Education and the Brain: A Bridge Too Far

j O HN T. BR UE R

Educational Researcher, Vol . 26, No. 8, pp. 4-16 place, that indirectly link brain function with educational
practice. There is a well-established bridge, now nearly 50
years old, between education and cognitive psychology.
rain science fascinates teachers and educators, just as There is a second bridge, only around 10 years old, between

B it fascinates all of us. When I speak to teachers about

applications of cognitive science in the classroom,
there is always a question or two about the right brain ver-
cognitive psychology and neuroscience. This newer bridge
is allowing us to see how mental functions map onto brain
structures. When neuroscience does begin to provide use-
sus the left brain and the educational promise of brain- ful insights for educators about instruction and educational
based curricula. I answer that these ideas have been practice, those insights will be the result of extensive traffic
around for a decade, are often based on misconceptions and over this second bridge. Cognitive psychology provides the
overgeneralizations of what we know about the brain, and only firm ground we have to anchor these bridges. It is the
have little to offer to educators (Chipman, 1986). Educa- only way to go if we eventually want to move between ed-
tional applications of brain science may come eventually, ucation and the brain.
but as of now neuroscience has little to offer teachers in
terms of informing classroom practice. There is, however, a The Neuroscience and Education Argument
science of mind, cognitive science, that can serve as a basic The neuroscience and education argument relies on and
science for the development of an applied science of learn- embellishes three important and reasonably well-estab-
ing and instruction. Practical, well-founded examples of lished findings in developmental neurobiology. First, start-
putting cognitive science into practice already exist in nu- ing in infancy and continuing into later childhood, there is
merous schools and classrooms. Teachers would be better a dramatic increase in the number of synapses that connect
off looking at these examples than at speculative applica- neurons in the brain. This synaptic proliferation (synapto-
tions of neuroscience. genesis) is followed by a period of synaptic elimination.
The teachers' questions arise out of the perennial interest Second, there are experience-dependent critical periods in
in the brain and neuroscience that has always existed at the the development of sensory and motor systems. Third, in
margin of educational research and reform discussions. Re- rats at least, complex, or enriched, environments cause new
cently, however, interest in how neuroscience might im- synapses to form.
prove education has moved from the margins to center The argument runs as follows. Starting in early infancy,
stage. Educators and education policy experts are the most there is a rapid increase in the number of synapses or neural
vocal enthusiasts. Educational writers, likewise fascinated connections in children's brains. Up to age 10, children's
by the brain but puzzled by the mind, have picked up on brains contain more synapses than at any other time in their
this enthusiasm. Over the past year, there have been nu- lives. Early childhood experiences fine-tune the brain's
merous books, journal articles, policy studies, and stories in synaptic connections. In a process that we might describe as
the media about how our emerging understanding of brain synaptic pruning, childhood experiences reinforce and
development and neural function could revolutionize edu- maintain synapses that are repeatedly used, but ~nip away
cational practice. 1 Neuroscientists, while interested in how the unused synapses. Thus, this time of high synaptic den-
their research might find application outside the laboratory sity and experiential fine-tuning is a critical period in a
and clinic, are more guarded in their claims. Often they are child's cognitive development. It is the time when the brain
puzzled by the neuroscientific results educators choose to is particularly efficient in acquiring and learning a range of
cite, by the interpretations educators give those results, and skills. During this critical period, children can benefit most
by the conclusions educators draw from them. from rich, stimulating learning environments. If, during
This article examines those results, interpretations, and this critical period, we deprive children of such environ-
conclusions-a set of claims that I will call the neuroscience ments, significant learning opportunities are lost forever. As
and education argument. The negative conclusion is that one popular article put it, "with the right input at the right
the argument fails. The argument fails because its advocates
are trying to build a bridge too far. Currently, we do not
know enough about brain development and neural func-
tion to link that understanding directly, in any meaningful, JoHN T. BRUER is president of the James S. McDonnell Founda-
defensible way to instruction and educational practice. We tion, 1034 S. Brentwood Blvd., Suite 1850, St. Louis, MO 63117;
may never know enough to be able to do that. The positive phone 314-721-2068; e-mail [email protected]. He specializes in
conclusion is that there are two shorter bridges, already in cognitive science and the philosophy of science.

4 EDUCATIONAL RESEARCHER
time almost anything is possible," but "if you miss the win- tion points for helping children develop knowledge and
dow you're playing with a handicap" (Begley, 1996, p. 56). skills, positive attitudes toward learning, healthy behav-
Educators appeal to this argument to support a number iors, and emotional attachments of powerful and endur-
of claims. E. D. Hirsch Jr. (Hirsch, 1996) uses it to argue that ing significance. If these opportunities are squandered, it
Jerome Bruner was actually correct to claim that any subject becomes progressively more difficult and more expensive
can be taught effectively in some intellectually honest form to make up for the deficits later on. (Carnegie Task Force,
1996, p. 10)
to any child at any stage of development. According to
Hirsch, Bruner's claim now "represents the current think- What's wrong with this? In its synopsis, Years of Promise
ing in mainstream neurobiology. 'Nature' is actually telling cites two neuroscience articles and a keynote address on
us something very different from the message carried by brain development given by a science journalist. These are
the phrase 'developmentally appropriate.' What nature is the only references to the neuroscience literature in the en-
really saying about much learning much of the time is 'the tire report. Yet, it contains hundreds of citations to the cog-
earlier the better'" (p. 223). For Hirsch, neuroscience nitive, developmental, and social psychology literature.
proves that "developmentally appropriate" are dirty This latter literature, not the neuroscience, provides evi-
words. dence for the report's significant claims about the impor-
The claim that children are capable of learning more at a tance of early childhood. And, unfortunately, it has been
very early age, when they have excess synapses and peak primarily the neuroscience angle that commentators have
brain activity, is one of the more common ones made in the seized on in their secondary discussions of the report.
neuroscience and education literature. Neuroscience im- When I received a telephone inquiry from a journalist about
plies that if information is presented in ways that fit each the report, she wanted to know what I would advise par-
child's learning style, children are capable of learning more ents about choosing a preschool based on what neuro-
than currently believed (Education Commission of the science tells us about brain development. My answer was
States, 1996, p. vi). On this same evidence, other articles brief: "Based on neuroscience, absolutely nothing."
urge that children begin the study of languages, advanced We can't choose preschools based on neuroscience. Nor
mathematics, logic, and music as early as possible, possibly can we look to neuroscience as a guide to improved educa-
as early as age three or four. Parents should realize that they tional practice and policy. Our fascination with the brain
have a "golden opportunity to mold a child's brain. And leads us to overlook and underestimate what behavioral
that calls for a full-court press during the early years-that science can already provide to improve policy and practice.
is, a rich child-care environment without undue academic The neuroscience and education argument may be rhetori-
stress" (Viadero, 1996, p. 32). Parents should become deeply cally appealing, but scientifically, it's a bridge too far. To see
involved in their children's early education because "when why, let's review what neuroscientists do know about
brain activity is high, parents have a unique opportunity to synaptic growth, critical periods, and enriched environ-
foster a love of learning" (Abelson, 1996, p. 1819). One jour- ments.
nalist claims that, ideally, "at age 2'12 or 3, children would
start at Montessori school, where the educational program Synaptogenesis
comes closer to matching neurological findings than any I At birth, both nonhuman and human primate brains con-
know" (Beck, 1996, p. 23). tain synapses that connect brain cells into circuits.
The neuroscientific evidence shows, according to a vari- Neonates have slightly fewer synaptic connections than do
ety of educators, that there is a critical period for learning in adults. However, early in postnatal development, the in-
early childhood that is somehow related to the growth and fant brain begins to form synapses far in excess of adult
pruning of synapses. The ages for this critical period vary- levels. This process of synaptic proliferation, called synap-
birth to 3 years, birth to 6, birth to 10, 3 to 10. Educators cite togenesis, continues over a period of months that varies
this evidence to explain why some early childhood pro- among species. This period of synaptic overproduction is
grams are more successful than others. Developmental neu- followed by a period of synaptic elimination or pruning.
robiology can explain why Head Start programs fail to re- This experience-dependent pruning process, which occurs
sult in sustained improvements in children's IQs. Head over a period of years, reduces the overall number of
Start begins too late in children's critical learning period to synaptic connections to adult, mature levels, usually
rewire their brains (Begley, 1996, p. 56; Viadero, 1996, p. 33). around the time of sexual maturity for the species. The ma-
The neuroscience and education argument figures most ture nervous system has fewer synaptic connections than
prominently, however, in reports and policy studies, partic- were present during the developmental peak. It is the pat-
ularly to argue for the importance of early childhood edu- tern, rather than simply the number, of these connections
cation (Carnegie Task Force, 1996; Education Commission that form the mature brain's neural circuitry and that sup-
of the States, 1996; U.S. Department of Education, 1996). port normal brain function.
Among these reports, the Carnegie Task Force report, Years Most of what we know about synaptogenesis and synap-
of Promise, has deservedly received the most attention. Early tic pruning comes from animal research, primarily from ex-
in that report, there is a synopsis of developmental neuro- periments on cats and monkeys. The original demonstra-
biology that formulates the neuroscience and education ar- tion of overproduction and loss of synapses dates to 1975,
gument. Based on that argument, the report identifies the when Brian Cragg found that in the cat visual system the
years from 3 to 10 as a critical period in child development. number of synapses per neuron first increased rapidly and
This is a primary theme in the report: then gradually decreased to mature levels (Cragg, 1975a,
[T]he age span from three to ten [is] absolutely crucial for 1975b). The neuroscience and education argument, how-
children's optimal learning and development. These years ever, more typically cites the later work of Goldman-Rakic
offer families, communities, and schools critical interven- and Rakic on synaptogenesis in rhesus monkeys (Goldman-

