56: Experience and the Refinement of Synaptic Connections

• Development of Human Mental Function Is Influenced by Early Experience

  • Early Experience Has Lifelong Effects on Social Behaviors

  • Development of Visual Perception Requires Visual Experience

• Development of Binocular Circuits in the Visual Cortex Depends on Postnatal Activity

  • Visual Experience Affects the Structure and Function of the Visual Cortex

  • Patterns of Electrical Activity Organize Binocular Circuits in the Visual Cortex

• Reorganization of Visual Circuits During a Critical Period Involves Alterations in Synaptic Connections

  • Reorganization Depends on a Change in the Balance of Excitatory and Inhibitory Inputs

  • Postsynaptic Structures Are Rearranged During the Critical Period

  • Thalamic Inputs Are Also Remodeled

  • Synaptic Stabilization Contributes to Closing the Critical Period

• Segregation of Retinal Inputs in the Lateral Geniculate Nucleus Is Driven by Spontaneous Neural Activity In Utero

• Activity-Dependent Refinement of Connections Is a General Feature of Circuits in the Central Nervous System

  • Many Aspects of Visual System Development Are Activity-Dependent

  • Auditory Maps Are Refined During a Critical Period

  • Distinct Regions of the Brain Have Different Critical Periods of Development

• Critical Periods Can Be Reopened in Adulthood

• An Overall View

The human nervous system is functional at birth: Newborn babies can see, hear, breathe, and suckle. However, the capabilities of human infants are quite rudimentary compared to those of other species. Wildebeest calves can stand and run within minutes of birth, and many birds can fly shortly after they hatch from their eggs. In contrast, a human baby cannot lift its head until it is 2 months old, cannot bring food to its mouth until it is 6 months old, and cannot survive without parental care for a decade.

What accounts for the delayed maturation of our motor, perceptual, and cognitive abilities? One main factor is that the embryonic connectivity of the nervous system, discussed in Chapters 52 through 55, is only a "rough draft" of the neural circuits that exist in our adult selves. After birth, embryonic circuits are refined by sensory stimulation—our experiences. This two-part sequence—genetically determined connectivity followed by experience-dependent reorganization— is a common feature of mammalian neural development, but in humans the second phase is especially prolonged.

At first glance this delay in human neural development might seem dysfunctional. Although it does exact a toll, it also provides an advantage. Because our mental abilities are shaped largely by experience, we gain the ability to custom fit our nervous systems to our individual bodies and unique environments. It could be argued that it is not the large size of the human brain but rather its experience-dependent maturation that makes our mental capabilities superior to those of other species.

The plasticity of the nervous system in response to experience endures throughout life. Nevertheless, periods of heightened susceptibility to modification, known as sensitive periods, occur at particular times in development. In some cases the adverse effects of deprivation or atypical experience early in life cannot easily be reversed by providing appropriate experience at a later age. Such necessary periods of development are referred to as critical periods. As we shall see, new discoveries are blurring the distinction between sensitive and critical periods, so we will use the term critical periods to refer to both.

Critical periods can most easily be appreciated from the perspective of behavior—the capacity to perceive the world around us, to learn a language, or to form strong social relationships. A 5-year-old child can quickly and effortlessly learn a second language, whereas a 15-year-old adolescent may become fluent but is likely to speak with an accent, even if he lives to be 90 years old. Likewise, when a cataract that deprives a child of vision is removed in early childhood, there are no lasting effects of the cataract, but when the surgery occurs at the age of 10 years, she is unlikely ever to have normal visual acuity. In each of these cases relevant experience must occur within a critical period if behavior is to develop normally.

The process of learning is also influenced by critical periods. One of the most striking illustrations of a lifelong behavior established during a critical period is imprinting, a form of learning in birds. Just after hatching, birds become indelibly attached, or imprinted, to a prominent moving object in their environment, typically their mother. The process of imprinting is important for the protection of the hatchling. Although the attachment is acquired rapidly and persists, imprinting can only occur during a critical period soon after hatching—in some species only a few hours. Thus postnatal development of the nervous system can be viewed as a series of critical periods.

We begin this chapter by examining the evidence that early experience shapes a range of human mental capacities, from our ability to make sense of what we see to our ability to engage in appropriate social interactions. To illustrate the neural basis of these experiential effects we describe the role of experience in the development of the visual system in experimental animals. Studies of the development of the visual system have provided our most detailed understanding of how experience shapes neural circuitry throughout the brain. We will see that experience is needed to refine patterns of synaptic connections and to stabilize these patterns once they have formed. Finally, we will consider recent evidence that critical periods may be less restrictive than once thought; in some cases they can be extended or "reopened".

Understanding experience-dependent plasticity and the extent to which critical periods can be re-opened in adulthood has important practical consequences. First, much educational policy is based on the idea that early experience is crucial, so it is important to know exactly when a particular form of enrichment will be optimally beneficial. Second, medical treatment of many childhood conditions, such as congenital cataracts, is now predicated on the idea that early intervention is imperative if long-lasting deficits are to be avoided. Third, there is increasing concern that some behavioral disorders, such as autism, may be caused by impaired reorganization of neural circuits during critical periods. Finally, the possibility of reopening critical periods in adulthood is leading to new therapeutic approaches to neural insults such as stroke that previously were thought to have irreversible consequences.

