, 2005). But how
should we characterize neural self-organization at the developmental scale? Like other
developmental processes, brain development constitutes the increasing speciWcation of
structure, permitting activities that are more organized, more eYcient, and more complex.
For this process to be described as self-organizing in a strong sense (unlike, say, the devel-
opment of one’s nose), neural development must achieve forms that were initially indeter-
minate, based on the (real-time) activities of the brain itself. There are parts of the brain
that change little from infancy to old age. However, the cerebral cortex changes massively:
the cortex is continually reorganized through the activities of daily living (Nelson, 2000),
and the laying down of cortical structure is highly individualistic rather than prespeciWed
(Johnson, 1998). Developmental change in the cortex constitutes a literal reorganization of
the connections (or synapses) between neurons, and these connections evolve and stabilize
based on the activities of the neurons themselves. Hence, cortical development truly is self-
organizing, and the massive changes that take place in the cortex alter activity and struc-
ture across all other brain regions that interact with it. In this sense, the entire brain self-
organizes with development.
M.D. Lewis / Developmental Review 25 (2005) 252–277 255
There are several other principles of developmental self-organization that can be
applied to the brain. First, the formation of structure through self-organization means that
systems lose degrees of freedom as they develop. This concept can appear diYcult, but it is
really very simple. When you move into a new house and look at the empty living room,
you can imagine many ways for that room to take shape. Once the movers have put down
the furniture, there are fewer possible rooms that can be imagined. Then, when the sofa is
placed against a particular wall, there are fewer still; and Wnally when the carpet is laid
down and the book shelf put up, there may be only one conWguration that “works,” despite
ongoing minor variations. All developmental processes work this way. A 4-year old child
has not yet developed a learning style, but that same child at 14 has already crystallized
habits for attending, learning, and remembering. Personality development epitomizes the
loss of degrees of freedom with development. Parents cannot predict what their child will
be like as an adult. But meet an adult friend after several years and she will be much as you
remembered her in college. The degrees of freedom in her personality development have
been mostly used up. Note that this does not imply a loss of Xexibility in real time! There
are many ways to play a baseball game, but the rules of the game change little from season
to season. In the cortex and limbic system, the stabilization of developmental structure is
achieved through synaptic sculpting. As will be detailed later, synaptic structure is self-per-
petuating and hence increasingly determinate (as opposed to indeterminate). Moreover,
synaptic pruning consolidates developmental stability, by getting rid of the vast, proliferate
web of under-used Wbers that provided all those degrees of freedom in the Wrst place. By
getting rid of what is not being used, pruning stamps a special kind of permanence on the
synapses that remain.
A related principle is what I call cascading constraints (Lewis, 1997). The emergence of a
structure (e.g., a schema, skill, or belief) at any point in a developmental sequence con-
strains the characteristics of the structure to emerge next, and so on. Learning habits that
emerge in elementary school constrain the academic functioning of the child in high
school, and study habits in high school constrain ambitions and accomplishments in uni-
versity. Putting the sofa against one wall constrains the available locations for the book-
shelf. The idea of cascading constraints helps make sense of developmental trajectories.
Structures appearing early in development limit the possible features of later structures,
then development selects from among those features, further narrowing the path of possi-
bilities. Cascading constraints are ubiquitous in brain development. For example, pro-
longed stress in infancy leads to excessive glucocorticoid activity, resulting in synaptic
reduction and even cell death in the hippocampus and other structures. The resulting
decrements in self-regulation, memory, positive mood, and other functions can never be
completely undone, but they can be partly ameliorated by maternal nurturing in a subse-
quent phase of development (Heim & NemeroV, 2001; see Nelson, 2000, for a review).
These cascading constraints, self-organizing in their own right, constitute a set of markers
along the route of a developmental trajectory, and each plays its part in fashioning and
reWning that trajectory.
A Wnal principle is that of developmental transitions. Developmentalists with a
dynamic systems orientation have shown that transitions are periods of increased Xuctu-
ation or instability (Ruhland & van Geert, 1998; van Geert and Steenbeek, this volume;
van der Maas & Molenaar, 1992). Behaviors become unmoored from their entrenched
habits, a variety of new forms proliferate for a while (temporarily increasing degrees of
freedom developmentally), then some subset of those forms stabilizes, providing new
256 M.D. Lewis / Developmental Review 25 (2005) 252–277
habits for the next stage of development. An example is the proliferation of sentence
types shown by toddlers when they Wrst learn to talk at about 18–24 months of age. Fol-
lowing that period, speech settles down to more predictable forms, based on the domi-
nant linguistic environment. Developmental transitions are explained as reorganizations
of the structure of the system, and this is nowhere clearer than in brain development. As
detailed later, a developmental transition may take the form of a wave of synaptogenesis
followed by a wave of pruning. However, pruning itself can reorganize the brain, by forc-
ing it to utilize more eYcient routing for its synaptic traYc. The increased variability in
thought and behavior in early adolescence has been associated with massive pruning at
this age.
In this section, I have outlined principles of self-organization that can be applied to
any domain of development and demonstrated their instantiation in neural develop-
ment, at least with broad brush strokes. However, to appreciate individual diVerences
in neural development in Wner grain, it is Wrst necessary to describe how the brain works
in real time. The following section provides an account of brain function that is
consistent with a dynamic systems view, whereby conWgurations of neural coherence
self-organize rapidly, in the moment, with each cognitive or behavioral act. Following
Freeman and Tucker, I will suggest that emotion is a critical arbiter of neural self-orga-
nization in real time, and hence the forge for neural change and stabilization over
development.
Brain function in real time
The brain is the ultimate self-organizing system. In the cerebral cortex alone, approxi-
mately 20 billion cells, each with thousands of connections, provide a massive population
of interacting units in a state of continuous Xux. Despite its potential for immense noise,
chaos, or disorder, this system converges rapidly to highly ordered, synchronous states (e.g.
Thompson & Varela, 2001). Each of those states taps enormous cooperativity across the
elements in this system. In fact, this cooperativity can take the form of literal phase syn-
chrony, whereby the waveform of voltage Xuctuations in one part of the brain is locked in
step with the waveform produced by another part of the brain. Corresponding with neural
self-organization, but at a diVerent level of description, the components of cognition and
attention can be said to converge and form into coherent thoughts and plans. The various
sensory, motor, and executive systems become linked, working memory becomes engaged,
actions are selected and reWned, and so forth. Some scientists have studied the parallels
between neural coherence and cognitive coherence (e.g. Engel, Fries, & Singer, 2001;
Skarda & Freeman, 1987; Thompson & Varela, 2001), and most studies of neural coher-
ence indeed focus on phase synchrony in the cerebral cortex (usually at the frequency
range of gamma oscillations). However, as neuroscientists become increasingly interested
in emotion, they have begun to examine coherence or synchrony across subcortical as well
as cortical systems (e.g. Kocsis & Vertes, 1994; Paré, Collins, & Pelletier, 2002). They have
discovered evidence for spontaneous coupling or synchrony (at the theta frequency range)
across brainstem, hypothalamic, limbic, and paralimbic systems when animals are motiva-
tionally engaged (see Lewis, 2005, for a review & synthesis). Thus, we can assume that neu-
ral self-organization in real time is embodied by phase synchrony across multiple systems,
providing a mode of communication by which diverse brain regions can couple together to
form a coherent whole.
M.D. Lewis / Developmental Review 25 (2005) 252–277 257
To think about neural synchronization across multiple subsystems, and to establish its
importance for emotion as well as cognition, I will review the functional anatomy (with
some major pieces missing) of each of the hierarchical levels of the neuroaxis. A more com-
plete analysis would include bodily processes as well, though these go beyond the scope of
the present article. We can roughly divide the brain into four levels, each more advanced
and appearing later in evolution than the previous one. Fig. 1 provides a sketch of the hier-
archical arrangement of these levels. Fig. 2 provides anatomical detail for many of the sys-
tems included.
Fig. 1. Vertical integration across four levels of the neuroaxis. This sketch highlights the bidirectional Xow of
information that integrates functioning over the entire brain.
Fig. 2. Some of the major anatomical structures of the human brain (Lewis, 2005; used by permission).
258 M.D. Lewis / Developmental Review 25 (2005) 252–277
1. The brain stem. The shaft of nerve tissue at the core of the brain (divided into midbrain,
pons, medulla) contains sets of nuclei for programmed responses to internal and exter-
nal events. These nuclei control relatively primitive, packaged response patterns (e.g.,
defensive and attack behavior, vigilance, feeding, freezing, sexual behaviour, facial
expressions), each highly independent and stimulus-bound, and many of which go back
to our reptilian ancestors. Brainstem systems orchestrate emotional behavior even in the
absence of higher brain systems. For example, animals without a forebrain display
“sham rage”, which has the behavioral appearance of rage. Panksepp (1998) argues that
there is nothing “sham” about this rage: it exempliWes a basic emotion system function-
ing without cortical inhibition. He emphasizes that partially independent brainstem
(and striatal) circuits can be identiWed for anger, fear/anxiety, love/attachment, interest/
excitement, sadness, joy, and sexual desire—hence the brain stem is the seat of many
basic emotions, and the behavioral propensities it orchestrates may be considered the
action tendencies discussed by emotion theorists (e.g. Frijda, 1986). Critically, the brain
stem and nearby structures also produce a variety of neuromochemicals (e.g., dopamine,
norepinephrine) that modulate activity in the cortex and virtually all other brain sys-
tems. Many of these chemicals also aVect bodily systems, such that bodily responses are
prepared to correspond with brain changes.
2. The hypothalamus. In higher animals, the actions of many brainstem systems are coordi-
nated by or synchronized with activity in the hypothalamus, which sits just above them.
