THURSDAY, April 1, 2004
Session 3: Neuroscience, Brain, and Behavior II:
Emotional and Cognitive Development in Children
Jerome Kagan, Ph.D.,
Daniel and Amy Starch Research Professor of Psychology,
Elizabeth S. Spelke, Ph.D.,
Professor of Psychology, Harvard University
CHAIRMAN KASS: This afternoon's session, the
first of two, is titled "Neuroscience, Brain, and Behavior
II: Emotional and Cognitive Development in Children."
And I've been asked to say at least a sentence or two
about, well, to be blunt, what's going on here. This
will not be long, and we'll have more to say about what's
going on here in the last session when we're talking amongst
ourselves, but I remind everybody that the purpose of today's
session agreed to last time was that before we took up and
searched for various ethical or social or philosophical issues
raised by advances in neuroscience and psychology, we ought
to learn some of the basic facts.
And the purpose of these discussions is to lay the groundwork
for anything further that we would do. The morning was on
the neuroscience. This afternoon is on the side of psychology.
There are no axes here, and there are no agendas, other
than getting us informed about the current state of knowledge
about the developing brain and the developing mind and behavior
of children. So if anybody is impatient, just soak up this
knowledge. It's terrific stuff.
The relation of the brain and the mind, of the activities
of molecules and synapses to mentation, never mind consciousness,
is, as everybody knows, a venerable question, a deep philosophical
issue that has occupied and vexed and challenged the best
minds since classical antiquity, and one simply has to mention
the names of Lucretius and Aristotle to show you how old these
controversies are. There are idealists; there are dualists;
there are compatiblists; there are epiphenomenalists.
Yet even as those sort of prize questions continue to be
discussed and debated, even people who are committed to fully
neurochemical and mechanistic account of all mental and behavioral
activity recognize the prime importance of studying the mental
and behavioral phenomena in their own right and on their own
level, leaving for later any attempts to connect the domains
of psychology, the study of the psyche, and the domain of
neuroscience, the science of the brain.
So with no prejudice regarding this deferred questions about
the relation of mind and behavior to the brain, we also want
to know not only about the neural development, normal development
of the brain and nervous system in children, but the normal
and also abnormal development in all of its variations of
the emotional and temperamental side of human development,
and of the cognitive capacities and activities of children,
and that is the theme for this afternoon's discussion,
and we're very fortunate to have two colleagues from Harvard's
Department of Psychology, Jerome Kagan, who is the Daniel
and Amy Starch Research Professor of Psychology at Harvard
University, and Elizabeth Spelke, who is Professor of Psychology,
also at Harvard University.
Dr. Kagan is going to speak about the temperament and affective
side, and Professor Spelke will speak about the cognitive
side, and I think the procedure is we will let each of them
make their presentations, perhaps with small questions of
clarification after each talk, and then the general discussion
Thank you both very much for coming down and being with
us, and we're delighted to have you here and look forward
to the presentations.
Professor Kagan, would you like to start?
PROFESSOR KAGAN: Yes. Can you hear?
Thank you very much for inviting me, and let me also follow
Professor Jessell's suggestion that interrupt me if there
are any questions that you have during the presentation.
And I promise to hold it to 40 minutes, ten of three, so
that you can hear Dr. Spelke and have time for discussion.
Unlike the area you heard this morning, the growth of the
brain, which is, you know, mid-volume, the systematic work
on human temperament is about 50 years old. Since American
psychology was committed to a behaviorism that did not want
to acknowledge biology, and although there have been many
essays going back to Hippocrates on temperament, there is
no empirical work.
So this is a field that is in Chapter 1, and therefore,
we can't show off the wonderful findings you heard this
morning, but we can give you a scaffolding.
Also, you should appreciate that in biology when a speaker
says "dendrite," everybody knows what he means.
But when you get to psychological concepts, people have different
understandings, and so I will try to give you the understanding
that I'm using. Remember the meaning of words, as Virginia
Wolf said, is a function of how they're used.
So the concept of temperament as it is used today in the
Western world means variation within humans in mood and behavior
that is biologically based. My own view is that it should
be restricted to inherited variation, but there are those
who say any variation, even variation caused prenatally.
Now, the analogy would be we talk about breed differences
in dogs. So some people like Rhodesian ridgebacks. Some
people like cocker spaniels. They belong to the same species.
Their behaviors are different, and we say those are breed
When we talk about humans, that is the domain we're
talking about, but we use the word "temperament."
My own view is that most of the temperamental variation—one should never say "all" in the life sciences—that most of the temperamental variation will be due to
inherited differences in neurochemistry. There are over 150
molecules discovered. Many more may be discovered. We all
have the same molecules, but we differ in their concentrations,
and we differ in the density receptors, those proteins that
sit on the neurons. How dense are those receptors.
And we differ in the distribution of those receptors. Thomas
Insel, who now is the Director of NIMH, provides us with a
perfect example of what we mean by a neurochemical variation
that causes a big difference in behavior. So let me use it,
and it will help you understand the work on humans.
The vole is a very small rodent. It looks like a mouse.
Now, there's one strain of voles called prairie voles.
They pair bond. Once the male and female mate for six hours
they will never mate again with anyone else.
The montane vole that shares 99.99 percent of its genes
with the prairie vole doesn't pair bond. Now, that is
a dramatic difference in behavior, and Insel in 20 years of
really elegant research finds that the main cause of that
difference is a change.
Remember a gene is a strip of DNA, but in front of each
gene is an area called the promoter region which governs the
DNA. Well, in the promoter region for two molecules, one
is called vasopressin and the other is called oxytocin, secreted
during sexual intercourse incidentally, both of these molecules;
that that explains the difference.
So tiny, tiny genetic differences, not even in the DNA,
but in the promoter region for the DNA.
Now, I wish I could tell a story. Here are some of the
molecules, and remember there are over 150. At the moment
they look relevant. So children could be born with differences
in opioid concentrations or the receptors for opioids.
For example, right here in our neck is the structure called
the medulla. All pain, information from your heart and gut
come up through your body, and they have to pass through that
Well, supposed some individual was born with a light set
of receptors for opioids. Then they would experience pain
and muscle strain more easily than others.
GABA is an inhibitory molecule prevalent throughout the
entire brain and its job is to mute excitability. I'm
going to say in a moment that some infants are very irritable,
extremely irritable. It looks like a temperamental trait
that expresses itself in certain behaviors later in life.
It could be that what these children inherit, these very
irritable infants is a failure, a compromised function in
Dopamine is a powerful molecule. Every time any one of
us anticipate a trip, a holiday, a good meal tonight at 6:30,
dopamine pours out of our central nervous system. When a
rat is about to get food it wants, it pours out dopamine from
its source in the brain.
Now, individuals, and it is believed by many that there's
a subtle surge of pleasure when one is looking forward to
seeing Rome for the first time and you're there. People
differ in their hedonic tone, in the amount of pleasure they
take from experience. It is not beyond reason that someone
some day will find that dopamine is playing a role.
Notice there's no determinism here. I'm going to
use words like "enable," "determine a role,"
Norepinephrine is a very important molecule. If you're
listening hard for a signal that your wife is coming home
because it's midnight. Norepinephrine acts on sensory
neurons so that you hear the signal you're interested
in and not the noise or when you're trying to hear a conversation
several feet away. So children are extremely vigilant to
change, any subtle change, and some children seem oblivious.
Perhaps norepinephrine makes a contribution there.
And finally, just to have a flavor for these, corticotropin
releasing hormone is often secreted but not always when one
is under stress, and I'm sure most of you know that all
of us when we're under stress secrete a small hormone
called cortisol from our adrenal cortex, and variation in
corticotropin releasing hormone could play a role here.
Now, I wish we could go to a book and look up everything
we've learned about neurochemistry and now talk about
human temperaments. One day, but not today.
So those of us who study temperaments must begin with behavior.
That would be like medicine 250 years ago when one knew nothing
about what's happening in the immune system, and so the
patient tells you, "I have an ache in my body. I itch
in my arm," and so on. You go to the surface, and one
day, of course, we will tie together the behavior with the
Now, many psychologists have been studying the temperamental
traits, and here are four that look like they might be temperamental,
that is, due to inherited variation in this neurochemistry.
I mentioned irritability. Some infants are extremely irritable,
and the data indicate that that persisted the first year.
Now, you stop being extremely irritable when you're one
and two years of age, but investigators who have followed
such infants find that they are different at five, six, and
seven years of age. Some children are very active and show
very high levels of muscle tension, and we'll be talking
about that in a moment.
Believe it or not, in the first six months some infants
smile a lot, and you'll see in about ten minutes I regard
this as a very powerful temperamental trait in humans. Children
who smile a lot in the first six months spontaneously tend
to preserve a more sanguine view of life through adolescence.
Some infants don't smile at all. As a matter of fact,
in the laboratory they'll show a frown on their face,
and they tend to be more pessimistic children later in life.
Now, in order to concretize this discussion, let's tell
two stories. I'm going to tell you one story. It's
the story that I've been writing for 25 years, but it
only reflects two temperaments of the many. There are going
to be thousands of temperaments.
If there are 150 molecules and they vary in their concentration
and receptor density, remember from your algebra how many
combinations of 150 things you can have. There are going
to be millions of temperaments, and even if half were not
functional, you're going to have a very large number,
some being very rare like the Unabomber or Mozart, and some
I'm going to talk about two common temperaments only
because the work on them is more extensive than others, and
they're relatively frequent, but not that they are necessarily
the most important temperaments, and it has to do with reaction
In every vertebrate species, within every vertebrate species,
fish, birds, cats, mice, monkeys, and of course, humans, there
are some members that react to novelty and unfamiliarity by
become immobile if you're an animal, freezing if you're
a rat, not exploring an unfamiliar area if you're a mouse
or if you're a child, closing down and exploring the situation
before you assimilate it and move forward.
And other children are just the opposite. Now, in animals,
this is extremely heritable. You can breed it. You can breed
quail, rats, mice, and Steve Suomi of NIH believes even breed
monkeys so that after 20 generations you have either very
timid or very bold monkeys, mice or rats.
Now, it is believed by the work of many scientists that
deserve a great deal of credit, it looks like the amygdala,
which is a small structure right in back of your temporal
lobe, tiny, shape of an almond. It's very important because
whenever you present novel stimuli to any animal and you record
from neurons in the amygdala, those neurons respond.
And so if you presented a novel event to a monkey and you
had electrodes down the amygdala, it would respond, but if
you kept on presenting the stimulus, the neurons would stop
responding. The amygdala response is a novelty, and you can
see why this is important.
You're a monkey out in the savannah, and you're
munching your bananas, and suddenly an odd sound occurs.
The amygdala fire, makes you alert, and then you decide whether
you're going to flee or it's an unimportant event
and you keep on eating your bananas.
Now, I have been interested for many years in timid versus
bold children, and so to abbreviate ten or 15 years of work,
we decided that perhaps these were temperamental traits traceable
to infancy, and so we began a study of 500 healthy, middle
class, four month old infants.
Now, why did I restrict it? Because if you want to study
temperament, you have to eliminate all of the other causes
of possible timidity: a mother took drugs during her pregnancy;
drank too much; smoked cigarettes; drank coffee. And so you
want mothers who cared about their pregnancy and gave birth
to healthy babies at term.
That means that if one did the study I'm about to describe
on infants born to compromised pregnancies we might get different
results. Okay? So these are healthy babies born at term.
Now, the reason why the amygdala is important is that if
you stimulate the amygdala of a cat or a monkey, you get limb
movements and you get distress cries, and of course, human
infants will display those two responses.
So here is the central idea behind the work. This is a
schematic of the amygdala. Vision audition and touch come
into this area of the amygdala, send their information up
to this nucleus, and then out to the body to produce tension,
immobility, a rise in heart rate, a rise in blood pressure
or other biological consequences.
Assumption: if some infants were born with a chemistry
unknown at the moment that rendered the amygdala excitable
to unfamiliar events, then they should show a lot of motor
activity and crying when you present these unfamiliar events.
If you were born with a different chemistry, then you should
show low motor activity and not be very distressed, and we
call those infants high and low reactive.
So after testing 500 infants by showing them unfamiliar
mobiles of different colored elements moving in front of them,
listening to speech on a tape with sentences like, "Hello,
baby. How are you today. Thank you very much for coming,"
or presenting a cotton swab dipped in butyl alcohol to their
nostril, olfactory, auditory, visual, what you see is that
20 percent of the babies are very different from all other
babies. They begin to thrash. They arch their backs. They
become very aroused motorically and cry. They should have
a more excitable amygdala.
