Instruction versus Selection
The 10,000 or so synapses per cortical neuron
are not established immediately. On the contrary,
they proliferate in successive waves from birth to
puberty in man. . . . One has the impression that the
system becomes more and more ordered as it receives
"instructions" from the environment. If the
theory proposed here is correct, spontaneous or
evoked activity is effective only if neurons and
their connections already exist before interaction
with the outside world takes place. Epigenetic
selection acts on preformed synaptic substrates. To
learn is to stabilize preestablished synaptic
combinations, and to eliminate the surplus.
--Jean-Pierre Changeux[1]
The most complex object yet discovered anywhere in
the universe is the organ that fills the space
between our ears. Although weighing only about 1300
to 1500 grams (three to four pounds), the human brain
contains over 11 billion specialized nerve cells, or neurons,
capable of receiving, processing, and relaying the
electrochemical pulses on which all our sensations,
actions, thoughts, and emotions depend.[2] But it is not the sheer number of
neurons alone that is most striking about the brain,
but how they are organized and interconnected. And to
understand how neurons communicate with each other we
first must consider their typical structure.
Although there are many different types of neurons,
almost all of them share certain common features as
portrayed in figure 5.1. The cell body, or soma,
contains the nucleus of the neuron, which in turn
houses a complete set of the organism's genes. The
nucleus is surrounded by cytoplasm, the chemical
"soup" of the cell that contains the
organelles essential to the neuron's functioning and
metabolism. In these respects, neurons are similar to
other cells throughout the body, except for the fact
that unlike most other cells they rarely divide to
reproduce new neurons.
The ways in which neurons are specialized to carry
out their communicative function is made evident by
closer examination of the appendages they sport, that
is, their dendrites and axons. The dendrites can be
likened to a bushy antenna system that receives
signals from other neurons. When a dendrite is
stimulated in a certain way, the neuron to which it
is attached suddenly changes its electrical polarity
and may fire, sending a signal out along its single
axon where it may be picked up by the dendrites of
other neurons.[3]
Considering the small size of the neuron's body, the
length of an axon can be considerable, up to several
meters in the neck of the giraffe. Thus the firing of
one neuron can influence the firing of another one a
considerable distance away.
For one neuron to influence another, the two must
be connected, and this is accomplished by junctions
called synapses (figure 5.2). These synaptic
junctions usually connect the axon of one neuron with
the dendrites of another, a typical neuron in the
cortex of the human brain having about 10,000
synapses. The synapses therefore constitute an
exceedingly complex wiring system that surpasses by
many orders of magnitude the complexity of even the
most advanced supercomputers. It is this organization
of connections both within the skull and to more
distant sense organs and muscles that gives the brain
its amazing abilities. Indeed, it is widely believed
today by neuroscientists, psychologists, and even
philosophers that all of the knowledge the human
brain contains--from being able to walk to the
ability to perform abstract scientific and
mathematical reasoning--is a function of the
connections existing among the neurons.
How this unfathomably complex organization allows
us to perceive, behave, think, feel, and control our
environment presents us with what may be the most
striking puzzle of fit we have yet encountered. The
puzzle actually has three aspects. First, we must
consider how over millions of years the primitive
nervous system of our early ancestors evolved into an
organ that has made it possible for the human species
to become the most adaptable and powerful organism on
the planet--living, thriving, and modifying the
environment (both intentionally and unintentionally)
from the tropics to the polar regions, and perhaps
soon in outer space and on other planets.[4] Second, we must
understand how it is possible for the intricate
structure of the brain to develop from a single
fertilized egg cell. Finally, we must try to
comprehend how the mature brain is able to continue
to modify its own structure so that it can acquire
new skills and information to continue surviving and
reproducing in an unpredictable, ever-changing world.
In this chapter we will consider the research and
theories that are beginning to provide answers to
these questions. Indeed, the 1990s has been referred
to as "the decade of the brain," as
scholars and scientists in fields from philosophy to
molecular neurobiology focus their energies on
understanding humankind's ultimate inner frontier.
Neurons are quite distinct from other body cells
in ways that make them suited to their specialized
role of signal processing and communication, but it
is not too difficult to see how they could have
evolved from less specialized cells. All living cells
are surrounded by a cell membrane that separates the
special chemical composition of its interior from
that of the external world. This difference in
chemical composition results in a small electrical
potential between the inside and outside of the cell,
in much the same way that a voltage exists between
the two sides of a battery. When a part of a cell's
membrane is disturbed in a certain way, it loses its
electrical potential, becoming depolarized at the
site of the disturbance. This sudden change in
electrical potential can itself be a disturbance,
causing additional depolarizations along the membrane.
In most cells, such depolarization would not spread
far, certainly not to neighboring cells. But a few
changes in the shape and arrangement of cells (in
just the way that neurons are fashioned) permits
depolarization to propagate quickly from one neuron
to the next, and allows it to travel quickly as an
electrochemical signal from one end of an animal to
the other.
An example of a simple nervous system is provided
by the jellyfish (or Medusa). The jellyfish's
nervous system forms an undifferentiated network and
serves primarily to coordinate the animal's swimming
motions. Since the jellyfish's skirt must open and
contract in a coordinated manner for the animal to
move through the water, its nervous system serves as
a simple communications network making it possible
for all parts of the skirt to open repeatedly and
then contract at the same time.