The hypothalamus controls the internal milieu, including the organs and vascular sys-
tems, partly through its output to the autonomic nervous system via direct axonal path-
ways and partly through the release of hormones into the blood. It also receives
information from these systems in return, thus functioning as a central regulator of
bodily responses to relevant environmental events. At the same time, the hypothalamus
complements the neurochemical output of the brain stem. It produces neuropeptides
that set body and brain systems into coherent goal-directed states, such as territorial
aggression, scavenging for food, courting and mating, and so forth. Each neuropeptide
has parallel eVects on body and brain. For example, the endogenous opiates inhibit var-
ious physiological processes to protect the animal from physical stress while providing
analgesia and soothing through their actions in paralimbic areas (areas at the interface
between the cortex and limbic system). Neuropeptide-induced states are organized at a
higher level than the more elementary and diVuse modes elicited by brainstem neuro-
modulators. Moreover, the extended release of neuropeptides and the comprehensive
action orientations they eVect may help maintain lasting emotional states or moods
(Panksepp, 1998; Potegal, Hebert, DeCoster, & MeyerhoV, 1996).
3. The limbic system. This is a rough semicircle of structures that grew out of the dienceph-
alon and evolved profoundly in mammals. These structures mediate learning and mem-
ory, whereas lower structures control perception and action according to Wxed
“programs” that required no learning. This is a critical distinction, especially when it
comes to development. Limbic and higher structures may be considered “open” in that
they change with development on the basis of experience, whereas lower structures are
considered “closed” because they change little or not at all (Panksepp, 1998). The pro-
gression from sensory input to motor output is slowed down in the limbic system, so
that responses can be Wt more precisely to the learned aspects of situations (Tucker,
Derryberry, & Luu, 2000). According to Tucker and colleagues, this slowdown corre-
sponded with the evolutionary advent of emotions, whose motivational force works by
M.D. Lewis / Developmental Review 25 (2005) 252–277 259
maintaining the focus of attention and action rather than by rapidly instigating some
“Wxed action pattern.” Indeed, the limbic system mediates emotional states that orient
attention and action to whatever is presently meaningful. The amygdala, a key limbic
structure, tags neutral stimuli with emotional content (LeDoux, 1995; Rolls, 1999),
thereby creating chains of associations based on emotional experiences. Connections
from the amygdala to lower (hypothalamic and brainstem) structures activate motiva-
tional response systems given current stimulus events, and connections from the amyg-
dala up to the cortex entrain perception and attention to these events. While the
amygdala mediates emotional memory, the hippocampus organizes episodic memory
and allows for the monitoring of one’s movements through space and time. The hippo-
campus interacts with the prefrontal cortex (possibly through the mediation of the cin-
gulate cortex—Barbas, 2000) to allow volitional cognitive activities to recruit speciWc
memories in the service of planning and strategizing. The amygdala and hippocampus
appear to function in synchrony, such that emotional associations pull for explicit mem-
ories as animals pursue plans to achieve their goals.
4. The cerebral cortex. The layers of the cortex surround the limbic system, and the
recently evolved cells that inhabit these layers are the locus of what we normally call
cognition, perception, and attention. In the cortex, the time between stimulus and
response appears to be greatly stretched out (Tucker et al., 2000). Inputs from the world
and potential actions connect with each other through a matrix of associations, compar-
isons, synthesis across modalities, planning, reXection, and sometimes, but not always,
conscious control. These operations take time, and emotions maintain a coherent orien-
tation to the world during that period of time. For example, deliberate action is guided
by attention to alternative plans, and anticipatory attention is constrained by emotions
concerning the pursuit of particular goals. Thus, cortically mediated actions are func-
tional, not only at the level of some phylogenetically ancient blueprint, but also at the
level of a continuously reWned model of the world, achieved by selecting, comparing,
and pursuing particular plans while integrating the information fed back by the world.
The cortex is also a key system for the cognitive control of emotional responses—often
referred to as “emotion regulation.” In particular, the prefrontal regions execute sophis-
ticated perceptual and cognitive activities (including attention, monitoring, decision-
making, planning, and working memory) that are recruited by (and that regulate) the
emotional responses mediated by the amygdala and lower structures (Barbas, 1995;
Bechara, Damasio, & Damasio, 2000; Davidson & Irwin, 1999).
There are two cortical systems that are especially important for integrating the cognitive
and emotional aspects of psychological functioning: the anterior cingulate cortex (ACC)
and the orbitofrontal cortex (OFC). Both regions serve as interfaces between the prefrontal
cortex and limbic system. They are therefore called “paralimbic,” and they appear to medi-
ate cognitive activities relevant to emotional states (Barbas, 2000; Rolls, 1999). As shown
in Fig. 2, the ACC is located on the medial surfaces of the PFC. ACC activation has been
associated with monitoring and evaluating potential actions, monitoring and resolving
conXicts, and selective attention more generally (Carter et al., 2000; Gehring, Goss, Coles,
Meyer, & Donchin, 1993; van Veen, Cohen, Botvinick, Stenger, & Carter, 2001). The exec-
utive system mediated by the dorsal ACC is characterized by voluntary choice and is cen-
tral for directed attention and for learning (Frith, Friston, Liddle, & Frackowiak, 1991;
van Veen et al., 2001). However, the more ventral regions of this complex system have been
260 M.D. Lewis / Developmental Review 25 (2005) 252–277
closely linked with emotional processing. The OFC, on the ventral surface of the PFC (see
Fig. 2), appears to encode and hold attention to threatening or rewarding aspects of the
environment (Rolls, 1999). Such processes are thought to extend or build onto the more
basic conditioning functions of the amygdala (Cardinal, Parkinson, Hall, & Everitt, 2002).
The OFC is responsive to changes in the hedonic valence of anticipated events (Hikosaka
& Watanabe, 2000; Rolls, 1999), and it is activated when “implicit appraisals” of motiva-
tionally relevant situations are held in mind (Schore, 1994). Its downstream connections to
the amygdala, hypothalamus, and brain stem are also integral to emotion, and its activity
has frequently been implicated in the activation and regulation of emotional states (David-
son, Putnam, & Larson, 2000; Hariri, Mattay, Tessitore, Fera, & Weinberger, 2003; Lév-
esque et al., 2003).
Mechanisms of neural change in real time
How do these neural structures interact with each other as the brain goes about its busi-
ness in real time? The hierarchy of brain levels is often construed in terms of domination or
control of lower levels by higher levels. Indeed, the cerebral cortex subordinates the more
primitive functions of the limbic system, which subordinates functions in the brain stem.
However, as emphasized by Tucker et al. (2000), the downward Xow of control and modu-
lation is reciprocated by an upward Xow of synaptic activation and neurochemical stimula-
tion (see Fig. 1). The brain stem and hypothalamus entrain limbic structures by means of
neuromodulators and neuropeptides, locking in perceptual biases and associations, and
they also recruit cortical activities to ancient mammalian and even reptilian agendas. Prim-
itive agendas and requirements thus Xow up the neuroaxis from its most primitive roots at
the same time as executive attention, planning, and knowledge subordinate each lower
level by the activities of the cortex. If not for the bottom–up Xow, the brain would have no
energy and no direction for its activities. If not for the top–down Xow, recently evolved
mechanisms for perception, action, and thought would have no control over bodily states
and behaviour. It is the reciprocity of these upward and downward Xows that links sophis-
ticated cognitive processes with basic motivational mechanisms in a rapid process of syn-
chronization called “vertical integration” (Tucker et al., 2000). Vertical integration is
hypothesized to occur whenever a signiWcant change in internal or external events triggers
an emotion and thus demands the initiation of a cognitive or motor response. Moreover, as
noted earlier, this convergent activity may be mediated by phase synchrony (e.g., in the
theta band), as diverse systems begin to resonate with each other at the same or related fre-
quencies, thus maximizing their communication.
The principle of vertical integration helps explain how multiple neural structures
become coordinated in a coherent conWguration—a uniWed whole—in a matter of
moments. This is the essence of real-time self-organization. But it is important to brieXy
describe the connections between neural structures that make this possible. Two kinds of
connections can be distinguished: the Xow of information from one neuron to the next
through normal synaptic processes, and the action of neuromodulators and neuropeptides
that enhance or inhibit these events.
Synaptic communication takes place between neurons at all levels of the nervous sys-
tem. Axons from a sender neuron release “intrinsic” neurotransmitters (primarily gluta-
mate and GABA) that cross the synapse and either increase or decrease the voltage of the
receptor neuron. This change in voltage increases or decreases that neuron’s tendency to
M.D. Lewis / Developmental Review 25 (2005) 252–277 261
Wre. Excitatory neurons increase the Wring tendency (which translates to the Wring rate) of
the receptor neuron, while inhibitory neurons decrease it. These inXuences are always jux-
taposed with the inXuences of many other neurons, such that the impact of one neuron on
another is relatively small. However, neurons belong to populations that work together,
and the compilation of excitatory or inhibitory signals from populations of neurons have a
powerful impact on the Wring rate of receptor neurons. Thus, Wring rates of neuronal popu-
lations inXuence the Wring rates of neurons downstream from them, which inXuence the
Wring rates of neurons further downstream, and so on, in a synaptic chain, and such
changes in Wring rates are the fundamental currency of neuronal communication. In addi-
tion, the tendency for neurons to fall into phase synchrony with each other allows their
inXuences to compile more eYciently, and that is why phase synchrony constitutes a basic
mode for neuronal coherence and self-organization.
The kind of neurotransmitter that supports normal synaptic transmission is manufac-
tured in each neuron and functions locally at nearby synapses. However, neurotransmitters
and neuropeptides manufactured in the brain stem and hypothalamus are released in
larger volumes, at many synapses simultaneously, far from their sites of origin. These neu-
rochemicals, which I will lump together under the term “neuromodulators,” travel to sites
all over the brain, including all limbic, striatal, and cortical areas (Fuster, 1996; Izquierdo,
1997). There they alter the inXuence of one neuron on the next, modulating Wring rates
upward or downward. Thus, the eVects of a small number of cells in the brain stem and
hypothalamus radiate outward to many synaptic sites, producing a one-to-many eVect—a
fountain of modulation. These eVects are global rather than local, and they provide a key
mechanism by which motivational concerns inXuence cognitive and perceptual processes.
