SEPTEMBER / OCTOBER 2017 | MIND.SCIENTIFICAMERICAN.

COM

Artificial
Intelligence
Becomes
More Human
The science of child learning
is transforming AI

Mental Health Smell Test • This Is Your Brain on Baby • Mind-Boosting Marijuana
FROM THE EDITOR
From What to Whom?
The phrase “artificial intelligence,” or AI, seems to be on everybody’s lips these
days—along with a lot of questions. After a spurt of development in the 1950s and
1960s, AI languished for a time. Now advances such as machine learning are driving
it into multiple fields of human endeavor, from transportation to medicine to finance.
But applying AI successfully for all of us means improving it. And one way to do that,
writes researcher Alison Gopnik in our cover story, “An AI That Knows the World Like
Children Do,” is to teach its networks to learn like a child. The challenge of helping
machines develop like youngsters will provide computer scientists with challenges
for years—and the rest of us with some societal conundrums. Should we worry?
“Natural stupidity can wreak far more havoc than artificial intelligence,” Gopnik
writes. Amen.

Elsewhere in the issue, we look at more mysteries of the mind—from why the brain
likes lying politicians to how thought is a community effort to new solutions for
motor neuron scourges such as amyotrophic lateral sclerosis (Lou Gehrig’s disease)

COVER BY BRYAN CHRISTIE; THIS PAGE: SPENCER HEYFRON

and Parkinson’s.

As ever, we’d love to hear what you think.

Mariette DiChristina
Editor in Chief

2
CONTENTS
News
21 An AI That Knows the World 4 Smell Test May Sniff Out Oncoming Parkinson's
Like Children Do and Alzheimer's
8 Marijuana May Boost, Rather Than Dull,
the Elderly Brain
12 Scientists Surprised to Find No Two Neurons
Are Genetically Alike
16 Why It's So Hard to Keep a Secret
19 Brain Activity during Sleep Can Predict When
Someone Is Dreaming

Features
21 An AI That Knows the World Like Children Do

SIMON PRADES
29 You Do Not Think Alone
34 New Drugs for Lou Gehrig's Disease Head
29 You Do Not Think Alone New Drugs for Lou Gehrig's Disease 34 for Clinical Trials
Head for Clinical Trials 43 Post-Truth: The Dark Side of the Brain
49 In Baby Mode

Opinion

JEREMY WILSON
56 Cracking the Popularity Code
60 Is the U.S. Education System Producing a
Society of “Smart Fools”?
Post-Truth: The Dark Side of the Brain 43
64 Can Stimulating a Nerve in the Ear Make You
a Whiz in Mandarin Class?
67 Why You Shouldn't Tell People about
JAMIE GRILL GET T Y IMAGES