NOVEMBER 1997 5
Rakic, 1987; Rakic, 1995). This work found that in rhesus togenesis occurs very early in the human visual cortex, but
monkeys, synaptic density (the number of synapses per in the frontal cortex, it appears to occur later and the prun-
unit volume of brain tissue) reaches maximal levels two to ing process takes longer. In the frontal cortex, synaptic den-
four months after birth and appears to do so simultane- sities do not stabilize at mature levels until mid- to late
ously in all areas of the cerebral cortex. Then pruning be- adolescence. This brain area, once thought not to be of
gins. For rhesus, synaptic densities gradually decline to much interest, is now thought to _b e the brain area respon-
adult levels at around three years of age, the time of sexual sible for planning, integrating information, and maintain-
maturity for that species. ing executive control of cognitive functions. Thus, what
In reviewing this work, readers outside the field should neuroscientists know about synaptogenesis does not sup-
be aware of its complexity and the methodological issues port a claim that zero to three is a critical period for humans.
involved. For example, measuring the number of synapses Whatever the time course of synaptogenesis in humans,
per neuron provides a more readily interpretable measure if it has relevance for child development and education, we
of synapse formation and loss than does synaptic density. must be able to associate this neurodevelopmental change
Between birth and adulthood, the human brain increases in with changes in infants' behavior and cognitive capacities.
volume by a factor of three or four. Thus, if the number of What kinds of learning and development do neuroscien-
synapses at birth remained constant, there would be a tists think occurs during this time?
three- to four-fold drop in synaptic density due entirely to When neuroscientists discuss the behavioral correlates of
changes in brain volume during development. Readers synaptogenesis, they often cite changes in the behavior and
should also be aware that whichever of these measures a cognitive capacities of monkeys. Again, this is not surpris-
study uses, we are measuring the aggregate number of ing because most of their research is on monkeys. When
synapses at any point in time. The measures reflect the they extrapolate from the animal research to human infants,
number added less the number lost between the times of they typically rely on the same set of examples (Chugani,
measurement. We know from other studies that different Phelps, & Mazziotta, 1987; Goldman-Rakic, 1987; Hutten-
classes of neurons in the same brain region gain and lose locher & de Courten, 1987). At the time synaptogenesis be-
synapses at different rates (Boothe, Greenough, Lund, & gins, at around 2 months of age, human infants start to lose
Wrege, 1979), and the same neurons can be adding synapses their innate, infantile reflexes. At age 3 months, when
in one part of their dendritic field, while losing them in an- synaptogenesis is well under way in the visual cortex, in-
other part (Greenough & Chang, 1985). Thus, even the best fants can reach for an object while visually fixating on it. At
measurements of synapses per neuron are only partial re- 4 to 5 months, infants' visual capacities increase. At 8
flections of synapse loss and gain. Brain development at months, infants first show the ability to perform working
this level is a complex process indeed, and the studies we memory tasks, such as Piaget's A-not B and delayed-re-
have to date give us only approximations to what is actu- sponse tasks. In these tasks, the infant watches while the ex-
ally happening in the brain. perimenter places an object that interests the infant in one
These difficulties aside, occasionally, one sees claims in of two hiding wells in front of the infant. The experimenter
the educational literature that the "critical period" in hu- covers both wells with identical covers, and for a period of
mans may be as early as from birth to age three years (Edu- 1 to 10 seconds, the experimenter prevents the infant from
cation Commission of the States, 1996). If based on neuro- looking at or moving toward the correct well. Then the in-
science, this claim makes two assumptions. First, it assumes fant is allowed to reach for the object. In order to make a
that the time course of synaptogenesis is the same for hu- correct response, the infant must remember where the ob-
mans as it is for rhesus monkeys. Second, it assumes that the ject was hidden. In A-not B, the experimenter continues to
period of synaptic excess is the critical period for learning. place the object in the same well until the infant makes sev-
Unfortunately, there is comparatively little data on syn- eral correct responses in a row. Delaye<;l-response tasks are
aptogenesis in humans. Counting synapses in slices of exactly the same, except that where the object is hidden
monkey or human brain tissue is slow, tedious work. Fur- varies randomly on each trial. Between 8 and 12 months of
thermore, human studies are more difficult than animal age, the time delay at which infants can succeed at this task
studies because researchers can only obtain specimens of increases steadily from a few seconds to 10 to 12 seconds.
brain tissue for study at autopsy. What data there are sug- By 18 to 24 months of age, a time after which synaptogene-
gest that synaptogenesis in humans follows a different time sis has peaked at least in the visual cortex, children begin to
course. The human neonate has approximately 2.5 x 108 use symbols, start to speak in sentences, and show spurts in
synapses per 100 mm3 of gray matter. In the visual cortex, vocabulary acquisition.
there is a rapid increase in the number of synaptic connec- These examples are all significant developmental mile-
tions at around 2 months of age, which reaches a peak at 8 stones that no doubt depend on brain development. We do
to 10 months. Then there is a steady decline in synaptic den- know that these milestones are correlated with synaptoge-
sity until it stabilizes at around 3.5 x lOS synapses I 100 mm3 nesis (at least in the visual cortex), but that is all we know.
at around age 10 years. Synaptic density in the visual cor- Educators should note two things, however. First, in all
tex remains at this level throughout adult life (Hutten- these examples, increases in synaptic density are correlated
locher, 1990). with the initial emergence of skills and capacities. These
Unlike the monkey, where synaptogenesis appears to skills and capacities continue to improve after synaptic den-
occur simultaneously across all regions of the brain, the lim- sities begin to regress to adult, mature levels. Some of these
ited human data suggest that changes in the number of skills and capacities continue to improve after synaptic den-
synapses per neuron or changes in synaptic density in our sity reaches mature levels. Thus, the most we can say is that
species may vary among brain areas. However, we have de- synaptogenesis may be necessary for the initial emergence
tailed data on only two regions of the human brain. Synap- of these abilities and behaviors, but it cannot account en-