Forty percent, twice as much, are just the opposite. They
don't move. They lie there. They move an arm. They
Now we call the first group high reactive and the second
group low reactive, and now briefly I'm going to show
you what happens if you follow these infants through 11 years
of age. Okay?
So you bring these infants back at 14 and 21 months for
two hours with their mother and they encounter unfamiliar
events. Nothing threatening, no snakes, no mice, just people
they don't know, a clown, people dressed in clown costumes,
robots that move, novel events that are not obviously dangerous.
Some children do not become very frightened. Some cry and
clutch their mother. I didn't bring my laser so if you
follow me, you see the high reactives are in the light color,
the low reactives in the dark purple. So at 14 months those
who are high reactive at four months were more fearful. At
21 months they were more fearful.
At seven and a half years the IRB of the university, in
order to make children cry, you have to do things that are
unethical. So we don't do that.
Now, with age what happens is that these traits become internalized,
and so rather than cry you begin to show the traits of the
introvert. You don't smile or talk easily with a stranger.
So if you're interviewed by an examiner for two hours,
you talk less. You see at seven and a half years you smile
You remember I said that smiling is very sensitive, and
based on observations of children and interviews with the
mothers and the teachers, you're more likely to have anxious
symptoms. Now, notice I'm not calling this child as having
an anxiety disorder. This is a child who doesn't want
to sleep over at a friend's house, needs night light on,
asks their mother will they be kidnapped, is their mother
going to die. They're afraid of large dogs or insects.
And you'll see that 45 percent of this group who are
high reactive had anxious symptoms. Remember only 20 percent
of the group was high reactive. So that's twice as many
as you'd expect. Only ten percent of low reactives had
symptoms, while 40 percent of the sample. So that's much
less than you would expect.
So these children are developing in a way that one would
anticipate, given the assumption about their amygdala. Okay?
Now, by the time they're 11 years of age, a lot of them
aren't shy anymore. What happens with increasing age
is that you get an increasing dissociation between your behavior,
what you have called your persona, and what's going on
inside. Your grandmother knew it as "don't judge
a book by its cover."
But we believe that they retain their biology. So we have
to measure their biology, and I apologize for doing this quickly.
There are four measures that based on the research of many
scientists one would reasonably assume should characterize
the high reactive children at 11 years, but not the low reactives,
and they are: One is to have greater activation in the right
hemisphere than the left. For example, if you take a newborn
baby and put lemon juice on its tongue, you get right hemisphere
activation. If you put sugar water on its tongue, you left
Now, there are exceptions here, but as a rough rule the
right hemisphere is more active under states of uncertainty
and adversiveness, the left hemisphere under the complimentary
state. So we measured that using EEG.
Let's do the behavior first. When they come in at 11
years of age, we watch their behavior, and if we combine their
behavior at seven years and ten years of age, and here's
the important result, 40 percent of the children who had been
high reactive made many low comments. You see the turquoise
blue bar on the left, and very low smiles, 40 percent, while
less than ten percent of the low reactives did.
While if you said who talked and smiled a great deal, it's
just the opposite. So, in other words, temperament constrains
what you will become. Forty percent are honest to their early
temperament, but only six or seven percent cross over. So
that means temperament doesn't determine what you will
be, but its power is to prevent a high reactive infant from
becoming an extremely social, exuberant child, while a low
reactive temperament constrains that child from becoming an
extremely timid and subdued child. We'll return to this
in a moment.
Now, here are the data on hemisphere activation. The high
reactives are in pink, and the low reactives in yellow, and
on the left side of the graph is showing right hemisphere
activation, and on the right side of the graph is left hemisphere
activation. So follow the pink line, and you'll see that
43 percent of the high reactives showed greater right hemisphere
activation, and as you move towards the left fewer and fewer
high reactives, while the yellow line, only 18 percent of
low reactives were right hemisphere active and they moved
So there's the first prediction. AT 11 years of age,
you can do better than chance at predicting hemisphere activation
at 11 years of age from what they were like at four months.
Now, the next measure is more direct. In the system for
hearing, when you hear any sound it goes through a series
of ganglia, first your basilar membrane in your ear. Then
it goes through a series of nuclei and ends up in the mid-brain
in a structure called the inferior colliculus.
And if you present clicks to an infant or a child, if you
get a series of brain waves from those clicks, every infant
in Massachusetts is tested this way to insure that it can
hear. Now, notice pk V. That is the evoke potential from
the inferior colliculus. It occurs in about six milliseconds.
Now, here's why that's important. The amygdala,
our friendly amygdala, sends projections down to the colliculus,
pk V, but not to any other structure before it, and that means
that if you had an excitable amygdala, you should show a larger
pk V, a larger wave V than if you were a low reactive.
I hope that's clear. So the 11 year olds wore earphones,
and they heard clicks for 90 seconds, and sure enough, as
we expected, the children who had been high reactive at four
months had large wave forms, especially when the loudness
was 70 decibels and you were measuring it on the opposite
side of where the information came in.
So there's our second prediction, and that one does
support the notion that high reactives have a more excitable
The third, many scientists over the last 35 years have made
an important discovery. Whenever you're presented with
a visual or auditory event that surprises you, you have a
very distinctive wave form. If you were sitting in a laboratory
with EEG electrodes on and you heard the following, "Washington,
D.C. is a vegetable," that on the word "vegetable,"
you would have a wave form like that.
But if you heard "Washington, D.C. is a city,"
there would be no wave formed. So whenever you're surprised
by an event you don't expect, you get a very distinctive
Now, remember what I said about high reactive infants.
They are very sensitive to unexpected events. So we then
hope to see that at 11 years of age, the high reactives would
show larger, evoke larger event related potentials as you
saw in the last slide to scenes that were totally unfamiliar.
Go to the far right where it says "invalid."
These are ecologically invalid scenes, none of which are dangerous.
For example, a baby's head on an animal's body or
a car in midair or a chair on one leg.
While for the frequent, if you go over to the far left where
it says "frequent," that's a fire hydrant being
shown 70 percent of 169 trials.
So there's our third prediction affirmed. Right hemisphere
activation, a larger wave V, and a large event related potential
to surprising scenes.
And the last measure, the amygdala sends projections to
the sympathetic nervous system, and therefore, one should
show greater sympathetic tone in the cardiovascular system,
and we measured that in several ways, and so the term sympathetic
means that you have greater priming of the circulatory vessels
and the heart, while the vagal system means you have less
priming because it is mediated by the parasympathetic system,
and as you can see, about 65 percent of the high reactives
at age 11 were sympathetic compared with about 38 percent
of the low reactives and the opposite for vagal tone.
Now, let's put it together. A temperamental bias constrains
what you will become. Let me skip that, and here is the final
line. About one in four high reactives and one in four low
reactives combined expected behavior with biology versus one
of 20 who do not.
That means that if outside that door there were 100 adults
and I said—they're 20 years old—I said they're
high reactive infants, every one of them. If I predict that
they will not be exuberant, bold, highly sociable 20 year
olds who show right hemisphere—they wouldn't show right
hemisphere activation; they wouldn't show a big wave V;
they wouldn't show sympathetic tone; I'm going to
be right 95 percent of the time.
If you say they're going to be quiet introverts who
have high sympathetic tone in a large wave V, you'll be
right 20 percent of the time.
And of course, the same thing shows for the environment.
If I say, "I have this beautiful girl born to nurturing
parents in a lovely suburb of Boston who went to good schools,
she's 25 years old. Tell me about her," you will
be more correct if you what she is not.
She's probably not a drug addict. She's probably
not a prostitute. She's probably not on welfare, but
what else? Who did she marry? What did she major in? What
will she do? You have no idea.
And that's an important message which we have in psychology,
tend to think of the environment and biology as deterministic,
and we should begin to think of it rather as constraining
rather than deterministic.
Because of time, let me go to some of the implications because
I don't want to take much time from Professor Spelke.
Here are some implications.
Individuals with similar public profiles can differ in the
origins of those profiles. Current psychiatry, 99 percent
of all you read in the papers about the epidemiology of psychiatric
illness is based only on interviews. They never assess the
biology of the person.
And so two people can say that they worry. They're
worried about the war or they're worried about terrorists,
but one person has an extreme emotional reaction, and once
psychiatrists begin to add not necessarily these measures,
biological measures to the interview, then you will see all
of the prevalence figures for mental illness change because
right now they're based on one source of evidence.
Second, the ancients understood a temperamental bias does
not imply that will is impotent with respect to a behavior.
Remember the ancients said that temperaments control your
mood. They don't control your action.
So I can feel anxious, but I can control my tendency to
avoid. I can feel angry easily, but I can control my impulse
to strike, and so on. And so we're in a dangerous period
because biological determinism is so popular in having the
public believe, well, after all, if this is genetic, then
why should I be held responsible for my behavior.
The third implication will come in about 20 or 25 years
from now because there will be variation in temperaments across
reproductively isolated populations. When mutations occur
and they change the shape of your eyes, the color of your
hair, whether you're vulnerable to spina bifida or not,
nature doesn't stop there. Obviously the genes that separate
reproductively isolated populations are going to affect the
neurochemistry, too, and so we're going to discover in
a quarter century that there are temperamental differences
among Asians, Africans, aborigines of Australia, Europeans.
No question about it.
Although the work is preliminary, it is pretty clear now
that Asian and European infants differ dramatically. There
are three studies: Daniel G. Freedman, Caudill, Michael Lewis,
Asian infants tend to be very low arousal, whether they
are born in America or born in Beijing, while European infants
are much more active, more easily distressed, and those look
like temperamental differences, and I'm sure when we are
less self-conscious—unfortunately we are—about ethnic
differences, right now it's focused unfortunately on school
performance and IQ and that will vanish, but there will be
temperamental differences. I think that the benevolent consequence
of that is that we'll recognize that every reproductively
isolated group, as is true for humans, is true for animals,
has a special set of advantages and a special set of disadvantages,
and that's the way it goes, and that will be, I think,
But there will be differences in risk for particular moods
and psychological symptoms. There's an Asian psychiatrist
in Los Angeles who has written several papers that, for example,
Asian patients in Los Angeles require half the dose of Prozac
or Valium than Caucasian patients with exactly the same psychiatric
So we return to voles again. This research has just begun.
I must confess to you as I now stop that I'm surprised
by these results. I wouldn't have expected them.
When I was a graduate student, I was very hostile to the
role of biology in human affairs. I was convinced that most
of the variation among 20 year olds was a function of what
happened within the walls of their family, but I've been
dragged to this conclusion by their data, even though as I've
shown you the role of the environment is powerful for temperament
constrains rather than determines.
Thank you very much.
CHAIRMAN KASS: Let's agree to have just a couple
of minutes for questions of clarification, although the two
papers I think might be best discussed at the end.
Diana Schaub, Diana.
DR. SCHAUB: You spoke of ethnic and racial differences.
Are there sexual differences also? Do girls tend to be more
high reactive than boys?
PROFESSOR KAGAN: Right. Surprisingly, there's
no difference at four months. High reactive and low reactive
infants, it's equal on the sexes.
Now there's an interesting story. Under age seven or
eight more high reactive girls are timid, shy, introverted
than boys. But by adolescence, it has changed. Now, I think
this may be—here's my hypothesis.
At 15, it's just the opposite, and I think it's
because girls in this culture are gentler with timid girls
than boys are. Boys are very cruel, and so at 15, which is
the age we're studying now, high reactive boys who have
not lost their persona, they are very shy, very frightened,
Our girls are garrulous, have many friends, and my own belief
is that that's all environmental. It's because girls
are less cruel toward a girl who is initially timid in her
But there is no difference at four months.
CHAIRMAN KASS: Jim Wilson.
PROFESSOR WILSON: Thank you very much, Jerry.
You said toward the end of your remarks, speaking, I think,
of adult populations, that psychiatrists should gather biological
evidence. What kind of biological evidence, and how do they
PROFESSOR KAGAN: I would say I think it would be
very useful if as part of the psychiatric examination, I'm
thinking of something that could be done in the office. You
could easily gather sympathetic and vagal tone, easily. Given
the fact that many are associated with the hospital, you could
order an EEG and get right and left hemisphere dominance,
yeah, and I think that would help the diagnosis.