Early Experience Has Lifelong Effects on Social Behaviors

The conclusion that certain social or perceptual experiences are important for human development was first arrived at through studies of children who had been deprived of these experiences early in life. In rare cases children abandoned in the wild and later returned to human society have been studied. As might be expected, these children were socially maladjusted. Surprisingly, however, the defects proved to be generally irreversible.

In the 1940s the psychoanalyst René Spitz provided more systematic evidence that early interactions with other humans are essential for normal social development. Spitz compared the development of infants raised in a foundling home with the development of infants raised in a nursing home attached to a women's prison. Both institutions were clean and both provided adequate food and medical care. The babies in the prison nursing home were all cared for by their mothers, who, although in prison and away from their families, tended to shower affection on their infants in the limited time allotted to them each day. In contrast, infants in the foundling home were cared for by nurses, each of whom was responsible for several babies. As a result, children in the foundling home had far less contact with other humans than did those in the prison's nursing home.

The two institutions also differed in another respect. In the prison nursing home the cribs were open, so that the infants could readily watch other activities in the ward; they could see other babies play and observe the staff go about their business. In the foundling home the bars of the cribs were covered by sheets that prevented the infants from seeing outside. In reality, the babies in the foundling home were living under conditions of severe sensory and social deprivation.

Groups of newborn infants at the two institutions were followed through their early years. At the end of the first 4 months the infants in the foundling home fared better on several developmental tests than those in the prison nursing home, suggesting that intrinsic factors did not favor the infants in the prison nursing home. But by the end of the first year the motor and intellectual performance of the children in the foundling home had fallen far below that of children in the prison nursing home. Many of the children in the foundling home had developed a syndrome that Spitz called hospitalism and is now often called anaclitic depression. These children were withdrawn and displayed little curiosity or gaiety. Moreover, their defects extended beyond the emotional and cognitive. They were especially prone to infection, implying that the brain exerts complex controls over the immune system as well as behavior. By their second and third years, children in the prison nursing home were similar to children raised in normal families at home—they were agile, had a vocabulary of hundreds of words, and spoke in sentences. In contrast, the development of children in the foundling home was still further delayed—many were unable to walk or to speak more than a few words.

More recent studies of other similarly deprived children have confirmed these conclusions and shown that the defects are long-lasting. Longitudinal studies of orphans who were raised for several years in large impersonal institutions with little or no personal care, then adopted by caring families, have been especially revealing. Despite every effort of the adoptive parents, many of the children were never able to develop appropriate, caring relationships with family members or peers (Figure 56–1A). More recent imaging studies have revealed defects in brain structure correlated with, and presumably due to, this deprivation (Figure 56–1B).

As compelling as these studies on humans are, it is difficult to derive definitive conclusions. An influential set of studies that extended the analysis of social behavior to monkeys was carried out in the 1960s by two psychologists, Harry and Margaret Harlow. The Harlows reared newborn monkeys in isolation for 6 to 12 months, depriving them of contact with their mothers, other monkeys, or people. At the end of this period the monkeys were physically healthy but behaviorally devastated. They crouched in a corner of their cage and rocked back and forth like autistic children (Figure 56–1C). They did not interact with other monkeys, nor did they fight, play, or show any sexual interest. Thus a 6-month period of social isolation during the first 18 months of life produced persistent and serious disturbances in behavior. By comparison, isolation of an older animal for a comparable period was found to be without such drastic consequences. These results confirmed, under controlled conditions, the critical influence of early experience on later behavior. For ethical reasons, these studies would not be possible today.

Development of Visual Perception Requires Visual Experience

The dramatic dependence of the brain on experience and the ability of that experience to shape perception is evident in people born with cataracts. Cataracts are opacities of the lens that interfere with the optics of the eye but not directly with the nervous system; they are easily removed surgically. In the 1930s it became apparent that patients who had congenital binocular cataracts removed after the age of 10 years had permanent deficits in visual acuity and had difficulties perceiving shape and form, a condition called amblyopia. In contrast, when cataracts that develop in adults are removed decades after they form, normal vision returns immediately.

Likewise, children with strabismus (crossed eyes) do not have normal depth perception (stereopsis), an ability that requires the two eyes to focus on the same location at the same time. They can acquire this ability if their eyes are aligned surgically during the first few years of life, but not if surgery occurs later in adolescence. As a result of these observations, congenital cataracts are now usually removed, and strabismus is corrected surgically, in early childhood. Over the past few decades researchers have elucidated structural and physiological underpinnings of these critical periods.

Because sensory experience of the world is transformed into patterns of electrical activity in the brain, one might imagine that electrical signals in neural circuits affect the brain's circuitry. But is this true? And if it is true, what changes occur and how does activity trigger them?

Our most detailed understanding of these links comes from studies of the neural circuits that mediate binocular vision. The key figures in the early phases of this work were David Hubel and Torsten Wiesel, who undertook a set of studies on cats and monkeys to investigate how experience affects the structural and functional organization they had delineated (Figure 56–2).

Visual Experience Affects the Structure and Function of the Visual Cortex

In one influential study Hubel and Wiesel raised a monkey from birth to 6 months of age with one eyelid sutured shut, thus depriving the animal of vision in that eye. When the sutures were removed it became clear that the animal was blind in the deprived eye. They then performed electrophysiological recordings from cells along the visual pathway to determine where the defect arose. They found that retinal ganglion cells in the deprived eye, as well as neurons in the lateral geniculate nucleus that receive input from the deprived eye, responded well to visual stimuli and had essentially normal receptive fields.