6 EDUCATIONAL RESEARCHER
tirely for their continued refinement (Goldman-Rakic, irreversible consequences of early sensory deprivation pro-
1987). Some other form of brain maturation or change must vide a vivid, compelling image to underscore the over-
contribute to this ongoing development. Some other neural whelming importance of early childhood education.
mechanism must operate to support the significant learning In their experiments, Hubel and Wiesel sutured shut one
that takes place after synaptogenesis and pruning cease. eye of a cat or kitten, depriving the eye of all visual input.
Second, note that all these examples are examples of the They were particularly interested to determine the effect of
emergence or changes in sensory, motor, and working visual deprivation on what were assumed at that time to be
memory functions. The development of vision, tactile dis- innate cortical structures, ocular dominance columns. To
crimination, movement, and working memory are devel- test for the development of ocular dominance columns,
opmentally significant. However, these are not abilities and Hubel and Wiesel recorded how neurons in the animals' vi-
skills children learn in school or go to preschool to acquire. sual cortex responded to visual stimuli presented to the
Normal children in almost any environment acquire these normal eye and to the sutured eye after it was reopened. To
capacities at approximately the same age-children in af- simplify, let us assume that they recorded neural activity·in
fluent suburbs, children in destitute inner cities, children in the eat's left visual cortex. In a normal adult cat, the right
rural-pastoral settings throughout the world. No doubt, in (contra-lateral) eye activated around 20% of the cells from
some way, the development of these capacities supports fu- which they recorded. The left (ipsi-lateral eye) activated
ture learning. However, we have no idea, certainly no idea around 15% of the recorded cells. Around 65% of the cells
based on neuroscientific research, how the emergence of responded to input from either eye. Monocular deprivation
these capacities relates to later school learning or to the for periods between 3 and 16 months in adult cats had no
acquisition of culturally transmitted knowledge and skills. effect on this pattern of ocular dominance. The cells re-
Synaptogenesis is a significant neurodevelopmental pro- sponding exclusively to one or the other eye tended to
cess that occurs in a variety of mammalian species, most occur in alternating patches that neuroscientists refer to as
likely for good evolutionary reasons. We do know from an- "columns."
imal models that experience affects the pattern of synaptic Kittens responded differently. Newborn kittens deprived
connections and that it is the pattern, not just the number, of visual input in one eye for two to three months after birth
of connections that matters for normal brain function. How- did not form normal ocular dominance columns. In these
ever, our current understanding of synaptogenesis can tell kittens, 85% of the cells from which Hubel and Wiesel
educators little, if anything, about what kinds of early child- recorded responded only to visual input from the open eye
hood, preschool, or learning experiences might enhance and approximately 15% of the cells responded to neither
children's cognitive capacities or their educational out- eye. In a second experiment, Hubel and Wiesel deprived
comes. Given what we know about this complex develop- kittens of visual input in one eye for two to three months
mental process, it is premature, at best, to draw highly spe- and then performed a reverse suture, opening the deprived
cific educational conclusions and recommendations from eye and suturing the open eye. They let the cats navigate
this knowledge. their normal (laboratory) environment for periods up to a
year. They found that this reverse-closure operation had no
Critical Periods effect on recovery of either ocular dominance or visual
Research on critical periods has been prominent in devel- function. Over 90% of the cells from which they recorded re-
opmental neurobiology for over 30 years. This research has sponded only to stimulation of the initially open eye. In
shown that if an animal's sensory and motor systems-that both these experiments, Hubel and Wiesel reported that the
is, systems like vision or tactile discrimination-are to de- kittens remained functionally blind in the initially deprived
velop normally, then the animal must have certain kinds of eye and could not use the eye to guide their behavior. Sub-
experiential input at specific times during its development. sequent research suggested that some recovery is possible
William Greenough, a neuroscientist at the University of depending on the specific period of deprivation and the cir-
lllinois, provides a useful way to think about this develop- cumstances following deprivation (LeVay, Wiesel, & Hubel,
mental phenomenon: It is as if evolution has resulted in 1980).
neural systems that expect to find certain kinds of stimuli in Already in 1972, however, there were indications that
the environment in order to fine-tune their performance some recovery was possible. K. L Chow and D. L. Stewart
(Greenough, Black, & Wallace, 1987). (Chow & Stewart, 1972) deprived kittens of visual input to
In discussing critical periods, articles making the neuro- one eye for a period of one year after birth. They then did
science and education argument often refer to the Nobel the reverse closure for an additional year. Rather than just
prize-winning research of David Hubel and Tors ten Wiesel. letting the kittens navigate the laboratory environment,
They studied how visual deprivation affects the develop- however, they subjected the kittens to a training regimen
ment of cats' visual systems. In their 1965 article, Hubel and that forced them to use the initially deprived eye. Kittens
Wiesel wrote, "In kittens, monocular or binocular depriva- forced to use the initially deprived eye showed some re-
tion by lid suture for the first three months of life leads to covery. Chow and Stewart found that around 15% of the
virtual blindness, marked morphological changes in the lat- cells in the visual cortex responded to stimulation of the ini-
eral geniculate body, and a severe deterioration of innate tially deprived eye, around 65% of the cells responded to
cortical connections" (Wiesel & Hubel, 1965, p. 1071). They stimulation of the initially open then deprived eye, and
also showed that the same or longer periods of complete vi- around 10% of the cells responded equally to stimulation of
sual deprivation had no such effects on the visual system of either eye. This is not a normal ocular dominance pattern,
adult cats, nor on their ability to use the deprived eye, when but it represents an improvement over the recovery re-
it was subsequently reopened, to guide their behavior. For ported in Wiesel and Hubel' s article. Also, after the training
some educators and education writers, these rather drastic, regimen, the kittens regained sufficient function in the de-

NOVEMBER 1997 7
prived eye to use it to guide their behavior. With appropri- sitivity to deprivation, there seems to be a third phase of the
ate training, kittens recovered function in the deprived eye critical period. During this phase, the system remains suffi-
well after the critical period originally described by Hubel ciently plastic that appropriate sensory experience can com-
and Wiesel. pensate for deprivation and the animal can regain near-nor-
The neuroscience and education literature reports Wiesel mal function. With appropriate training and therapy, at the
and Hubel's finding in statements like this: "When they re- appropriate time, cats, monkeys, and humans can recover
moved the patches several weeks later, they found the kit- near-normal visual function following periods of depriva-
tens were blind in that eye .... The cats never did develop tion. How long critical periods and their phases last de-
vision in those once-patched eyes" Oones, 1995, p. 26). This pends on the specific function and on the maturational
both oversimplifies and misrepresents what we now know timetable for the brain areas that support the function.
about critical periods in neural development. Thus, to use our understanding of critical periods for ther-
Hubel and Wiesel launched an extremely important re- apeu tic purposes requires both identifyi ng specific compo-
search program in developmental neurobiology that con- nent functions within a system like vision and possessing
tinues to this day. Over the past 30 years, this research pro- detailed know ledge about the maturation and develop-
gram, engaging hundreds of neuroscientists, has advanced ment of particular brain areas. Neuroscientists are begin-
our u nderstanding of critical periods. (See, for example, ning to understand what some of these functions and areas
Daw, 1995.) Neuroscientists now understand that critical might be for vision, but we know relatively little about crit-
periods and synaptogenesis I synaptic pruning are related. ical periods for other sensory and motor systems.
Neural systems, particularly highly acute systems like vi- As a result, all this very interesting neuroscience provides
sion, have evolved to depend on the presence of ubiquitous little guidance or insight for educators. Critical periods are
environmental stimuli to fine-tune their neural circuitry. related to synaptogenesis. As with synaptogenesis, we have
These environmental stimuli maintain and re-enforce evidence for the existence of critical periods only for com-
synapses that are repeatedly used to process them, while ponent functions within sensory and motor systems and in
other synapses wither. This results in highly sensitive sen- humans for components of language. Currently, we do not
sory systems. Generally, critical periods coincide with the know if critical periods do or do not exist for culturally
period of excess synapse formation . Critical periods for transmitted knowledge systems-reading, arithmetic-
neural systems end at or about the time when synaptic den- that children acquire through informal social interaction
sities in the brain areas supporting that system stabilize at and formal school instruction. We do not know what role
mature levels. During this time, some neural systems, like synaptogenesis plays, if any, in the acquisition of these
vision, are particularly sensitive to the presence or absence skills. Given our current state of neuroscientific under-
of general kinds of stimuli. Neuroscientists also know that standing, however, we should be skeptical of claims that
there are different critical periods for specific functions . For attempt to generalize from what we know about critical
example, within the visual system, there are different criti- periods in brain development to critical periods for the ac-
cal periods for ocular dominance, visual acuity, binocular quisition of culturally transmitted knowledge.
function, and stereopsis (Daw, 1995). The human language If, as some neuroscientists think, over evolutionary time,
function also seems to have several critical periods. Based primates and other mammals evolved to rely on environ-
on behavioral, not neuroscientific evidence, the critical pe- mental features to fine-tune highly sensitive neural sys-
riod for phonology begins in infancy (Kuhl, 1994) and most tems, then these features are ubiquitous and available to
probably ends around age 12. There also appears to be a
any organism that inhabits any reasonably normal environ-
lengthy critical period for acquiring syntax that ends at
ment. Greenough calls this the "experience-expectant plas-
around age 16. In contrast to phonology and syntax, there
ticity" of sensory and motor systems (Greenough, Black, &
is no critical period for learning the lexicon. Our ability to
Wallace, 1987). The expected experiences must be present
acquire new vocabulary continues throughout our lifetimes
during certain developmental periods, but the expected ex-
(Neville, 1995).
Neuroscientists know that it makes little sense to speak periences are of a very general kind-patterned visual
of a critical period for vision or for any other sensory sys- input, the ability to move and manipulate objects, noises,
tem. Nor do they any longer interpret the critical period the presence of speech sounds. These kinds of stimuli are
phenomenon as "a window nature temporarily throws available in any child's environment, unless that child is
open then slams shut." Rather, they now tend to interpret abused to the point of being raised in a sensory-deprivation
critical periods in terms of subtle, possibly gradual, changes chamber. In short, experience-expectant brain plasticity
in brain plasticity-changes in the brain's ability to be does not depend on specific experiences in specific envi-
shaped and changed by experience that occur over the life- ronments, and for this reason, does not provide much guid-
time of the animal. ance in choosing toys, preschools, or early child-care poli-
Neuroscientists now also think that for each specific func- cies. The experiences children need to develop fundamen-
tion of a sensory system, like stereopsis in the visual system, tal sensory-motor and language skills occur in any normal
there are three distinct phases within the critical period environment. This makes sense from an evolutionary per-
(Daw, 1995). First, there is a time of rapid change during spective and from reflection on the cultural diversity in
which a function, like stereopsis, quickly reaches its mature child-rearing practices around the world. If infants really
processing level. This is followed by a second phase. Dur- needed highly specific experiences to become normal
ing this phase, if the animal does not continue to receive ap- adults, the human race would be extinct. Cultural varia-
propriate sensory input from the environment, the system tions in child rearing suggest that there are many equally
is still sufficiently plastic that deprivation can result in de- successful way to provide the normal environment needed
terioration or loss of that function. After the period of sen- for brain development.