Worms are the simplest organisms to have a central
nervous system, which includes a distinct brain that
is connected to groups of neurons organized as nerve
cords running along the length of its body. This more
complicated nervous system allows worms to exhibit
more complex forms of behavior. An anterior brain
connected to a nerve cord is the basic design for all
organisms with central nervous systems, from the
earthworm on the hook to the human on the other end
of the fishing rod. But although we can discern a
separate brain in worms, it is not the case that the
brain is the sole "commander" of the animal
that the rest of the nervous system and body obeys.
Indeed, even with its brain removed, worms are able
to perform many types of behaviors, including
locomotion, mating, burrowing, feeding, and even maze
learning.[5]
As we move to insects we find increased complexity
in all aspects of the brain and nervous system. So-called
giant fiber systems (also found to some extent in
worms and jellyfish) that allow rapid conduction of
nerve impulses connect parts of the brain to specific
muscles in legs or wings. Such connections permit the
cockroach to dart away as soon as it senses the
moving air preceding a quickly descending human foot.
The brain itself is typically divided into three
specialized segments, the protocerebrum, the
deutocerebrum, and the tritocerebrum. In addition,
insects possess a greater variety of sensory
receptors than any other group of organisms,
including vertebrates, that are sensitive to the
odors, sounds, light patterns, texture, pressure,
humidity, temperature, and chemical composition of
their surroundings. The concentration of these
sensory organs on the insect's head provides for
rapid communication with the tiny yet capable brain
located within.
Although minuscule by human standards, the range
of abilities made possible by insect brains is
impressive. These creatures show a remarkable variety
of behaviors for locomotion, obtaining food, mating,
and aiding the survival of their offspring. They can
crawl, hop, swim, fly, burrow, and even walk on water.
The female wasp hunts down a caterpillar, paralyzes
it with her venom, and then lays its egg on the
motionless prey so that her offspring will have a
fresh and wholesome meal immediately after hatching.
Leafcutter ants harvest leaves and bring them into
their nest where they use them to cultivate indoor
gardens of edible fungus. Honeybees live in social
communities where there is a strict division of labor,
and where foodgathering worker bees perform a special
dance to communicate the location and richness of
food sources to their hivemates. It is the evolution
of their brains, together with the complementary
evolution of their other body parts, that make
insects the most abundant multicellular organisms on
our planet.
The brain becomes both much larger and still more
complex as we move to vertebrates such as fish,
amphibians, and reptiles. The spinal cord, now
protected within the vertebrae of the backbone, has
become primarily a servant of the brain, a busy two-way
highway of communication with fibers segregated into
descending motor pathways and ascending sensory ones.
The brain itself is now composed of a series of
swellings of the anterior end of the spinal cord (the
brain stem), the three major ones making up the three
major parts of the vertebrate brain: the hindbrain,
midbrain, and forebrain. From the hindbrain sprouts a
distinctive structure, the "cerebellum" (Latin
for "little brain").
Among mammals, the brain keeps its three major
components, but with two new structures. The
neocerebellum ("new cerebellum") is added
to the cerebellum, looking much like a fungal growth
at the base of the brain, and the neocortex ("new
cortex") grows out of the front of the forebrain.
In most mammals, these new additions are not
particularly large relative to the brain stem. In
primates they are much larger, and in the human they
are so large that the original brain stem is almost
completely hidden by this large convoluted mass of
grey neural matter. In keeping with this remarkable
increase of neocerebellar and neocortical tissue,
humans enjoy the largest ratio of brain weight to
body weight of any of earth's creatures.
It is not possible to know exactly why the human
brain evolved as it did, but consideration of the
structural evolution of the brain and results of
comparative research on human and nonhuman brains
provides some useful clues. It is now believed that
during the long evolution of our brain, nervous
systems changed in four principal ways. First, they
became increasingly centralized in
architecture, evolving from a loose network of nerve
cells (as in the jellyfish) to a spinal column and
complex brain with impressive swellings at the
hindbrain and forebrain. This increasingly
centralized structure also became increasingly hierarchical.
It appears that newer additions to the human brain
took over control from the previous additions and in
effect became their new masters. Accordingly, the
initiation of voluntary behavior as well as the
ability to plan, engage in conscious thought, and use
language depend on neocortical structures. Indeed,
the human neocortex can actually destroy itself if it
wishes, as when a severely depressed individual uses
a gun to put a bullet through his or her skull.
Second, there was a trend toward encephalization,
that is, a concentration of neurons and sense organs
at one end of the organism. By concentrating neural
and sensory equipment in one general location,
transmission time from sense organs to brain was
minimized. Third, the size, number, and variety of
elements of the brain increased. Finally, there was
an increase in plasticity, that is, the brain's
ability to modify itself as a result of experience to
make memory and the learning of new perceptual and
motor abilities possible.