In contrast, cells in the visual cortex were fundamentally altered. In the cortex of normal animals most neurons are binocularly responsive. In animals that had been monocularly deprived for the first 6 months, most cortical neurons did not respond to signals from the deprived eye (Figure 56–3). The few cortical cells that were responsive were not sufficient for visual perception. Not only had the deprived eye lost its ability to drive most cortical neurons, but this loss was permanent and irreversible.

Hubel and Wiesel went on to test the effects of visual deprivation imposed for shorter periods and at different ages. They obtained three types of results, depending on the timing and duration of the deprivation. First, monocular deprivation for a few weeks during the first 2 postnatal months led to loss of cortical responses from the deprived eye that was reversible after the eye had been opened, especially if the opposite eye was then closed to encourage use of the initially deprived eye. Second, monocular deprivation for a few weeks during the next months also resulted in a substantial loss of cortical responsiveness to signals from of the deprived eye, but in this case the effects were irreversible. Finally, deprivation in adults, even for periods of many months, had no effect on the responses of cortical cells to signals from the deprived eye or on visual perception. These results demonstrated that the cortical connections that control visual perception are established within a critical period of early development.

Are there structural correlates of these functional defects? To address this question we need to recall three basic facts about the anatomy of the visual cortex. First, inputs from the two eyes remain segregated in the lateral geniculate nucleus. Second, the geniculate inputs carrying information from the two eyes to the cortex terminate in alternating columns, termed ocular dominance columns. Third, lateral geniculate axons terminate on neurons in layer IVC of the primary visual cortex; convergence of input from the two eyes on a common target cell occurs at the next stage of the pathway, in cells above and below layer IVC (Figure 56–2).

To examine whether the architecture of ocular dominance columns depends on visual experience early in postnatal life, Hubel and Wiesel deprived newborn animals of vision in one eye and then injected a labeled amino acid into the normal eye. The injected label was incorporated into proteins in retinal ganglion cell bodies, transported along the retinal axons to the lateral geniculate nucleus, transferred to geniculate neurons, and then transported to the synaptic terminals of these axons in the primary visual cortex. After closure of one eye the columnar array of synaptic terminals relaying input from the deprived eye was reduced, whereas the columnar array of terminals relaying input from the normal eye was expanded (Figure 56–4). Thus sensory deprivation early in life alters the structure of the cerebral cortex.

How are these striking anatomical changes brought about? Does sensory deprivation alter ocular dominance columns after they have been established, or does it interfere with their formation? It is now clear that the mature pattern of ocular dominance columns in monkeys is not achieved until 6 weeks after birth (Figure 56–5). Only at this time do the terminals of fibers from the lateral geniculate nucleus become completely segregated in the cortex (Figure 56–6). Because the inputs are not well segregated at the time that visual deprivation exerts its effects, we can conclude that the deprivation perturbs the segregation of the inputs. We will see later that the remodeling of thalamic axonal arbors contributes to the perturbation of cortical columns.

Patterns of Electrical Activity Organize Binocular Circuits in the Visual Cortex

What determines the extent of ocular dominance columns? The crucial factor may be the existence of minor differences in the proportion of inputs from each eye that converge on common target cells at birth. If by chance the fibers conveying input from one eye are initially more numerous in one local region of cortex, those axons may have an advantage.

How might this occur? An attractive idea, based on a theory first proposed in the 1940s by Donald Hebb, is that connections are strengthened when pre- and postsynaptic elements at a synapse are active together. In the case of binocular interactions, neighboring axons from the same eye tend to fire in synchrony because they are activated by the same visual stimulus at any instant. The synchronization of their firing means that they cooperate in the depolarization and excitation of a target cell. This cooperative action maintains the viability of those synaptic contacts at the expense of the noncooperating synapses.

Cooperative activity promotes branching of axons and thus creates the opportunity for the formation of additional synaptic connections with cells in the target region. At the same time, the strengthening of synaptic contacts made by the axons of one eye will impede the growth of synaptic inputs from the opposite eye. In this sense fibers from the two eyes may be said to compete for a target cell. Together, cooperation and competition between axons ensure that two populations of afferent fibers will eventually innervate distinct regions of the primary visual cortex with little local overlap.

Competition and cooperation are not simply the outcome of neural activity per se, or of differences in absolute levels of activity among axons. Instead, they appear to depend on precise temporal patterns of activity in the competing (or cooperating) axons. The principle was dramatically illustrated by Hubel and Wiesel in a set of studies that examined stereoscopic vision—the perception of depth. The brain normally computes depth perception by comparing the disparity in retinal images between the two eyes. When the eyes are improperly aligned, this comparison cannot be made and stereoscopy is impossible. Such misalignments occur in children who are "cross-eyed," or strabismic. As noted, this condition can be surgically repaired, but unless the surgery occurs during the first few years of life, the children forever remain incapable of stereoscopy.

Hubel and Wiesel examined the impact of strabismus on the organization of the visual system in cats. To render cats strabismic, the tendon of an extraocular muscle was severed in kittens. Both eyes remained fully functional but misaligned. Inputs from the two eyes that converged on a binocular cell in the visual cortex now carried information about different stimuli in slightly different parts of the visual field. As a result, cortical cells became monocular, driven by input from one eye or the other but not both (Figure 56–7). This finding suggested to Hubel and Wiesel that disruption of the synchrony of inputs led to competition rather than cooperation, so that cortical cells came to be dominated by one eye, presumably the one that had dominated it very slightly at the outset.