8 EDUCATIONAL RESEARCHER
 Thus, we should be wary of arguments that use this neu- Greenough and his colleagues showed that even the brains
roscientific evidence in arguments for highly specific early of mature, adult rats form new synapses in response to new
childhood environments, experiences, and policies. Despite experiences. Unlike critical-period phenomena, the ability
what we see in the policy literature and read in the news- to create new synapses in response to new experiences
papers, as far as this developmental process is concerned, it seems to persist throughout the animal's life span.
matters little, if at all, whether the child is at home with Greenough's work suggests that there is a second kind of
Mom or in a Montessori preschool. brain plasticity. Whereas synaptogenesis and critical peri-
However, our understanding of critical periods does ods figure in experience-expectant plasticity, Greenough
have one important implication for early childhood care: It characterizes synaptic growth in complex environments
is exceedingly important that parents and teachers identify as experience-dependent plasticity. Experience-dependent
and treat children's sensory problems--cataracts, eye mis- plasticity allows an organism to acquire knowledge that is
alignment, chronic inner ear infections-as early as possi- specific to its own environment. It allows an organism .to
ble. Even binocular focal disparity and severe astigmatism learn about features of the particular environment that it in-
may have lasting effects. Normal fine-tuning cannot occur habits, environmental features that are not ubiquitous for
if the child cannot see, hear, or feel the ubiquitous environ- the entire species. For example, an animal must learn where
mental stimuli. And after the sensory problem is fixed, we to find water, food, and shelter in its environment. It must
must make sure--just like the kittens who were trained to learn to recognize significant conspecifics-its mother, its
use their once-deprived eye-that appropriate therapeutic siblings, the dreaded alpha male. Humans also have to
experiences are available so that children can regain normal learn these kinds of things, along with the specifics of our
function. surrounding culture. We must also learn the particular, spe-
Neuroscientific research on critical periods supports an cific features of our native languages. Among the particu-
educational moral or policy recommendation about the im- lar, specific features that vary widely depending on the
portance of diagnosing and treating children's sensory sys- sociocultural niche we inhabit is vocabulary. Our brain's ex-
tems. It gives us relatively little specific guidance about perience-dependent plasticity allows us to acquire knowl-
how to design early childhood learning environments. edge of these specifics throughout our lives.
Thus, research on the effects of complex environments on
Enriched Environments and Synaptic Growth the brain is exceedingly interesting and important because
The third theme in the neuroscience and education argu- it does begin to link learning with synaptic change and
ment makes a claim, based on neuroscientific grounds, for brain plasticity. It points to a kind of brain plasticity that is
the importance of enriched, stimulating early childhood en- present throughout the animal's lifetime. This kind of plas-
vironments. In support of this claim, proponents of the ar- ticity allows the animal to Jearn from experience. It allows
gument often cite the research of William Greenough and the animal to acquire knowledge about an environment and
his colleagues (Greenough et al., 1987). Greenough raises to use that knowledge to solve novel problems that arise in
rats in various environments and studies the effects of these that environment. It allows the animal to become more ex-
environments on synapse formation in the rats' brains. In a pert in negotiating its environment. This kind of brain plas-
series of experiments, he raised rats in what he calls "a com- ticity, unlike synaptogenesis and critical periods, might
plex environment." In a complex environment, several rats eventually provide a neural basis for the informal and for-
live together in large cages, filled with toys and obstacles. mal learning that goes on in our sociocultural environ-
Greenough calls these environments complex-not en- ments, including our schools.
hanced or enriched-because he intends that a complex en- This presents an intriguing possibility and one that is se-
vironment mimic rats' natural, wild environment. Complex ductive for both educators and journalists. However, edu-
environments are certainly enriched compared to typical cators should be cautious in interpreting this work and in
laboratory rearing conditions. Usually laboratory rats are thinking about how it might inform policy and practice. A
housed in individual cages or small group cages with no recent journalistic treatment of this research illustrates three
toys. Greenough has found that rats raised in complex en- things educators should keep in mind when considering
vironments are superior to their more austerely lab-reared the implications of research on complex environments.
mates on some learning tasks, like learning to run mazes. On the cover of the October 1996 issue of Tennis USTA, a
He and his colleagues have also found that rats raised in publication of the United States Tennis Association, the
complex environments have 20% to 25% more synapses per question, "Can tennis build brain power?," appeared under
neuron in the visual cortex. The rats also have more a picture of Gardner Mulloy. Mulloy is 82 years old and the
synapses per neuron in other brain areas, but the differ- winner (at last count) of 108 USTA Championships. A cap-
ences are not as large as those found in the visual cortex. tion described him as "Fit and feisty at 82." An article inside
However, Greenough is careful in interpreting his find- titled "Tennis and the Brain" briefly reviewed the research
ings. First, he argues that the synapse formation that occurs on what happens to rats reared in complex environments
as a result of living in a complex environment is largely not and gave an affirmative answer to the question on the
a critical-period phenomenon. The kind of brain plasticity cover. Sports that involve strategy stimulate brain-cell
that arises from rearing in a complex environment seems to growth and stave off the brain's aging process, the writer
rely on a neural mechanism that is very different from the concluded. Gardner Mulloy-fit, feisty, and 82-is living
pruning mechanism that gives rise to critical periods. Rear- proof.
ing in a complex environment, on Greenough's interpreta- Although educators and neuroscientists might want to
tion, causes new synapses to form in response to new and dispute the particulars of this journalistic treatment, the
varied experiences. Second, although his original experi- Gardner Mulloy and tennis example provides some useful
ments were done on newborn rats, in subsequent studies, reminders in thinking about complex environments. First,

NOVEMBER 1997 9
complex environment research is often cited as evidence for the brain to development, learning, and instruction. We
the critical importance of early childhood environments, simply do not know enough about how the brain works to
particularly in the years from birth to three. Gardner Mul- draw educational implications from changes in synaptic
loy provides a useful reminder that the research does not morphology. We do not know how synaptic change sup-
support such a simple conclusion. Although the effects of ports learning. There is a gaping chasm between our un-
complex environments may occur more readily in younger derstanding of what happens to synapses as a result of ex-
animals, the effects do occur throughout the life span. Older perience and what happens or should happen in preschool
animals learn, too. How we learn might change as we de- or third grade. The neuroscience and education argument
velop and mature, but the research does indicate that we attempts to bridge this chasm by drawing educationally rel-
learn and that the brain remains plastic throughout our evant conclusions from correlations between gross, unana-
lives. If so, we should be wary of arguments from the effects lyzed behaviors-learning to read, learning math, learning
of complex environments to the conclusion that there languages-and poorly understood changes in brain struc-
should be a selective educational focus on children's earli- ture at the synaptic level. This is the bridge too far. Our
est years. emerging understanding of the brain may eventually be
Second, we should note that Ceramics magazine could able to contribute to education, but it will require us, at least
have run the same article, "Ceramics and the Brain," with initially, to take a different, less direct route, a route that
an accompanying profile of Beatrice Woods "Throwing and links brain structures with cognitive functions and cogni-
thriving at 104." Muscle Magazine, Field and Stream, and, tive functions with instructional goals and outcomes.
possibly1 Country Living could run similar articles featuring
the appropriate mature practitioner. Not only are we life- Mind, Brain, and Education
long learners, but we can learn a myriad of things, all of If we cannot build the neuroscience and education bridge,
which no doubt affect our brains, all of which increase our but are interested in how brain structure supports cognitive
brain power if we are willing to equate brain power with function, we can pursue a more promising strategy that in-
synaptic growth. The reminder here is that the neuroscien- volves traversing two existing spans. The first connects ed-
tific evidence points to the existence of a general neural ucational practice with cognitive psychology, and the sec-
mechanism that contributes to life-long brain plasticity and, ond connects cognitive psychology with brain science.
presumably, to learning. Although this is intriguing, the Cognitive psychology is the study of mind and mental
finding provides little insight into how to teach tennis, function, a study not necessarily concerned with brain
ceramics, reading, or algebra. This research does not yet structure and function. Cognitive scientists attempt to dis-
provide much guidance in our attempts to answer the fun- cover the mental functions and processes that underlie ob-
damental educational question: How should we design in- served behavior. They attempt to analyze those functions
struction-how should we design complex, pedagogical into even more el~mentary cognitive operations. For exam-
environments-to optimize learning in any domain for ple, cognitive scientists analyze reading into a set of com-
children or adults? ponent cognitive skills that include word recognition,
Finally, that the article appeared in a tennis magazine in grammatical processing, text modeling, and metacognitive
itself provides a useful reminder. Tennis is an upscale, monitoring. In turn, they analyze word recognition into
largely middle-class activity. However, we should remind more elementary cognitive operations of initial encoding-
ourselves that we should be careful about drawing infer- forming a visual representation of the printed word-and
ences from the existence of a general neural mechanism to lexical access-determining if the visual representation
what subjects and skills children, or adults, should learn. In matches a word in the reader's language. In an educational
appealing to this research, advocates move too easily from context, this analytic method helps us understand the com-
"complex" to "enriched," where "enriched" is very much in ponent processes, skills, and knowledge structures that
the eye of the beholder, often reflecting the beholder's cul- underlie expertise in domains like reading, mathematics,
tural and class values. Rich, complex environments tend to writing, and science. Cognitive psychology already has a
include what the authors value and exclude what they justified claim to be the basic science of learning and teach-
abhor-Sesame Street but no other television, music lessons, ing (Bruer, 1993) and has contributed to the design of effec-
athletics of the right kind, early math instruction, attending tive instructional tools (McGilly, 1994). Although some
the right preschool, and having the right toys. Complex, en- might wish that more educators traveled it, there already is
riched environments for humans end up having many of a bridge between cognitive psychology and educational
the features of upper-middle class, urban, and suburban practice.
life-Gardner Mulloy and Mister Rogers' neighborhoods. We This same analytic method also supports a bridge be-
may have reasons to prefer Latin to Ebonies, Mozart to tween cognitive psychology and one area of neuroscience--
Buddy Guy, tennis to bowling, and suburbs to inner cities, cognitive neuroscience. Cognitive neuroscientists work at
but, we should remind ourselves, neuroscience does not the mind-brain interface, at the interface between biological
provide the reasons. and behavioral science. One thing they do as biological sci-
Despite its popularity, the neuroscience and education ar- entists is use brain imaging and recording techniques to
gument does not offer much support for the conclusions capture and analyze brain activity. As behavioral scientists,
and recommendations its advocates attempt to draw from they use the methods and models of cognitive psychology
it. What we know about synaptogenesis, critical periods, to identify and analyze cognitive functions that guide
and complex environments cannot provide much guidance human behavior. These cognitive analyses and models
for educational policy, classroom practice, or early child- allow cognitive neuroscientists to formulate informative,
hood education. The primary reason the argument fails is testable hypotheses about how brain structures implement
that it attempts to link what happens at the synaptic level in the mental functions that underlie learning and intelligent