For example, acetylcholine has often been linked to motivational enhancement of atten-
tion through the activation of cortical neurons (Gu, 2002). Dopamine activates approach
and exploratory behavior through its eVects on synapses in the orbitofrontal cortex and
striatum (e.g. Depue & Collins, 1999). Opiates have multiple eVects on diVerent brain
regions, but their overall impact is to make the animal less responsive to dangers in the
environment. Critically, these neuromodulatory eVects are often triggered by emotional
associations or perceptions mediated by limbic and paralimbic structures. For example,
amygdala responses mediating emotional associations send activation down “descending”
pathways to brainstem and hypothalamic centers, which in turn produce their global
eVects through “ascending” Wbers to these and other brain regions. In this way, emotions
greatly aVect information processing throughout the entire brain.
Finally, the Xow of information among neuronal populations generally assumes the
structure of feedback loops. Receptive populations of neurons send axons back to the neu-
rons from which they receive information, both locally to their immediate neighbours and
distally to structures far away across the brain. This design feature of neural tissue is criti-
cal because it promotes self-organization and vertical integration. Feedback circuits tend
to amplify small changes in a self-enhancing manner, and then, as more elements are
recruited to a particular pattern, to settle down into steady states of activation. Thus, mes-
sages from one neuronal population can recruit other assemblies to “join along,” eliciting
massive change given small perturbations. However, once enough populations of neurons
participate in the new pattern, they exert a stabilizing inXuence on each other, such that the
whole pattern endures for a prolonged period of time. Critically, the amygdala and para-
limbic cortex (ACC and OFC) are at the crossroads of all important synaptic traYc along
the neuroaxis. And it is these systems, intimately involved in processing emotion, that gen-
262 M.D. Lewis / Developmental Review 25 (2005) 252–277
erally trigger the fountain of neuromodulation originating from the brain stem and hypo-
thalamus. As a result, emotional processes enhance and extend neural feedback.
SpeciWcally, incipient emotional responses are capable of promoting sudden changes in
global neural patterning, causing a rapid switch in appraisals; emotional responses then
lead to global stabilization, causing appraisals to become entrenched for seconds, minutes,
or even hours. That is why emotions grab our attention, direct our thoughts and percep-
tions, and hold them in place until some activity (either physical or mental) intervenes and
reduces emotional activation.
The role of emotion in neural self-organization thus functions as a double-edged sword.
On one hand, self-augmenting feedback, orchestrated by limbic and paralimbic structures
with the help of ascending neuromodulators, promotes synaptic activity and hence initiates
synaptic change. In this respect, emotional processes yield novel synaptic conWgurations.
On the other hand, self-stabilizing feedback, orchestrated by the same structures and neu-
romodulators, but lasting longer and recruiting additional subsystems, consolidates pat-
terns of synaptic activity and hence minimizes synaptic change. In this respect, emotional
processes are central to the maintenance of synaptic patterning. Thus, emotional processes
cut both ways: they generate synaptic change and they maintain synaptic sameness. Earlier
I suggested that developmental self-organization relies on four principles: progressive reor-
ganization based on the system’s own recursive activities, increasing self-speciWcation and
predictability (loss of degrees of freedom), cascading constraints that narrow the range of
future options, and transitional phases. In the following sections I will show how the dou-
ble-edged sword of synaptic change and stabilization serves each of these principles,
thereby carving out individual trajectories of development.
Mechanisms of neural change in development
Some aspects of brain development are highly universal. For example, increased myeli-
nation and pruning of prefrontal cortex in early adolescence is thought to usher in the
capacity for abstract thought. However, some aspects of brain development are highly
individualistic. Even the localization of language varies among individuals, and functions
such as emotion regulation show immense variety in regional activation even in young chil-
dren (e.g., lateral asymmetries in prefrontal activation in response to emotion-eliciting
events, Fox & Davidson, 1987, 1991). As discussed in more detail later, functional diVeren-
tiation increases with development. A greater number of viable trajectories become possi-
ble and variability becomes increasingly conspicuous. What remains a mystery, however, is
the growing inertia that each developmental path accumulates over time—the eerie man-
ner in which developing humans become increasingly crystallized versions of themselves.
Why does this occur? A “brain’s eye view” of development shows that change and consoli-
dation of synaptic patterning is responsible for all psychological development—both the
laying down of normative functions such as vision, object permanence, and theory of mind,
and the laying down of individual pathways, for example the continuous branching of
unique trajectories of personality development. Work by Johnson, Nelson, and their col-
leagues provides an excellent inroad to understanding the neural basis of normative devel-
opment (e.g. Johnson, 2001; Nelson, 2000). In this article, I focus instead on the
development of individual variations.
Synaptic elaboration (proliferation and strengthening) and synaptic pruning are the two
forces that sculpt neural networks and thus shape development. These forces are deeply
M.D. Lewis / Developmental Review 25 (2005) 252–277 263
complementary, because synaptic elaboration increases the use of some synapses over oth-
ers while pruning gets rid of synapses that are under-used. In this section, I go on to
describe these and related processes as instruments of self-organization, highlight their
dependence on motivational mechanisms, and extend a model of normative cortical devel-
opment to a discussion of emerging individual diVerences.
Earlier I reviewed the dynamic systems principle that the Xow of activity among the ele-
ments of a system lays down traces, changing the structure of the elements and their con-
nections, and thereby enhancing the probability that the same patterns of activity would
recur on future occasions. Hebb (1949) applied a similar principle to explain brain plastic-
ity and learning, whereby the co-activation of neurons produced structural changes at syn-
apses between them, increasing their probability of becoming co-activated in the future. An
important class of candidate mechanisms for this kind of learning includes long-term
potentiation (LTP). In LTP, particular frequencies or durations of Wring of the pre-synap-
tic neuron produce long-term chemical changes in the post-synaptic neuron, permanently
altering the structure of the synapse to potentiate transmission in future. For this to occur,
the post-synaptic neuron must be activated, glutamate from pre-synaptic terminals must
travel to a speciWc class of receptors in the post-synaptic neuron, and then protein synthesis
must take place for a period of time (up to hours). It is generally thought that these events
must also take place across occasions for changes in synaptic structure to endure. As a
result of these events, it takes less activation to produce the same response in the post-syn-
aptic neuron on future occasions. Amazingly, such changes in synaptic sensitivity are the
basis of all learning and development, both normatively and individually. But what is the
role of emotion in this fundamental mechanism of change?
In fact, emotion and motivation Wgure strongly in LTP and related mechanisms, not
only in “social” learning but in all learning. The state of excitability of the receptive neuron
and the time course of its activation are crucial determinants of LTP. For many authors
this implies that the neurochemical excitation that accompanies emotional states is essen-
tial for synaptic modiWability and learning (Freeman, 1995; Post et al., 1998; Tucker, 2001).
Research demonstrates that neuromodulator arousal facilitates LTP (e.g. Centonze, Pic-
coni, Gubellini, Bernardi, & Calabresi, 2001; Izquierdo, 1997; Izumi & Zorumski, 1999)
and that neuropeptide action consolidates synaptic change and enhances memory forma-
tion (Adamec, Kent, Anisman, Shallow, & Merali, 1998; Flood, Baker, Hernandez, & Mor-
ley, 1990). The amygdala may be critical to memory consolidation in various systems,
because of its facilitation of brainstem/hypothalamic neurochemical release (Packard &
Cahill, 2001) and its direct projections to the hippocampus (Hamann, Ely, Grafton, &
Kilts, 1999). Moreover, phase synchrony between the amygdala and hippocampus is
thought to mediate emotional inXuences on memory formation and consolidation (Paré
et al., 2002). ACC activation, which I have described as integrating cognitive and emo-
tional aspects of attention, is thought to be critical for learning new contingencies (Cardi-
nal et al., 2002; Gemba, Sasaki, & Brooks, 1986). These and related Wndings suggest that
events that are not emotionally signiWcant may not maintain arousal or attention long
enough for learning to take place (cf. Gallagher & Holland, 1992; Lewis, 2005; Rolls &
Treves, 1998). Finally, LTP has been observed in limbic, paralimbic, striatal, and cortical
structures, but lower brain systems do not show plasticity of this kind. As a result, the
information that consolidates through LTP must derive from the attentional and evalua-
tive contents of the cortex and its limbic underpinnings, and these, as already discussed, are
highly inXuenced by present emotional states.
264 M.D. Lewis / Developmental Review 25 (2005) 252–277
But what exactly is self-organizing about synaptic shaping? The key point is that syn-
apses are strengthened the more they are used, and they are more used the more they are
strengthened. Moreover, emotional arousal functions as a catalyst for this cycle, enhancing
use within occasions and hence strengthening across occasions. This fascinating mechanism
constitutes a feedback cycle that works in developmental time rather than real time, or,
more precisely, that links real time and development. Change initiates more change and in
this sense is self-augmenting. This is how new habits get a leg up. (And note that even
learning to count to ten is a new habit, fueled by emotional states of interest and excite-
ment.) But as long as many of the same synaptic activities recur across occasions—as they
must, given synaptic strengthening—change is self-stabilizing, which means that new hab-
its quickly begin to gain inertia, and hence to win out over competing forms that are yet to
come along. This is the double-edged sword of change and stabilization to which I referred
earlier. And it is the means by which individual developmental paths self-organize—that is,
diVerentiate in unpredictable ways, based on novel structure, and then consolidate, crystal-
lizing that structure. More generally, the same double-edged sword cuts trenches in the
Xowerbed when it rains. Recursive activity creates new trenches, almost haphazardly, but
then bestows on them the mantle of permanence just because they are there. Recursive
modiWcation is thus the “blind gardener” of diverging trajectories (a takeoV on Richard
Dawkins’ “The Blind Watchmaker”). However, there is at least one important diVerence
between brains and Xowerbeds. Despite its increasing tendency to Xow into established
channels, the rainwater still has other options for its Xow. In developing brains, pruning
eventually gets rid of the other options, so the pathways that are strengthened and elabo-
rated become the only game in town.
Pruning is also a mechanism of synaptic shaping—and thus developmental change— in
and of itself. Pruning is often thought of as a complement of synaptic elaboration, because
it does indeed get rid of under-used synapses, thus entrenching patterns created by synaptic
elaboration. In this sense, pruning brings up the rear: synaptogenesis builds new roads
through the jungle, synaptic strengthening (e.g., via LTP) paves them, making them more
eYcient, and pruning gets rid of the under-used dirt roads that have now become obsolete.