These physiological studies led investigators to test whether pharmacological blockade of electrical activity in retinal ganglion cells could affect neural connectivity in the visual system. Activity was blocked by injecting each eye with tetrodotoxin, a toxin that selectively blocks voltage-sensitive Na⁺ channels. Signals from the two eyes were generated separately by direct electric stimulation of the bilateral optic nerves. In kittens ocular dominance columns are not established if activity in retinal ganglion neurons is blocked before the critical period of development. When the two optic nerves are stimulated synchronously, ocular dominance columns still fail to form. Only when the optic nerves are stimulated asynchronously are ocular dominance columns established.

If the development of ocular dominance columns indeed depends on competition between afferent fibers, might it be possible to induce the formation of columns where they normally are not present, simply by establishing competition between two sets of axons? This radical possibility was tested in frogs, where retinal ganglion neurons from each eye project only to the contralateral side of the brain. In normal frogs afferent fibers from the two eyes do not compete for the same cortical cells, so there is no columnar segregation of afferent inputs. To generate competition a third eye was transplanted early in larval development into a region of the frog's head near one of the normal eyes. The retinal ganglion neurons of the extra eye extended axons to the contralateral optic tectum. Remarkably, axon terminals from the transplanted and normal eyes segregated, generating a pattern of regular alternating columns (Figure 56–8).

This finding suggested that competition between two sets of afferent axons for the same population of cortical neurons drives their segregation into distinct target territories. The columnar segregation of retinal inputs in the frog brain is dependent on synaptic activity, presumably at the synapses between retinal axons and tectal neurons. Thus neural activity has powerful roles in fine-tuning visual circuits.

The pioneering work of Hubel, Wiesel, and their colleagues showed that early experience is a critical prerequisite for the emergence of normal structure and function in the visual cortex. However, despite four decades of research, many questions about the cellular and molecular mechanisms that underlie the critical period have remained unanswered. Hubel, Wiesel, and their disciples studied cats and monkeys, in which manipulations of cells and molecules are difficult. Recently, many investigators have begun to address these issues in mice because they are more amenable to mechanistic analysis.

Reorganization Depends on a Change in the Balance of Excitatory and Inhibitory Inputs

When one eye is closed or patched, the cortex no longer receives input from that eye. If the loss of input is prolonged for a few days before the eye is reopened, the cortex cannot be activated by input from the formerly deprived eye. Nor can the cells of the cortex be activated by direct stimulation of the corresponding optic nerve.

What converts this early loss of functional input into a permanent alteration of functional capability? One idea is that thalamic axons carrying information from the deprived eye lose their ability to activate cortical neurons. A decrease in efficacy of the thalamocortical synapse may contribute to this effect, but this is not the whole story. In cats and monkeys each thalamic axon carries input from only one eye (Figure 56–2). Because loss of responsiveness to the deprived eye occurs only if the other eye remains active, one might imagine that the earliest changes would occur where inputs from the two eyes converge on binocular cells. Local circuits connect the targets of the monocular thalamic inputs in layer IV to the binocular neurons in layers II/III and V. The first physiological changes following closure of one eye occur in layers II/III and V, not in layer IV. This implies that the loss of cortical responsiveness to the deprived eye results from a circuit alteration rather than from a simple loss of input.

What changes in cell function account for these changes in circuitry? At least three have been proposed. First, excitatory synapses within the primary visual cortex may decrease in strength, perhaps by undergoing long-term depression (LTD) (see Chapter 66). In fact, direct recording from cortical neurons shows that LTD does occur soon after eye closure. Second, inhibitory synapses may become stronger, leading to a net decrease in the level of excitation of cortical neurons by inputs from the closed eye. Third, an increase in inhibition within the cortex may alter network properties in a more subtle way, such as tuning the circuit to favor LTD.

The involvement of inhibition has been demonstrated by studies in the binocular region of mouse visual cortex. In mice, as in cats and monkeys, closure of the contralateral eye during the critical period for ocular dominance markedly shifts the preference of binocular neurons to inputs from the ipsilateral eye (Figure 56–9). Closure before or after the time of this normal critical period, however, fails to alter the preference of the neurons. Physiological studies have shown that monocular deprivation potentiates inhibitory feedback onto the neurons that the deprived eye normally excites. Moreover, the critical period in which monocular deprivation elicits changes in preference can be advanced using a genetic method to enhance gamma-aminobutyric acid (GABA) signaling (Figure 56–10). Conversely, the period in which monocular deprivation enhances the preference for ipsilateral eye input can be delayed by delaying GABA signaling (Figure 56–10). Thus a balance of intracortical excitation and inhibition is required for reorganization during this critical period.

What explains the change in responsiveness? After one eye is closed responsiveness to the open eye increases. This increase involves enhanced efficacy of synaptic transmission, a process resembling long-term potentiation (see Chapters 66 and 67). Inputs firing together with the majority of their neighbors become stronger, whereas inputs that are improperly matched to the majority pattern become weaker. This mechanism of reorganization conforms to Hebb's theory.

Postsynaptic Structures Are Rearranged During the Critical Period

What are the structural correlates of the physiological changes that result in altered responsiveness of the visual cortex to input from the closed and open eyes? Particular attention has been paid to dendritic spines as potential sites of plasticity.