YUJI SAKAI GET T Y IMAGES Your Dreams

3
 IMAGESBAZAAR GET T Y IMAGES

NEWS

Smell Test May Sniff Out Oncoming Parkinson´s and Alzheimer´s

How an unlikely clue could help solve the mystery of two devastating disorders
4
S
ight and hearing get all the glory, but
the often overlooked and underap-
preciated sense of smell—or prob-
lems with it—is a subject of rapidly growing
interest among scientists and clinicians
who battle Alzheimer’s and Parkinson’s
diseases. Impaired smell is one of the ear-
liest and most common symptoms of both,
and researchers hope a better understand-
ing will improve diagnoses and help un-
lock some of the secrets of these incur-
able conditions.
The latest offering from the burgeon-
ing field is a review published in June in
Lancet Neurology. It proposes neurotrans-
mitter dysfunction as a possible cause of
smell loss in a number of neurodegenera-
tive diseases, including Alzheimer’s and
Parkinson’s.
More than 90 percent of Parkinson’s pa-
tients report some level of olfactory dys-
function. And because problems with smell
progress in Alzheimer’s, nearly all of those
diagnosed with moderate to severe forms
of the illness have odor-identification is-
sues. “It’s important, not just because it’s
novel and interesting and simple but be-
cause the evidence is strong,” says Davan-
gere Devanand, a professor of psychiatry
and neurology at Columbia University, who
was not an author of the paper. His most
recent paper on the subject, a review, was
published in the American Journal of Geriat-
ric Psychiatry in December.
Studies have shown impaired smell to
be even stronger than memory problems as
a predictor of cognitive decline in currently
healthy adults. It is especially useful for
forecasting the progression from mild cog-
nitive impairment (MCI) to full-blown Alz-
heimer’s. According to the Alzheimer’s As-
sociation, approximately 15 to 20 percent
of people older than 65 have MCI. About
half of them go on to develop Alzheimer’s,
Devanand says—and the sooner they are
identified, the earlier doctors can begin in-
terventions, including treatment with the
few existing Alzheimer’s drugs.
Among Parkinson’s patients, loss of
smell occurs more often than the most
common type of tremor. It can precede the
classic motor signs of the disease by sever-
al years, and olfactory tests have been
shown to do a better job than motor func-
tion tests when it comes to distinguishing
Parkinson’s patients from healthy control
subjects. In fact, not having trouble with
smell is often enough to rule out the dis-
ease. “If a person scores very well on a
smell-identification test, then you can be
pretty sure he or she is not going to have
Parkinson’s, at least within the next four
years,” says neurologist and researcher G.
Webster Ross of the Veterans Affairs Pacific
Islands Health Care System.
As eager as they are to raise awareness
of the connection between olfactory dys-
function and neurodegenerative disease,
experts in the field are also quick to point
out that problems with smell are common
in the general population and do not al-
ways indicate a higher risk of developing
Parkinson’s or Alzheimer’s. The num-
ber-one cause of long-lasting or even per-
manent smell loss is the common head
cold, says Richard Doty, primary founder
and director of the Smell and Taste Center
at the University of Pennsylvania and au-
thor of the review. Doty, a professor of oto-
rhinolaryngology, invented the University
of Pennsylvania Smell Identification Test
(UPSIT), a 40-odor scratch-and-sniff test
that has become the gold standard for di-
agnosing olfactory dysfunction. Additional
causes include other types of viral infec-
tions, a history of smoking, head injuries
and getting older. Indeed, according to
5
Doty, about half the U.S. population be-
tween the ages of 65 and 80 have some de-
monstrable loss of smell. If subjects are
older than 80, that number goes up to about
75 percent.
Once colds and other common causes are
ruled out, family doctors and general practi-
tioners might order a smell test and start to
consider the rarer and far more serious pos-
sibilities. “But it has to be done in the con-
text of a complete medical history and a
proper evaluation,” Devanand says. “You
can’t just randomly give it.” For neurolo-
gists such as Devanand and Ross, the results
of a smell test can initiate a workup or help
confirm a diagnosis. “It’s really a subgroup
of the population where this is useful for de-
tecting or predicting future cognitive de-
cline or Parkinson’s,” Doty says. Impaired
smell is just one of several early, nonmotor
signs that Ross looks for when he faces a po-
tential Parkinson’s case. Others include
chronic constipation (which he says may oc-
cur as much 12 years prior to diagnosis) as
well as slow reaction time on a computer
test and excessive daytime sleepiness.
Separate from its link to neurodegener-
ative disease, olfactory dysfunction has
been associated with an increased risk of
death, independent of other causes, includ-
ing dementia, in four studies of older adults.
Whereas the reasons for the effect remain
unclear, a large study led by Devanand sug-
gests “safety and nutritional issues associ-
ated with the loss of smell”—the inability
to smell smoke, for example, and lack of in-
terest in food from loss of taste—could be
contributors. In a 2014 study at the Univer-
sity of Chicago, researchers wrote, “We be-
lieve olfaction is the canary in the coal mine
of human health.”
The simplicity and low cost of a smell
test—$26.95 retail, before additional doc-
tor or hospital charges—make it especially
appealing in neurology, a field dominated
by positron-emission tomography scans,
dopamine transporter single-photon emis-
sion computed tomography imaging and
other expensive technologies. Because it is
so economical, Devanand says researchers
should incorporate olfaction testing into
studies that look at other Alzheimer’s risk
factors in large populations. He has applied
for a grant to do just that in an upcoming
Columbia study involving 3,000 subjects.
He would also like to see more studies that
directly compare olfaction testing with
costly and invasive procedures such as
brain scans and spinal taps.
In addition to its importance as an early
and reliable warning of Alzheimer’s and
Parkinson’s, impaired smell offers a win-
dow into the underlying mechanisms of the
two diseases. “If somebody has very mild
Alzheimer’s and dies,” Devanand says, “on
autopsy, we will see the pathology of Alz-
heimer’s—the tangles and plaques—in the
olfactory bulb.” The presence of the dead,
snarled nerve cells and sticky protein frag-
ments characteristic of Alzheimer’s in the
main smell-processing structures of these
deceased patients could suggest the dis-
ease gets its start in the olfactory system.
There is some evidence for this, Devanand
says, “but it’s not overwhelming.” Similar-
ly, Doty says there is an argument to be
made for calling the olfactory system the
origin of Parkinson’s but adds this is “prob-
ably an oversimplification.”
Such caution and uncertainty are hardly
surprising at this point, given the complex-
ity of the anatomy involved. The olfactory
system comprises six million to 10 million
receptor cells (of which there are nearly 400
6
different types) and links to multiple brain
structures and neural pathways, including
those involved in memory, emotion and
movement. Doty draws on more than 30
years of research to describe the role faulty
brain chemistry, particularly that involv-