One way of understanding the evolution of the
human brain is to see it as the addition of higher
and higher levels of control. We will see in chapter
8 that the function of animal and human behavior can
be understood as the control of perceptions, with
perceptions corresponding to important aspects of the
environment. For a sexually reproducing organism to
survive and leave progeny, it must be able to control
many different types of perceptions, that is, sensed
aspects of its environment. At a minimum, it must be
able to find food, avoid enemies, and mate. But as
life evolved, the environment of our ancestors became
more complex due to increasing numbers of competing
organisms. So it would have been of considerable
advantage to be able to perceive and control
increasingly complex aspects of this environment. The
bacterium E. coli can control its sensing of
food and toxins only in a primitive way; organisms
with more complex brains are able to sense and
control much more complex aspects of their
surroundings.
This capacity for increased environmental control
is nowhere more striking than in our species. Using
the advanced perceptual-behavioral capacities of our
brain together with our culturally evolved knowledge
of science and technology, we can visit ocean floors,
scale the highest peaks, and set foot on other worlds.
(The role that language is believed to have had in
the evolution of the human brain will be considered
separately in chapter 11.) But can the most complex
human abilities and mental capacities be explained by
natural selection? Our brain has certainly not
changed appreciably over the last couple of hundred
years, and yet we can solve mathematical, scientific,
technological, and artistic problems that did not
even exist a hundred years ago. So how could natural
selection be responsible for the striking abilities
of today's scientists, engineers, and artists?
This is actually the same problem that troubled
Alfred Russel Wallace, as mentioned in chapter 3. It
will be recalled that Wallace, despite being an
independent codiscoverer of natural selection, could
not, for example, imagine how natural selection could
account for Africans' ability to sing and perform
European music, since nothing in their native
environment could have selected for such an ability.
Consequently, for him the brain could only be a
creation provided to us by God. We now know that in
his embrace of this providential explanation, Wallace
failed to realize that natural selection can lead to
new abilities unrelated to those that were originally
selected.
To use an example from technological evolution,
the first personal computers were used to perform
financial calculations in the form of electronic
spreadsheets. However, these same machines with the
proper software could also be used for word
processing, telecommunications, computer games, and
many other purposes, even though they were not
originally designed with these functions in mind. A
classic example of this phenomenon of functional
shift in biological evolution is the transformation
of stubby appendages for thermoregulation in insects
and birds into wings for flight.[6] In the same way, selection
pressure was undoubtedly exerted on early hominids to
become better hunters. The ability to understand the
behavior of other animals and organize hunting
expeditions must have been very important in the
evolution of our species. And the increasingly
complex and adapted brain thus selected would have
made other skills possible, such as making tools and
using language, traits that in turn could become
targets for continued natural selection. This
transformation of biological structures and behaviors
from one use to another was given the unfortunate
name of preadaptation by Darwin, unfortunate since it
can too easily be misunderstood to imply that somehow
evolution "knows" what structures will be
useful for future descendants of the current
organisms.
American evolutionary paleontologist Stephen Jay
Gould provided a better term for this phenomenon--exaptation.
He made a major contribution to our understanding of
evolution by insisting that we distinguish adaptation,
the evolutionary process through which adaptedly
complex structures and behaviors are progressively
fine-tuned by natural selection with no marked change
in the structure's or behavior's function, from exaptation,
through which structures and behaviors originally
selected for one function become involved in another,
possibly quite unrelated, function. Exaptation makes
it difficult if not impossible to understand why our
brain evolved as it did. Although the brain allows us
to speak, sing, dance, laugh, design computers, and
solve differential equations, these and other
abilities may well be accidental side effects of its
evolution. As Gould and his associate Vrba cautioned:
. . . current utility carries no automatic
implication about historical origin. Most of what
the brain now does to enhance our survival lies
in the domain of exaptation--and does not allow
us to make hypotheses about the selective paths
of human history. How much of the evolutionary
literature on human behavior would collapse if we
incorporated the principle of exaptation into the
core of our evolutionary thinking?[7]
But although we may never know the actual events
and specific selection pressures responsible for our
brain power, we have no scientific reason to believe
that evolution could not have fashioned our brain
through natural selection. The fact that living
organisms today have nervous systems and brains
ranging from quite simple to amazingly complex is
compelling evidence that our brain evolved through
forgotten ancestors in progressive stages from simple
to complex. And somehow, as a part of this
evolutionary process, that most remarkable and
mystifying of all natural phenomena came into being--human
consciousness.
From this evolutionary perspective, one might be
led to conclude that our brain in all its striking
adapted complexity is an inherited legacy of
biological evolution. That once evolved it is
thereafter provided to each individual by good
old natural selection, specified in all its fine
detail in the genome and transmitted through the
generations from parent to offspring.
This type of genetically providential thinking of
course is selectionist from the viewpoint of
biological evolution, but nonetheless providential at
the level of the individual organism. It can be seen
in the pioneering work on brain development and
function of Roger Sperry for which he shared a Nobel
prize in 1981. This research in the 1950s involved
disturbing the normal location of nerve fibers in the
developing brains of fish and rats. For example,
nerve fibers that normally connect the top part of
the fish's retina with the bottom part of the brain,
called the optic tectum, were surgically removed and
reconnected to the top part of the optic tectum.