Spines are small protrusions from the dendrites of many cortical neurons on which excitatory synapses form. They are dynamic structures, and their appearance and loss are thought to reflect the formation and elimination of synapses. Spine motility is especially marked during early postnatal development, and increases in spine dynamics and number have been associated with changes in behavior.

Striking alterations in the motility and number of dendritic spines on neurons in the mouse visual cortex are observed following closure of one eye. Two days after eye closure in young mice the motility and turnover of dendritic spines on neurons in the visual cortex increases, suggesting that synaptic connections are beginning to rearrange (Figure 56–11). A few days later the number of spines begins to change—the number of spines on the apical dendrites of pyramidal neurons initially decreases but after longer periods of deprivation increases again.

These alterations in spine motility and number can be correlated with three known features of the critical period. First, rather than occurring in layer IV, the changes occur primarily in superficial and deep layers of the cortex, where binocular cells lie. Second, they occur only in the portion of the visual cortex that normally receives binocular input. Third, they fail to occur following eye closure in adult mice (Figure 56–11).

Together these results support a model that links spine dynamics with critical period plasticity. According to this model, spine motility increases before physiological changes in responsiveness can be recorded. This condition may result from the imbalance of inputs to binocular neurons from the open and closed eyes, and it may reflect the first stages in synaptic rearrangement. In turn, the loss of spines, and presumably of synapses, corresponds in time and space to the loss of input from the closed eye and may provide a structural basis for the permanence of this loss. The later growth of new spines occurs as or after responsiveness to the open eye increases, and may underlie the adaptive rearrangement that permits the cortex to make the best use of the input available to it.

Thalamic Inputs Are Also Remodeled

How are local changes in spines related to the large-scale structural changes in ocular dominance columns shown in Figure 56–4? When developing axons from the lateral geniculate nucleus first reach the cortex, the terminal endings of several neurons overlap extensively. Each fiber extends a few branches over an area of the visual cortex that spans several future ocular dominance columns. As the cortex matures, axons retract some branches, expand others, and even form new branches (Figure 56–12A).

With time each geniculate neuron becomes connected almost exclusively to a group of neighboring cortical neurons within a single column. The arbors become segregated into columns through the pruning or retraction of certain axons and the sprouting of others. This dual process of axon retraction and sprouting occurs widely throughout the nervous system during development.

What happens after one eye is closed? Axons from a closed eye are at a disadvantage and a greater than normal proportion retract. At the same time, axons from the open eye sprout new terminals at sites vacated by fibers that would otherwise convey input from the closed eye (Figure 56-12B). If an animal is deprived of the use of one eye early during the critical period of axonal segregation, the normal processes of axon retraction and outgrowth are perturbed. In contrast, if an animal is deprived of the use of one eye only after the ocular dominance columns are almost fully segregated, axons conveying input from the open eye actually sprout collaterals in regions of the cortex that they had vacated earlier (see Figure 56–6).

Initially it was believed that rearrangements of thalamocortical axons in monocularly deprived animals caused the changes in cortical responsiveness to the open and closed eyes. We now know, through electrophysiological recording and assays of spine dynamics described above, that physiological changes and synaptic alterations precede these large-scale axonal rearrangements. So rather than causing the physiological changes, axonal remodeling may contribute to making these changes enduring and irreversible. The question then becomes: How do alterations in synaptic structure and function within the cortex lead to alterations in the input?

One idea is that activity regulates the secretion of neurotrophic factors by cortical neurons. Such factors may then regulate survival of some neurons at the expense of others (see Chapter 53) or promote the expansion of some axonal arbors at the expense of others. One such factor, brain-derived neurotrophic factor (BDNF), is indeed synthesized and secreted by cortical neurons. Administering excess BDNF, or interfering with its receptor trkB, modifies the formation of ocular dominance columns. Nevertheless, interpreting the actions of these trophic factors is not straightforward. BDNF and trkB signaling affect the cortex in many ways, including enhancing the growth of thalamocortical axons. BDNF can also speed the maturation of inhibitory circuits, which, as noted above, can influence plasticity. It remains unclear whether BDNF is a specific catalyst of the competition that preferentially promotes expansion of some arbors.

Synaptic Stabilization Contributes to Closing the Critical Period

A hallmark of critical periods is that the time interval in which experience affects the development of neural circuits is limited. What determines the timing of this remarkable biological response? To answer this question we must ask what opens the critical period and what closes it.

Opening of critical periods could result from maturation of systems required for plasticity. As we have mentioned, genetic studies indicate that maturation of inhibitory circuits in the visual cortex is required for initiation of the critical period for binocular development. Likewise, synapses that are modified by competitive interactions between the eyes, or by synchronous activity, may not undergo long-term potentiation or depression until they reach a sufficient level of maturity.

The factors that terminate the critical period have been studied in more detail. Since synapses and circuits are labile during critical periods, an obvious idea is that stabilizing factors bring this period of heightened plasticity to a close. The cellular and molecular landscape of the cortex changes in many ways as the brain matures, and several of these alterations may play roles. One parameter is the state of myelination of axons, which occurs around the time the critical period closes. Formation of myelin creates physical barriers to sprouting and axonal growth. Moreover, as discussed in detail in Chapter 57, myelin contains factors such as Nogo and myelin-associated glycoprotein that actively inhibit growth of axons. In mutant mice lacking Nogo or one of its receptors, NogoR, the critical period remains open into adulthood, suggesting that the appearance of these receptors normally contributes to closing the critical period (Figure 56–13).