Follow us on Twitter
ing the neurotransmitter acetylcholine,
might play in the smell impairment that
accompanies neurodegenerative disease.
For his part, Ross wonders if perhaps the @sciam
nasal mucosa—the layer of mucus lining twitter.com/sciam
the inside of the nose—“may be a conduit
through which some environmental toxin
can gain access to the brain,” potentially
triggering Parkinson’s and disrupting
smell along the way.
There may not be answers yet, but more
researchers are asking questions about
the relation between smell and neurode-
generative diseases. “In the past,” Deva-
nand says, “most neurologists thought,
‘Maybe there’s something there statisti-
cally in a paper, but it’s a bit flaky.’ ” Those
days are gone, he says, thanks to consis-
tent results from researchers in the U.S.
and Europe—findings that confirm the im-
portance of an unlikely clue in the fight
against devastating disorders. 
 —David Noonan


 CHRISTINA HEMPFLING GET T Y IMAGES

NEWS

Marijuana May Boost, Rather Than Dull, the Elderly Brain

Senior mice treated with THC improved on learning and memory tests
8
P ̏ We know the endogenous
icture the stereotypical pot smoker:
young, dazed and confused. Mari-

cannabinoid system is very

juana has long been known for its
psychoactive effects, which can include
cognitive impairment. But new research

dynamic; it goes through

published in June in Nature Medicine sug-
gests the drug might affect older users very
differently than young ones—at least in

changes over the life span.˝

mice. Instead of impairing learning and
memory, as it does in young people, the
drug appears to reverse age-related de-
clines in the cognitive performance of el-
derly mice.
Researchers led by Andreas Zimmer of
the University of Bonn in Germany gave
low doses of delta-9 tetrahydrocannabinol,
or THC, marijuana’s main active ingredi-
ent, to young, mature and aged mice. As ex-
pected, young mice treated with THC per-
formed slightly worse on behavioral tests
of memory and learning. For example, after
receiving THC, young mice took longer to
learn where a safe platform was hidden in a
water maze, and they had a harder time rec-
ognizing another mouse to which they had
previously been exposed. Without the drug,
mature and aged mice performed worse on
the tests than young ones did. But after the
elderly animals were given THC, their per-
 —Ryan McLaughlin, Washington State University
formances improved to the point that they
resembled those of young, untreated mice.
“The effects were very robust, very pro-
found,” Zimmer says.
Other experts praised the study but cau-
tioned against extrapolating the findings
to humans. “This well-designed set of ex-
periments shows that chronic THC pre-
treatment appears to restore a significant
level of diminished cognitive performance
in older mice, while corroborating the op-
posite effect among young mice,” wrote Su-
san Weiss, director of the Division of Extra-
mural Research at the National Institute
on Drug Abuse, who was not involved in
the study, in an e-mail. Nevertheless, she
added, “while it would be tempting to pre-
sume the relevance of these findings [ex-
tends] to aging humans ... further research
will be critically needed.”
When the researchers examined the
brains of the treated elderly mice for an ex-
planation, they noticed that neurons in the
hippocampus—a brain area critical for
learning and memory—had sprouted more
synaptic spines, the points of contact for
communication between neurons. Even
more striking, the gene-expression pattern