Despite this modification, the nerve grew back to its
normal position in the brain. Similar experiments
carried out by other researchers on rats indicated
that fibers that innervate muscles also "knew"
to which muscle they should be attached and made
their proper connections despite surgical
disturbances. This led Sperry to conclude that the
connections of the nervous system are completely
specified in the organism's genes. As his former
student Michael Gazzaniga explains:
In the original Sperry view of the nervous
system, brain and body developed under tight
genetic control. The specificity was accomplished
by the genes' setting up chemical gradients,
which allowed for the point-to-point connections
of the nervous system.[8]
But there is a vexing problem with the notion that
the genome provides complete information for the
construction of the nervous system of humans and
other mammals. It is estimated that just the human
neocortex alone has about 1015 (one followed by 15
zeros, or one thousand million million) synapses.[9] Since the human
genome has only about 3.5 billion (3.5 x 109)
bits of information (nucleotide base pairs), with 30%
to 70% of these appearing silent,[10] some neural and molecular
scientists have concluded that our genes simply do
not have enough storage capacity to specify all of
these connections, in addition to including
information on the location and type of each neuron
plus similar information for the rest of the body.
The problem is not unlike trying to save a document
made up of 100 million characters on a computer disk
that can hold only 1.4 million characters. As
Changeux noted:
Once a nerve cell has become differentiated it
does not divide anymore. A single nucleus, with
the same DNA, must serve an entire lifetime for
the formation and maintenance of tens of
thousands of synapses. It seems difficult to
imagine a differential distribution of genetic
material from a single nucleus to each of these
tens of thousands of synapses unless we conjure
up a mysterious "demon" who selectively
channels this material to each synapse according
to a preestablished code! The differential
expression of genes cannot alone explain the
extreme diversity and specificity of connections
between neurons. [11]
Additional understanding of the relation between
the genome and the nervous system can be gained by
considering Daphnia magna. Commonly referred
to as the water flea or daphnid, this small fresh-water
crustacean is familiar to many aquarium owners since
it is relished by tropical fish. But what makes the
daphnid interesting for our current purposes is that
when the female is isolated from males, she can most
conveniently reproduce by the asexual process of
parthenogenesis, giving birth to genetically
identical clones. In addition, the daphnid has a
relatively simple nervous system that facilitates its
study. If its genome completely controlled the
development of its nervous system, it should be the
case that genetically identical daphnids should have
structurally identical nervous systems. However,
examination of daphnid eyes using the electron
microscope reveals that although genetically
identical clones all have the same number of neurons,
considerable variation exists in the exact number of
synapses and in the configurations of connections
leading to and away from the cell body of each neuron,
that is, the dendritic and axonal branches. As we
move to more complex organisms, the variability of
their nervous systems increases. This provides clear
evidence that the structure and wiring of the nervous
system are not the result of following a detailed
construction program provided by the genes.
How then is the brain able to achieve the very
specific and adapted wiring required to function in
so many remarkable ways? For example, how does a
motor neuron know to which particular muscle fiber it
should connect? How is a sensory neuron in the visual
system able to join itself to the correct cell in the
visual cortex located in the occipital lobe of the
brain? If this detailed neuron-to-neuron connection
information is not provided by the genes, whence does
it come?
The first clues to solving this puzzle go back to
1906 when it was observed that in embryonic nerve
tissue, some neurons did not stain well and appeared
to be degenerating and dying.[12] Since it had been assumed that
in a developing embryo, nerve cells should be increasing
in number and not dying off, this finding was
somewhat surprising. But nerve cell death in the
developing nervous system has since been observed
repeatedly. The extent to which it occurs was
dramatically demonstrated by Viktor Hamburger. He
found that in a certain area of the spinal cord of
the chicken embryo over 20,000 neurons were present,
but that in the adult chicken only about 12,000, or
60%, of these cells remained.[13] Much of this neuronal death
occurs during the early days of the embryo's
existence. Nerve cells continue to expire thereafter,
albeit at a slower pace.
A particularly striking example of neuronal
elimination in development involves the death of an
entire group of brain cells:
Most frequently, neuron death affects only
some of the neurons in a given category. However,
in one case . . . a whole category of cells dies.
These particular neurons of layer I, the most
superficial layer of the cerebral cortex,
characteristically have axons and dendrites
oriented parallel to the cortical surface rather
than perpendicular to it, like the pyramidal
cells. These cells were first observed in the
human fetus but have since been found in other
mammals. Purely and simply, they disappear in the
adult.Changeux (1985, p. 217).[14]
But the death of obviously useless brain cells
cannot account for the specific connections that are
achieved by the remaining neurons. For example, the
visual cortex of cats and monkeys has what are called
ocular dominance columns within a specific
region known as cortical layer 4. In any one column
of this brain area in the adult animal we find only
axons that are connected to the right eye, while in
the neighboring column are located only axons with
signals originating from the left eye. So not only
must the axons find their way to a specific region of
the brain, which can be quite far from where their
cell bodies are located, they must also find a
specific address within a certain neighborhood.