Another possible agent of closure is the perineuronal net, a web of glycosaminoglycans that wraps certain classes of inhibitory neurons. These nets form around the time that the critical period closes. Infusion of the enzyme chondroitinase, which digests perineuronal nets, maintains plasticity. Thus critical periods may close once molecular barriers to synaptic growth and rearrangement come into play.

Why should there be an end to critical periods? Would it not be advantageous for the brain to maintain its ability to remodel into adulthood? Perhaps not—the ability of our brain to adapt to variations in sensory input, to gradual physical growth (eg, increases in the distance between the eyes affecting binocular correspondence), and to various congenital disorders is a valuable asset. At an extreme, if one eye is lost it is advantageous to devote all available cortical real estate to the remaining eye. Conversely, one would not want wholesale reorganization, possibly accompanied by loss of skills and memories, if vision through one eye were lost temporarily in adulthood due to disease or injury. So enhancing plasticity during a critical period may represent an adaptive compromise between flexibility and stability.

Some of the principles of development of the visual cortex we have discussed also govern the development of the lateral geniculate nucleus. The arbors of retinal ganglion cells from the two eyes are segregated into alternating layers in this nucleus, much as the projections from this nucleus are segregated in alternating ocular dominance columns in the visual cortex (Figure 56–14).

In both structures individual axons at first form terminals in multiple domains (layers in the geniculate nucleus, columns in the cortex). Later the terminals become segregated by a process of refinement. As in the cortex, the segregation of the inputs from each eye can be disrupted by applying tetrodotoxin to the optic nerves, indicating that activity is essential for segregation. In the lateral geniculate nucleus, however, the segregation of inputs is complete before birth.

Since segregation of retinal inputs in the lateral geniculate nucleus occurs before birth, vision cannot drive the neural activity essential for segregation. It turns out that the axons of retinal neurons are spontaneously active in utero, well before the eyes open. Neighboring ganglion cells fire in synchronous bursts that last a few seconds, followed by silent periods that may last for minutes. Sampling the activity of retinal ganglion neurons across the entire retina revealed that these bursts propagate across much of the retina in a wave-like manner (Figure 56–15). This pattern of ganglion cell activity appears to be coordinated by excitatory inputs from amacrine cells in the overlying layer of the retina (see Figure 26–3).

The spontaneous, synchronous firing of a select group of ganglion neurons excites a local group of neurons in the lateral geniculate nucleus. Such synchronized activity appears to strengthen these synapses at the expense of other nearby synapses. The fact that retinal ganglion neurons are spontaneously active in utero is an important clue to the development of many pathways in the brain that are refined before they have the chance to respond to environmental stimulation.

We have seen that neural activity is critical for segregating axons from the two retinas into distinct layers in the lateral geniculate nucleus and then into distinct columns in the visual cortex. Is this a special case, or does activity also affect maturation elsewhere in the visual system, and even in other parts of the brain? Studies of many systems show that activity-dependent control of refinement is a general property of neural circuits in the mammalian brain.

Many Aspects of Visual System Development Are Activity-Dependent

One well-studied example of activity-dependent development in the visual system is the sharpening of the topographic distribution of retinal ganglion cell axons onto their central targets, a topic we introduced in Chapter 54. Molecular cues such as ephrins guide axons to appropriate sites in the tectum, but they are not sufficient to form the refined visual map.

Histological and physiological studies have found that the map formed initially in the superior colliculus (the equivalent of the optic tectum of lower vertebrates) is coarse, and that individual retinal ganglion cell axons have large, overlapping arbors. These axonal arbors are later pruned to their mature size, resulting in a more restricted and precise field of termination. If retinal activity is inhibited, only the initial coarse map forms.

Is it the pattern of activity or activity itself that is important in visual map formation? Put another way, is activity simply "permissive" for refinement, or is it "instructive," determining exactly which axons win or lose the competition? Many experiments show that the latter idea is closer to the truth.

In one study the accuracy of the retinotectal map was assessed in fish raised in a tank illuminated only by brief flashes from a strobe light. A control group was raised in a normal laboratory environment. The total light intensity presented to the fish was similar under both conditions, but the resulting pattern was very different. In control fish the images fell haphazardly on various parts of the retina as the fish swam around their tanks. This input produces local synchronous activity of the sort generated by the waves of spontaneous activity described above—neighboring ganglion cells tend to fire together, but there is little correlation with the firing patterns of distant ganglion cells. In these fish the map becomes precise. In contrast, stroboscopic illumination synchronously activates nearly all of the ganglion cells, and as a consequence the retinotectal map remains coarse.

Presumably, the tectum determines which retinal axons are near neighbors by judging which ones fire in synchrony, much as activity patterns in the lateral geniculate nucleus or visual cortex determine which axons carry signals from the same eye. This information is then used to refine the topographic map, through mechanisms similar to those in the cortex. When all of the axons fire in synchrony, the tectum cannot determine which axons are neighbors; refinement fails, and the map remains coarse.