9
in the hippocampi of THC-treated aged
mice was radically different from that of
untreated elderly mice. “That is something
we absolutely did not expect: the old ani-
mals [that received] THC looked most sim-
ilar to the young untreated control mice,”
Zimmer says.
The findings raise the intriguing possi-
bility that THC and other “cannabinoids”
might act as antiaging molecules in the
brain. Cannabinoids include dozens of bio-
logically active compounds found in the
Cannabis sativa plant. THC, the most highly
studied type, is largely responsible for mar-
ijuana’s psychoactive effects. The plant
compounds mimic our brain’s own marijua-

OKSANA SMITH GET T Y IMAGES

nalike molecules, called endogenous can-
nabinoids, which activate specific receptors
in the brain capable of modulating neural
activity. “We know the endogenous canna-
binoid system is very dynamic; it goes
through changes over the life span,” says
Ryan McLaughlin, a researcher who studies
cannabis and stress at Washington State
University and was not involved in the cur-
rent work. Research has shown that the can-
nabinoid system develops gradually during
childhood, “and then it blows up in adoles-
cence—you see increased activity of its en-
zymes and receptors,” McLaughlin says.
“Then as we age, it’s on a steady decline.”
That decline in the endogenous canna-
binoid system with age fits with previous
work by Zimmer and others showing that
cannabinoid-associated molecules become
more scant in the brains of aged animals.
“The idea is that as animals grow old, simi-
lar to in humans, the activity of the endog-
enous cannabinoid system goes down—and
that coincides with signs of aging in the
brain,” Zimmer says. “So we thought, ‘What
if we stimulate the system by supplying
[externally produced] cannabinoids?’ ”
That idea does not seem so outlandish,
considering the role of cannabinoids in

10
maintaining the body’s natural balance,
says Mark Ware, a clinical researcher at Mc-
Gill University, who was not part of the
study. “To anyone who studies the endo-
cannabinoid system, the findings are not
necessarily surprising, because the system
has homeostatic properties everywhere we
look,” meaning its effects may vary de-
pending on the situation. For example, a
little marijuana may alleviate anxiety, but
too much can bring on paranoid delusions.
Likewise, cannabis can spark an appetite in
cancer patients but in other people may
produce nausea. Thus, the detrimental ef-
fects seen in young brains, in which canna-
binoids are already plentiful, may turn out
to be beneficial in older brains that have a
dearth of them.
These chemicals also work to maintain
order at the cellular level, McLaughlin says.
“We know the endogenous cannabinoid
system’s primary function is to try to pre-
serve homeostasis within a given brain cir-
cuit. It works like an internal regulator;
when there’s too much [neuronal] activity,
cannabinoids suppress activity to prevent
neurotoxicity.” Restoring that protection
might help safeguard the brain against cel-
lular stress that contributes to aging. “A
critical takeaway of this study is that they
used low doses,” Ware says, considering
that different doses could have entirely dif-
ferent effects. It would be difficult, if not
impossible, to translate the dose they used
in mice to a human equivalent, “but it’s
clear we’re not talking about vast amounts.
We don’t know what would happen with
higher doses.”
Scientists do not know exactly how mar-
ijuana affects older adults, in part because
they have been focused squarely on young-
er people, who are thought to be at greatest
risk. “Because of the public health concern,
research has had a very strong focus on mar-
ijuana’s effects in adolescence,” Ware says.
But although young people make up the
largest group of cannabis users, their rate of
use has remained relatively stable over the
past decade even as the drug has become
increasingly available. Meanwhile use
among seniors has skyrocketed as the drug’s
stigma has faded. A March study showed
that in people aged 50 to 64, marijuana use
increased nearly 60 percent between 2006
and 2013. And among adults older than 65,
the drug’s use jumped by 250 percent.
The researchers do not suggest seniors
should rush out and start using marijuana.
“I don’t want to encourage anyone to use
cannabis in any form based on this study,”
Zimmer says.
Older adults looking to medical canna-
bis to relieve chronic pain and other ail-
ments are concerned about its side effects,
Ware says: “They want to know, Does this
cause damage to my brain? Will it impair
my memory? If these data hold up in hu-
mans, it may suggest that [THC] isn’t likely
to have a negative impact if you’re using
the right dose. Now the challenge is thrown
down to clinical researchers to study that
in people.”
Zimmer and his colleagues plan to do
just that. They have secured funding from
the German government, and after clearing
regulatory hurdles, they will begin testing
the effects of THC in elderly adults with
mild cognitive impairments.
 —Stephani Sutherland
 11
 SOLVOD GET T Y IMAGES