However, synaptic pruning can work to sculpt neural circuits on its own. Selectionist theo-
ries (e.g., Changeux, Edelman) posit an initial repertoire of synaptic overproduction, or else
continued overproduction, which is whittled down by experience or other factors to a
greatly-reduced but much more eYcient network of connections. Changeux and Dehaene
(1989) propose a “Darwinian” model stipulating an endogenously determined overabun-
dance of synapses, some of which regress and disappear whereas others endure and
become stabilized (see Johnson, 1998, for a review). Thelen and Smith (1994) review this
account in some detail because it so nicely matches their basic developmental tenet: that
self-organizing processes in development rely on selection from initial variability. Less
research is available on the speciWc mechanisms of pruning, and it is not known to what
extent emotional processes contribute to pruning. However, there are several possibilities.
First, some mechanisms of synaptic change diminish rather than potentiate synaptic trans-
mission, including long-term depression (LTD). Like LTP, LTD is inXuenced by neuro-
modulation as well as by the frequency and duration of synaptic activity, and it is thus
bound to be inXuenced by emotion. Second, Schore (2003a) has developed an elaborate
model based on evidence that emotional stress dysregulates neural activity in infancy. By
truncating prefrontal activation during social exchanges, excessive stress results in the
shutdown of circuits for processing social information, leading to lifelong tendencies for
M.D. Lewis / Developmental Review 25 (2005) 252–277 265
blunting socioemotional behavior and avoiding novelty (Schore, 2003a). Pruning could be
one contributor to this shutdown. Third, and conversely, we can infer that more normal
levels of emotion contribute to pruning simply by strengthening synaptic pathways that
end up not getting pruned.
One particularly inXuential account of developmental change (Greenough & Black,
1992) points to two diVerent kinds of complementarity between elaboration and pruning,
depending on whether the change is normative or driven by individual experiences. In the
Wrst case, experience-expectant change involves the overproduction of synapses in anticipa-
tion of a species-general class of experiences that are particularly important to “learn” at a
certain stage of development. Leading up to this developmental window, the overproduc-
tion of synapses allows for maximal sensitivity to the impact of this class of experiences—
for example the development of stereoscopic vision in the Wrst few months of life and of
speech perception in the second half of the Wrst year. The formation of synaptic connec-
tions in a particular region of cortex encodes the most systematic or salient experiences
within the new domain, and then pruning gets rid of the left-over synapses. With experi-
ence-expectant learning, synaptogenesis is prespeciWed, and experience—though of a par-
ticular sort—is responsible for selecting the connections that will remain for a lifetime. The
other kind of relation between synaptic elaboration and pruning is called experience-depen-
dent. In this case, synapses are formed in response to experience rather than in anticipation
of experience. Synapses grow as needed to incorporate relations inherent in new experi-
ences, gleaned from the environment, but potentially unique to the individual learner. This
growth process is then followed by pruning which crystallizes the most relevant
associations and gets rid of those which are extraneous or unreliable. As a result of experi-
ence-dependent synaptic change, developmental trajectories can be characterized as “a
quiltwork of small blooms that may further regress on individual schedules” (Greenough
& Black, 1992, p. 175). Note that experience-expectant change lends itself to a model of
sensitive periods in neural development, whereas experience-dependent change is indepen-
dent of any normative developmental timing.
The impact of motivational factors on both experience-expectant and experience-depen-
dent synaptic change has not been systematically evaluated. However, it seems likely that
experience-expectant changes are mediated by endogenous diVerences in neuromodulator
systems that may be associated with temperament, and that experience-dependent changes
make use of neuromodulator activity to promote and establish synaptic networks encod-
ing experiences that are emotionally compelling (Collins & Depue, 1992). Collins and
Depue model a sequence of synaptic shaping guided by individual diVerences in neuro-
modulator activity. First, a greater endogenous endowment of dopamine-producing cells
augments synaptic shaping in cortical regions devoted to reward-seeking during sensitive
periods (experience-expectant learning). The resultant greater connectivity in synaptic net-
works in these regions, for these individuals, would then enhance the capacity for (experi-
ence-dependent) learning about particular rewards, which would again rely on dopamine
reception. In this scenario, primitive impulses to seek rewards create and elaborate a lattice
of maximal speciWcity over several phases of development, demonstrating the intrinsic role
of motivational processes in neural self-organization. For example, an energetic and socia-
ble (and dopamine-rich) infant may not only become a language user earlier in develop-
ment, but may also use his linguistic Xuency to entertain and engage adults, thereby
increasing opportunities for social rewards and social learning. Alternatively, dopamine-
supported surgency during the “terrible twos” may facilitate skills for coercing others to
266 M.D. Lewis / Developmental Review 25 (2005) 252–277
satisfy one’s frustrated goals, leading to the consolidation of an assertive or aggressive per-
sonality style.
The notion of experience-expectant change implies normative periods of synaptic prolif-
eration and synaptic pruning. These periods may serve not only as sensitive periods for
normative acquisitions, but also as phases of pronounced plasticity in which developmen-
tal trajectories can be established or altered. For example, Schore (1994, 2003b) has written
a great deal on the establishment of infant attachment patterns based on the plasticity of
the right orbitofrontal cortex in the Wrst year of life. Changes in grey-matter volumes are
the principal means for studying normative trends in synaptic proliferation and pruning.
Yet, age-related increases and decreases in grey-matter volumes vary tremendously across
the cortex (Gogtay et al., 2004). For example, synaptic density in the visual cortex is at its
peak toward the end of the Wrst year and then begins to decline steadily toward adult levels.
However, grey-matter volumes in some parts of the PFC (e.g., dorsolateral PFC) continue
to increase up to pre-adolescence, with maximum thickness occurring around 11–12 years
of age, followed by a decline that lasts into early adulthood (Giedd et al., 1999). Levels of
glucose metabolism, which may also reXect synaptogenesis and pruning, rise and fall in
frontal cortex on a roughly parallel timeline (Chugani, 1994). It thus appears that very
rapid pruning in the PFC—up to 10,000 synapses per second (Spear, 2000)!—Wnally over-
takes synaptic elaboration by early adolescence. Indeed, the transition to adolescence has
been described as a massive reorganization of PFC allowing increasing eYciency in the cir-
cuitry of decision-making, self-regulation, and abstract thinking (Steinberg et al., in press).
However, this may also be a period in which individual diVerences in interpersonal capabil-
ities, emotion regulation skills, and personality structure are at peak sensitivity and most
vulnerable to environmental inXuence (Granic & Patterson, in press). The consolidation of
these characteristics over the years of adolescence (the Wnal phase of prefrontal pruning)
may represent a loss of degrees of freedom in individual development that can never be
regained.
Modeling the nature and timing of synaptic change is crucial for understanding
developmental pattern formation. But it does not provide an overarching picture of neural
self-organization. Such a picture may be glimpsed in Johnson’s (1999, 2000) “interactive
specialization” approach. This model explains the development of cortical organization on
the basis of interconnections within and among cortical regions. The Wrst claim is that the
specialization of cortical regions (e.g., areas of the fusiform gyrus for face processing) is not
built in. Rather, it develops through usage, speciWcally the interaction of cortical cells with
each other and with other sources of sensory information (e.g., the thalamus). Given slight
intrinsic diVerences in the properties of diVerent cortical tissues, patterns of feedback
among cells are biased toward the processing of particular types of information (e.g., lin-
guistic information, visuospatial information). These diVerences in eYciency set up inter-
regional competition. Consequently, particular functions gravitate to regions that mediate
them most eYciently. The second claim is that the emergence of new skills depends, not on
the recruitment of new regions of cortex, but on the coupling of activities across regions
that are already functioning. Thus, new cognitive acquisitions are mediated by the func-
tional integration of regions that have already become specialized through competitive
processes. With respect to both claims, on-line neuronal interaction leads to network elab-
oration, giving rise to new capabilities in normal development. Although Johnson is not
speciWc on this point, it would seem that this process must depend in large part on the
mechanisms of synaptic shaping discussed earlier. What is unique to this model, however,
M.D. Lewis / Developmental Review 25 (2005) 252–277 267
is that the dedication of cortical activities to particular functions emerges within regions
initially and then across regions subsequently. Thus, the actual real-time activity of inter-
acting neurons is responsible for the initial organization and progressive reorganization of
the cortex.
Johnson’s model embodies several key principles of developmental self-organization.
The Wrst is the emergence of structure through recursive interactions among the elements
of the system. Though Johnson sometimes refers to his model as constructivist in nature,
the emergence of higher-order organization or complexity through recursive processes is
central to a self-organizational account (e.g. Lewis, 2000b). Second, the model speciWes that
structural reorganizations are typical of development, with new skills subserved by new
patterns of interaction among existing structures. Structural reorganization is a key feature
of developmental self-organization, as highlighted earlier. Third, while Johnson is more
interested in normative than individual pathways, Johnson, Halit, Grice, and KarmiloV-
Smith (2002) speculate that developmental disorders such as autism and Williams’ syn-
drome result from the progressive emergence of unique patterns of cortical specialization
rather than preexisting brain anomalies. They propose that these disorders result from the
laying down of unique conWgurations of intra- and inter-regional interactions that build on
themselves over the time course of development. This kind of emergent trajectory exempli-
Wes cascading constraints, resulting in the loss of degrees of freedom, and I will revisit these
notions in greater detail later. Finally, Johnson et al. (2002) claim that their model of corti-
cal specialization is consistent with evidence that the same functions are subserved by
smaller, more localized regions of cortical activation as children develop. Indeed, neuroim-
aging research has shown that the same tasks elicit less cortical activation in adults than in
children (Casey et al., 1997; Durston et al., 2002), and our own research (Lewis, Lamm,
Segalowitz, Zelazo, & Stieben, in press) reveals decreasing magnitudes of cortical activity
across Wve developmental waves from age 7 to 16 in a task tapping emotion regulation. We
also Wnd increasing “frontalization” (localization of function to prefrontal regions) with
age, consistent with other research (e.g. Bunge, Dudukovic, Thomason, Vaidya, & Gabrieli,
2002; Rubia et al., 2000). Thus, Johnson’s claims about increasing specialization seem to
apply very neatly to emotion regulation capacities.