CHAIRMAN KASS: Other questions? Michael Sandel.
PROFESSOR SANDEL: This is a simple minded
question, Jerry, but what was the question that drew you to
this? Were you trying to figure out what makes some kids
shy and some kids bold?
PROFESSOR KAGAN: No.
PROFESSOR SANDEL: What was the animating
question for you?
PROFESSOR KAGAN: The animating question was this.
My first job was at the Fels Research Institute in Yellow
Springs, Ohio, in the campus of Antioch College. I inherited
a corpus of data that has been gathered since 1929, and Howard
Moss and I studied the adults, and he rated the children,
and only one trait was stable from the first two years of
life to adulthood, and it was these two traits, but I didn't
understand it in 1957.
But I was thinking about it, but I resisted. Then Zowazo,
Richard Kiersley, and I were doing a study of the effect of
day care. Remember years ago, in 1978, the Congress was going
to vote day care, and we were certain day care was bad for
infants. So we got an NIH grant, and we began to study in
the South End the role of day care on young infants.
But we needed political protection. Things were very bad
then if you remember, and so the Chinese Christian Church
said, "We'll protect you, but you must take Chinese
American infants from Boston's Chinatown."
So we had half Chinese American infants and there it was,
and that's when I saw that these children were temperamentally
so different. Then I remembered the Fels, and of course,
I had been studying children for 40 years, and I realized
I was resisting the notion of temperament. I was resisting
And my resistance collapsed, and that's why I did this.
CHAIRMAN KASS: Peter Lawler.
DR. LAWLER: You seem to have, as far as I can understand,
covered some of the wisdom of Machiavelli, right? Some people
are impetuous; some people are cautious. These are natural
things. We really can't change them very readily, and
which is better sort of depends upon circumstances, and the
point of education is to rein in the destructive aspects of
one or the other.
But then the last page of the very fine chapter that you
gave us you said what we now have to do through education
we will eventually be able to do through pills. Do you really
think this is so?
PROFESSOR KAGAN: Oh, no. I'm sorry. I hope
you didn't misinterpret the last page. This book says
in our technological culture both types have advantages.
There's a line, I think, in the chapter which says if
we ask T.S. Eliot the day after he won the Nobel prize, you
know, you—if you read his biography, he was a very inhibited
child—you know, would you have wanted your mother to give
you medicine, he would have said no, because he wouldn't
have become a playwright.
We need people who like to work alone, bench scientists,
programmers, and of course, that's where our high reactives
drift, and we need statesmen and surgeons and trial lawyers,
and I think the account is balanced.
If I can tell an anecdote that Steve Suomi told me, on the
island of Cayo Santiago, which lies off Puerto Rico, there
are 1,000 Rhesus monkeys. No one lives there, but graduate
students code their behaviors. They know which ones are timid,
And one spring, Steve—
PARTICIPANT: (Speaking from an unmiked
PROFESSOR KAGAN: Sorry? No, the monkeys. Sorry.
In one year two juvenile males died of starvation. They
were the timid ones. Because food is put out once a year,
and they wait, and if you wait too long, you die of starvation.
Two bold monkeys died because they attacked two large alpha
males, and so four males died, two from one temperament, two
from another, and the account is balanced.
We don't tell our mothers of high reactors that, you
know, this is a trait you should change. In an earlier study,
one of our most inhibited boys said to an examiner at nine
or ten, to the question what do you want to be, he said, "I
want to be a scientist, physicist."
He paused, and he said, "I like being alone."
That boy is getting a Ph.D. in physics from Berkeley this
year. He's going to be a productive member of our society.
So in a society as diverse as ours, it is not obvious that
one of these temperamental types has advantages and you wish
to change it.
DR. LAWLER: So although you try to be nonjudgmental
here at the end, you actually are quite judgmental on this.
That is, nature has given us a pretty good deal. We shouldn't
mess with it that much.
PROFESSOR KAGAN: Yeah, I think in our society, which
is mobile and youth leave their homes and so many jobs require
dealing with strangers and risks with minimal uncertainty,
at this historical moment, not in colonial times, there probably
is a slight advantage to the low reactive, slight, yeah.
CHAIRMAN KASS: Let's just take one more, and
then we can hold the rest of the conversation. Gil, do you
want something short?
PROFESSOR MEILAENDER: I think so.
You used the words "bold" and "timid"
with respect to temperament. Moralists in speaking about
virtues sometimes use words like "courageous" and
"cowardly." What would be the relationship do you
think between your words "bold" and "timid"
and the moralists talk about "courage" and "cowardice"?
PROFESSOR KAGAN: I'm glad you asked that. Very
different. When I use the word "bold" and "timid,"
I mean what is your initial reaction to a challenge or a novel
event. Is the initial tendency to psychologically freeze,
encase it, or go forward?
That has got nothing to do with defending your values, and
as a matter of fact, the interviews at age 15 are revealing
that it's the high reactives who are more likely to defend
Let me end by reminding you of that wonderful paragraph
in Portrait of an Artist. Remember Stephen Daedalus
is Joyce, who is a frightened, timid boy, and remember his
mother. He's talking to a friend, and he says, "I'm
not going to Easter mass. I don't believe in it."
He says, "But your mother wants you to go."
And then he says, "No, I shall be a hypocrite."
He says, "So what? You don't believe. Go to please
And Stephen says, "I can't."
That's often the reaction, so that in terms of courageously
defending your beliefs, I have the inkling—and I'll
stop—that those are the high reactives. They're often
ideologically more courageous.
CHAIRMAN KASS: Let's hold the rest of the discussion
until we have Dr. Spelke's presentation.
PROFESSOR SPELKE: While we're waiting for the
PowerPoint to come up, can I just ask if anyone has a laser?
That would be great. If not, I'll make do.
CHAIRMAN KASS: Do we have one or not?
PROFESSOR SPELKE: If not, it's okay. I can
Thank you for inviting me today, My task is both exciting
and impossible. There's no way that in 40 minutes I'm
going to really be able to convey to you what we've been
able to learn about children's cognitive development,
but I do want to try to talk about three aspects of the work
that has been going on.
First, I want to tell you a little bit about the methods
that have been developed over the last 50 years or so for
addressing questions that people have asked for 2,000-plus
years about the origins of our understanding of the world,
the experiences of young infants, the states of knowledge
of infants and how our knowledge grows and changes with development.
These questions are very old, but research strategies for
addressing them have really emerged in the last 50 years,
and I want to introduce you, in particular, to two research
strategies that have been important and will do so in talking
about space perception.
Then I want to turn to some substantive topics and talk
about two core systems of knowledge that I think these strategies
have given us evidence for in young infants. One is a system
for representing and reasoning about the physical world, particularly
solid, manipulable objects, and the other a system for representing
and reasoning about people.
And then my final substantive topic, I want to turn from
the systems of knowledge that infants have to some central
systems of knowledge that they appear to lack and that children
appear to construct over the course of the preschool years
from about age two to about five.
And although there's many such interesting systems,
because of time limits I'll just give you one example
and talk about the construction of natural number concepts,
and then if there's time at the end and you'll permit
me, I wanted to lay out just a couple of themes that I thought
might suggest questions that you would want to consider in
So beginning then with the origins of space perception,
as I said, this is a very old question, but until relatively
recently, "recently" being around the mid-1950s
or so, there weren't systematic attempts to answer it
because there seemed to be this insurmountable obstacle.
In order to answer these questions, we needed to understand
what human infants experience about the world.
Yet human infants have extremely limited abilities to act
on the world, to communicate with other people, to convey
to us what their experience is. And that seemed to place
a major obstacle in the path of pursuing this research.
Well, I think the first serious progress in overcoming this
obstacle came about around the end of the 1950s through the
work of Eleanor J. Gibson and her collaborators, and the crucial
research strategy that she employed was a comparative strategy.
She started with incidental observations on newborn goats,
that if you took a goat just born and put it on a flat surface,
it would start walking around, but if instead you put it on
a tiny stool, it wouldn't move. It would freeze and not
move off the stool, and Gibson wondered what role visual information
might be playing for the goat in controlling this behavior.
So she designed the now-famous visual cliff. This is a
simple apparatus where there's a center board that you
put an animal on, and two plexiglass surfaces immediately
below the center board. So immediately below the center board
on both sides are two surfaces that will support the animal,
which if the animal reaches out and touches them, they'll
feel that they're rigid surfaces of support.
However, on one side, she put—if I had a pointer, I would
point to the right there—she put a visual pattern directly
below the plexiglass. So the surface not only was solid.
It also looked solid.
Whereas on the other side, she put that pattern much further
away. So it looked as if there was no surface of support
there, even though there was one, and what she found was that
newborn goats placed on the center board would readily go
scampering off on the visually shallow side and avoid the
visually deep side.
Now, of course, this is useful to goats, since they live
on mountains and need not to fall off them, but this observation
raised the question, is this an ability that we'll only
see in animals living in those kinds of environments or will
we see them more generally?
And to address that, Gibson repeated the visual cliff experiments
on a wide range of different animals, including rats and kittens
and human infants.
The findings are easy to summarize. For any terrestrial
animal that she studied, at the point at which the animal
becomes capable of locomoting on its own, which is at birth
for some animals; it's after a few weeks or months of
birth for other animals; for human infants, it's about
six to eight months after birth. At whatever point an animal
begins locomoting on its own, they would locomote over the
visually shallow side and avoid the deep side.
The next question Gibson asked then was: is each of these
abilities in different animals due to a distinct mechanism
or, rather, is there a single general mechanism for perceiving
depth and using depth to guide locomotion that evolved in
some distant ancestor common to all of these animals and,
therefore, the same mechanism at work across animals?
To address that question, Gibson did a whole further series
of studies, which I don't have time to describe, looking
for each animal at the signature limits for cliff avoidance.
That is to say, the conditions under which they would succeed
in avoiding a visual cliff and the conditions under which
they would fail, in which she found across this range of animals
was common limits across the animals, providing evidence that
common mechanisms were at work in all of these animals.
Finally, a question arises. What about for those animals
that don't locomote at birth and that require some weeks
or months of postnatal experience before they start engaging
in visually-guided locomotion? What role does experience
play for those animals?
Well, this is a question that one can't ethically ask
for human infants, but one can ask for other animals by doing
controlled rearing experiments, and Gibson did a whole series
of experiments where she reared rats or kittens in darkness
or with nonspecific visual stimulation or with specific experience
And to make a long story short, what she found was that
in those animals this ability developed independent of any
specific learning about the effects of cliffs. In some animals
nonspecific visual stimulation was necessary for the development,
probably for the reasons that Dr. Jessell talked about earlier
today, but no animal needed to learn to avoid the drop-off.
So the conclusion from all of these studies is that depth
perception is innate in the sense that it develops independently
of specific learning in mammals, including humans, given the
evidence that common mechanisms are at work across these animals.
Nevertheless, there's two general related limitations
to this line of work. It doesn't allow—when you study
systematic, coordinated behaviors that only emerge late in
human infancy, you don't have the possibility of studying
the actual development in humans of these abilities until
those behaviors emerge.
And the second limitation, this comparative method works
great for studying the development of abilities that we share
with other animals. It's less clearly applicable to studying
developments that are uniquely human.
So let me introduce the second research strategy that has
been important for this field. There are a number of people
responsible for it, but I think one watershed set of studies
were provided by the psychologist Robert Fantz, again, in
the late '50s. He focused on the fact that human infants
from the moment of birth do engage in at least one highly
systematic and controlled behavior. They look around the
world, and they will show systematic tendencies to look at
some things more than others.
So drawing on this observation, Fantz developed the preferential
looking method, which you see a picture of here. What you
have is a baby lying on its back. It's looking at two
visual displays side by side, and between the displays is
a peep hole. Back in the '50s you didn't have a video
camera, I guess, to do this, and a person looking through
the peep hole and just judging which of the two displays infants
And he found that down to newborns infants would show systematic
visual preferences under certain conditions, and two particularly
interesting ones. First, they would tend to look longer at
novel arrays. So if you presented an infant over a whole
series of trials with, say, two red circles on both sides
of the peep hole, and then after they had seen those for a
while you presented one red circle and one green square, the
infants would tend to look longer at the new display, showing
that they were showing some form of memory and visual discrimination,
and I'll come back to that novelty preference later.