NEWS

Scientists Surprised to Find No Two Neurons

Are Genetically Alike
The genetic makeup of any given brain cell differs from all others. That realization may
provide clues to a range of psychiatric diseases
12
T
he past few decades have seen in-
tensive efforts to find the genetic
roots of neurological disorders,
from schizophrenia to autism. But the
genes singled out so far have provided only
sketchy clues. Even the most important ge-
netic risk factors identified for autism, for
example, may account for only a few per-
cent of all cases.
Much frustration stems from the reali-
zation that the key mutations elevating
disease risk tend to be rare because they
are less likely to be passed on to offspring.
More common mutations confer only small
risks (although those risks become more
significant when calculated across an en-
tire population). There are several other
places to look for the missing burden of
risk, and one surprising possible source has
recently emerged—an idea that overturns a
fundamental tenet of biology and has many
researchers excited about a completely new
avenue of inquiry.
Accepted dogma holds that—although
every cell in the body contains its own
DNA—the genetic instructions in each cell
nucleus are identical. But new research has
now proved this assumption wrong. There
are actually several sources of spontaneous
mutation in somatic (nonsex) cells, result-
ing in every individual containing a multi-
tude of genomes—a situation researchers
term somatic mosaicism. “The idea is
something that 10 years ago would have
been science fiction,” says biochemist
James Eberwine of the University of Penn-
sylvania. “We were taught that every cell
has the same DNA, but that’s not true.”
There are reasons to think somatic mosa-
icism may be particularly important in the
brain, not least because neural genes are
very active.
A paper published on April 28 in Science
by a group founded two years ago—the
Brain Somatic Mosaicism Network
(BSMN)—outlines a research agenda for
using new technologies to explore the ge-
netic diversity found in each cell and to in-
vestigate what links, if any, tie such muta-
tions to a variety of neurological condi-
tions. “The field was abuzz with interest in
exploring mosaicism, but there was no
money,” says Thomas Lehner, director of
the Office of Genomics Research Coordina-
tion at the National Institute of Mental
Health, which is now devoting $30 million
in funding to the BSMN over the first three
years, two of which have elapsed.
The consortium consists of 18 research
teams at 15 U.S. institutions with access to
repositories of postmortem brain tissue
taken from healthy people and others
with schizophrenia, autism, bipolar disor-
der, Tourette’s syndrome or epilepsy. Each
team is tackling different samples. “There’s
a lot of new technology application and
development involved, and a ton of data
that will become a resource,” Lehner says.
“We also wanted to understand if there’s
an association with new technology, so we
encouraged researchers to include brain
banks of individuals with various neuro-
logical conditions.”
Studies that preceded the consortium
have confirmed mosaicism is commonplace.
One report estimated there may be hun-
dreds of changes in single letters of genetic
code (single-nucleotide variants, or SNVs)
in each neuron in the mouse brain. Another
found more than 1,000 in human neurons.
These findings suggest somatic mosaicism
is the rule, not the exception, with every
neuron potentially having a different ge-
nome than those to which it is connected. A
primary cause of somatic mutations has to
do with errors during the DNA replication
that occurs when cells divide—neural pro-
13
genitor cells undergo tens of billions of cell
divisions during brain development, prolif-
erating rapidly to produce the estimated 80
billion neurons in a mature brain. The im-
age of each cell carrying a carbon copy of
the genetic material of all other cells is start-
ing to fade—and for good reason. Genetic
sequencing does not normally capture the
somatic mutations in each cell. “You get a
sort of average of the person’s genome, but
that doesn’t take into account any brain-spe-
cific mutations that might be in that per-
son,” says lead study author Michael McCon­
nell of the University of Virginia.
A 2012 study found somatic mutations
in the brains of children with hemimegal-
encephaly, a developmental disorder in
which one hemisphere is enlarged, causing
epilepsy and intellectual disability. The
mutations were found in brain tissue but
not always in blood or in cells from unaf-
fected brain areas and in a small fraction
(around 8 to 35 percent) of cells from af-
fected areas. Such studies, showing somat-
ic mutations can cause specific populations
of cells to proliferate and lead to cortical
malformations, have researchers wonder-
ing whether somatic mutations may also
play a role in more complex conditions.
Mature neurons stop dividing and are
among the longest-living cells in the body,
so mutations will stick around in the brain.
“In the skin or gut, cells turn over in a
month or week, so somatic mutations
aren’t likely to hang around unless they
form cancer,” McCon­nell says. “These mu-