The ability of axons to connect to the appropriate
regions of the brain during development has been
studied in careful detail since the beginning of this
century. Axons grow in the brain like the stem of a
plant. At the end of the growing axon is found a
growth cone which was described by Spanish
neuroscientist Ramón y Cajal in 1909 as "a sort
of club or battering ram, possessing an exquisite
chemical sensitivity, rapid amoeboid movements, and a
certain driving force that permits it to push aside,
or cross, obstacles in its way . . . until it reaches
its destination."[15]
Although the exact mechanisms by which this is
accomplished are still unknown, it appears that the
growth cone is sensitive to certain chemicals along
its path that are released by its target region. In
this way visual system axons originating in the
lateral geniculate nucleus find their way to cortical
layer 4 in the occipital lobe of the brain in much
the same way that a police bloodhound is able to
sniff out the escaped prisoner hiding in an Illinois
cornfield.
But although these growth cones lead their axons
to the proper region of the brain (or muscle in the
case of motor neurons), they cannot lead them to the
precise target addresses. For a particular growth
cone, it appears that any cell of a particular type
will serve as a target. Indeed, in the newborn cat,
ocular columns receive axons from both eyes, not just
from one or the other, as in the adult brain. For
this final and important fine-tuning to be achieved (on
which stereoscopic vision depends), many of the
original terminal connections of the axon must be
eliminated. In the case of vision, all axonal
connections from the wrong eye are eliminated, and
those from the correct eye are retained. In the case
of motor systems that initially have many-to-many
connections between motor neurons of the spinal
column and muscle fibers (that is, many motor neuron
axons connected to same muscle fiber, and many muscle
fibers connected to the same axon), the mature animal
possesses a much more finely ordered system with each
muscle fiber enervated by one and only one motor
neuron. The mammalian nervous system changes from
birth to maturity from a degenerate system having
many redundant and diffuse connections, to a much
more finely tuned system that makes both adaptedly
complex behaviors and perceptions (such as
stereoscopic vision) possible.
So now the question naturally arises, how does the
nervous system know which connections to retain and
which to eliminate? The work of David Hubel and
Torsten Wiesel in the 1970s (both of whom shared a
Nobel prize with Sperry in 1981) provided the first
clue. They conducted their ground-breaking
experiments by closing the lid of one eye of newborn
cats, and found that even one week without sight
altered the connections of the eyes to layer 4 of the
occipital cortex. Axons carrying nervous signals from
the closed eye made fewer connections with the cortex,
whereas axons from the open eye made many more
connections than was normal. This suggested that
visual system axons compete for space in the
visual cortex, with the result of the competition
dependent on the amount and type of sensory
stimulation carried by the axons. Subsequent research
by others using drugs to block the firing of visual
system neurons, as well as artificial stimulation of
these neurons, showed that it is not neural activity
per se that results in the selective elimination of
synapses, but rather that only certain types of
neural activity result in the retention of certain
synapses, while all others are eventually eliminated.
In a sense, then, cells that fire together
wire together. The timing of the action-potential
activity is critical in determining which
synaptic connections are strengthened and which
are weakened and eliminated. Under normal
circumstances, vision itself acts to correlate
the activity of neighboring retinal ganglion
cells, because the cells receive inputs from the
same parts of the visual world.[16]
The dependence of the development of the visual
system on sensory stimulation would seem to indicate
that the fine-tuning of its connections would have to
wait until the birth of the animal when it is
delivered from the comforting warm darkness of the
womb to the cold light of day. However, recent
evidence suggests that this fine-tuning actually
begins to take place in utero. Prenatal development
appears to depend on spontaneous firing of retinal
cells that do not depend on light stimulation from
the external world. Similar endogenous patterns of
activity may also exist in the spinal cord, and may
refine the synaptic connections of motor systems as
well.[17]
Nonetheless, interactive postnatal experience of
the external world is required for normal
development of senses and nervous systems in mammals.
Cats who have one eye sewn shut at birth lose all
ability to see with this eye when it is opened
several months later. The same applies to humans.
Before the widespread use of antibiotics, eye
infections left many newborn infants with cloudy
lenses and corneas that caused functional blindness,
even though their retinas and visual nervous systems
were normal at the time of birth. Years later a
number of these individuals underwent operations to
replace their cloudy lenses and corneas with clear
ones, but it was too late. Contrary to initial
expectations, none of these people was able to see
after the transplant.[18]It
was simply not known at the time that early visual
experience was essential to the normal maturation of
the brain's visual circuitry. Similarly, some
children are born with a wandering eye that does not
fixate the same part of the visual field as the
normal eye, and other children have one eye that is
seriously nearsighted or farsighted; in both cases,
the retina of the abnormal eye must be provided with
clear visual stimulation, usually by age four years,
or it will become functionally blind since its
connections to the brain's vision centers will be
eliminated in favor of the normal eye.
We thus see that the normal development of the
brain depends on a critical interaction between
genetic inheritance and environmental experience. The
genome provides the general structure of the central
nervous system, and nervous system activity and
sensory stimulation provide the means by which the
system is fine-tuned and made operational. But this
fine-tuning does not depend on adding new components
and connections in the way that a radio is assembled
in a factory, but rather it is achieved by eliminating
much of what was originally present. It is as if the
radio arrived on the assembly line with twice as many
electrical components and connections as necessary to
work. If such an overconnected radio were plugged in
and turned on, nothing but silence, static, or a hum
would be heard from its speaker. However, careful
removal of unnecessary components and judicious
snipping of redundant wires would leave just those
components and connections that result in a
functioning radio. This snipping is analogous to the
elimination of synapses in the human brain as part of
its normal development.