It's not incompatible with the novelty weariness when
you present an entirely new situation that babies show under
The second finding was that infants looked longer at three
dimensional displays than at two dimensional displays, and
that's actually the study that's being pictured here.
What a baby is being shown is a sphere, three dimensional
sphere on one side of the peep hole and a flat disk on the
other side, and the babies looked reliably longer at the sphere.
Now, this may seem to show that babies perceive depth, but
actually it doesn't because there's two different
accounts one could give of that visual preference, as the
psychologist Richard Held pointed out some 20 years later.
One way to illustrate these two different accounts is to
consider some experiments that Held did himself. Now, these
experiments were looking, testing for the development of a
certain kind of depth perception, not the only kind, but the
one whose neural mechanisms Professor Jessell was talking
about this morning, perception of depth from stereopsis.
And for these studies, Held put stereoscopic goggles on
infants such that each of the two eyes was seeing a different
display, different pair of visual displays.
One of the displays that was projected to both eyes projected
exactly the same array to both eyes, and for an adult with
normal stereoscopic depth perception, that looks like an array
of flat stripes.
The other display presented stripes that were two patterns
that were slightly offset with respect to each other, to the
two eyes, and when you put on stereoscopic goggles and look
at that display, you will see stripes arrayed in depth.
Now, the first thing that Held found was that at about four
months of age, between three and four months of age, infants
start looking longer at the display that we adults see as
stripes arranged in depth than at the display that we see
But Held pointed out there's two very different stories
you could tell here about this preference. On the one hand,
maybe infants are also seeing depth just like we do.
On the other hand, it's possible that when infants look
at the display on the right, they just see a pair of double
images, whereas when they look at the display on the left,
they see a pair of coincident imagines that fuse into a single
imagine, and so a very different experience could be underlying
their preferential looking relative to our depth perception.
So how do you get around that problem?
Well, what Held suggested, and this is the critical research
strategy that I think has been followed in most of the work
that has been done since this time looking at perceptual and
cognitive development, what he suggested relies on 100 years
of work on the type of physics of depth perception in adults,
work that reveals that we adults perceive depth from stereoscopic
information only under a very restricted and well defined
set of conditions.
So, for example, in a display like this, we'll see depth
if the edges are slightly offset from each other, but not
if they're very far offset from each other. We'll
perceive depth if the edges are vertical, but not if they're
horizontal. We'll perceive depth if we look through the
stereogram with the glasses on, but not if we take the glasses
And so Held suggested we can look across this range of conditions
at when infants do and don't show the preference for the
display on the right.
And what he found was that infants' preferences lined
up down the line with adults' perception of depth. Infants
preferred the display on the right, starting at about four
months under all and only the conditions in which adults perceived
depth. So that common set of signature limits suggests that
infants aren't just responding to double images. Rather,
they've got the same mechanism that we have and it's
working in the same way, giving us reason to attribute to
infants the same kind of experience of a three-dimensional
world at that point in development as we have as adults.
So to conclude this first section, what this work put together
suggests is an answer to a 2000 plus year long debate. The
tendency to look out into the world and see a stable array
of surfaces at a distance from ourselves and stable, coherent
three-dimensional arrangements appears to develop largely
independently of experience. We seem to be built to perceive
The mechanisms by which we do so are not unique to humans.
They're shared with other animals. And the mechanisms
by which we do so as adults have a long developmental history
in us. We see the same mechanisms at work in human infants.
And these continuities, ontogenetic and phylogenetic, provide
a set of tools that we can use for shedding light on infant's
perception capacities and also on some of their cognitive
So what I want to do now is use these tools and take you
very briefly through some of the highlight conclusions of
research looking at what infants understand about inanimate
objects and what they understand about people.
I'll be particularly fast in talking about objects because
this work is relatively older and some of it was in some of
the supporting materials that there circulated to all of you,
but basically starting as using Gibson's approach focusing
on adaptive behaviors on objects and particular in these studies
the adaptive behavior would be reaching out for objects and
manipulating them and also studies using Fantz's approach
focusing on preferential looking and in particular the tendency
to look longer at novel events. These two lines of studies
have both been conducted to ask, what do infants see when
you present them with an array of objects. And they converge
on the same set of conclusions.
One conclusion is that infants have capacities that we also
have as adults, starting as young as about 2 months of age,
maybe younger, though. It's hard to do these studies
when infants are younger. Capacity is to take a continuous
array of surfaces and break it into units. And the boundary
of those units generally coincide with the boundaries that
we take to be the objects, in a scene.
So, for example, if you present a baby with two objects
sitting on top of one another, one of which slides over the
other but remains in contact with it throughout that motion,
present it until babies are bored with it, and then reach
out and lift up the top object and either it moves by itself
or the two objects move together, babies look longer if the
two move together, suggesting that even though the objects
were in contact throughout the time the babies were becoming
bored with them, they were nevertheless perceiving a boundary
Other studies have asked whether babies are able to interpolate
parts of objects that are hidden from view. If you see an
object whose top and bottom are visible and its center is
hidden behind another object, can babies ever extrapolate
that connection between the two and see a connected object?
And again the answer is they can, if you bore them with a
display like this, a rod moving together behind a block, never
enough for its center to come into view. They will be relatively
bored if you then take the block away and show them a complete
rod, suggesting that that's not new to them, that's
what they were seeing before and relatively more interested
if you show them a rod with a gap in the center.
Further studies have asked whether babies are able to represent
objects as continuing to exist when they move fully out of
view. And these studies show that they can, under certain
conditions, and that they keep track of objects that move
in and out of view in accord with the basic principle that
objects are going to move on spaciotempororally connected
paths. They're not going to jump from one place in time
So, for example, the study of ours took four months old
infants looked just at the—I'll just describe the case
on the right. Four-month-old infants and presented within
an array where there's two screens, an object moves behind
one screen. Then there's a pause. And an object that
looks just like it moves out into view from behind the other
Now, adults looking at that will infer that there's
two objects in that event behind the screen because one thing
couldn't move from far on the left to far on the right
without traversing the space between the two.
To see if babies perceive that, we again bore them with
the event I just described and then remove the screens and
alternately show them arrays with one versus two objects.
They look longer when you show only one, suggesting that they
like us, perceived two objects in that scene.
And one last example, studies have asked whether babies
are able to make inferences about mechanical relationships
between objects. And in particular, if babies infer that
objects will act on each other when and only when they come
into contact. This is one of the earliest experiments that
was done by, I think, an undergraduate at the time. It may
have been an undergraduate of Jerry's, William Ball.
Here's the events that he presented to infants. There's
a screen. There's an object that's partly hidden
behind the screen and another object that moves behind the
screen, and when it's fully out of view, the object that's
initially half hidden and stationary starts to move, okay.
Now adults look at that and infer that the first object hit
the second and set it into motion. But actually, the event
is physically consistent both with that possibility and with
the second. The screen is big enough that the first object
could have stopped short of the second object, and it could
have started moving on its own.
So to see whether babies made the same inference as adults,
we bored with the top event, and when took the screen away
and then indeed they looked longer at the event where the
first object stopped short of the second, suggesting that
they inferred that the two would come into contact.
Well, putting all this work and other related work together,
I proposed and others have proposed that babies starting at
about two to four months have a system for building representations
of objects that accord with three general spaciotemporal constraints
on object motion.
They build objects that are cohesive, that is, bodies that
are internally connected and move relative to one another,
that are spaciotemporally continuous, that is, that these
are things that bodies can't do. They move on connected
paths, but they don't jump from one place to another and
their paths also don't intersect, such that two things
occupy the same space at the same time. And they move when
and only when they come into contact, there's no action
at a distance.
But interestingly, this is not to say that infants perceive
objects under all the conditions that adults do. My colleague
Susan Carey has discovered some interesting limits to infants'
abilities to perceive objects. If you present them with situations
where spaciotemporal information for properties like cohesiveness
and continuity do not dictate where the boundaries of objects
are. For example, you place a toy duck on top of a toy cup,
presenting no relative motion between the two, no motion at
all in the scene, infants' perception of the boundary
of those objects is indeterminate.
Now recently there's been a lot of beautiful work, some
of it on the island of Cayo Santiago that Jerry was talking
about asking whether this same system of representation exists
in nonhuman primates and also asking whether the same system
exists in adults. And like the strategies of Gibson and Held,
the way of asking whether the same system exists is to ask
do we see the same abilities and the same limits. Well, to
just simply give you the answer, since I don't have time
to take you through the studies, the answer appears to be
yes in both cases. This system exists in adult rhesus monkeys.
It also exists in human adults when you test us under conditions
where we're not able to use our specific knowledge about
the world, our language, other strategies that infants lack
to track things through time. We show the same patterns of
success and failure as infants do.
Well, let me turn to the other core system that I wanted
to lay out for you. This, I think, is a system for reasoning
about persons, and there have been many signs from studies
of infants over the last 20 years or so that the system exists
and is at least as important for infants as the system for
One source of evidence for this system comes from some studies
conducted by the psychologist John Morton and Mark Johnson
with newborn infants. They took brand new babies, put them
on the lap of an experimenter, and presented them with simple
schematic oval-shaped patterns, either a pattern representing
a face or one of a number of other kinds of displays. And
when the baby was looking at the pattern, they slowly moved
it to the left or right, and the measure was how far would
the baby follow the pattern, how far could they move it before
the baby would lose interest, turn away, no longer track it?
And what they found was that babies would track for the face
pattern at birth reliably longer than for the patterns that
clearly were not face-like and not reliably, but somewhat
longer when they tracked a simplified face pattern as well.
And this study and others suggests initial sensitivity to
the structure of the human face.
Here's another ability that comes in by about two to
three months of age from studies by Bruce Hood. In these
studies, infants view the face of a person on a computer screen
looking straight at them and then, I don't know if you
can see it up here, but the person's eyes turn either
to the left or to the right, and right after that the person
disappears and an object appears either on the left or on
the right. Now no matter which side the object appears on,
the babies turn to look at it. But when Hood measured how
fast babies turned to look at it, it turned out that they
turned to look to the object faster if the person's eyes
had—if it appeared on the same side where the person's
eyes had turned than if it had appeared on the opposite side.
So from about two to three months of age, babies are sensitive
to human gaze, and they're following the gaze of a person
who's looking straight at them to an object that's
facilitating their attention to other objects.
Well, what about people's actions? Do babies form any
sensible representations of other people's actions? Well,
this is something that psychologist Amanda Woodward has been
studying for the last five years or so through some simple
ingenious preferential looking experiments. In these experiments,
she shows a baby two objects and a person whose hand reaches
repeatedly for one of the two objects, so for example, for
a given baby it might be the ball. After the baby has gotten
bored with looking at this, she then switches the positions
of the two objects and has the hand alternately reach to one
versus the other.
Here's the reason for doing the experiment. The question
is, when the baby sees the person reach for an object how
do they encode what the person is doing? Do they think the
person is just like an inanimate object that would move from
one position in space to another position in space or do they
encode the action as goal-directed, directed to the goal,
in this case reaching for the ball? Well, if they encoded
the action as simply a movement through space, then this should
be the event that's most similar to it. But if they encoded
it as an action directed at a goal, then when the two objects
move, it's the new action to the new position to the old
goal that should be seen as more similar, and infants'
looking time was consistent with that second possibility.
Okay? Infants looked longer when the goal changed than when
the physical trajectory of the action changed.
PROFESSOR SANDEL: Could I just ask you, the
test, though, is looking longer?
PROFESSOR SPELKE: Right.
PROFESSOR SANDEL: How do you know what that
means? How do you know whether they're looking longer
because they're making sense of it or they're looking
longer because it doesn't make sense?
PROFESSOR SPELKE: Okay, in all of these studies,
the pattern of data internal to the study answers that question,
okay? So for example, you can have conditions in which you
don't change the positions of the objects. Right, and
so if they're looking longer to something that's familiar
or sensible, they look longer to the old motion than to the
new, but that's not what they do. Does that make sense?
So I mean this is a situation that pits two kinds of novelty
against each other and when we see they look longer to this,
we're assuming they're going to look longer to novelty.
PROFESSOR SANDEL: That's my question.
Why do you assume that looking longer corresponds to novel—the perception of novelty?
PROFESSOR SPELKE: Right, because you can also do
studies where instead of pitting two kinds of novelty against
each other, you can pit an event which by any description
is more novel against an event which by any description is
more familiar. So suppose, for example, instead of switching
the positions of these two objects, we just left them where
they were, let the babies get bored with one and then alternately
show reaching to the old one versus reaching to the new.