10 EDUCATIONAL RESEARCHER
behavior. Fortunately, cognitive analyses and models pro- search on numerical cognition. The first piece of research
vide candidate mental functions that occur at levels of tem- addresses an educational problem. Sharon Griffin, Robbie
poral (seconds to milliseconds) and spatial (millimeters) Case, and Bob Siegler applied the methods of cognitive psy-
resolution that best exploit the power of current imaging chology to analyze the cognitive skills and knowledge chil-
and recording technologies. In contrast. synaptic change oc- dren must have to succeed in learning elementary arith-
curs at spatial resolutions on the order of ten thousandths metic (Griffin, Case, & Siegler, 1994). They found that the
of a millimeter. At this much lower level of temporal and ability to do numerical comparisons-which is bigger, 5 or
spatial resolution, cognitive neuroscientists are beginning 7?-is one such skill. They also found that some children
to identify the neural correlates and circuits that underlie from low-SES homes may not acquire this skill before en-
specific cognitive functions. tering school, but with appropriate instruction, they can
Among brain imaging technologies, the best known and acquire it. Their work is but one example of a bridge that ex-
most widely used are Positron Emission Tomography (PET) ists between cognitive psychology and instruction. The
and functional Magnetic Resonance Imaging (fMRI). PET second piece of research is from the field of cognitive neu-
can measure changes in cerebral blood flow, oxygen uti- roscience. Using a model that analyzes the numerical com-
lization, and glucose utilization that are linked to neural ac- parison skill into its subcomponents, Stanislas Dehaene
tivity. fMRI measures changes in the ratio of oxygenated to conducted a series of brain-recording experiments to trace
deoxygenated hemoglobin that, through a longer chain of the neural circuitry involved in making such comparisons
reasoning, can also be linked to neural activity. Both allow (Dehaene, 1996). This is an example of a second bridge that
cognitive neuroscientists to make images of these changes is now being rapidly built between cognitive psychology
in normal human brains as subjects perform cognitive and systems neuroscience, a bridge that will help us under-
tasks. PET and fMRI have a spatial resolution in the mil- stand how brain structures implement cognitive functions .
limeter range, but a temporal resolution of, at best, seconds. Cognitive psychology, a behavioral science committed to
These methods allow us to see how cognitive tasks change the analysis and scientific study of our mental capacities, is
brain activity at the level of cortical columns to cortical not only making fundamental contributions to educational
maps, brain structures that contain millions of synapses. practice and our understanding of the brain, but also pro-
However, because these technologies have relatively poor vides the intermediate level of analysis we need if we are
temporal resolution, they can tell us little about the timing ever to link brain and education.
and sequencing of the component processes in a cognitive
task. For example, consider skilled reading. A skilled reader The First Bridge: Instruction to Cognition
fixates and processes one word every 250 milliseconds. In a First, let's consider how cognitive psychology has con-
quarter of a second, a skilled reader can identify a word in tributed to the solution of an educational problem. Improv-
a text, assign the word a meaning and grammatical role, in- ing mathematics instruction is an acknowledged educa-
tegrate the word into a grammatical structure for the sen- tional problem for American schools. In the most recent
tence it occurs in, and incorporate an interpretation of that such study, the Third International Mathematics and Sci-
grammatical structure into a mental model of the "gist" of ence Study, U.S. eighth-grade students scored below the in-
the text. Thus, imaging techniques like PET and fMRI can ternational average in mathematics. In 20 of the 41 countries
helping us localize, one by one, areas of brain activity that surveyed, the average performance was significantly better
underlie the various cognitive components of reading; they than the average performance of U.S. students. In 13 coun-
cannot give us a picture, as yet, of how these areas interact tries, average performance was not significantly different
during reading. The limited temporal resolution (at best, from that in the U.S. The U.S. average student performance
seconds) of these technologies cannot tell us much about in mathematics wa~ significantly better than that of only 7
the temporal dynamics of the brain processes that underlie nations (Mathematics Achievement, 1996). Furthermore,
skilled reading because in skilled reading, too much hap- the National Assessments of Educational Progress (NAEP)
pens in one or two seconds. in mathematics reveals an additional, serious problem
Brain recording techniques like electroencephalography within the United States. Since the first NAEP math assess-
(EEG), event-related potentials (ERP), and magnetoen- ment in 1978, Black and Latino/ a students have 'improved
cephalagraphy (MEG) measure the electric or magnetic consistently over the years on this assessment, while scores
fields that neural activity generates at the scalp surface. for White students have remained relatively stable.
These methods have a temporal resolution in the millisec- Nonetheless, minority students still score significantly
ond range, but a spatial resolution of only centimeters. below their White majority counterparts. On the 1994
These techniques allow accurate timing of changes in brain NAEP math assessment, the average mathematics score for
activity during a cognitive task, but can localize that activ- White 9-year-olds was 237, compared to 212 for Black 9-
ity with a precision only in the range of tens to hundreds of year-olds and 210 for Latino/ a 9-year-olds (Campbell,
millimeters, often only to the level of hemispheric regions. Reese, O'Sullivan, & Dossey, 1994, p. 55). On tests of nu-
Thus, using cognitive models and analyses in imaging and merical concepts, significant numbers of children attending
recording experiments, cognitive neuroscientists can map kindergarten in low-income, inner-city communities did
elementary cognitive operations, occurring on a time scale not demonstrate number knowledge comparable to that of
between milliseconds and seconds, onto brain structures their middle-income peers (Case & Griffin, 1990). These ini-
that range in size from hemispheric regions (centimeters) to tial differences tend to increase, rather than disappear, dur-
cortical columns (millimeters). ing the elementary school years (NSF Survey, 1988). If it
The following example illustrates how cognitive psy- were possible to eliminate these differences at school entry,
chology can begin to link educational questions with cog- many fewer students would be at risk for math failure and
nitive neuroscience. The example presents two pieces of re- many more might successfully complete the elementary