It is clear from this discussion that the cortex is highly plastic and malleable. But to
what degree do subcortical structures change with experience? The hippocampus and
amygdala show neuronal branchings and synapse formation. Like the cortex, these organs
are functional only insofar as they are able to change structurally in response to experience.
They are designed for structural elaboration so that mammals can anticipate and interact
Xexibly with a changing world on the basis of learning. However, sublimbic levels of the
neuroaxis (e.g., the hypothalamus and brain stem) are not designed for learning of this sort.
Their contribution is to anchor the Wne-tuned interface with the world, elaborated in the
cortex and limbic system, with the age-old requirements and functions of the body. This is
not to say that cells in these structures do not change at all with activity. Hypothalamic
synapses change functionally in response to over- and understimulation and become more
or less responsive to physiological inputs, resulting in modiWcations of neuromodulator
release and other functional changes. As far as is known, brainstem nuclei show little if any
plasticity. Their functions are set in place prenatally and change little in response to experi-
ence. As reviewed earlier, brainstem nuclei can be considered the seat of basic emotions
such as anger and fear. When stimulated by perceptual, cognitive, and memory functions
mediated by the cortex and limbic system, these groups of cells orchestrate behavioral
268 M.D. Lewis / Developmental Review 25 (2005) 252–277
impulses and neurochemical activities that implement the action tendencies ascribed to
basic emotions (e.g. Panksepp, 2003).
Thus, the brain consists of multiple systems that are highly diverse in their structure and
function and yet become rapidly synchronized in real time, when situations are emotion-
ally compelling. In developmental time, however, these diVerent systems show very diVer-
ent patterns. The cortex and limbic system go through massive structural organization and
reorganization in response to experience. They literally self-organize with development.
These changes are partly determined by normative schedules of synaptic shaping and func-
tional specialization and partly by the playing out of unique interactions between the indi-
vidual and the world. However, the basic wiring of the hypothalamus and brain stem
changes little or not at all with experience. Thus, these systems provide a kind of anchor or
fundamental constraint to developmental self-organization. They ensure that, whatever
corticolimbic patterns are sculpted by experience, they continue to be bound by emotional
agendas to serve organismic goals and requirements. This would imply that corticolimbic
patterns supporting normative acquisitions, such as language, and individual acquisitions,
such as aggressive behaviour, shyness, and suspicion, are constrained by the requirements
of food, mating, protection, exploration and so forth embodied in basic emotion systems.
Yet the question remains: how are these constraints implemented? I suggest that they are
implemented through the requirements of vertical integration—that is, synchronization
across multiple levels of the neuroaxis—in real time. Corticolimbic conWgurations that sta-
bilize in real time do so because they provide a way for thought, behavior, and emotion to
work together in pursuit of a viable goal or strategy. What doesn’t “work” doesn’t
cohere—not even for a moment. Because the hypothalamus and brain stem are key players
in vertical integration, only synaptic activities that serve emotional requirements cohere in
the Wrst place. Hence, it is only those patterns that become articulated and consolidated,
through synaptic elaboration and sculpting, in development. To return to the rainwater in
the garden, many diVerently shaped rivulets may appear, leading to many diVerent patterns
of trenches over time. However, all viable rivulets must satisfy one basic requirement—that
the water Xows downhill. The following section maps out a model of individual diVerences
in personality and social development based on these premises.
Self-organizing individual diVerences
In the previous section, I described the shaping of synaptic networks through elabora-
tion and pruning, normative timelines of developmental plasticity, and Johnson’s model of
increasing cortical specialization, complexity, and consolidation. I also proposed a role for
emotional contributions to each of these processes and noted their implications for indi-
vidual diVerences. In this Wnal section, I speculate on the neural bases of self-organizing
individual diVerences in socioemotional and personality development. This account incor-
porates aspects of a model that was Wrst introduced several years ago (Lewis, 2000a), but it
sticks closer to what we know about the brain. The fundamental proposition is simple: cor-
ticolimbic conWgurations that last longer and recur more frequently in real time are those
that become entrenched in development. In other words, the cognitive–emotional states we
inhabit the most, in childhood and adolescence, sculpt the neural parameters that deter-
mine who we are as persons for the rest of our lives. I have already discussed the processes
of elaboration and pruning that translate real-time synaptic conWgurations into develop-
mental structure. These processes depend on emotional activation in a variety of ways, also
M.D. Lewis / Developmental Review 25 (2005) 252–277 269
discussed earlier. To review, emotional states constrain the patterning of synaptic activities
in real time, the neurochemical correlates of emotion are necessary for synaptic strengthen-
ing and memory consolidation, and the motivational functions of lower brain systems pro-
vide a relatively constant anchor for self-organizing processes in development.
Neuroscientists are just beginning to map out the neural underpinnings of personality.
Some personality diVerences have been attributed to diVerences in neuromodulator sys-
tems such as the ventral tegmental area, a source of dopamine for motivated attention
(Depue & Collins, 1999). Others have been associated with diVerences in the amplitude or
latency of cortical responses to stimuli. For example, event-related potentials (ERPs) that
tap response inhibition and self-monitoring are larger than usual for anxious, obsessive
individuals (Gehring, Himle, & Nisenson, 2000) and smaller than usual for undersocialized
individuals (Dikman & Allen, 2000). Because these ERPs are thought to originate in the
ACC, and because they generally appear in motivationally relevant conditions (Luu &
Tucker, 2002), these diVerences have been hypothesized to reXect individual diVerences in
emotion regulation (Lewis & Stieben, 2004). In a recent fMRI study, higher scores on
extraversion and neuroticism correlated with greater activation in various cortical regions,
and in the amygdala in the case of extraversion, in response to positive and negative stimuli
respectively (Canli et al., 2001). Finally, neural correlates of mood and behavior disorders
have been studied through lesion studies as well as imaging research. Davidson and col-
leagues (e.g. Davidson, 1998) have related asymmetries in frontocortical activation to
depressive tendencies, and they have hypothesized, based on these results, that left-frontal
activation is crucial for overcoming negative emotional states mediated by the right PFC.
Fox (1992) discovered similar associations between lateral asymmetry and avoidant behav-
ioral styles in infants, suggesting a neural correlate of early diVerences in temperament.
One particularly interesting set of studies Wnds that individuals suVering with anxiety and/
or depression show increased activation of ventral prefrontal cortex (e.g., OFC) (Drevets &
Raichle, 1998). In contrast, individuals with reactive aggression tend not to activate ventral
prefrontal regions such as the OFC (Blair, Colledge, & Mitchell, 2001). All of these studies,
and many others not reviewed here, come to a similar conclusion: the neural underpinnings
of sociopersonality and clinical diVerences involve regions that are centrally involved in
emotion or its regulation. However, these studies have almost nothing to say about the
development of these individual diVerences. How might this be conceptualized?
Let us take as a model problem the Wnding that increased ventral prefrontal activation
appears to accompany anxiety and depression. Neuroimaging research indicates that anx-
ious or depressed individuals utilize ventral prefrontal regions such as the OFC and ventral
ACC on various neuropsychological tasks more than normal controls. For example, a vari-
ety of studies have shown that cerebral blood Xow, measured by PET technology, is higher
in the orbitofrontal cortex, and higher in ventral ACC when controlling for volume diVer-
ences, in depressed than normal subjects. This research area is complex, and it includes
some contradictory Wndings. However, there are three particularly important discoveries to
emphasize. First, successful treatment (e.g., with antidepressive drugs) can decrease activa-
tion in ventral PFC (Drevets, 1999; Mayberg et al., 1999). In other words, individual diVer-
ences in ventral activation are functionally malleable, and not completely entrenched, even
in depressed subjects. Second, normal subjects who undergo sad or anxious mood induc-
tion, and anxious or obsessive subjects exposed to anxiety induction, both show increased
activity in these regions (see Drevets & Raichle, 1998, for a review). Thus, negative emotion
helps to induce this pattern of cortical activation for both normal and atypical
270 M.D. Lewis / Developmental Review 25 (2005) 252–277
populations. Third, increased ventral prefrontal activity corresponds with decreased dorsal
prefrontal activity (including dorsolateral PFC and dorsal ACC) for depressive subjects,
and this pattern can also be reversed by successful treatment (e.g. Bench, Frackowiak, &
Dolan, 1995; Mayberg et al., 1999). Moreover, as with ventral overactivation, dorsal under-
activation can be induced experimentally when normal subjects are exposed to anxiety-
provoking stimuli.
The tradeoV between ventral and dorsal prefrontal activation has been of interest to
neuroscientists of many stripes. Dorsal prefrontal regions, including the dorsolateral PFC
and the dorsal ACC, are known to mediate “cool” cognitive processes involved in tasks
that require comparing alternatives, selecting among potential strategies, working memory,
and so on (Zelazo & Mueller, 2002). Conversely, ventral prefrontal regions such as the
OFC and ventral ACC are commonly recruited for “hot” cognitive tasks, that require
judgements with emotional consequences, and for establishing a response mode (approach
or withdrawal) in reaction to immediate emotional contingencies such as threat or reward
(Rolls, 1999). Moreover, these ventral regions are densely connected with the amygdala,
hypothalamus, and brain stem structures, all of which are involved in diVerent levels of
emotion processing. Thus, depressive or anxious individuals somehow turn oV cortical sys-
tems that would permit the calm evaluation of present circumstances and turn on systems
that hold attention to the emotional aspects of the environment, surely amplifying and
extending their anxiety-related appraisals. Why do they do it?
At Wrst glance, the answer may appear simple. Heightened anxiety recruits these cortical
regions in real time, through vertical integration. Vertical integration ensures compatibility
in the activation patterns of regions up and down the neuroaxis. If the brain stem and
hypothalamus are presently mediating anxiety states, then they would entrain the activity
of cortical regions that mediate appraisals of threat and prepare the organism for escape.