Auditory Maps Are Refined During a Critical Period

Activity-dependent refinement is not unique to the visual system. Activity is essential for shaping patterns of connectivity in most regions of the central nervous system. In the auditory system, for example, inputs form orderly tonotopic maps in the cochlear nucleus, the inferior colliculus, and the auditory cortex, such that neurons responding best to low frequencies lie at one edge of each structure, while neurons tuned to high frequencies are at the other. These maps underlie our sense of pitch. In addition, neurons vary in their sensitivity to sounds sensed by the contralateral and ipsilateral ears, and this discrepancy helps us determine the point in horizontal space from which a sound arises (see Chapter 31).

These auditory patterns are analogous to the retino-topic maps of the visual system. And, like their visual analogs, these auditory connections can be refined or modified by experience. Thus, although tonotopic maps can form in silence, exposure of an animal to "white noise" impairs refinement of the tuning curves of neurons in the inferior colliculus, while exposure to specific frequencies enhances representation of those frequencies in the map.

Neural circuit reorganization not only optimizes processing of information in one sensory modality but also brings information of multiple sensory modalities into register. Studies on barn owls have provided insight into how auditory and visual maps are coordinated during a critical period. During the day owls use vision to localize their prey—mice or other small rodents—but at night they rely on auditory cues, and at dusk both sensory channels are used. The localization of sound must be precise if owls are to succeed in finding prey, and it is intuitively obvious that the visual and auditory cues for the same location need to be consistent.

Auditory localization in owls, as in people, results in large part from computation of the temporal difference between a sound arriving at the two ears. This difference is only a few tens of microseconds, as expected from calculations based upon the speed of sound and the width of the head. Remarkably, the auditory system is sensitive to these extremely short interaural time differences (ITDs) and can calculate prey position from them (Figure 56–16). Moreover, many neurons in the optic tectum with receptive fields centered on a particular location are also tuned to ITDs that correspond to sounds emitted from that same point in space. The registration is imprecise at early stages but becomes progressively more precise during early adolescence as a consequence of the animal's experience.

Crucial insight into how this registration occurs came from experiments in which prisms were mounted over the eyes of owls of different ages. The prisms shifted the retinal image horizontally so that the visual map in the tectum reflected a world systematically displaced from its "actual" orientation. This change abruptly disrupted the correspondence between visual and auditory receptive fields. Over the next several weeks, however, the ITD to which tectal neurons responded optimally, ie, their auditory receptive fields changed until the visual and auditory maps came back into register (Figure 56–17). Thus the visual map instructs the auditory map.

Further experiments showed that this reorganization resulted from rewiring of connections between two deeper auditory nuclei (Figure 56–18). When prism goggles were placed on young owls, changes in ITD tuning were fully adaptive, in that the animals compensated completely for the effects of the prisms. In contrast, goggles placed on mature owls (older than 7 months of age) had little effect. Thus reorganization of this auditory projection occurs optimally during a critical juvenile period.

Distinct Regions of the Brain Have Different Critical Periods of Development

Not all brain circuits are stabilized at the same time. Even within the visual cortex the critical periods for organization of inputs differ among layers. As an example, the neural connections in layer IVC of the visual cortex of the monkey are not affected by monocular deprivation by the time the animal is two months old. In contrast, connections in the upper and lower layers continue to be influenced by sensory experience (or lack of it) for almost the entire first year after birth. Critical periods for other features of the visual system, such as orientation tuning, occur at different developmental stages (Figure 56–19A).

The timing of critical periods also varies between brain regions (Figure 56–19B). The adverse consequences of sensory deprivation for the primary sensory regions of the brain are generally fully realized early in postnatal development. In contrast, social experience can affect the intracortical connections over a much longer period. These differences may explain why certain types of learning are optimal at particular stages of development. For example, certain cognitive capacities—language, music, and mathematics—usually must be acquired well before puberty if they are to develop at all. In addition, insults to the brain at specific early stages of postnatal life may selectively affect the development of certain perceptual abilities and behavior.

By definition critical periods are limited in time. However, it is now clear that they are less sharply defined than previously thought. Extending or reopening critical periods in adulthood could increase brain plasticity and make it possible to facilitate recovery from strokes and other insults that inactivate discrete regions of the nervous system.

Some of the best evidence for the view that critical periods can be extended comes from studies on matching of auditory and visual maps in owls. In initial experiments the realignment of auditory and visual maps following displacement of the visual field with prism goggles was largely restricted to an early sensitive period. However, three strategies dramatically enhance binaural tuning plasticity in adult owls.

First, when adult owls that had worn goggles as adolescents are refitted with the goggles, the auditory map again shifts to align with the new visual map (Figure 56–20A). In contrast, in adult owls that had not worn the goggles as adolescents, the use of goggles has little effect on the organization of the auditory map. Thus the events of map rearrangement during the normal critical period must leave a neural trace that permits rearrangement later in life. In fact, in the owls that wore prisms in early life, axonal projections to auditory nuclei that were normally pruned were maintained, providing a structural basis for the reorganization in adulthood.

A second method for inducing late plasticity is to displace the retinal image in small steps by having the owl wear a series of prism spectacles of progressively increasing strength. Under these conditions adjustment of the auditory map is typically two- to threefold greater than the response to a single large displacement of the retinal image (Figure 56–20B).

The third technique is to allow owls to hunt live prey. In earlier experiments on plasticity in the optic tectum of owls, the animals were housed and fed under standard laboratory conditions. However, when adult prism-wearing owls are allowed to capture live mice under low light conditions for 10 weeks, they exhibit far greater plasticity than owls fed dead mice (Figure 56–20C), albeit less than that exhibited by juvenile owls that did not hunt. So hunting increases the plasticity of binaural tuning in adult owls. This finding provides a dramatic demonstration that behavioral context affects the ability of the nervous system to reorganize. Whether this effect results from increased sensory information, attention, arousal, motivation, or reward needs to be resolved.