NOVEMBER 1997 11
school mathematics curriculum. How might we change the Griffin and her colleagues evaluated the Number Worlds
arithmetic and number skills curricula in kindergarten and curriculum in Worcester, MA, public schools (Griffin &
first grade to eliminate these differences? Cognitive and de- ase, 1993). Using a diagnostic test for number knowledge,
velopmental psychologists have helped us answer this they identified a group of entering kindergarten students
question. who scored below criterion on the test and were thus
Over the past 20 years, cognitive and developmental psy- judged at risk for failing their first formal arithmetic in-
chologists have delineated the basis and emergence of nu- struction. These students all attended neighborhood
merical cognition. Humans, like other animals, possess an schools in Worcester, and most of them were from low-in-
innate, preverbal sensitivity to quantity. Building on this come homes. Half of these children received the Number
innate sensitivity, by age two to three years, human infants Worlds curriculum and half, the control group, received the
begin to learn the number names in their native language school system's standard early math curriculum. Griffin
and to use these number names in counting-like activities also identified a group of students who attended a city-
and games. By age three or four years, most children can wide magnet school and who scored well on the diagnostic
compare two small numbers for size and reliably determine test. These students were a normative group, who could be
which is larger and which is smaller. Soon after, they are expected to achieve at or above levels one could reasonably
able to coordinate their counting and comparing skills to in- expect in a public school system.
vent counting strategies for solving simple arithmetic prob- On entering kindergarten, the Number Worlds and con-
lems. Among these invented strategies are the well-known trol group students scored significantly below the expected
counting-up, counting-on, and min strategies. Before enter- kindergarten score of 9 to 11 points on the number knowl-
ing school, most children also learn Arabic numerals and edge test. The students in the normative group scored at or
become adept at using their invented strategies to learn above the criterion on kindergarten entry. After one year of
basic number facts . Most children seem to acquire these Number Worlds instruction, scores on the number knowl-
rudiments of numerical cognition informally, before they edge test for the Number Worlds students did not differ sig-
begin formal arithmetic instruction in kindergarten or first nificantly from normative group's scores. Students in the
grade. control group, however, still lagged significantly behind.
Robbie Case, Sharon Griffin, and Bob Siegler character- After an additional year of Number Worlds in grade one,
ized this informally acquired number knowledge as the the treatment and normative groups were again indistin-
central conceptual structure that is a necessary prerequisite guishable, scoring at or above the criterion on the number
for learning formal arithmetic in the early elementary knowledge test. The control group still lagged significantly
grades (Griffin, Case, & Siegler, 1994). Children who pos- behind.
sess this central conceptual structure have what Case, Grif- One might argue that this evaluation identified children
fin, and Siegler call a "mental number" and the rudiments on the basis of poor performance on the number knowledge
of number sense. Children who have this conceptual struc- test, taught them skills needed to pass the test, and then
ture know the number names in their language, know that tested them again with the same instrument. Other assess-
when one counts it is necessary to assign one number name ments, however, showed that the Number Worlds children
and only one number name to each object, and know that also improved significantly compared to the control group
the number names occur in a fixed order. They know that students on tasks and problems usually encountered in first
the number names refer to set sizes and know that as one grade arithmetic. In fact, the Number Worlds children
moves up (down) through the number names in the count- scored significantly higher than even the normative group
ing process, each number name refers to a set that has one on solving word problems, expressing numerical relations
more (less) object in it then the previously named set. They in number stories, and doing successive mental arithmetic
also understand that the Arabic numerals are alternative operations. In another series of tests, the three groups re-
written symbols they can use to name set sizes. ceived the first 20 questions from a test of number skills
Griffin and Case's "Rightstart" curriculum (more re- used in a crossnational comparative study (Stevenson, Yee,
cently renamed "Number Worlds") attempts to teach this & Stigler, 1986). On this test, the Number Worlds children
conceptual structure-to teach number sense-to kinder- outperformed the normative group by a considerable mar-
garten, first, and second-grade students (Griffin, Case, & gin, and their performance exceeded or equaled that of
Siegler, 1994). The curriculum meets the special needs of Japanese students on this test. After two years in this cog-
children who may not have acquired the mental number nitively based curriculum, students known to be at risk for
line before entering school. Most children acquire this con- failing early arithmetic performed at world-class levels on
ceptual structure informally through interactions with par- this test of number skills.
ents and siblings before they enter kindergarten. Children The Number Worlds curriculum, now undergoing fur-
who have not acquired it require formal instruction to do ther refinement and evaluation in Worcester and in other
so. The kindergarten Number Worlds curriculum explicitly cities, is an example of a how a detailed understanding of
teaches the number-word sequences from 1 to 10 and 10 to numerical cognition can help solve an educational problem.
1, the one-one correspondence rule fundamental to count- It is based on a cognitive model that specifies the compo-
ing, and that numbers name set sizes. It teaches the incre- nent skills and pieces of number knowledge children must
menting and decrementing rules on the number line-i.e., possess and coordinate if they are to have number sense
that moving up (down) the number line in increments of 1 and succeed at early mathematics instruction. The Number
is equivalent to adding (subtracting) 1. The curriculum Worlds curriculum builds a bridge between cognitive psy-
teaches that the Arabic numerals are alternate names for the chology and educational practice.2
numbers. Finally, it explicitly teaches children how to com- The detailed cognitive model provided an informed basis
pare numbers for size. for the diagnostic number knowledge test Griffin and Case

12 EDUCATIONAL RESEARCHER
used in the Number Worlds project. It allowed them to de- In his experiment, right-handed college students had to de-
termine which component numerical skills and knowledge cide if a number flashed on a computer screen was larger or
facets children had or had not acquired before entering smaller than five, then press a key to indicate their response.
kindergarten. Individual test items tapped students' Dehaene manipulated three independent factors, where
knowledge of the number names, their ability to count, each factor was assumed to influence processing within
their ability to compare numbers for size, and their ability only one of the model's stages. For the stimulus identifica-
to solve simple arithmetic problems (presumably by coor- tion stage, he contrasted subjects' performance when given
dinating their counting and comparing skills). In pilot stud- Arabic (1, 4, 6, 9) versus verbal notation (one, four, six,
ies to norm this test, the research team discovered, to its sur- nine). For the magnitude comparison stage, he compared
prise, that there were significant differences in children's subjects' performance on close (4, 6 and four, six) versus far
test performance based on their families' socioeconomic (1, 9 and one, nine) comparisons to the standard 5. His rea-
status. Among entering kindergarten students, there were son for choosing this factor is the well-established distance
relatively small differences in the ability to count, compare, effect (Moyer & Landauer, 1967). The distance effects shows
and add sets of visible, physically present objects. However, that it takes subjects longer, and they make more errors,
when children were asked to solve similar problems men- when asked to compare numbers that are close in numeri-
tally, using only verbal statements of the problem, lower- cal value than when asked to compare numbers that are far-
SES children scored significantly below higher-SES children. ther apart in numerical value. In Dehaene's experiment,
This difference was most pronounced on the numerical half the comparisons were close comparisons and half were
comparison task. On questions like, "Which number is far comparisons, a factor that should affect only the magni-
bigger, 5 or 4?," 96% of high-SES children could answer cor- tude comparison stage. Finally, on half the trials, subjects
rectly, but only 18% of low SES-children could do so (Grif- had to respond "larger" using their right hand and
fin, Case, & Siegler, 1994, p. 31). Numerical comparison was "smaller" using their left hand, and on half the trials,
the component skill that most distinguished the high-SES, "larger" with their left and "smaller" with their right. This
math-ready children from the low-SES, at-risk children. The factor should influence reaction times only for the motor
Number Worlds curriculum taught numerical comparison preparation and execution stage.
explicitly and was most successful in imparting this skill to When Dehaene analyzed subjects' reaction times on the
the at-risk children, who, for some reason, had failed to ac- numerical comparison task, he found that the overall me-
quire it at the "normal" age seen in most middle-class chil- dian (correct) reaction time was around 400 milliseconds.
dren. The cognitive psychologists' analysis of numerical Subjects needed less than one half second to decide if a
cognition into its component skills revealed that numerical number was greater or less than 5. Furthermore, he found
comparison plays a crucial role in children's ability to ben- that each of the three factors had an independent influence
efit from formal arithmetic instruction on school entry. on reaction time. Reactions to Arabic stimuli were 38 mil-
liseconds faster than those for verbal notation, far compar-
The Second Bridge: Cognition to Neural Circuitry isons were 18 milliseconds faster than close comparisons,
Numerical comparison is but a single, albeit important, skill and right-hand responses were 10 milliseconds faster than
within numerical cognition. It is at this level of analysis, left-hand responses. Finally, the three factors had an addi-
however, that cognitive neuroscientists are beginning to un- tive effect on subjects' total reaction times, just as one would
derstand how neural structures and brain circuits imple- expect if subjects were using the serial-processing model.
ment cognitive processes. Using cognitive models and brain Dehaene's experiment, however, went beyond the typi-
recording techniques, they can begin to trace the neural cir- cal cognitive experiment that would have stopped with the
cuitry involved in a skill like numerical comparison. analysis of reaction times. Dehaene also recorded event-re-
Cognitive psychologists have further analyzed numerical lated potentials (ERPs ), while his subjects performed the
comparison into its subcomponents and developed cogni- number comparison task. His ERP system measured elec-
tive models of this process. One simple model analyzes nu- trical currents emerging from the scalp at 64 sites, currents
merical comparison into three stages. First, there is an iden- that presumably were generated by the electrical activity of
tification stage in which the subject identifies the input large numbers of nearby neurons. Recall that ERPs have rel-
stimuli. Second, there is a magnitude comparison stage in atively poor spatial resolution, but relatively precise tem-
which the subject judges which of the stimuli is larger or poral resolution. Significant changes in the electrical activ-
smaller. Third, there is a response stage in which the subject ity recorded at each of the 64 sites as subjects compared
prepares and executes a verbal or motor act to indicate his numbers might give general indications about where the
or her answer to the comparison question. This is a serial neural structures were in the brain that implemented the
model of numerical comparison. It assumes that three three processing stages. The ERPs' more precise temporal
stages occur one after the other and that processing in a resolution might indicate the time course of the three pro-
later stage does not influence processing in an earlier stage. cessing stages. Together, the spatial and temporal data
Cognitive psychologists have methods to assess the va- would allow Dehaene to trace, at least approximately, the
lidity of a serial stage model that rely only on behavioral neural circuitry that is active in numerical comparison. A
data. One of these methods is the Additive Factors Method. cognitive model together with brain recording techniques
If a serial model is correct, then altering a factor that affects created the possibility of mapping sequences of elementary
only one stage of the process-here, either identification, cognitive operations onto their underlying neural struc-
comparison, or response-should influence subjects' reac- tures and circuits.
tion times only for that processing stage. This first significant ERP effect Dehaene observed oc-
Stanislas Dehaene (Dehaene, 1996) designed such an ex- curred 100 milliseconds after the subjects saw either the
periment to test this serial model of numerical comparison. Arabic or verbal stimulus. This change in brain activity was