However, emotions do not always recruit appraisals. Sometimes, perhaps most of the time,
appraisals of particular aspects of the world recruit emotions, and in fact this is the party
line in emotion theory (Lazarus, 1999; Scherer, 1999). In keeping with a dynamic systems
perspective, I would revise this story to say that tendencies to appraise situations in partic-
ular ways induce or augment incipient emotional states that feed back with cognitive pro-
cesses which enhance or stabilize these appraisals. Hence, a tendency to process
information ventrally rather than dorsally would begin to elicit anxiety in otherwise neu-
tral or at least novel situations, limbic and lower-brain circuits mediating this anxiety
would become increasingly activated, and this activation would feed back to ventral corti-
cal systems that increasingly attend to the anxiety-relevant aspects of the present situation
(Lewis, 2005). Feedback and coupling among these cortical and subcortical systems can
thus explain the emergence of anxious appraisals in real time. But how do these tendencies
become habitual for characterologically anxious or depressive individuals? An answer is
provided by the phenomenon of recursive synaptic modiWcation, or what I have referred to
as developmental feedback (feedback between real-time processes and developmental pro-
cesses). Each time vertical integration turns on ventral prefrontal regions, and conse-
quently turns oV dorsal prefrontal regions, synaptic shaping in the cortex increases the
probability of similar activation patterns on future occasions. Because ventral systems
have dense reciprocal connections with the amygdala and lower brain systems, recurrent
Wring patterns in the amygdala would ensue from the same epoch of vertical integration,
leading to synaptic shaping in the amygdala that parallels the sculpting of ventral cortical
networks. The coupling of activation across these systems in real time would be very likely
M.D. Lewis / Developmental Review 25 (2005) 252–277 271
to elicit negative emotion, and this emotional coloring would maintain stable patterns of
activation in real time and enhance synaptic shaping over developmental time. Thus, ven-
tral networks that monitor threat would become more and more elaborated, and more
readily called upon to assess uncertain situations. They would also hold an increasing
advantage over dorsal networks for responding to novel situations.
Let us look a little more closely at the development of these corticolimbic habits for
anxious or depressed individuals. Synaptic shaping over development would serve to elab-
orate and strengthen threat-related circuits in ventral prefrontal cortex, but it would do so
at the expense of alternative circuits in dorsal regions—circuits that would remain less
sculpted and less eYcient as a result. The on-line synchronization of ventral and dorsal
regions might in turn remain more diYcult, especially in the presence of negative emotion.
This description is close to Schore’s (2003a) model of orbitofrontal consolidation in
circumstances of insecure attachment, when high levels of stress reduce the Xexibility of
networks for appraising and responding to interpersonal challenges. The evolution of a
multi-site ventrofrontal-limbic “infrastructure” would also epitomize Johnson’s notion of
interactive specialization. Following Johnson, this infrastructure could be said to support
the development of a new “skill”—in this case, a cognitive set for expecting and being pre-
pared for the worst. Within the cortex itself, this infrastructure might couple orbitofrontal
systems for directing attention and preparing a response with posterior circuits attuned to
aspects of the environment rich with threat-related information (e.g., the fusiform gyrus for
attending to facial information). Johnson does not discuss interactions with limbic struc-
tures, but ventral prefrontal activity would likely recruit amygdala circuits mediating
threat-related associations as well. The notion of corticolimbic coupling was introduced by
Tucker (1992). He hypothesized a state of “resonance” between the cortex and limbic sys-
tem whenever gist-like cognitive appraisals recruited emotional meaning. This idea fore-
shadows an account of vertical integration based on phase synchrony (in the theta band)
across the entire neuroaxis (Lewis, 2005). In any case, individual corticolimbic “infrastruc-
tures” simply imply that cortical specialization consolidates along with limbic proclivities
for expressing and regulating emotional responses.
I have suggested that a consolidating infrastructure across corticolimbic circuits
becomes increasingly ingrained over occasions through emotional enhancement and syn-
aptic shaping, mediating the self-organization of an anxious or depressive disposition over
development. This picture of developing individual diVerences is consistent with a loss of
degrees of freedom, as personality or clinical patterns become more deeply entrenched in
neural tissue. However, this somewhat pessimistic account should also be tempered by con-
sideration of resilience in development. For some individuals, corticolimbic plasticity
remains a potent antidote to consolidating habits of appraisal and behavior. In fact, Bar-
bas (1995) suggests that the plasticity of paralimbic systems is the key to both risk and
resilience, supporting both lifelong Xexibility and vulnerability to psychiatric disorders.
This picture is also consistent with the principle of cascading constraints. The increasing
elaboration or consolidation of one set of circuits reduces the viability of another set. This
diVerentiation helps determine whether subsequent experiences will be inXuential or incon-
sequential, or whether they will be processed as threats or opportunities. Thus, as in any
self-organizing system (and in contrast to entropic systems), diVerences are created rather
than used up, and each newly established diVerence fans out to subsequent diVerences
down the line. Finally, this picture remains compatible with the principle of developmental
transitions. Let us imagine that the consolidation of an orbitofrontal “bias” emerges at one
272 M.D. Lewis / Developmental Review 25 (2005) 252–277
of three ages: middle childhood, early adolescence, or late adolescence. We could make sen-
sible predictions that this processing habit would remain in place for several years if it
started in middle childhood, would be highly transitory if it appeared in early adolescence,
and would endure for a lifetime if it emerged in late adolescence. Experience-expectant
waves of synaptic elaboration and synaptic pruning would result in a diVerent balance
between sensitivity and entrenchment for diVerent regions of cortex at each of these ages.
In this way, the roadmap of individual developmental pathways will always be synchro-
nized with a normative timeline of synaptic plasticity.
Finally, I have used data on ventral overactivation to demonstrate competition between
cortical regions. This was intended to show how a personality or clinical disposition could
consolidate through the recursive elaboration of circuits in one region at the expense of
another. However, I think this represents a very coarse example of developing individual
diVerences. Brain science is nowhere close to identifying diVerences in the wiring of small
subregions—populations of cells within the OFC or ACC—that could mediate more subtle
diVerences in appraisal styles. At an even Wner grain of analysis, the content of individual
appraisals, or conceptual categories for making sense of the world (e.g., associations
among women, Wre, and dangerous things, following LakoV, 1987), should also correspond
to dominant patterns of network activation established through experience. However, the
same principles should apply. At any spatial scale, the elaboration and eYciency of partic-
ular synaptic networks will have grown at the expense of alternative networks, through
repeated experiences, thus increasing the probability that these networks would be
recruited on future occasions. In young children, these synaptic competitions, and the full
throttle of emotional activation that supports them, may ingrain particular tendencies until
other powerful learning experiences reorganize the synaptic landscape. Luckily, fear of the
dark and suspicion of the opposite sex are usually among these transient tendencies. How-
ever, by late adolescence, when pruning has surpassed synaptic growth in all prefrontal
regions, such reversals should be increasingly diYcult to achieve. By this age, the overall
reduction in corticolimbic plasticity will have entrenched a repertoire of interpretations,
beliefs, and habitual emotional responses, many of which will endure for a lifetime. Yet the
cortex maintains substantial Xexibility long beyond childhood. This Xexibility resides in
ongoing opportunities for synaptic reorganizations, as exempliWed by the reduction of ven-
tral prefrontal activation in adults treated for depression. Moreover, recent research has
demonstrated the growth of new neurons in the human hippocampus, well into adulthood
and middle age, providing additional possibilities for reconWguring cortical networks (e.g.
Gould, Tanapat, Hastings, & Shors, 1999). Nevertheless, as emphasized by Freeman (1995)
and Tucker (2001), emotions are necessary for learning, change, and plasticity at any age.
Thus, at any age, strong emotions may be required for retooling the machinery of person-
ality that was set in place by strong emotions in the Wrst place.
To conclude, the divergence and consolidation of developmental trajectories demon-
strate that brains do in fact self-organize at a developmental time scale. DiVerences in brain
development are dependent, in part, on initial conditions (i.e., temperament and other
endogenous factors). But much of this divergence is a product of experience, and here expe-
rience is a short form for the ongoing stream of activity by which the brain teaches itself
what to learn. Because each episode of real-time cognitive–emotional activity leaves some
degree of synaptic change in its wake, we can say that brains develop by elaborating and
extending the outcomes of their own activities. And synaptic alterations are recursive,
which means that these activities tend to repeat themselves, forming lineages of individual
M.D. Lewis / Developmental Review 25 (2005) 252–277 273
patterning that progressively elaborate their own emergent themes. These features of neu-
ral development epitomize self-organization in natural systems. Moreover, because brain
activities also change the interpersonal environment (e.g., aggression promotes isolation,
independence promotes mastery and admiration), this sequence of self-elaboration occurs
in the context of a social world that becomes progressively more shaped to the features of
the individual brain. If our minds were not inscribed in Xesh, we would not have to worry
about the properties of complex dynamic systems. But our minds are greatly dependent on
our brains, and brains are designed by evolution to self-organize rapidly under the sway of
experiences and the emotions that color them. Therefore, to understand developing minds,
we need to understand developing brains, and the principles of self-organization provide a
foundation for doing so.
References
Adamec, R., Kent, P., Anisman, H., Shallow, T., & Merali, Z. (1998). Neural plasticity, neuropeptides and anxiety
in animals: Implications for understanding and treating aVective disorder following traumatic stress in
humans. Neuroscience and Biobehavioral Reviews, 23, 301–318.
Barbas, H. (1995). Anatomic basis of cognitive–emotional interactions in the primate prefrontal cortex. Neurosci-
ence and Biobehavioral Reviews, 19, 499–510.
Barbas, H. (2000). Connections underlying the synthesis of cognition, memory, and emotion in primate prefrontal
cortices. Brain Research Bulletin, 52, 319–330.
Bechara, A., Damasio, H., & Damasio, A. R. (2000). Emotion, decision making and the orbitofrontal cortex. Cere-
bral Cortex, 10, 295–307.
Bench, C. J., Frackowiak, R. S. J., & Dolan, R. J. (1995). Changes in regional cerebral blood Xow on recovery from
depression. Psychological Medicine, 25, 247–251.