tations are going to be in your brain forev-
er.” This could alter neural circuits, thereby
contributing to the risk of developing neu-
ropsychiatric disorders. “In psychiatric dis-
ease, we don’t know that much yet, and
that’s largely the goal [to find an answer],”
McCon­nell says. “It’s a good hypothesis,
but it’s going to require this big, multiteam
effort to really address it.” To investigate,
the consortium will sequence brain DNA
from control and patient samples. “Before
you can get to your destination, you have
to have a map, and this is going to help
build that map of somatic mutations that
have potential for influencing neural func-
tioning and disease,” says Eberwine, who
was not involved in the new research. “So
this consortium is critically important for
neuroscience.”
One question to be explored is whether
genes associated with a brain disorder may
harbor somatic mutations. The fact that
specific genes explain just a small propor-
tion of cases may be because researchers
have been looking only in the germ line
(sex cells), McCon­nell says. “Maybe the per-
son doesn’t have the mutation in their germ
line, but some percentage of their neurons
have it.” Somatic mosaicism may also con-
tribute to neural diversity in general. “It
might explain why everybody’s different—
it’s not all about the environment or ge-
nome. There’s something else,” says neu-
roscientist Alysson Muotri of the Universi-
ty of California, San Diego, who is not part
of the consortium. “As we understand more
about somatic mosaicism, I think the con-
tribution to individuality as well as the
spectrum [of symptoms] you find in, for ex-
ample, autism, will become clear.”
Somatic mutations can occur in multi-
ple circumstances. They may emerge during
DNA replication or from DNA damage
(caused by free radicals or environmental
stresses), combined with imperfect repair
machinery. In addition to SNVs, mutations
known as indels, involving insertions and
deletions of small DNA sequences (typical-
ly tens of nucleotides), also occur frequent-
ly. Larger, rarer mutations include struc-
tural changes in chromosomes, in either
14
the form of gains or losses of whole chro-
mosomes or copy number variants (CNVs),
in which the number of repetitions of long
chunks of DNA (covering multiple genes) is
altered. Within genomes, there are also mo-
bile genetic elements that act almost like
parasites, jumping around or making copies
of themselves and inserting themselves
elsewhere in the genome, seemingly to en-
sure their survival. These strange entities
are an active field of research in their own
right: they are important here because they
can cause somatic mutations, including a
type known as mobile genetic element in-
sertions, or MEIs. They are switched on in
the same way as genes involved in produc-
ing new neurons, making them especially
active in the brain during development.
The paper outlines three methods for
studying these mutations. The first in-
volves using technologies to sequence a
whole genome from bulk brain tissue. This
technique can detect many variants, but
the rarest types are diluted by the mass of
cells in bulk tissue. “Large CNVs and mo-
bile elements are much more difficult to
detect in bulk tissue than SNVs,” McCon-
nell says. Also, this method cannot reveal
how mutations vary between cell types.
This can be partly solved using a technique “This is supposed to initiate an important
known as sorted pools, which sorts out neu- area of research,” Lehner says. “We hope it
rons from other unwanted cell types. The will give us a landscape of mosaicism in the
most important recent advance that will brain and insights into the contribution of
aid the consortium, however, is the advent mosaicism to mental disorders, but I don’t
of technologies that allow the genomes of expect to have all the answers.” Such in-
individual cells to be sequenced. “By going sights may ultimately lead to the discovery
into single cells, we can compare [what we of new genetic targets for treating a range
find] to the neighboring cell and say, ‘Aha, of hard-to-treat disorders.
they’re different!’ That’s the advance that “This is exploratory research; we’re learn-
allows us to really move forward,” Muotri ing about the phenomenon,” Muotri says.
says. “I’m very excited—this is the begin- How important it will be is not clear at this
ning of something completely new in biol- stage, but “by figuring out how it works, we
ogy and neuroscience.” may reveal new therapeutic opportunities.”
The project is funded until 2020 and will  —Simon Makin
make all data publicly available—and for
some results, that should be in 12 to 24
months. “Around 10,000 sequencing data
sets will be generated, and we’ll be making
that available in a database for the scientif-
ic community to dig in more deeply,” Mc-
Connell says. There are also plans to col-
laborate with other NIMH initiatives, in-
cluding BrainSpan, which maps gene
expression during brain development, and
psychENCODE, which is mapping the brain
epigenome (environmentally driven modi-
fications of DNA that influence gene activ-
ity without changing the genetic code).
 15
 IRENETINTA GET T Y IMAGES