The process by which brain connections change over
time as maturing animals interact with their
environments has been studied in detail by
psychologist William Greenough of the University of
Illinois at Urbana-Champaign. Using sophisticated
techniques for determining the numbers and densities
of neurons and synapses in specific regions of the
rat's brain, he and his associates found that during
the first months of the rat's life a rapid spurt in
the growth of synapses occurs regardless of the
amount or type of sensory experience.[19] This period of synaptic "blooming"
is followed by a sharp decline in the number of
synapses. That is, an elimination or "pruning"
of synapses then takes place based on the activity
and sensory stimulation of the brain, and ultimately
results in the configuration of connections
characteristic of the mature rat's brain. Greenough
refers to this initial blooming and pruning of
synapses as "experience-expectant" learning,
since the initial synaptic overproduction appears to
be relatively independent of the animal's experiences.
It is as though the brain is expecting important
things to be happening during the first weeks and
months of life, and is prepared for these experiences
with an overabundance of synapses, only a fraction of
which, however, will be selectively retained.
The work of Greenough and his associates is
limited to rats and monkeys, but their findings have
much in common with those of Peter Huttenlocher of
the University of Chicago who counted the synapses in
specific regions of the brains of humans who died at
various ages. Huttenlocher found that:
The increase in synaptic density plus
expansion of total cortical volume leave no doubt
that the postnatal period is one of very rapid
synaptogenesis in human frontal cortex. By age 2
years, synaptic density is at its maximum, at
about the same time when other components of
cerebral cortex also cease growing and when total
brain weight approaches that of the adult.
Synaptic density declines subsequently, reaching
by adolescence an adult value that is only about
60% of the maximum.[20]
This wealth of synapses is thought to be
responsible for the striking plasticity of the
immature brain that permits the learning of skills
that can be learned only with much greater difficulty
or not at all by the already pruned adult brain. We
already saw how immature animals and children are
unable to develop normal vision if they are not
exposed to a sharply focused visual world during this
period of brain development. It has also been
repeatedly observed that although many adults
initially may make quite rapid progress in learning a
foreign language, young children appear to have an
important advantage over adults in being able to
master the sounds of languages. Canadian child
language researchers Janet Werker and Richard Tees
observed that children younger than one year appear
able to distinguish between the speech sounds used by
any human language. By age 12 months, however, they
begin to lose the ability to discriminate between
sound contrasts that are not used in the language
they hear every day. So whereas all normal infants
can distinguish between the two related but distinct
sounds represented by the letter t in Hindi,
those who hear only English quickly lose this ability,
and Hindi-speaking children retain it.[21] The work of Werker and Tees
therefore provides important human behavioral
evidence that is consistent with the view that normal
brain development involves the loss of synaptic
connections, which results in the loss of certain
skills as the brain approaches its adult form.
A sensitive period for the acquisition of a first
language was demonstrated by the plight of Genie, an
American girl who was brutally isolated from all
normal human interaction until she was found at age
13 years, and who never subsequently developed normal
language abilities.[22]
There is striking evidence that the immature,
overconnected brain is also better suited than a
mature one to acquiring second languages and sign
languages.[23]
Taken together, these findings paint a picture of
the developing brain that contrasts sharply from the
genetic providentialism favored by Sperry. Instead of
the brain unfolding according to a genetically
specified blueprint, we see instead a process of
selection by which overly abundant neuronal
connections are eliminated through a weeding-out
process, leaving only those connections that permit
the animal to interact successfully with its
environment.
The mammalian brain appears most adaptive during
the early postnatal period, and continues to adapt
and learn from new experiences throughout its adult
life. During the 1960s and 1970s a series of studies
offered impressive evidence that rats grew thicker
brains and new synapses when they were placed in
complex and challenging environments. These findings
were consistent with the then-popular belief that
learning and memory in mature mammals (as opposed to
the brain development of immature animals) were additive
processes involving the formation of new synaptic
connections or the strengthening of already existing
ones. The influential Canadian psychologist Donald
Hebb assumed that "the changed facilitation that
constitutes learning" was the result of "the
growth of synaptic knobs."[24] Similarly, Sir John C. Eccles,
who shared a Nobel prize in 1963 for his research on
the transmission of nerve impulses, believed that
memory and learning involved "the growth . . .
of bigger and better synapses."[25]
However, it was also suggested that more than just
adding synapses was involved in learning. One of the
first to propose that subtractive brain
changes could be involved in adult learning and
memory was J. Z. Young, who in 1964 posited that such
learning could be the result of the elimination
of neuronal connections.[26]Several
years later J. S. Albus hypothesized that "pattern
storage must be accomplished principally by weakening
synaptic weights rather than by strengthening them,"[27] and Richard
Dawkins speculated that the selective death of
neurons could underlie the storage of memories.[28]
But how could a subtractive process of neuron
elimination be involved in learning and memory? It is
particularly difficult to understand how the learning
of a new skill, such as riding a bicycle or speaking
a foreign language, or acquisition of new memories,
such as learning the words to a poem or song, could
be made possible by loss of synapses. We saw in the
development and maturation of the brain that synaptic
connections that are rarely used are weakened or
eliminated, whereas those in active neural pathways
are retained and perhaps strengthened. This
subtractive process makes sense when dealing with an
overwired, immature brain that may have close to
twice as many synapses as it will have as an adult.