Blair, R. J. R., Colledge, E., & Mitchell, D. G. V. (2001). Somatic markers and response reversal: Is there orbito-
frontal cortex dysfunction in boys with psychopathic tendencies? Journal of Abnormal Child Psychology, 29,
499–511.
Bressler, S. L., & Kelso, J. A. S. (2001). Cortical coordination dynamics and cognition. Trends in Cognitive Sci-
ences, 5, 26–36.
Bunge, S. A., Dudukovic, N. M., Thomason, M. E., Vaidya, C. J., & Gabrieli, J. D. E. (2002). Immature frontal lobe
contributions to cognitive control in children: Evidence from fMRI. Neuron, 33, 301–311.
Canli, T., Zhao, Z., Desmond, J. E., Kange, E., Gross, J., & Gabrieli, J. D. E. (2001). An fMRI study of personality
inXuences on brain reactivity to emotional stimuli. Behavioral Neuroscience, 115, 33–42.
Cardinal, R. N., Parkinson, J. A., Hall, J., & Everitt, B. J. (2002). Emotion and motivation: the role of the amyg-
dala, ventral striatum, and prefrontal cortex. Neuroscience and Biobehavioral Reviews, 26, 321–352.
Carter, C. S., MacDonald, A. M., III, Botvinick, M. M., Ross, L. L., Stenger, V. A., Noll, D., et al. (2000). Parsing
executive processes: Strategic vs. evaluative functions of the anterior cingulate cortex. Proceedings of the
National Academy of Sciences, 97, 1944–1948.
Casey, B. J., Trainor, R. J., Orendi, J. L., Schubert, A. B., Nystrom, L. E., Giedd, J. N., et al. (1997). A developmen-
tal functional MRI study of prefrontal activation during performance of a Go-No-Go task. Journal of Cogni-
tive Neuroscience, 9, 835–847.
Centonze, D., Picconi, B., Gubellini, P., Bernardi, G., & Calabresi, P. (2001). Dopaminergic control of synaptic
plasticity in the dorsal striatum. European Journal of Neuroscience, 13, 1071–1077.
Changeux, J.-P., & Dehaene, S. (1989). Neuronal models of cognitive functions. Cognition, 33, 63–109.
Chugani, H. T. (1994). Development of regional brain glucose metabolism in relation to behavior and plasticity.
In G. Dawson & K. W. Fischer (Eds.), Human behavior and the developing brain (pp. 153–175). New York:
Guilford.
Collins, P. F., & Depue, R. A. (1992). A neurobehavioral systems approach to developmental psychopathology:
Implications for disorders of aVect. In D. Cicchetti (Ed.), Rochester symposium of developmental psychopathology
Vol. 4. University of Rochester Press.
Davidson, R. J. (1998). AVective style and aVective disorders: Perspectives from aVective neuroscience. Cognition
and Emotion, 12, 307–330.
274 M.D. Lewis / Developmental Review 25 (2005) 252–277
Davidson, R. J., & Irwin, W. (1999). The functional neuroanatomy of emotion and aVective style. Trends in Cogni-
tive Sciences, 3, 11–21.
Davidson, R. J., Putnam, K. M., & Larson, C. L. (2000). Dysfunction in the neural circuitry of emotion regulation:
A possible prelude to violence. Science, 289, 591–594.
Depue, R. A., & Collins, P. F. (1999). Neurobiology of the structure of personality: Dopamine, facilitation of
incentive motivation, and extraversion. Behavioral and Brain Sciences, 22, 491–569.
Dikman, Z. V., & Allen, J. J. B. (2000). Error monitoring during reward and avoidance learning in high- and low-
socialized individuals. Psychophysiology, 37, 43–54.
Drevets, W. C. (1999). Prefrontal cortical-amygdalar metabolism in major depression. Annals of the New York
Academy of Sciences, 877, 614–637.
Drevets, W. C., & Raichle, M. E. (1998). Reciprocal suppression of regional cerebral blood Xow during emotional
versus higher cognitive processes: Implications for interactions between emotion and cognition. Cognition and
Emotion, 12, 353–385.
Durston, S., Thomas, K. M., Yang, Y., Ulug, A. M., Zimmerman, R., & Casey, B. J. (2002). A neural basis for the
development of inhibitory control. Developmental Science, 5, F9–F16.
Edelman, G. M. (1987). Neural Darwinism. New York: Basic Books.
Elman, J. L., Bates, E. A., Johnson, M. H., KarmiloV-Smith, A., Parisi, D., & Plunkett, K. (1996). Rethinking
innateness: A connectionist perspective on development. Cambridge, MA: MIT Press.
Engel, A. K., Fries, P., & Singer, W. (2001). Dynamic predictions: Oscillations and synchrony in top–down pro-
cessing. Nature Reviews Neuroscience, 2, 704–716.
Flood, J. F., Baker, M. L., Hernandez, E. N., & Morley, J. E. (1990). Modulation of memory retention by neuro-
peptide K. Brain Research, 520, 284–290.
Fogel, A. (1993). Developing through relationships: Origins of communication, self, and culture. Chicago: University
of Chicago Press.
Fox, N. A. (1992). Frontal brain asymmetry and vulnerability to stress: Individual diVerences in infant tempera-
ment. In T. M. Field, P. M. McCabe, & N. Schneiderman (Eds.), Stress and coping in infancy and childhood (pp.
83–100). Hillsdale, NJ: Erlbaum.
Fox, N. A., & Davidson, R. J. (1987). Electroencephalogram asymmetry in response to the approach of a stranger
and maternal separation. Developmental Psychology, 23, 233–240.
Fox, N. A., & Davidson, R. J. (1991). Hemispheric specialization and attachment behaviors: Developmental pro-
cesses and individual diVerences in separation protest. In J. L. Gerwirtz & W. M. Kurtines (Eds.), Intersections
with attachment (pp. 147–164). Hillsdale, NJ: Erlbaum.
Freeman, W. J. (1995). Societies of brains: A study in the neuroscience of love and hate. Hillsdale, NJ: Erlbaum.
Frijda, N. H. (1986). The emotions. Cambridge: Cambridge University Press.
Frith, C. D., Friston, K., Liddle, P. F., & Frackowiak, R. S. (1991). Willed action and the prefrontal cortex in man:
A study with PET. Proceedings of the Royal Society of London: Series B—Biological Science, 244(1311),
241–246.
Fuster, J. M. (1996). The prefrontal cortex: Anatomy, physiology and neuropsychology of the frontal lobe (3rd ed.).
New York: Raven Press.
Gallagher, M., & Holland, P. C. (1992). Understanding the function of the central nucleus. In J. P. Aggleton (Ed.),
The amygdala (pp. 307–321). New York: Wiley.
Gehring, W. J., Goss, B., Coles, M. G. H., Meyer, D. E., & Donchin, E. (1993). A neural system for error detection
and compensation. Psychological Science, 4, 385–390.
Gehring, W. J., Himle, J., & Nisenson, L. G. (2000). Action monitoring dysfunction in obsessive-compulsive disor-
der. Psychological Science, 11, 1–6.
Gemba, H., Sasaki, K., & Brooks, V. B. (1986). “Error” potentials in limbic cortex (anterior cingulate area 24) of
monkeys during motor learning. Neuroscience Letters, 70, 223–227.
Giedd, J. N., Blumenthal, J., JeVries, N. O., Castellanos, F. X., Liu, H., Zijdenbos, A., et al. (1999). Brain develop-
ment during childhood and adolescence: A longitudinal fMRI study. Nature Neuroscience, 2, 861–863.
Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D., Vaituzis, A. C., et al. (2004). Dynamic mapping
of human cortical development during childhood through early adulthood. Proceedings of the National Acad-
emy of Sciences, 101(21), 8174–8179.
Gould, E., Tanapat, P., Hastings, N. B., & Shors, T. J. (1999). Neurogenesis in adulthood: a possible role in learn-
ing. Trends in Cognitive Sciences, 3, 186–192.
Granic, I., & Patterson, G. R. (in press). Toward a comprehensive model of antisocial development: A dynamic
systems approach. Psychological Review.
M.D. Lewis / Developmental Review 25 (2005) 252–277 275
Greenough, W. T., & Black, J. E. (1992). Induction of brain structure by experience: Substrates for cognitive
development. In M. R. Gunnar & C. A. Nelson (Eds.), Minnesota symposium on child psychology, Developmen-
tal behavioral neuroscience (Vol. 24, pp. 155–200). Hillsdale, NJ: Erlbaum.
Grossberg, S., & Somers, D. (1991). Synchronized oscillations during cooperative feature linking in a cortical
model of visual perception. Neural Networks, 4, 453–466.
Gu, Q. (2002). Neuromodulatory transmitter systems in the cortex and their role in cortical plasticity. Neurosci-
ence, 111, 815–835.
Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdala activity related to enhanced memory for
pleasant and aversive stimuli. Nature Neuroscience, 2, 289–293.
Hariri, A. R., Mattay, V. S., Tessitore, A., Fera, F., & Weinberger, D. R. (2003). Neocortical modulation of the
amygdala response to fearful stimuli. Biological Psychiatry, 53(6), 494–501.
Hebb, D. O. (1949). The organization of behavior. New York: Wiley.
Heim, C., & NemeroV, C. B. (2001). The role of childhood trauma in the neurobiology of mood and anxiety disor-
ders: Preclinical and clinical studies. Biological Psychiatry, 49, 1023–1039.
Hikosaka, K., & Watanabe, M. (2000). Delay activity of orbital and lateral prefrontal neurons of the monkey
varying with diVerent rewards. Cerebral Cortex, 10, 263–271.
Izquierdo, I. (1997). The biochemistry of memory formation and its regulation by hormones and neuromodula-
tors. Psychobiology, 25, 1–9.
Izumi, Y., & Zorumski, C. F. (1999). Norepinephrine promotes long-term potentiation in the adult rat hippocam-
pus in vitro. Synapse, 31, 196–202.