NEWS

Why It´s So Hard to Keep a Secret

The real reason has more to do with mental meandering than actively hiding information
16
I
t is no secret that we all have secrets.
Maintaining them can be draining, but The Top 10 Categories of Things We Keep Secret
not for the reason most researchers

CREDIT: AMANDA MONTAÑEZ; SOURCE: “THE EXPERIENCE OF SECRECY,” BY MICHAEL L. SLEPIAN ET AL., IN JOURNAL OF PERSONALIT Y AND SOCIAL PSYCHOLOGY. PUBLISHED ONLINE MAY 8, 2017
have long assumed. A new study redefines
Over three studies (a subset of 10), researchers surveyed 600 participants about commonly
“secrecy” itself and offers a novel explana-
secret behaviors they had experienced and to what degree they had kept each one hidden.
tion for its known link to depression, anxi-
The graph shows data for the 10 behaviors most frequently kept totally secret.
ety and poor overall health. The research-
ers suggest that secrecy is primarily the in-
tention to conceal information, regardless Number of Study Participants
of any active concealment around others. 0 100 200 300 400 500 600
And that hurts us by making us feel inau-
thentic, even when we are alone. Thoughts of cheating on partner Secret from
everyone
Michael Slepian, a psychologist at Co-
Sexual behavior
lumbia Business School, and his colleagues Secret from
recently reported their findings in a paper Having lied to someone some people
published online in the Journal of Person- Romantic desires about someone Secret once
ality and Social Psychology. In six studies, while single not anymore
Number of Study Participants
they queried a total of 1,200 Americans via Violation of someone’s trust Never kept
0 100 200 300 400 500 600
the Internet—as well as 312 people pic- secret
Theft
nicking in New York City’s Central Park— Thoughts of cheating on partner Secret from
regarding 38 categories of behavior or Nonsexual infidelity (such as flirting) everyone
Sexual behavior
identity that are often kept secret. For five Secret from
Ambition or goal some people
of these studies, respondents said they Having lied to someone
were currently hiding information in about Romantic desires about someone Family details Secret once but
13 of these categories on average (includ- while single not anymore
Financial details
ing about five for which they had a secret Violation of someone’s trust Never kept
they were keeping from everyone). The secret
Theft
most common secrets were extrarelation-
Nonsexual infidelity (such as flirting)
17
al thoughts (thinking about having rela-
tions with another person while already
in a relationship), romantic desire (while
being single) and sexual behavior

Looking for more than a booze cruise? Us too.