But how can it work for a mature adult brain that has
already been substantially whittled down by synaptic
pruning?
To illustrate this problem, imagine an adult
Spaniard learning English. To do this, the Spaniard
will have to learn to hear and produce certain sound
distinctions that are not used in Spanish, such as
the contrasts involved in ship versus sheep,
sue versus zoo, and watch versus
wash. The research of Werker and Tees would
lead us to predict that the Spaniard would not
initially be able to make these distinctions since
they are not made in the language he has heard and
spoken all his life. The synaptic connections
necessary for making these discriminations were
present when he was born, but we would expect them to
have been promptly pruned away since they were not
used in the language of his environment. It is
therefore not clear how any further pruning of
synapses would permit him to learn this aspect of the
English language.
Instead, it seems more likely that a process
involving the addition of new synapses, or at
least reorganizing current ones, would be necessary
for this learning to take place. But then we run into
the equally thorny problem of understanding how the
brain could ever know which new synapses to add or
modify! Surely, some combination of synaptic changes
should allow the Spaniard to learn English, since
many adults learn English and other languages, and
such learning must be the result of changes in the
synaptic connections of the brain. But just which new
combination of synapses will do the trick? At the
very least it would appear that the brain would
somehow have to try out a number of new combinations
and select the best ones. But to select the best ones,
a source of variation is necessary, perhaps not
unlike the initial variation of synaptic connections
present in the immature, overconnected brain.
A possible solution to this riddle was offered by
French neurobiologist Jean-Pierre Changeux in 1983.
In his book L'Homme Neuronal (published in
English in 1985 as Neuronal Man), Changeux
proposed a "Darwinism of the synapses"[29] to account for the
development of the brain and the learning it
undergoes within its cultural environment.
According to this scheme, culture makes its
impression progressively. The 10,000 or so
synapses per cortical neuron are not established
immediately. On the contrary, they proliferate in
successive waves from birth to puberty in man.
With each wave, there is transient redundancy and
selective stabilization. This causes a series of
critical periods when activity exercises its
regulatory effect.[30]
In effect, he was suggesting that all adaptive
brain changes, or at least those occurring between
birth and puberty in humans, involve the elimination
of preexisting synapses, but that these preexisting
synapses were not necessarily all present at the same
time. From birth to puberty, Changeux hypothesized
that waves of synaptic growth would occur, with
subsequent experience serving to retain the useful
ones and eliminate the useless and redundant ones.
These waves of synaptic overproduction would provide
the source of variation on which synaptic selection
could operate. Such learning resulted in an absolute
increase in synaptic growth and numbers over time.
This growth was not constant, but was rather
envisioned as analogous to repeatedly taking two
steps forward--randomly adding new synapses--followed
by one step backward--eliminating the useless ones
just added.
Changeux provided no hard evidence for his
hypothesis that synaptic variation in the form of
overproduction would precede the elimination of
synapses as part of the brain's restructuring to
permit the learning of new skills and acquisition of
new knowledge. But such evidence was found a few
years after the publication of his book. William
Greenough and his associates, whose work on the
maturational development of the rat's brain was noted
earlier, also conducted research on changes in the
brain induced by placing adult rats in special,
enriched environments. In one study this resulted in
a 20% increase (roughly 2000) in the number of
synapses per neuron in the upper layers of the visual
cortex.[31] Later
research showed that such dramatic increases in
synapses were not restricted to the rat's visual
cortex.[32]
These and other similar findings led Greenough's
group to propose that the waves of synapse
proliferation first described by Changeux could be
elicited by the complex demands placed on the adult
brain in a new, challenging environment. These
researchers referred to this process as "experience-dependent"
development since it depends on the environment
triggering the formation of new synaptic growth on
which the selective process can act.[33]
Greenough's conception of how the adult brain is
able to learn new skills and form new memories offers
an appealing solution to the problem concerning the
additive and subtractive processes underlying the
adult's brain adaptation to new environments.
According to this theory, experience-dependent
learning combines both additive and subtractive
processes. The additive component involves the
blooming of new synapses in response to the animal's
attempt to control aspects of a new, complex
environment. Although the brain does appear to know
what part of itself has to be involved in this new
synapse-construction project, it need not (indeed,
could not) know which particular connections to make.
By forming a large variety and number of new
connections, the brain can select the combinations
that work best, in the same way that the immature,
developing brain retains useful connections from its
initial oversupply of synapses. The long-term result
is an overall addition to the number of synapses. But
the actual selection process that fine-tunes the
connections is a subtractive one in which the useful
connections are selectively retained and less useful
ones eliminated. Although clear evidence exists for
synaptic increase in learning, as I write this we
still have no such evidence in mature learning for an
overproduction of synapses that are then pruned away.