Infants will look longer at the new one, and so that's
suggesting that in this situation when they're bored,
they will tend to prefer the more novel event. Did that make
PROFESSOR SANDEL: I thought the question—
PROFESSOR SPELKE: Let me—
PROFESSOR SANDEL: Please, so ahead.
PROFESSOR SPELKE: Let me keep going because I think
in principle that is a real concern and in practice the pattern
of results across the studies addresses it.
The one last thing I want to tell you about the Woodward
studies is that this effect is specific to people. When she's
repeated these experiments with inanimate objects moving towards—and the inanimate object she used was superficially in some
ways similar to a human hand and arm. It was a stick with
a sponge like deformable thing on the end of it and she bores
babies with the stick moving to the ball and then switches
the positions of the two objects. She does not get a preference
for moving to the same goal, suggesting that this propensity
to understand actions as goal-directed is applied to people
in infants of this age and I should have said these are five-month
old infants. It's not applied to inanimate objects.
Finally, there are other studies, infant sensitivity to
human actions, that show that babies are sensitive to relationships
between what they themselves do and what other people do.
The most famous studies, also were very controversial for
a long time, were conducted by Andy Meltzoff some 25 years
ago and showed that when, for example, he sticks out his tongue
at an infant, infants will reliably not really stick out their
tongue in kind, but act in a way that a blind observer looking
at it would judge to be more like sticking out your tongue
than like other events that the model engaged in.
We now know after many years of controversy that this ability
in newborn infants is extremely limited, but it is real and
it leads to much more dramatic abilities to attend to the
actions of other people and reproduce their actions on objects
later in infancy.
I have just one example here to share with you. This is
also from a study that Meltzoff conducted with much older
infants. Starting at about nine months of age, if an infant
views a person acting in a particular way on an object and
then that object is presented to the infant, the infant will
tend, if possible, to reproduce the action that they saw the
person perform. In this study, which is with infants that
are somewhat older, they're about 14 months. Meltzoff
went a step further and asked what will happen if an infant
sees a person attempting to do something but failing, and
in this study the person is attempting to take this little
barbell and pull it apart and failing to pull it apart. What
he found was that when he then gives the barbell to the infants
they grab it and pull it apart. And again, this is specific
to people when they see the very same series of motions, but
it's a machine that's doing the actions, they don't
show that effect.
Okay, so these are all ways in which infants seem to be
able to make sense of the actions of other people. But we
can ask, do infants also apply their understanding of inanimate
objects to other people? Do they expect that people will
interact on contact, that people will exist and move continuously.
And some experiments have begun to ask this. In particular,
Woodward did a series of studies where she took up the old
experiment by Bill Ball showing that babies infer that two
inanimate objects that move in succession come into contact
and asked, will they make the same inference for people?
So for this study she presented infants with videotaped events
of real people or real large complicated inanimate objects,
things like potted plants and high chairs, not the objects
that I've diagrammed there, conducting the same experiment
that I described earlier.
What she found is that as in the previous experiments, when
the objects were inanimate, infants inferred that when two
objects moved in succession, they came into contact, but when
the objects were people, they did not. Okay? No inference
that the first person slammed into the second to set them
Finally, very recent set of studies being conducted at Yale
in the lab of Paul Bloom have asked whether infants infer
that people's motion is subject to continuity. And here
they took our method for studying continuity that I already
described to you, two screens and an object that moves behind
one and then another object that appears behind the other
and they found, like us, that when the objects are inanimate
infants infer that if there's been discontinuous motion
there must have been two objects. Interestingly though, they
don't infer that in the case of people, even though it's
true, right? A person can't move from the first place
to the second without passing through the space in between,
but infants don't infer that people's behavior is
subject to that constraint.
So to summarize what I've told you about infants' understanding
of inanimate object motion and human action, I think we see
evidence here for two quite distinct systems of knowledge.
In the case of inanimate objects, infants reason about object
motion in accord with a principle of no action at a distance,
contact mechanics, if you will. They infer that such objects
are not goal-directed. They don't show goal-directed
inferences in the Woodward kinds of studies.
In the case of people, you see just the reverse. People
are predicted to act in relation to goals and not in relation
to proximal mechanical forces.
Similarly, tracing people over time may involve making inferences
about their intentionality. Certainly babies seem to be sensitive
to intentionality, and it doesn't involve making inferences
about continuity, whereas for inanimate objects the reverse
would seem to be the case.
So it looks like we have two distinct systems here. Now
I've raised the question that I can't, unfortunately,
answer about whether these systems show continuity over phylogeny
and also over ontogeny. In the case of studies of nonhuman
primates, there are a number of studies now of nonhuman primates
showing that pieces of the abilities that we see in infants
are shared by other animals, but there's currently a debate
in the field as to whether the entire set of abilities that
we see in human infants are part of our primate heritage of
whether any of them are unique to us. I think that's
a question we have to leave on the table for the moment.
The decisive studies on nonhuman primates just haven't
In the case of ontogeny, I want to throw out a speculation
to you. This is an interesting group, I think, in which to
do this because I believe you may have already heard presentations
from some neuroscientists saying there's really only one
system of causal relationships that underlies all behavior.
Human beings are just complicated machines. Our minds and
brains operate in accord with the same mechanical principles
as inanimate objects. And if one takes these claims seriously,
it would seem to suggest that people can come to overcome
this notion that people and inanimate objects are fundamentally
different from one another.
My own experience though, for what it's worth, is that
colleagues who say these things to me about human action tend
to agonize over personal decisions as much as anybody else
does, tend to experience moral indignation. I think that
this notion that people choose their actions and could have
acted otherwise is actually very deeply ingrained within us,
and I think we're seeing the origins of it in these studies
Okay, I want to completely shift for the remainder of my
time and talk about some capacities that we don't see
in other animals for sure and that we also don't see in
human infants but that we do see in most children at the time
that formal education begins, a set of distinctively human
concepts and abilities—one of the abilities which I didn't
put up here is the capacity for language—that allow us to
communicate with one another, to use symbols and to use abstract
concepts that are at the basis of all science and technology
and much of modern life.
Now, what I want to suggest in the case of all of these
concepts is first of all that infants lack them; second of
all, that these concepts aren't explicitly taught to anyone.
They develop spontaneously in children, most children between
the ages of about 2 and 5; third, that their development depends,
in part, on experience; and fourth, that their development
is absolutely necessary for formal education, that a child
who hasn't developed these fundamental concepts will be
at serious disadvantage in formal educational setting.
Now if I had hours and hours I would try to tell you about
all of these. Since time is very limited, I want to try to
flesh out these claims by looking just at one set of concepts,
natural number concepts. And what I will try to do in my
last few minutes here is begin by telling you very quickly
about two core systems for capturing numerical information
that we do find in infants and nonhuman primates, as well
as in us adults, one for dealing with small numbers, the other
for dealing with large approximate numerocities.
Then I want to talk about the construction of the uniquely
human natural number concepts over the course of the preschool
years and then finally, I want to give you a bit of evidence
that this construction depends, in part, on experience. So
the first core knowledge system that serves as a building
block for children's number concepts I've already
introduced you to in talking about objects. This is a system
for representing small numbers of objects and all I want to
point out here is that it has some of the properties of our
system of natural number. So this is a system, for example,
that research by Karen Wynn has shown babies are able to use
to do something like compute the effects of adding an object
to a scene or taking an object away from a scene. If you
place an object on the stage, this is again a preferential
looking experiment, 5-month-old infants; place an object on
a stage, cover it by a screen, add a second object to the
scene and then ask infants, in effect, how many objects are
there by lowering the screen and presenting either the right
or wrong number of objects. Infants will look longer if you
present the wrong number, and they do that both in this one
plus one kind of problem and also in other similar problems.
Infants can also use representations of small numbers of
objects to make numerical comparisons, and these are studies
that have been explored most thoroughly with older infants
about 12-month-olds. Research in Susan Carey's lab has
taken 12-month-old infants and presented them with graham
crackers, putting one plus one equals two graham crackers
into one box. Two minus one equals one graham cracker into
the other box, pushing the boxes apart and encouraging the
infants to crawl to them. And infants will tend to crawl
to the box that has the larger number of graham crackers,
showing that it can both represent the numbers in each box
and compare those numbers.
But as always with infant research, the limits to infants'
abilities are as interesting as the abilities themselves and
in particular, in these situations we see two general limits.
The first is a domain limit. You see these abilities when
you present babies with solid, manipulable objects. You don't
see them when you present them with nonsolid substances or
many other kinds of perceptible entities and the second is
a set size limit. You see these abilities when you present
up to three objects, but when you present more than three
objects babies fall apart.
So here is the results, for example, of the box choice study,
one versus two graham crackers, they go to two. Two versus
three, they go to three. But if you then test with three
versus four or even four versus eight, they're choosing
at chance between the two. They're not able to keep track
of more than about three objects.
Now we see the same limits in monkeys. This is research
by Mark Hauser and the same limits in human adults when you
prevent us from counting or otherwise verbally encoding the
displays, suggesting this is the system that shows considerable
The second system has been revealed through experiments
using an even simpler novelty preference method. For example,
a method that's been used a lot in studies of speech perception
where you take an infant and present them with the sequence
of sounds coming out of one of two side speakers, simply measure
how interested they are in the sound sequence by seeing how
long they will turn their head in the direction of the speaker.
Then you can test for their abilities to discriminate different
numbers by familiarizing infants to a set of different sequences
all presenting the same number and then testing them with
new sequences alternately presenting the same number of a
different number. I hope this is clear.
So what's illustrated here, half the babies are bored
with four sound sequences, with four sounds; half with sequences
of eight and then everybody is tested with sequences of four
and eight, and you see will they turn their head to the speaker
longer when they hear a new number?
And very quickly, the findings, the study has been done
with many different numerical discriminations. The findings
are at about six months of age, the first age at which you
can use this method, babies do show abilities to discriminate
on the basis of number. They'll discriminate sequences
of four from sequences of eight, for example. Their number
discriminations are imprecise. So if you repeat the experiment
to test discrimination of four from six, they fail. And what
determines whether they succeed or fail is the ratio of the
two numerocities. So a baby who succeeds with 4 versus 8
will succeed with eight versus 16. If they fail with 4 versus
6, they'll fail with 8 versus 12.
And finally, as you test older babies, you find that the
critical ratio for discrimination narrows. A 9-month old
in particular, will succeed with a two to three ratio that
the 6-month-old fails with.
Now one can vary the kinds of events or displays presented
to infants to see whether this is an abstract system of number
representation by instead of presenting sequences of sounds
you can present sequences of actions, a puppet that jumps
some number of times. You can also present visual spatial
arrays, arrays of dots, of one or another numerocity. The
findings from those studies line up perfectly with the findings
from the studies with sound sequences. If babies can discriminate
a given pair of numbers of sounds, they can also discriminate
that pair of numbers of dots or visible actions.
Now these large number representations show two signature
limits. One that I've already described is the ratio
limit on discrimination. The other limit I also described
earlier in talking about the first system. It's a tracking
limit. Babies are able to discriminate 8 events from 16 events
when the events occur in immediate succession in an object
that's continuously visible. But if they see one at a
time, four objects going into a box versus eight objects going
into a box one at a time, and they have to track each of these
individuals as it's hidden, they are not able to use this
large number system in that case. So the same limits have
been found in monkeys and also in human adults, people like
us, when we're prevented from counting or otherwise use
language to encode the displays, suggesting that the second
system is also showing continuity.
So it looks like we have evidence for two systems capturing
aspects of numerical information in infants, one focused on
small numbers, the other focused on large numbers, each system
showing successes and failures under a distinctive pattern
So the question then is what happens to these two systems,
as children learn verbal counting, as they interact with other
people, and as they develop the kinds of number systems that
they're going to need to use in school?
Now clearly, when children go to school, the assumption
is made that they've got one system of number concepts,
not two; that the system shows neither a set size limit nor
a ratio limit. Natural number concepts allow you to represent
whole numbers precisely with no clear upper bound. And that
each number concept refers to a set of numerically distinct
individuals. And the question is where did these representations
Well, I think some very interesting work that began again
with Karen Wynn and has been pursued by a number of different
investigators since then suggests that these number concepts
actually emerge well after children first learn the verbal
So in most American households, somewhere around age 2 or
two and a half, children start engaging in counting. I'm
sorry this is so distorted. There are supposed to be about
nine fish on this table, and if you take an average two year
old and say how many fish are here, the child will very likely
go through the routine of pointing at each of the fish one
at a time and going one, two, three, four, five, etcetera,
engaging in verbal counting.