Recent data suggest that critical periods can also be extended or reopened in mammals. There have even been reports that the best-studied critical period, the period for formation of ocular dominance columns, can be modified. Human amblyopia can be ameliorated in adulthood by training. In addition, new insights into the molecular mechanisms underlying critical periods in mice have suggested pharmacological strategies to extend or reopen critical periods. Encouraging results have already been obtained with drugs that alter inhibitory circuitry, destabilize perineuronal nets, or neutralize the growth inhibiting properties of myelin.

How can we reconcile the strong evidence for critical periods with the newer evidence for reorganization of circuitry in adults? The plasticity of critical periods can be distinguished from plasticity in adulthood by its magnitude and by the ease with which it is triggered. These differences result from two factors. First, from early postnatal life into adolescence the molecular environment in the brain is conducive to axonal growth, and cellular mechanisms are optimal for promoting the formation, strengthening, weakening, and elimination of synapses. Under these conditions circuits can undergo fundamental changes in their architecture and biochemistry in response to the animal's experience. Conversely, in mature circuits molecular and structural elements promote stability and impede plasticity.

Second, in a developing circuit no particular pattern of connectivity is firmly entrenched, so there is less to overcome. The connections specified by genetic determinants are less precise and the connections themselves are relatively weak. The patterns of neural activity that are stimulated by experience sharpen and even realign these patterns of connectivity. Once a pattern of connectivity becomes established, strong activation of the established circuit impedes the development of alternative wiring. This difference may help explain the special circumstances needed to trigger plasticity in adulthood. Circuits can be altered by passive exposure of animals to unusual environments during the critical period, whereas adult plasticity may require that the animal pay attention to the stimulus. In sum, experience during critical periods has a potent effect on circuits because the cellular and molecular conditions are optimal for plasticity and because the instructed pattern of connectivity does not have to compete with a long-existing pattern.

Connections among neurons in the mammalian brain arise from two fundamentally different developmental programs: molecular guidance cues and patterned neural activity. Molecular cues guide axons to target regions and initiate the formation of synaptic connections. Once synaptic contact is established, however, continued development depends on the coordination of neural activity between pre- and postsynaptic neurons.

In many instances activity is stimulated by sensory experience, allowing a circuit to refine itself to suit a particular environment. This remarkable ability allows the nervous system to be, in essence, individualized so that it performs optimally. Conversely, it would be maladaptive to allow experience to trigger large-scale reorganization of the nervous system in adults, once behaviors have been established and skills learned. Perhaps for this reason, refinement of neural circuits in response to environmental conditions is largely restricted to critical periods in early development.

Activity-dependent reorganization has been studied in greatest detail in the visual system, particularly in the development of a coordinated binocular view of the world. If afferent fibers carrying input from the same region of the retina of one eye converge on a common cortical neuron, they have an advantage. Their synchronous firing strengthens the synapses of all cooperating fibers, while the synapses of the noncooperating fibers from the other eye decline.

These functional changes lead to local structural changes, for example in dendritic spines, and eventually to large-scale structural changes, with some axon terminals sprouting new synaptic contacts and others withdrawing entirely. As a consequence, overlap of inputs from the two eyes is eliminated and the inputs from the two eyes become segregated into alternating columns in the cortex. If vision is lost in one eye, the cortex reorganizes to devote most of its space to the open eye: Responsiveness to the deprived eye decreases and columns of cells receiving input from the deprived eye shrink, while responsiveness to the open eye increases and its columns expand. One can think of this reorganization as adaptive in that it ensures that as much of the cortex as possible ends up being used.

Studies of monocular deprivation have also provided insight into mechanisms that restrict plasticity in particular circuits to critical periods. A proper balance of excitation and inhibition may be necessary to initiate the critical period. During this period the segregation of afferent fibers and establishment of ocular dominance columns is affected dramatically by changing the balance of activity in the fibers from the two eyes. After the critical period existing connections stabilize and are much less susceptible to such modification. Myelination of axons and formation of matrix-rich nets around interneurons contribute to the stabilization. Studies of the development of ocular dominance columns suggest how other, more complex sensory experiences early in development might change the circuitry and structure of the growing brain.

Because of the enduring nature of neural changes early in life, critical periods are times of both great opportunity and great vulnerability. If the environment to which an animal is exposed during a critical period is representative of the environment it will experience as an adult, plasticity shapes the connectivity of a circuit to make optimal use of cortical real estate and process information optimally. However, maladaptive stimulation during this period can lead to lifelong defects. Social deprivation in early childhood, for example, has lasting effects on social behavior. Some behavioral disorders such as autism may also result in part from aberrant or defective circuit maturation during critical periods.

It has long been clear that reopening of critical periods could, if properly controlled, have numerous benefits. Therapies could be designed to reorganize and retrain the brain to compensate for losses incurred by injuries and disease. Maladaptive reorganization in childhood could be remedied. It might even be possible for adults to learn new skills with some of the effortless efficiency that children exhibit during their critical periods. These possibilities remain science fiction, but it is heartening that recent studies have begun to reveal ways of extending or reopening critical periods in animals, by training regimens or in a few cases pharmacological intervention.