NOVEMBER 1997 13
not influenced by any of the experimental factors. It ap- liably and unambiguously support. In fact, Dehaene sug-
peared to occur in the right posterior portion of the brain. gests, the existence of such a bilateral system could explain
Based on this and other imaging and recording experi- some of the puzzling features about the patterns of lost ver-
ments, early activation in that part of the brain is most likely sus retained number skills neuropsychologists see in these
the result of the brain's initial, nonspecific processing of vi- patients. Second, this experiment suggests there is a brain
sual stimuli. area in the right hemisphere that is used in numerical com-
At approximately 146 milliseconds after stimulus pre- parison. This area might be the site of an abstract represen-
sentation, Dehaene observed a notation effect. When sub- tation of numerical magnitude, a representation that is in-
jects processed number words, they showed a significant dependent of our verbal number names and written num-
negative electrical wave on the electrodes that recorded ber symbols. This, too, runs counter to common neuropsy-
from the left posterior occipital-temporal brain areas. In chological wisdom. Neuropsychologists commonly hold
contrast, when subjects processed Arabic numerals, they that the left parieto-occipito-temporal junction, not the
showed a similar negative electrical wave on electrodes right, is the critical site for number processing because dam-
recording from both the left and right posterior occipital- age to this area in the left hemisphere causes acalculia. De-
temporal areas. This suggests that number words are haene's finding of right hemisphere involvement during
processed primarily on the left side of the brain, but that the comparison phase suggests that neuropsychologists
Arabic numerals are processed on both the left and right should look more carefully than they might have in the past
sides. at numerical reasoning impairments among patients who
To look for a distance effect and the timing and localiza- have suffered damage to the right posterior brain areas.
tidn of the magnitude comparison stage, Dehaene com- They might find, for example, patients who are able to read
pared the ERPs for digits close to 5 (4, four and 6, six) with Arabic numerals and perform rote arithmetic calculations,
the ERPs for digits far from 5 {1, one and 9, nine). This com- but who are unable to understand numerical quantities,
parison revealed a parieto-occipto-temporal activation in make numerical comparisons, or understand approximate
the right hemisphere that was associated with the distance numerical relations.
effect. This effect was greatest approximately 210 millisec- Dehaene' s work is just one example of how cognitive
onds before the subjects gave their responses. What is sig- neuroscience is advancing our understanding of how brain
nificant here, according to Dehaene, is that the timing and structures might support cognitive function. Cognitive neu-
distribution of the electrical currents was similar for both roscientists at numerous institutions are starting to trace the
Arabic digits and verbal numerals. This supports the claim, neural circuitry for other cognitive constructs and culturally
Dehaene argues, that there is a common, abstract, notation- transmitted skills. Several studies suggest that automatic
independent magnitude representation in the brain that we and controlled processing rely on distinct brain circuits
use for numerical comparison. To make a numerical com- (Raichle et al., 1994). Other studies show how attention can
parison, we apparently translate both number words and reorder the sequence in which component cognitive skills
Arabic digits into this abstract magnitude representation. are executed in a task: The areas of brain activation remain
Finally, Dehaene found a response-side effect that oc- the same, but the sequence in which the areas become ac-
curred approximately 332 milliseconds after the stimulus or tive changes (Posner & Raichle, 1994, ch. 6). We are begin-
equivalently 140 milliseconds before the key press. This ap- ning to understand the different brain systems that under-
peared as a substantial negative wave over motor areas in lie language processing and their developmental time
the brain. The motor area in the left hemisphere controls course (Neville, 1995). Using our rather detailed cognitive
movement of the right side of the body, and the motor area models of reading-particularly word recognition-PET,
in the right hemisphere controls movement of the left side fMRI, and ERP studies allow us to trace the neural circuitry
of the body. Thus, as expected, this negative wave appeared for early reading skills and to document the developmental
over the left hemisphere for right-hand responses and over course of this circuitry in children between the ages of 5 to
the right hemisphere for left-hand responses. 12 years (Posner, Abdullaev, McCandliss, & Sereno, in
Dehaene's experiment exemplifies how cognitive neuro- press).
scientists use cognitive theories and models in brain Imag- However, in most cases, we are still far from under-
ing and recording experiments. Well-designed, inter- standing how these results might contribute to advances in
pretable imaging and recording studies demand analyses of the clinic, let alone in the classroom. It is not yet clear how
cognitive tasks, construction of cognitive models, and use we move from results like these across the bridge to educa-
of behavioral data, like reaction times, to validate the mod- tional research and practice. The example does, however,
els. Experiments like these suggest how neural structures make two things clear. First, there is no way that we could
implement cognitive functions, tell us new things about possibly understand how the brain processes numbers by
brain organization, and suggest new hypotheses for further looking at children's classroom or everyday use of numbers
experiments. or by looking at math curricula. Second, there is no way we
Dehaene's experiment traces the approximate circuitry could possibly design a math curriculum based on De-
the brain uses to identify, compare, and respond to number haene' s results. It is the cognitive research, exemplified here
stimuli. The experiment reveals several new things about by the work of Griffin, Case, and Siegler, that creates both
brain organization that suggest hypotheses for further ex- of those possibilities.
periments. First, the experiment points to a bilateral neural When we do begin to understand how to apply cognitive
system for identifying Arabic digits. This is something that neuroscience in instructional contexts, it is likely that it will
one could not discover by analyzing behavioral data from first be of most help in addressing the educational needs of
normal subjects. Nor is it a finding neuropsychologists' special populations. Cognitive psychology allows us to un-
studies of patients with brain lesions and injuries could re- derstand how learning and instruction support the acquisi-

14 EDUCATIONAL RESEARCHER
tion of culturally transmitted skills like numeracy and liter- Marchese, T. (1996). The search for next-century learning. Wing-
acy. Cognitive psychology in combination w ith brain imag- spread Journal, 18(3). The Johnson Foundation.
Public Radio International. Gray matters: The developing brain. Avail-
ing and recording technologies also allows us to see how able at http://www.dana.org/dana/gray.html.
learning and instruction alter brain circuitry. It opens the Shanker, A. (1996, October 27). Building brains. New York Times Week
possibility of being able to see and to c.ompare these learn- in Review, p . 7.
ing-related changes in normal-versus-speciallearning pop- Shore, K. (1997). Rethinking the brain. New York: Families and Work
ulations. Such comparative studies might yield insights Institute.
White House Conference. (1997, April 17). Early childhood develop-
into specific learning problems and, more importantly, into ment and learning: What new research on the brain tells us about our
alternative, compensatory strategies, representations, and youngest children.
neural circuits that children with learning disabilities might Your child. (1997). Special edition of Newsweek.
2 Despite this demonstrated success, parents and teachers recently
exploit. These insights could in tum help us develop better
asked Case why one should bother teaching number sense. Wouldn't
instructional interventions to address specific learning it be more effective and beneficial, they suggested, to exploit the
problems. Mozart effect? These parents and teachers had read about the contri-
butions of brain science to education. (See Begley, 1996; Jones, 1995.)
The Mozart effect is the claim that when children exercise cortical neu-
Conclusion rons by listening to classical music (unfo rtunately not R&B or heavy
The brain does and should fascinate all of us, and we should metal), they are also strengthening brain circuits used for mathemat-
ics. 5o the neuroscience and education articles do have an audience
find advances in neuroscience exciting. As educators, we and do have repercussions for instruction and research.
should also be interested in how basic research might con- 3 The author thanks the editors and reviewers of Educational Re-