What Wynn showed, though, is that at this early point when
children are engaging in this activity, they don't have
the faintest idea what these words mean or what the activity
is about. And one way to show that is to ask them a simple
question. After this child has counted all nine fish—here's
a pair of questions that shows this. After the child has
counted nine fish, you ask the child would you put one fish
in the pond? And the child succeeds. That shows the child
understood the question, is motivated to answer correctly.
So now you ask would you put two fish in the pond? And
I can tell you the reaction of my son, aged two years or so
at the time that Karen conducted this experiment on—she
was conducting it at the time, and she came to visit us and
did it on him. He looked at her as if she had suddenly switched
to a foreign language. He didn't have the faintest idea
what she wanted, and he then grabbed a handful of fish and
put them in the pond. What her data showed was that at this
point there was no relation between the number asked for and
the number given, except that a child always gave one when
asked for one and always gave more than one when asked for
Now this state persists for about nine months. For about
nine months children are using all of these number words in
the verbal counting routine without understanding what any
of them mean. And then somewhere around age three and a quarter
or so, children learn the meaning of the word two and at that
point when you ask for two, you get two. When you ask for
three, you'll get a handful, but not one and not two.
And about another three months later, they learn what the
word three means, and then something magical happens, and
they figure out what the whole counting routine is about.
Now what could be going on here? I think in light of the
work on infants, we can make the following suggestion, that
in the initial step of learning verbal counting, children
figure out that the word one applies when you've got a
single object, a single individual to represent, and the system
for representing small numbers of individuals may support
They also know that the other number words apply when you've
got a set, a bunch of things, some nonspecific number of things.
When they learn the meaning of two, what they have to learn
is the word two applies just in case your system for representing
small numbers picks out an individual and another individual,
and your system for representing large numbers picks out a
set of things, a small set of things. And then three can
be learned in the same way, and then the child can't go
any further because the small number system, remember the
core system, has a limit of three.
But what the child can do at this point is discover that
the progression from two to three and the counting routine
involves two things, adding an individual to the set and increasing
the cardinal value of the set. And once they've got that
and can generalize that to the other number words, they've
worked out the meaning of the counting routine.
Now I've gone through all this because there's no
evidence for any of these developments in any nonhuman animal.
This is, I believe, a distinctively human achievement. And
what it consists of, what the child is doing with support
from parents but without any explicit instruction by anyone
is putting together representations from two core systems
with a system of number words and quantification that's
emerged through communication with other people through their
natural language and their culturally specific counting routine.
And I certainly don't have time to give you this evidence,
but there's a wealth of evidence now suggesting that the
same three systems are at work in us as adults when we use
and reason about natural number concepts.
So the last question I want to ask does experience play
a role in this construction? In one trivial sense it must.
The number words of English are different from the number
words other languages, and they have to be learned, but I
mean to be asking a deeper question. Do the concepts that
those words pick out emerge, does experience play any role
in the emergence of those concepts?
Now I think there's evidence from the research of two
educational psychologists, Robbie Case and Sharon Griffin,
that indeed experience does play a role. What they showed
first is that in the United States and Canada, although most
children from middle class families have developed these number
concepts by the time schooling begins, many children from
disadvantaged families have not. In particular, Case and
Griffin did a set of studies where first they would take children
and have them count and show that all of the children, these
are in kindergarten classes, all of the children could count
to a nice satisfyingly high number, and then they would take
numbers that were in the list that the child had produced,
him or herself, and simply ask questions like if I have five
apples and you have four apples, who has more applies, testing
their understanding that the word "five" picks out
a larger numerosity, a larger number than the word "four."
Now when they asked these questions to children from high
or middle income families, almost everybody can answer them.
When they ask them to children from economically disadvantaged
families, they got a much lower rate of success in kindergarten.
The next question was why would the kindergarten children
from the disadvantaged families be succeeding less? And one
possibility that they pursued was that these children may
just not be having the interactions in their homes that middle
class children are having and that lead children to make this
construction spontaneously on their own. He found, for example,
they found that in middle class families there was lots of
talk about number, lots of play with board games, rolling
dice and counting up moves and things like that, much less
of that in the disadvantaged homes.
So Case and Griffin designed a series of intervention studies
where they simply took kindergartners from disadvantaged backgrounds
who tested badly on these number concepts initially and played
a set of games with them, a set of board games, no explicit
teaching, but played games involving talking about number,
rolling dice, counting out moves and so forth and reported
dramatic improvements in the children's understanding
of number which were followed up in one year and three year
follow-ups with dramatic advantages in their formal mathematics
So to summarize this, it looks like preschool children construct
natural number concepts spontaneously without formal instruction,
as long as they're given supportive environments. Without
environmental support, it's not clear that these concepts
will develop on their own. Children who don't develop
them then would be at a disadvantage when formal schooling
starts, because that schooling presupposes that when a teacher
uses the word "three" and the child uses the word
"three", they're talking about the same thing.
This failure, though fortunately, the work of Case and Griffin
suggests, can be remedied.
Well, I think this general picture applies to many of our
most interesting uniquely human concepts, so I could have
given a different talk about mental state concepts, concepts
like beliefs and desires, propositional attitude concepts.
There's good evidence that infants don't understand
these concepts, but that most 4-year-olds do and that children
aren't explicitly taught what beliefs and desires are,
they figure it out on their own in supportive environments.
Similarly, for understanding symbols, young children may
look like they understand symbols when they point to a picture
of a cow and say "cow", but do they really understand
that that set of marks on paper is a representation of a cow,
a symbol that stands for a cow, not a cow itself. There's
evidence that that understanding is not present in infants,
that it develops over the years from two to three or so.
The research of Judy DeLoache; again, distinctively human
ability constructed by children without formal instruction,
but only in supportive environments, same for concepts of
works of art and tools. So in each case, I think these concepts
are developing in the ways that natural number concepts did.
And I think this suggests that this time from age two to five
is really a critical time in human development, for the development
of a whole host of uniquely human cognitive abilities, abilities
that mark us from other animals and also abilities that make
us capable of formal instruction.
So what I want to do, I'm probably over time, but I'm
just about done, is just end with three general themes and
suggestions that I think this work may support, at least that
I wish to offer to you to consider whether you think this
work might support them. The first theme goes back to the
thing you questioned me about, this preference for novelty.
I think it's the case that as early as we look in infancy,
we see that infants are motivated to learn. Now they're
motivated, in part, to learn on their own seeking out situations
that give them new information. But they're especially
motivated to learn from other people, and we see this in everything
from the following a person's gaze to the objects that
they're looking at, to reproducing their actions, to reading
through their actions to the intentions behind them.
But I think a possible implication of this work is that
while infants are built to learn, they're not built to
learn alone. They're built to learn in interaction with
other people who already are immersed in the culture into
which they will be growing. They need social partners to
do this, and I think arguably the best social partners to
do this would be one or both of the infant's parents.
In that context, I think you might consider whether our
society is well advised to give so many parents the cruel
choice between having to decide between the economic welfare
of their families, on the one hand, and their opportunity
to be there as a participating parent during the first year
of a child's life, on the other. Should we be requiring
single mothers on welfare to work when their children are
infants? Should we be requiring families that require two
incomes to support their children to be choosing between giving
their children that economic support and giving the children
the presence of a full-time parent in the first year? I think
that's one issue worth considering.
The second theme, which I just ended the substantive presentation
with, is that we should pay more attention to the years from
two to five, that this is a critical time for children's
cognitive development. This isn't a time when I think
we need to be putting children in schools or starting formal
education. It's a time when children are learning great
things on their own. But they're only learning those
things in supportive environments. And I think that suggests
a further duty that we may have to children that you may want
to consider in this panel, the duty, first of all, of understanding
what are these capacities that are developing in the years
from two to five, and what kinds of environmental support
are necessary for the development of them. The kinds of studies
that Case and Griffin have done for number need to be done,
I think, in other domains as well.
And second, to assure that children throughout our
country are able to grow in environments that provide the
resources that they need for that development.
The final theme is maybe the most speculative one, but for
me, I think it may go the deepest and it takes me back to
the first part of my talk in talking about systems of core
knowledge in infants. Now I think that a general conclusion
that's emerged over the last 50 years is that to a surprising
extent young infants, not always newborns, but 3-month-olds,
5-month-old infants share ways of conceiving the world with
us adults, that they experience the world in distinctive ways
that are much like the ways that we, as adults, experience
the world. And that's true, I think, in the case of space
and objects and other people, the three cases that I talked
But there's a way in which infants are radically different
from us. When we experience negative situations, we're
capable of acting to change those situations. When we find
our own resources for acting are limited, we can communicate
our needs to other people and seek help from them. But infants,
though they share many of our experiences and capacities,
radically lack the capacities to act that we have to ensure
that their environments are safe and healthy, that they're
cared for and that their needs are met. And I think that
gives us the most fundamental ethical responsibility of all,
vis-a-vis infants, and that is to care for these creatures
who are so like us in our experiences, but so unlike us in
their capacities to provide their own care.
CHAIRMAN KASS: Thank you very much, Dr. Spelke,
Professor Kagan. The floor is open for discussion.
PROFESSOR SANDEL: Thank you very much for
that great presentation which really covered an enormous amount.
I would like, if I could, to go back to the question about
the looking longer test of novelty, the preferential looking
test and what we can infer from it.
Thinking back to that two by two diagram that summarized
the study of children's perceptions of what makes inanimate
objects and people move, contact in the case of inanimate
objects, goals in the case of people, it's a lovely result,
so lovely that it's also a suspicious result because what
it discovers is that the way children perceive the world and
the laws of movement recapitulates the 17th century split
between explanation in the physical sciences and in the human
sciences where we concluded in the last 300 years that mechanism
governs the movement of inanimate objects, not teleology.
But the teleology or goal-directed explanations govern the
human sciences, the way people behave, and it just turns out,
lo and behold, that children naturally, so to speak, perceive
that relatively recent discovery, the way that the modern
science has bequeathed this split between the modes of understanding
in the physical and human sciences. So this is really to
explore that suspicion.
What's the warrant for inferring from the looking longer
test one thing rather than another? And here's another
experiment, and I wonder what you would read from these results.
You can imagine you could time, never mind children, even
adults, how long some adults might gaze at the sunset or the
rising of the sun as those events naturally occur, familiar
though they are, not novel. People gaze sometimes for long
periods of time out of appreciation or contemplation or who
knows what. And then suppose you, as the counter example,
you did an experiment of the kind that occurs in the movie,
The Truman Show. You remember in The Truman Show, unbeknownst
to Truman, he was the character, the Jim Carrey character.
He's living within an elaborately constructed television
set where everything, including the seasons and the weather
and the rising and the setting of the sun is governed by a
producer who's in a control room. And at a certain point
this fiction is maintained, but at a certain point the producer,
they're desperately looking for Truman who has escaped,
but it's the middle of the night. And they can't
find him, and so the producer decides he has to spoil the
fiction in order to bring up the lights. And he says "cue
the sun," and the sun rises in the middle of the night.
That would be novelty.
Now suppose you timed, you applied the looking longer test
to the sun that arose at that unnatural moment, unfamiliar
moment, would people, infants or would we—would you expect
that we would look longer at that unnatural, unfamiliar rising
of the sun, than we do when we sit in contemplation of a natural
sunrise? And how would you know if we looked longer at the
one rather than the other that one was novel and the other
PROFESSOR SPELKE: Right. Actually, in that case,
I think we probably would. If the sun suddenly rose in the
middle of the night, I think we'd all run outside and
look at it. However, I completely grant your general point.
And in fact, the reason I went through the whole discussion
of the Held experiments is that for any given behavior that
an infant chose, looking longer at one thing than at another,
we could tell multiple stories about the causes of that behavior.
That's absolutely true.
So one thing that I had to do in making this presentation
because I wanted to cover a lot of ground was give you the
results of single experiments, but it's when you look
across experiments that I think the following things emerge.
First of all, it is not the case that infants will absolutely
and under all circumstances prefer novelty. That's not
true any more than we as adults will always prefer novelty.