However, recent research has found evidence for an
overproduction of dendrites in mature rats during
readaptation of the brain after brain injury, which
at least suggests that synaptic overproduction may be
involved as well.[34]
These findings fit very nicely with the subtractive
synapse findings on brain maturation and provide a
solution to the mystery of how the brain could know
exactly which new synaptic connections to establish
to enable it to acquire new knowledge, skills, and
memories.
Although only a relatively small number of
neuroscientists have opted for a selectionist
approach to their research and theorizing, Changeux
and Greenough and their associates are not the only
ones whose research suggests that the adult brain
develops and learns through a process of cumulative
neural variation and selection. This theory has now
been embraced and given additional support by several
other leading neuroscientists. William Calvin refers
to the brain as a "Darwin machine" that
follows the plan "make lots of random variants
by brute bashing about, then select the good ones."[35] Gerald Edelman,
who shared a Nobel prize in 1972 for his research on
the chemical structure of antibodies in the immune
system, has contributed a remarkable outpouring of
books describing aspects of his "neuronal group
selection theory" of brain development and
learning through a selectionist process he refers to
as "neural Darwinism."[36]And noted psychologist and
neuroscientist Michael Gazzaniga, best known for his
ground-breaking research on humans with split brains,
recently embraced a selectionist account of brain
functioning and development.[37]
Current research is under way to determine whether
unambiguous physical evidence can be found for the
overproduction and elimination of newly formed
synapses in the adult brain in response to
environmental changes. Such a finding would place the
brain alongside the immune system as another striking
example of how cumulative variation and selection
processes during the lifetime of an organism make it
possible to adapt to complex, changing environments.
We have now seen that understanding both the
adapted and adaptive complexity of the human brain
involves finding answers to three questions: how did
the brain originate as a biological organ?; how does
it develop from a fertilized egg into a mature brain?;
and how is it able when mature to rewire itself to
learn from and adapt to changes in its environment?
Much more work must be done before we have
detailed answers to these questions. But substantial
progress has already been made as we move midway into
the "decade of the brain." To a large
extent this progress has consisted of rejecting
providential and instructionist explanations for
these puzzles of fit, and finding considerable
evidence and reason in favor of selectionist
explanations. The powerful process of cumulative
blind variation and selection working over millions
of years is not only the only reasonable theory for
the biological evolution of the brain, but we find
that it has surfaced again in a different but still
recognizable form as an explanation for the brain's
embryonic growth and continued development during its
relatively brief lifetime.
It is here, as Changeux remarked, that "the
Darwinism of synapses replaces the Darwinism of genes."[38] To close the
circle, it should be noted that a striking
consequence of the joint effects of among-organism
genetic and within-organism synaptic selection is the
brain's understanding both itself and the process of
selection that is responsible for its extraordinary
abilities.
[1]Changeux (1985,
p. 248; first emphasis added).
[2]Gazzaniga (1992,
p. 50).
[3]Not all
dendrites serve to excite the attached neuron. Some
are inhibitory in that they act to prevent the neuron
from firing.
[4]It has been said
that "human beings have caused greater changes
on earth in 10,000 years than all other living things
in 3 billion years. This remarkable dominance is
related to the development of the brain from the
minute cerebrum of simple animals to the complex
organ of about 1350 grams in man" (Sarnat &
Netsky, 1981, p. 279).
[5]Bullock (1977, p.
410).
[6]See Gould (1991a,
essay 9).
[7]Gould & Vrba
(1982, p. 13). See also Gould (1991b) for an
introduction to the concept of exaptation.
[8]Gazzaniga (1992,
p. 35).
[9]Changeux (1985,
p. 206).
[10]Eccles (1989,
pp. 1, 4
[11]Changeux (1985,
pp. 216-217).
[12]Hamburger (1975).
[13]Changeux (1985,
p. 217).
[14]Ramón y
Cajal (quoted in Changeux, 1985, pp. 212-213).
[15]Schatz (1992,
p. 63).
[16]Schatz (1992,
pp. 66-67).
[17]Gazzaniga (1992,
p. 37).
[18]Greenough
& Black (1992).
[19]Huttenlocher
(1984, p. 490).
[20]Werker &
Tees (1984).
[21]Curtiss (1977).
[22]See Johnson
& Newport (1989) and Newport (1994).
[23]Hebb (1949, p.
65).
[24]Eccles (1965;
quoted in Rosenzweig et al., 1979).
[25]Young (1964,
p. 285).
[26]Albus (1971;
quoted in Rosenzweig et al., 1979).
[27]Dawkins (1971).
[28]Changeux (1985,
p. 272).
[29]Changeux (1985,
p. 248).
[30]Turner &
Greenough (1985).
[31]Black &
Greenough (1986).
[32]Black &
Greenough (1986, p. 33). These experience-dependent
changes in the mature brain are contrasted with the
experience-expectant development of the maturing
brain, the latter taking advantage of the great
number of redundant connections already present in
the postnatal brain.
[33]Jones &
Schallert (1992, 1994); Schallert & Jones (1993).
[34]Calvin (1987,
1990).
[35]Edelman (1987,
1988, 1989, 1992).
[36]See Gazzaniga
(1992, chapter 2). Gazzaniga's innatist construal of
brain selectionism is critiqued at the end of chapter
15.
[37]Changeux (1985,
p. 272).