Johnson, M. H. (1998). The neural basis of cognitive development. In W. Damon (Ed.), Handbook of child psychol-
ogy (pp. 1–49). New York: Wiley.
Johnson, M. H. (1999). Into the minds of babes. Science, 286, 247.
Johnson, M. H. (2000). Functional brain development in infants: Elements of an interactive specialization frame-
work. Child Development, 71, 75–81.
Johnson, M. H. (2001). Functional brain development in humans. Nature Reviews Neuroscience, 2, 475–483.
Johnson, M. H., Halit, H., Grice, S. J., & KarmiloV-Smith, A. (2002). Neuroimaging of typical and atypical
develop ment: A perspective from multiple levels of analysis. Development and Psychopatholo gy, 14,
521–536.
Keating, D. P. (1990). Charting pathways to the development of expertise. Educational Psychologist, 25, 243–267.
Kocsis, B., & Vertes, R. P. (1994). Characterization of neurons of the supramammillary nucleus and mammillary
body that discharge rhythmically with the hippocampal theta rhythm in the rat. Journal of Neuroscience, 14,
7040–7052.
LakoV, G. (1987). Women, Wre, and dangerous things: What categories reveal about the mind. Chicago: University
of Chicago Press.
Lazarus, R. S. (1999). The cognition–emotion debate: A bit of history. In T. Dalgleish & M. Power (Eds.), Hand-
book of cognition and emotion (pp. 3–19). Chichester, England: Wiley.
LeDoux, J. E. (1995). Emotion: Clues from the brain. Annual Review of Psychology, 46, 209–235.
Lévesque, J., Eugène, F., Joanette, Y., Paquette, V., Mensour, B., Beaudoin, G., et al. (2003). Neural circuitry
underlying voluntary suppression of sadness. Biological Psychiatry, 53, 502–510.
Lewis, M. D. (1995). Cognition–emotion feedback and the self-organization of developmental paths. Human
Development, 38, 71–102.
Lewis, M. D. (1997). Personality self-organization: Cascading constraints on cognition–emotion interaction. In A.
Fogel, M. C. Lyra, & J. Valsiner (Eds.), Dynamics and indeterminism in developmental and social processes (pp.
193–216). Mahwah, NJ: Erlbaum.
Lewis, M. D. (2000a). Emotional self-organization at three time scales. In M. D. Lewis & I. Granic (Eds.), Emo-
tion, development, and self-organization: Dynamic systems approaches to emotional development (pp. 37–69).
New York: Cambridge University Press.
Lewis, M. D. (2000b). The promise of dynamic systems approaches for an integrated account of human develop-
ment. Child Development, 71, 36–43.
Lewis, M. D. (2005). Bridging emotion theory and neurobiology through dynamic systems modeling (target arti-
cle). Behavioral and Brain Sciences, 28, 169–194.
Lewis, M. D., Lamm, C., Segalowitz, S. J., Zelazo, P. D., & Stieben, J. (in press). Neurophysiological correlates of
emotion regulation in children and adolescents. Journal of Cognitive Neuroscience.
Lewis, M. D., & Stieben, J. (2004). Emotion regulation in the brain: Conceptual issues and directions for develop-
mental research. Child Development, 75, 371–376.
276 M.D. Lewis / Developmental Review 25 (2005) 252–277
Lutz, A., Lachaux, J.-P., Martinerie, J., & Varela, F. J. (2002). Guiding the study of brain dynamics by using Wrst-
person data: Synchrony patterns correlate with ongoing conscious states during a simple visual task. Proceed-
ings of the National Academy of Science USA, 99(3), 1586–1591.
Luu, P., & Tucker, D. M. (2002). Self-regulation and the executive functions: Electrophysiological clues. In A.
Zani & A. M. Preverbio (Eds.), The cognitive electrophysiology of mind and brain (pp. 199–223). San Diego:
Academic Press.
Magai, C., & Haviland-Jones, J. (2002). The hidden genius of emotion: Lifespan transformations of personality.
Cambridge: Cambridge University Press.
Mayberg, H. S., Liotti, M., Brannan, S. K., McGinnis, S., Mahurin, R. K., Jerabek, P. A., et al. (1999). Reciprocal
limbic-cortical function and negative mood: Converging PET Wndings in depression and normal sadness.
American Journal of Psychiatry, 156(5), 675–682.
Nelson, C. A. (2000). Neural plasticity and human development: The role of early experience in sculpting memory
systems. Developmental Science, 3, 115–136.
Packard, M. G., & Cahill, L. (2001). AVective modulation of multiple memory systems. Current Opinion in Neuro-
biology, 11, 752–756.
Panksepp, J. (1998). AVective neuroscience: The foundations of human and animal emotions. New York: Oxford
University Press.
Panksepp, J. (2003). At the interface of the aVective, behavioral, and cognitive neurosciences: Decoding the emo-
tional feelings of the brain. Brain and Cognition, 52, 4–14.
Paré, D., Collins, D. R., & Pelletier, J. G. (2002). Amygdala oscillations and the consolidation of emotional memo-
ries. Trends in Cognitive Sciences, 6, 306–314.
Post, R. M., Weiss, S. R. B., Li, H., Smith, M. A., Zhang, L. X., Xing, G., et al. (1998). Neural plasticity and emo-
tional memory. Development and Psychopathology, 10, 829–855.
Potegal, M., Hebert, M., DeCoster, M., & MeyerhoV, J. (1996). Brief, high frequency stimulation of the corticome-
dial amygdala induces a delayed and prolonged increase of aggressiveness in male Syrian golden hampsters.
Behavioral Neuroscience, 110, 1–12.
Prigogine, I., & Stengers, I. (1984). Order out of chaos. New York: Bantam.
Rolls, E. T. (1999). The brain and emotion. Oxford: Oxford University Press.
Rolls, E. T., & Treves, E. T. (1998). Neural networks and brain function. New York: Oxford University Press.
Rubia, K., Overmayer, S., Taylor, E., Williams, S. C. R., Simmons, A., Andrew, C., et al. (2000). Functional frontalisation
with age: mapping neurodevelopmental trajectories with fM RI. Neuroscience and Biobehavioral Reviews, 24, 13–19.
Ruhland, R., & van Geert, P. (1998). Jumping into syntax: Transitions in the development of closed class words.
British Journal of Developmental Psychology, 16, 65–95.
Scherer, K. R. (1999). Appraisal theory. In T. Dalgleish & M. Power (Eds.), Handbook of cognition and emotion
(pp. 637–663). Chichester, England: Wiley.
Schore, A. N. (1994). AVect regulation and the origin of the self: The neurobiology of emotional development. Hills-
dale, NJ: Erlbaum.
Schore, A. N. (2003a). AVect regulation and disorders of the self. New York: Norton.
Schore, A. N. (2003b). AVect regulation and the repair of the self. New York: Norton.
Skarda, C. A., & Freeman, W. J. (1987). How brains make chaos in order to make sense of the world. Behavioral
and Brain Sciences, 10, 161–195.
Spear, L. P. (2000). The adolescent brain and age-related behavioral manifestations. Neuroscience and Biobehav-
ioral Reviews, 24, 417–463.
Steinberg, L., Dahl, R., Keating, D., Kupfer, D. J., Masten, A. S., & Pine, D. (in press). The study of developmental
psychopathology in adolescence: Integrating aVective neuroscience with the study of context. In D. Cicchetti
(Ed.), Handbook of developmental psychopathology. New York: Wiley.
Thagard, P., & Verbeurgt, K. (1998). Coherence as a constraint satisfaction. Cognitive Science, 22, 1–24.
Thatcher, R. W. (1998). A predator–prey model of human cerebral development. In K. M. Newell & P. C. M.
Molenaar (Eds.), Applications of nonlinear dynamics to developmental process modeling (pp. 87–128). Mahwah,
NJ: Erlbaum.
Thelen, E. (1990). Dynamical systems and the generation of individual diVerences. In J. Colombo & J. Fagen
(Eds.), Individual diVerences in infancy: Reliability, stability, prediction (pp. 19–43). Hillsdale, NJ: Erlbaum.
Thelen, E., & Smith, L. B. (1994). A dynamic systems approach to the development of cognition and action. Cam-
bridge, MA: MIT Press/Bradford.
Thelen, E., & Ulrich, B. D. (1991). Hidden skills: A dynamic systems analysis of treadmill stepping during the Wrst
year. Monographs of the Society for Research in Child Development, 56, (Serial No. 223).
M.D. Lewis / Developmental Review 25 (2005) 252–277 277
Thompson, E., & Varela, F. J. (2001). Radical embodiment: Neural dynamics and consciousness. Trends in Cogni-
tive Sciences, 5, 418–425.
Tucker, D. M. (1992). Developing emotions and cortical networks. In M. R. Gunnar & C. Nelson (Eds.), Minne-
sota symposia on child psychology: Developmental behavioral neuroscience (Vol. 24, pp. 75–128). Hillsdale, NJ:
Erlbaum.
Tucker, D. M. (2001). Motivated anatomy: A core-and-shell model of corticolimbic architecture. In G. Gainotti
(Ed.), Handbook of neuropsychology: Emotional behavior and its disorders (Vol. 5, 2nd ed., pp. 125–160).
Amsterdam, Netherlands: Elsevier.
Tucker, D. M., Derryberry, D., & Luu, P. (2000). Anatomy and physiology of human emotion: Vertical integra-
tion of brainstem, limbic, and cortical systems. In J. C. Borod (Ed.), The neuropsychology of emotion. London:
Oxford University Press.
van der Maas, H. L. J., & Molenaar, P. C. M. (1992). Stagewise cognitive development: An application of catastro-
phe theory. Psychological Review, 99, 395–417.
van Veen, V., Cohen, J. D., Botvinick, M. M., Stenger, V. A., & Carter, C. S. (2001). Anterior cingulate cortex, con-
Xict monitoring, and levels of processing. Neuroimage, 14, 1302–1308.
Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience.
Cambridge, MA: MIT Press.
Zelazo, P. D., & Mueller, U. (2002). Executive function in typical and atypical development. In U. Goswami (Ed.),
Handbook of childhood cognitive development. Oxford: Blackwell.