Actually, one case I think in which it's least likely
to be true is cases involving human action where if other
people do things that are bizarre, we may avert our eyes from
it rather than look at it. So it's not at all the case
that one can take as a given whenever an infant looks longer
at something, that must be more novel to them.
However, within a series of experiments what you do in these
studies is the following. You start with an analysis of an
ability. Under what conditions would we, as adults, see something
as goal-directed or not. Then you make a set of focused predictions
about what you would see in an infant if they had the same
system for representing things as we have. And then you test
down the line whether those predictions hold. Now if you
get a coherent pattern across them, all involving when you
change something in a way that would lead an adult to say
that's funny, the goal just changed, the baby looks longer.
Across that series of studies, there isn't a coherent
explanation to give if the baby were preferring the familiar
one and there is a coherent explanation to give if they're
preferring the novel one. So that's one answer to your
The other answer to your question, though, is that both
in the case of research on infants' reactions to people
and in the case of research on infants' reactions to inanimate
objects, the examples I gave were all preferential looking
experiments. But in fact—actually, they weren't all.
In both cases, multiple methods have been used to converge
in on the same abilities. So for understanding infants'
representations of other people's actions, you can use
imitation tasks where they repeat people's actions and
ask is their repetition true to the goal or is it true to
superficial properties of the actions? And the answers you
get from those studies converge with the looking time studies.
Similarly, in the studies of object representations, you
can ask what do babies see as the bounded objects in a scene,
either using preferential looking with the assumption of a
novelty preference or by using reaching, and you see a convergence
across them. So I think in both cases it's the converging
findings of series of studies within each method in relation
to adult abilities and also series of studies across methods
that lead to these conclusions. And you're quite right,
that if you single out any single experiment, then you're
in the situation that Richard Held was in when he only had
the first experiment on the 4-month-old infants looking longer
at the side with double image stripes than single image, and
you don't know what the basis of the response is.
CHAIRMAN KASS: Bill Hurlbut.
DR. HURLBUT: If our question here, fundamentally,
is nativism versus empiricism, I want to ask you about the
sort of fullest form, that being the development of the moral
mind, because building on what Michael was just talking about,
I believe there are studies that indicate that infants will
look at things, certain things, longer than others based on
what you might call intrinsic value. So, for example, you
mentioned faces. There are studies that indicate that certain
faces rated as attractive faces by adults and across cultures
command the longer attention of infants. Now that's really
amazing. That means somehow—and this is quite early infants,
I believe. Some sort of pattern, and it's not just symmetry,
some pattern has an intrinsic value. It's built in, it
would sound like. And that's the first thing. I'd
like you to elaborate on that.
The second thing is not just intrinsic value, but the roots
of our relationship between awareness and action. You cited
Meltzoff's work of imitating faces. When you stop and
you ponder that capacity, that again requires that the infant
somehow has a pattern that it knows sensorially how to apprehend,
but also how to reproduce.
Now I know there are some findings that suggest there are
cells, Rizzolatti's so-called mirror cells that could
mediate this. But it would imply that quite high order constructions
that somehow there was already a connection between sensory
input and motor action. And of course, the concomitant of
that is that if there is motor action, even with a sensory
input, that the subjective states that accompany those muscle
actions would already be communicated in a kind of inter-subjectivity.
All this seems to add up to a foundation for the moral mind.
Can you just comment on all of that?
PROFESSOR SPELKE: Sure. On your last point first,
I think there's actually quite rich evidence that babies
are sensitive really early to the correspondences between
their own actions and other people's actions. One thing
I didn't talk about from the Woodward studies is that
if you look at the kinds of actions that babies can attribute
goals to, they develop hand in hand with the child's own
capacities to act. So it's at the point at which a child,
him or herself, will start pointing to objects, that they
will interpret a pointing action that they see another person
perform as goal directed by the Woodward test. So I think
that goes along with your line of thinking.
And you're also quite right that we don't know whether
this is unique to humans or not. There's clearly pieces
of it in monkeys as shown by the Rizzolatti work and other
work like that.
I'm going to have to punt on the morality question,
not because I don't think it's important, but because
I really think that the jury is out. There are hints of pieces
of what will become a system of moral reasoning that I think
we can discern in infants, but whether we have a full blown
system that's showing itself in a glimmer here and a glimmer
there, or whether a full blown system of moral sense is going
to require the kinds of constructions that a system of numerical
reasoning requires, constructions that children will put together,
perhaps without explicit teaching, perhaps learning by example
and by observing other people. I think we just don't
have the evidence at this point to say.
There's, of course, a long tradition of studying moral
development in children. I think unfortunately much of that
tradition has focused not on children's moral intuitions,
but on their justifications, and then what you find, not surprisingly,
is that young children who are pretty terrible at giving coherent
justifications for anything they believe, don't give very
sophisticated looking moral justifications either.
But the question of whether they have underlying intuitions,
like our intuitions as adults and whether there's a core
to morality is, I think, going to be a very interesting and
important question to pursue, and we don't have the answer
DR. HURLBUT: Can you say at least a little bit about
intrinsic value, like beauty in faces?
PROFESSOR SPELKE: Here I want to voice some of the
skepticism I think related to what Dr. Sandel was saying before.
We can say that a baby will look longer at one thing than
another, but do we know that that means that they're esteeming
it, that they're endowing it with value? I'm not
sure that we do. But maybe Jerry—I think this goes out
of my domain of sort of cold cognition and into emotion, and
maybe Jerry wants to speak to that?
PROFESSOR KAGAN: Babies prefer symmetry. I thought
Liz was going to say that actually, and that's why, they
prefer vertical symmetry so I agree with Liz. They don't
understand beauty or that that face is pretty, but they will
look at vertically symmetrical designs, and pretty faces are
Incidentally, that doesn't mean that I don't believe
that biologically human children don't inherit a foundation
for moral sense. They do somewhere in the second year, but
I don't believe they have it in the first six or eight
months of life. They get it later.
CHAIRMAN KASS: I have Janet, Ben Carson, Mike Gazzaniga
and then myself.
DR. ROWLEY: I actually have three questions, if
I may, two probably for Liz and one for Jerry.
So neither one of you actually talked about the change in
the complexity of the brain connections that develop over
time and focus really, you focus on early cognition which
is certainly important, but all of the evidence, of course,
which I know you would agree with, that the brain certainly
matures over a long period of time and so there are certain
kinds of skills that children should be exposed to and experiences
they should get relatively early, but then there are other
things that are going to evolve later, and trying to force
a child to learn some things at three or four is futile or
Would you say a little bit more just in general about conductivity,
and then I've got two more parts.
PROFESSOR SPELKE: I agree completely, and I think
one of the values of the work on cognitive development is
there's a huge gap between what we understand about learning
and cognition on a functional level and what we understand
about neural growth and conductivity on a structural level.
So I think that the beautiful work on neural development tells
us to expect that children will be optimally ready to learn
different things at different ages, but it doesn't in
itself yet tell us what things they're ready to learn
at what ages, and we need the behavioral work to do that.
And I quite agree with you. For example, based on the work
that I presented on number, I think it would be futile at
best and possibly harmful to be pushing infants or very young
children to be developing mathematical concepts before they
even have a system in place for representing them, and I think
there's many cases of that where we can use work on basic
cognitive development to make more informed decisions about
what children are ready to learn at different ages.
DR. ROWLEY: Now the other—one of your final concerns
or practical consequences of what you've described was that
it's very important for parents to be able to be with
their children for the first couple of years of life and you
weren't really in favor of child care at that point.
But the concern is that there are, unfortunately, a fair proportion
of families in the United States where the parents themselves
are not really in a position to give the children these kind
of early stimuli. And so I wonder if there isn't really
a place for child care and nursery schools at a very, very
early age for—at least some availability for these, some
portion of the population.
PROFESSOR SPELKE: Thank you for that clarifying
question. I did not mean to be arguing against child care.
I mean to be arguing that every infant needs to be growing
up with adults who are responsive to them, who have the time
to take to be interacting with them, to be attending to them,
that infant development doesn't happen in a vacuum. And
in many cases, it's the parents who would want to be playing
that role if it were economically possible for them. But
that is not to say that a biological parent is the only person
who can play that role.
DR. ROWLEY: And finally, the question for Professor
Kagan, when does a sense of right and wrong or moral judgment
really develop in children? It's partly a question that
Bill asked, but we didn't get into—when can you really
hold a child responsible and what kind of actions can you
hold them responsible for?
PROFESSOR KAGAN: Three stages in moral development
which I believe are universal. In the middle of the second
year, if you're late by, I don't know, 30 months,
but most kids by 2, understand the concepts right and wrong,
good and bad, in their language. They have a concept of prohibited
actions. That's not morality, but that's the beginning.
Between five and seven, and that's why the Church and
English Common Law and Freud and Piaget, I mean everyone knows
that a profound change in brain occurs between five and seven.
We don't know what it is. Now children have a more abstract
concept of good and bad and understand that going to school
is good and watching the cow from 9 until dinner is something
I should do. Now we hold them responsible. And the last
phase is at adolescence, again, I think a maturational change
occurs, of course, all supported by the environment.
Now one has, one looks for consistency and one looks for
consistency in your beliefs which 7-year-olds don't.
So the child of seven can hold these beliefs. My father is
a wonderful man. My father is much too harsh with my mother.
One of those has to go. That's the change in adolescence.
Now the adolescent seeks consistency among their moral premises,
and that's the last phase maturationally. Of course,
experience and culture—morality is like language. You're
given the capacity, and now you're culture teaches you
whether you're going to learn Swahili, French or Germany,
and your culture teaches you, with the exception, I think,
of unprovoked aggression. I have the intuition that that
is universal. But with that exception it teaches you what
is moral and what is immoral.
CHAIRMAN KASS: Ben Carson.
DR. CARSON: Thank you both for those illuminating
discussions. This morning we talked a little bit about the
neurophysiological and anatomical aspects of the developing
human brain and this afternoon more about some of the cognitive
social developmental issues. In both cases, there's an
implication that the nurturing environment plays a very significant
role, it provides very significant advantages. I was particularly
struck by the slide that indicated the numerical reasoning
at a 95 percent versus an 18 percent range in children in
nurturing environments versus those in lower socio-economic
The question is is there a point at which it is too late
to close the gap? If so, what is that, and secondly, how
do you explain late bloomers?
PROFESSOR SPELKE: Both really good questions. The
work of Case and Griffin is actually, I think, very encouraging
here because although the middle class children tend to develop
these concepts in the years from two and a half to four, their
intervention was aimed at 5-year-olds starting kindergarten
and it was not too late. The 5-year-olds in their program
did extremely well. That doesn't tell us is there a point
later on that is too late. There will be a cumulative problem
though if you wait too long which is that when kids start
elementary school, they're getting exposed to a whole
curriculum that presupposes that those concepts are in place.
So if the concepts aren't in place, even if the child
is biologically still capable of gaining them, they're
going to get further and further behind because what the teacher
is saying is just not going to be making any sense to them.
So I think just on those grounds one would want, these kinds
of building block concepts, one would want interventions that
put them in place at the time that formal education begins.
On the issues of late bloomers, I think there's lots
of reason to think that nature is flexible and there are many
paths to success and many different rates and patterns that
children will follow to get there, and I don't think it
follows from any of the work that Jerry or I talked about
today. In fact, certainly not what Jerry talked about, but
not the work that I talked about either, that there is one
royal road that one must progress down and one specific time
table to get there.
PROFESSOR KAGAN: There's a difference between
the ability to reason and where you are in the rank order.
And this is where you are in the rank order. A great moment
for me when I had a sabbatical leave 30 years ago was to sit
on the edge of Lake Atitlan in northwest Guatemala with illiterate
Mayan Indians who had no schooling and asking 12-year-olds
what would happen if the lake dried up and getting perfect
I agree with the thrust of Liz's talk. I mean, look
at the conditions under humans grow up. We have to have a
basic set of cognitive abilities, but in our societies, technological
societies, it doesn't make any difference how good your
reason, that's irrelevant. It's where you are in
the rank order because we can have just so many chiefs, and
that's the problem. We don't want to confuse that.
We don't want to confuse the class differences in rank
with the ability to reason. That would be very dangerous.
CHAIRMAN KASS: Mike Gazzaniga.
DR. GAZZANIGA: Are you sure we have the time?