tribute to and improve educational practice. However, we searcher for their many constructive comments and criticisms on earlier
should be wary of claims that neuroscience has much to tell versions of this article.
us about education, particularly if those claims derive from
the neuroscience and education argument. The neuroscience
and education argument attempts to link learning, particu- References
larly early childhood learning, with what neuroscience has
discovered about neural development and synaptic change. Abelson, P. H. (1996). Preparing children for the future. Science, 274,
Neuroscience has discovered a great deal about neurons and 1819.
Beck, J. (1996, October 17). A meeting of minds between neuroscien-
synapses, but not nearly enough to guide educational prac- tists and educators is first step in improving America's schools.
tice. Currently, the span between brain and learning cannot Chicago Tribune, Section 1, p. 23.
support much of a load. Too many people marching in step Begley, S. (1996, February 19). Your child's brain. Newsweek, 55-62.
across it could be dangerous. Boothe, R. G., Greenough, W. T., Lund, J. S., & Wrege, K. (1979). A
If we are looking for a basic science to help guide educa- quantitative investigation of spine and dendritic development of
neurons in visual cortex (area 17) of Maccaca nemistrina monkeys.
tional practice and policy, cognitive psychology is a much Journal of Comparative Neurology, 186, 473-490.
better bet. It already is helping us solve educational prob- Bruer, J. T. (1993). Schools for thought: A science of learning in the class-
lems and design better instructional tools. Cognitive psy- room. Cambridge, MA: MIT Press.
chology, in the hands of cognitive neuroscientists, is also Campbell, J. R., Reese, C. M., O'Sullivan, C., & Dossey, J. A. (1994).
NAEP 1994trends in academic progress. Washington, DC: National Li-
fundamental to our emerging understanding of how neural brary of Education, Office of Educational Research and Improve-
structures support and implement cognition functions. If, ment, U.S. Department of Education.
in the future, brain research does contribute to educational Carnegie Task Force. (1996). Years of promise: A comprehensive learning
practice, it will most likely do so via the indirect, two-bridge strategy for America's children. New York: Carnegie Corporation of
New York.
route, not the direct one espoused in the neuroscience and Case, R., & Griffin, S. (1990). Child cognitive development. In C.-A.
education argument. Hauert (Ed.), Developmental psychology: Cognitive, perceptuo-motor and
Looking to the future, we should attempt to develop an psychological perspectives (pp. 193-230). Amsterdam, The Nether-
interactive, recursive relationship among research pro- lands: Elsevier Science.
grams in education, cognitive psychology, and systems Chipman, S. (1986). Integrating three perspectives on learning. InS. L.
Friedman, K. A. Klivington, & R. W. Peterson (Eds.), The brain, cog-
neuroscience. Such interaction would allow us to extend nition, and education (pp. 203-229). San Diego, CA: Academic Press.
and apply our understanding of how mind and brain sup- Chow, K. L., & Stewart, D. L. (1972). Reversal of structural and func-
port learning. In the meantime, we should remain skeptical tional effects of long-term visual deprivation in cats. Experimental
about brain-based educational practice and policy, but look Neurology, 34, 409--433.
Chugani, H. T., Phelps, M. E., & Mazziota, J. C. (1987). Positron emis-
more carefully at what behavioral science already can tell sion tomography study of human brain functional development.
us about teaching, learning, and cognitive development. 3 Annals of Neurology, 22,487-497.
Cragg, B. G. (1975a). The development of synapses in kitten visual cor-
tex during visual deprivation. Experimental Neurology, 46, 445-451 .
Cragg, B. G. (1975b). The development of synapses in the visual sys-
Notes tem of the cat. Journal of Comparative Neurology, 160, 147-166.
Daw, N. (1995). Visual development. New York: Plenum Press.
1 In addition to the cited references, other prominent mentions of the Dehaene, S. (1996).The organization of brain activations in number
neuroscience and education argument include: comparison: Event related potentials and the additive-factors
Caine, R. & G. (1996). Making connections. Wingspread Journal, 18(3), method. Journal of Cognitive Neuroscience, 8(1}, 47~ .
The Johnson Foundation. Education Commission of the States. (1996). Bridging the gap between
Carnegie Task Force. (1994). Starting points: Meeting the needs of our neuroscience and education. Denver, CO: Education Commission of
youngest children. New York: Carnegie Corporation of New York. the States.
Clinton, B. (1997, February 4). President Clinton 's 1997 State of the Goldman-Rakic, P. S. (1987). Development of cortical circuitry and cog-
Union Address. nitive function. Child Development, 58, 601-622.
The Johnson Foundation. (1997, February 3). How a child's brain de- Greenough, W. T., Black, J. E., & Wallace, C. S. (1987). Experience and
velops. Time. brain development. Child Development, 58, 539-559.

NOVEMBER 1997 15
Greenough, W. T., & Chang, F.-L. C. (1985). Synaptic structural corre- Neville, H. J. (1995). Developmental specificity in neurocognitive de-
lates of information storage in mammalian nervous systems. In C. velopment in humans. In M. S. Gazzaniga (Ed.), The cognitive neuro-
W. Cotman (Ed.), Synaptic plasticity and remodeling (pp. 335-372). sciences (pp. 219-231). Cambridge, MA: MIT Press.
New York: Guilford. NSF Survey. (1988). Science, 241, 408.
Griffin, 5. A., & Case, R. (1993). Teaching number sense. Year 2 report to Posner, M. I., Abdullaev, Y. G., McCandliss, B. D., & Sereno, S. E. (in
the James S. McDonnell Foundation. St. Louis, MO. press). Anatomy, circuitry, and plasticity of reading. In J. Everett
Griffin, S. A., Case, R., & Siegler, R. S. (1994). Rightstart: Providing the (Ed.), Visual and attentional processes in reading and dyslexia. London:
central conceptual prerequisites for first formal learning of arith- Routledge.
metic to students at risk for school failure. InK. McGilly (Ed.), Class- Posner, M.l., & Raichle, M. E. (1994). Jmages of mind. New York: Scien-
room lessons: Integrating cognitive theory and classroom practice (pp. tific American Library.
25-49). Cambridge, MA: MIT Press. Raichle, M. E., Fiez, J. A., Videen, T. 0., MacLeod, A.-M. K., Pardo,
Hirsch, E. D., Jr. (1996). The schools we need and why we don't have them. J. V., Fox, P. T., & Petersen, S. E. (1994). Practice-related changes in
New York: Doubleday. human brain functional anatomy during nonmotor learning. Cere-
Huttenlocher, P. (1990). Morphometric study of human cerebral cortex bral Cortex, 4(1 ), 8-26.
development. Neuropsychologia, 28(6), 517-527. Rakic, P. (1995). Corticogenesis in human and nonhuman primates. In
Huttenlocher, P., & de Courten, C. (1987). The development of M.S. Gazzaniga (Ed.), The cognitive neurosciences (pp. 127-145). Cam-
synapses in striate cortex of man. Human Neurobiology, 6(1), 1-9. bridge, MA: MIT Press.
Jones, R. (1995, November 26). Smart brains: Neuroscientists explore Stevenson, H. W., Yee, S. Y., & Stigler, J. W. (1986). Mathematics
the mystery of what makes us human. American School Board Journal, achievement of Chinese, Japanese, and American children. Science,
p. 26. 231, 693--699.
Kuhl, P. (1994) Learning and representation in speech and language. U.S. Department of Education. (1996). Building knowledge for a nation of
Current Opinion in Neurobiology, 4(6), 812-822. learners: A fram ework for education research 1997. Washington, DC:
LeVay, S., Wiesel, T. N., & Hubel, D. H. (1980). The development of oc-
U.S. Department of Education, Office of Educational Research and
uJar dominance columns in normal and visually deprived monkeys.
Joumal of Comparative Neurology, 191, 1-51. Improvement.
Mathematics Achievement. (1996). Mathematics achievement in the mid- Viadero, D. (1996, September 18). Brain trust. Education Week, p. 31-33.
dle school years: lEA 's third i11ternationalmathematics and science report. Wiesel, T. N., & Hubel, D. H. (1965). Extentofrecovery from the effects
Chestnut Hill, MA: TIMSS International Study Center, CSTEEP, of visual deprivation in kittens. Journal of Neurophysiology, 28,
Campion Hall323, Boston College. 1060-1072.
McGilly, K. (1994). Classroom lessons: Integrating cognitive theory and
classroom practice. Cambridge, MA: MIT Press. Manuscript received March 10, 1997
Moyer, R. S., & Landauer, T. K. (1967). Time required for judgements Revision received August 1, 1997
of numerical inequality. Nature, 215, 1519-1520. Accepted September 8, 1997

16 EDUCATIONAL RESEARCHER