CHAIRMAN KASS: No, go ahead. Let's take just
a few more minutes because we don't want to lose again
the benefit of having some questions.
DR. GAZZANIGA: First of all, I hope everybody appreciates
how Jerry and Liz asked these fundamental questions of biology
in the most low tech imaginable way with looking longer at
a stimulus. You didn't get into the peek-a-boo experiments,
but they're as captivating. And sometimes I think we
should throw our brain scanners out and just give the field
to these two and have them answer questions for us. But once
you get past the critical period issue here and where you
might have said constructionist ideas were needed to bring
the kids along to get them up to the sort of the level of
conceptual development that you might just call baseline normal
level, but then you've introduced this idea of the social
context, so you really don't know what it is that's
going on. Something is going on that's needed in a group
Now let's imagine that we've got our group of kids
up to the baseline level. And now you're trying to introduce
other concepts where people start to differentiate in their
understanding of the world. So you want to teach Newtonian
physics versus our natural naive physics. What then? What
are the tricks you have up your sleeve? Are there tricks
up your sleeve that can bring along the next level of conceptual
thinking, and is there a level, a definable level where you
start seeing separations that are very hard to overcome on
any social group?
PROFESSOR SPELKE: Yes. One of the things that I
was sorry to have to leave out of my already too long talk
was the evidence that when adults reason about number or physics
or other things, we bring together the same core systems that
we see children using when they assemble these concepts in
the first place. I think the general answer to your question
is that at any point in the educational system the way education
works is that it builds new concepts out of old concepts and
that the way to get a child, a high school student, to understand
Newtonian mechanics is not to ignore their intuitive notions
of mechanics which are profoundly not Newtonian, right, there's
no action at a distance and so forth, not to ignore them but
to work with them, to work with them and to build on them
and then by connecting them to their intuitive number concepts,
to see that there's a problem with them. And that there's
another way of thinking about how objects move, building on
what you know about number, that does a better job than your
intuitive notions of how objects move, building on this kind
of Medieval or Aristotelian notion of things in motion and
giving forces to each other.
So I think at every step of the educational system, good
teachers intuitively figure out what people's pre-existing
concepts are and work with them. And because in many cases,
especially early on in the educational system, it can be extremely
difficult to intuit what the concepts are of a child when
they're different from yours. The research can be helpful
in helping teachers to understand and that can be helpful
for designing curricula that takes those concepts and build
CHAIRMAN KASS: I'm next in line. I want to
make a comment and then a question.
The comment is partly inspired by Mike's question and
your answer. I understand that the core knowledge is somehow
the foundation and one builds upon that, but there really
might be certain kinds of real discontinuities and in the
area of number you'd have the primary case because the
modern concept of number bows the difference between a multitude
and a magnitude so that you have a number line to which—a Greek would find it unintelligible.
A number is a discrete multiple and to have a line in which
each point is somehow a number and that you blow the difference
between the discrete and the continuous is strange. And you've
got to somehow overturn your fundamental idea of multitude
in order to acquire the modern idea of number, but that raises
some kind of question about the difference between the things
which are somehow naturally ours and the things which are
somehow acquired as a result of new conceptual schemes, and
some people have a hell of a time learning algebra which is
based really upon our Cartesian coordinate system.
The question and you can comment on that if you'd like,
but the question really has to do with the division of labor
in this discussion between the cognitive capacities, and let
me give to Jerry Kagan not just questions of temperament,
but also emotion, motivation and interest. And I wonder whether
this perfectly respectable division of labor doesn't introduce—isn't based on a kind of distortion in which the question
is whether crucial to cognition are not just the native capacities
for discrimination and awareness, but interest, desire, motivation
One could talk about the game playing way of remedying,
a board game way of remedying the difficulties in numerosity
either as the acquisition of cognitive ability or as actually
caring about the matter because there's winning and losing
involved and somehow certain kinds of things now come to the
fore. So I'm wondering whether this sort of bifurcation
of cognition and motivation or cognition, temperament, emotion
and drive, whether that's an accurate representation of
how we should be thinking about how children learn.
Aristotle's remark "all children by nature desire
understanding," that's somehow put all human beings,
but begins really in childhood. And I just wondered whether
you would comment on how the two sides of the street meet
in your own understanding of what we're talking about
PROFESSOR KAGAN: Yes, I'll be brief, and then
Liz might want to add. I don't think temperament has
anything to do with it, but emotion does. As Liz intimated
earlier, all humans as a species enjoy mastery. That's
not a new idea. They enjoy understanding and enjoy using
their talents at the next level of challenge. But we can't
run a society that way. That is, each society has a set of
requirements, and we know the requirements for technological
societies. They require reading, mathematics, writing essays.
Those aren't the tasks that come naturally. So we are
forced, we must say to children, I'm sorry, these are
the tasks that you have to master. They are less natural
to the human species.
Now we need motivation, special motivation, acquired motivation.
This is what Liz was driving at, why we need parents. And
because no human being, child or adult, will invest energy
at a task they do not believe they would be successful in.
An animal won't do it, and a human won't do it. So
once you have a child in grades one, two, three and four who
comes to the decision every child able to do this except those
with a damaged brain, that there's no way I'm going
to be in the top third of my class. There's no way that
I'm—because there's no idea of absolute skill.
It's always relative, right? The adults and the peers
determine what is called mastery. And so by grades four or
five you lose motivation. That is the boat we're in and
so the task is not to say yes, we won't teach you any
reading, what do you wish to do?
The task is to devote more effort to the children who are
behind. And now my last point, not just in America or Europe,
in every country, every country where this study has been
done, the lower the educational level of the parents, the
poorer the academic record, period. Class is everything.
There are studies. NIH spent $50 million on the effect
of various forms of day care, home care, home alone. This
is a very famous study. And when you look at the data, the
best predictor of cognitive performance in the second grade
is the class, and after that everything else has trivial variance.
This is the issue that we have to deal with. It's profound.
It's not just—it's our problem in America, seriously,
but it's a problem around the world, and we have to get—we have to communicate, we have to do more for those who
are less well educated in our society to get them to understand
the importance of the task requirements we set for children
in our culture. They don't understand it. Many of them
are fatalistic. That is a very difficult job.
Now speaking for myself, I can imagine no more important
task this nation might undertake, no more important task.
PROFESSOR SPELKE: Let me just add that I think—I agree with what Jerry said, but I think that in addition
to children having an intrinsic desire for mastery, there's
an intrinsic desire for connecting to other people for becoming
able to do the things that other people are doing and that
this is driving much of the observational learning and of
coming to gain the skills that mark the people in their community
and in their culture and that, on that basis I completely
agree with you.
Certainly in the preschool years, the division between cognition
and motivation or emotion which is useful for figuring out
how to organize talks is quite artificial, and any program
aimed at children to enhance their cognitive development is
going to need to be building on these needs intrinsic to children
to be connecting.
If I can just say one thing quickly, though, about the point
you made about number, one of the reasons that number is so
fascinating is that both in the history of mathematics and
in the education that children and high school students and
college students go through in studying mathematics now, we
see a progression of revolutions in the concept of number.
I think the first revolution is the one I talked about between
age two and a half and age four, where children go from having
two quite different notions of number, neither of which has
the power of the natural number system to constructing the
natural number system. And although this is partly an expression
of faith and only partly based on data, I also think that
if we turn to the later constructions, understanding of number
lines, constructing notions of rational numbers, real numbers
and so forth, the general story that one advances to a new
level of understanding by taking one's pre-existing understanding,
in this case we'd have to look at understanding of points
and lines and spatial intuitions, which I didn't talk
about at all, but which young children have. One takes these,
one connects them in new ways and often with the help of teachers
who can point out what connections are useful, what properties
of points and lines are useful in thinking about number?
You then get to a new level of development.
CHAIRMAN KASS: Briefly, Robbie and then Peter and
then we'll break.
PROFESSOR GEORGE: Thank you. Anyone who has been
a parent or even an older sibling knows what a pleasure it
is to observe the reactions that children have to various
things. It must be wonderful to be able to do that in a professional
capacity as well and be paid for it. And, of course, it was
very interesting material that you presented to us from so
My impression is that—I want to ask you about something
that hasn't been raised yet, which is research ethics.
My impression is that while anyone who works with human subjects
is working under fairly strict ethics rules, that when it
comes to working with children there are an even richer set
of ethics rules, and I wondered if you could say a little
bit about them, and I'm particularly interested to know
whether the ethical norms that apply in working with children
are truly salient or whether they're mostly or merely
symbolic. In other words, do we sacrifice some knowledge?
Do we sacrifice some advances for the sake of respecting those
norms that are meant to protect children?
PROFESSOR SPELKE: Yes, we do, and I think it's
an excellent thing that we do, and one can't subject children
and one shouldn't subject children to any of the controlled
rearing studies that one can do on other animals. One would
be very wary of subjecting children to any kind of stressful
situation. I mean, our own rule of thumb for our experiments
is the parents are always present during the studies. They
should be completely happy at every second for what's
going on, with what's going on in the study. It should
be the kind of events that go on in children's lives or
that the parents would want to see going on in children's
lives outside the lab. So I think the standards are very
high, particularly for research like mine where there is no
immediate benefit to that child. We're asking basic questions
about the development of human knowledge. We hope that the
answers to those questions will be a benefit to society, in
general, but we have no illusion that the particular child
who comes in for our study is going to benefit. So they better
have a very good time and experience no negative consequences
from being in the research.
PROFESSOR GEORGE: Now in foregoing some knowledge
for the sake of respecting ethical norms, the norms themselves,
I take it, cannot be supplied by science or by scientific
reasoning? Is that right?
PROFESSOR SPELKE: I don't know about the history,
how the ethical norms developed actually.
PROFESSOR GEORGE: In other words, are they themselves
the fruit of scientific inquiry or are they—do they come
from another source?
PROFESSOR KAGAN: No, they come from the consensus
of a society that one does not cause distress.
PROFESSOR GEORGE: And that itself is not a scientific—is it a rational concern?
PROFESSOR KAGAN: Moral choice.
PROFESSOR GEORGE: It's a moral choice, but one
that comes to science from the outside. Okay. thank you.
CHAIRMAN KASS: Peter's last comment.
DR. LAWLER: On the basis of what you said, couldn't
you argue that they could come from what we know about the
distinctive nature of children? They could come from science,
PROFESSOR KAGAN: I'm sorry, I didn't hear
DR. LAWLER: The moral norms, the ethical norms,
why couldn't they come from—what you two do in such
a great way is restore the idea of a distinctively human nature.
We're distinctive in many respects, but we're also
natural beings. We have these potentials which are actualized,
right? So given all that we know about the nature of children
in some ways as distinct beings, social beings who take a
joy in learning and everything you've said, why couldn't
the norms come from what we know through science?
PROFESSOR KAGAN: I hope you meant that to provoke
me. I hope I speak for Liz. Science is a wonderful thing.
It's one of the great, great human missions, and those
of us in science feel privileged.
Do not look to science for moral norms. Science tells us
that male primates are promiscuous, therefore we should change
the norms of adultery? No? No. Science tells us that most
boys are much better at spatial problems like geometry than
girls. Therefore, we probably should have sex segregated
classes for teaching geometry. A referendum next November
would be defeated by Americans who are wise. Wittgenstein
understood this. The ancients understood this. Jim's
written about this in his book The Moral Sense.
Morality is very special. Very special. It has to do
with the sentiment of community, and that lies outside science.
We want to say to the scientists the following. Thank you.
Those are very interesting facts, very interesting, but in
this instance I don't choose to implement them, and that's
neither silly nor dumb.
CHAIRMAN KASS: We're going to stop. Thank you
both, very, very much for a wonderful afternoon and also for
the work that you're doing. This is really eye opening
and very important foundational work that will, I'm sure,
benefit all of us.
To the Council Members, we're behind as we so often
are. We had left the last hour, it will now be closer to
half an hour for stock taking and discussion of development
here. I see there are motions to adjourn. We can't partly
because this Ethics Council is required by law to be instructed
about ethics. We had an ethics lesson the very first time
we met. I think we escaped by somebody's inattention.
I think we require these annually, do we? And someone forgot
to give us lessons in ethics last year, but someone will be
arriving at 5:15, so you can't leave. We'll take
15 minutes. We'll talk amongst ourselves. Thank you
to our guests.
(Off the record.)