(consumption of pornography, fanta-
sies, and so on). The graphic presents a
full breakdown of the most common
secrets; additional data are available at Go beyond the ordinary with SA Travel
www.keepingsecrets.org.
People said that when they were not
Onboard Seminars. World-Class Speakers. Insider Tours.
interacting with anyone, they thought
about their secrets about twice as often as
they actively concealed them in conversa-
tion. The more often their mind wandered
to a secret, the more they reported that it
damaged their well-being and the less
healthy they said they were. Surprisingly,
active concealment did not affect well-be-
ing at all—in contrast to previous assump-
tions. Four additional studies, all involv-
ing couples and conducted online, pro-
duced similar findings.

SA Travel: Mindful. Not Mindless.

If you must keep a secret, Slepian
suggests avoiding dwelling on it by practic-
ing mindfulness or by discussing the for-
GET INFO
bidden topic in anonymous online forums.
 —Matthew Hutson


 HERO IMAGES GET T Y IMAGES

NEWS

Brain Activity during Sleep Can Predict

When Someone Is Dreaming
Studying the dreaming brain offers a window on consciousness
19
T
o sleep, perchance to dream”—
Shakespeare may not exactly have
been talking about our nocturnal
journeys to another world, but that does not
make the phenomenon of dreams any less
mysterious or meaningful. Recent research
is expanding our understanding—and yield-
ing insights into consciousness itself.
Sleep provides science with a way to
study consciousness in all its various forms,
from vivid dreams to no awareness at all,
says neuroscientist Benjamin Baird. When
subjects are snoozing, researchers can iso-
late conscious experiences from the con-
founding influence of the senses.
In a recent attempt to peek inside the
dreaming brain, Baird and leading con-
sciousness expert Giulio Tononi, both at the
University of Wisconsin–Madison, and their
colleagues used scalp electrodes to record
sleepers’ brain waves via high-density elec-
troencephalography. They woke people at
frequent intervals to ask if they had been
dreaming and, if so, what their dreams were
about. One of their experiments garnered a
total of about 200 such waking accounts
from 32 subjects, and a second added some
800 more from a smaller group specially
trained in dream reporting.
The team identified a “hot zone” in a Finding that dream-related activity was fo-
posterior cortical region, near the back of cused toward the back of the brain was sur-
the head, where low-frequency brain waves prising, he says, because consciousness is
(linked to unconsciousness) decreased and widely thought to arise in frontoparietal
high-frequency activity rose when people regions. One limitation of the study is the
said they had been dreaming—regardless delay between a subject waking up and re-
of whether or not it was during rapid eye counting a dream. Ultimately, Koch says,
movement (REM) sleep. (Despite common “we want to get closer and closer to the ex-
belief, dreaming can occur during both REM perience itself.”
and non-REM sleep.) The findings were de-  —Tanya Lewis
scribed in Nature Neuroscience.
In a third experiment with seven sub-
jects, the scientists predicted with 87 per-
cent accuracy whether participants were
dreaming. Furthermore, brain-wave activi-
ty in certain brain regions was linked to
specific dream content—including loca-
tions, faces and speech. These same areas
are activated during waking experience.
“We didn’t actually try to predict [dream]
content in this study,” Baird notes, but he
says that would be an exciting “potential
direction.”
This approach represents “a really cool
and innovative paradigm,” says neuro-
scientist Christof Koch of the Allen Insti-
tute for Brain Science in Seattle, who was
not involved in the work and who serves on
Scientific American’s board of advisers.
 20
 AI
SIMON PRADES

An That Knows the World Like Children Do

Artificial intelligence has staged a revival by starting to incorporate
what we know about how children learn
By Alison Gopnik
21
I
f you spend much time with chil-
dren, you’re bound to wonder how
young human beings can possibly
learn so much so quickly. Philoso-
phers, going all the way back to Pla-
to, have wondered, too, but they’ve never
found a satisfying answer. My five-year-old
grandson, Augie, has learned about plants,
animals and clocks, not to mention dino-
saurs and spaceships. He also can figure out
what other people want and how they think
and feel. He can use that knowledge to clas-
sify what he sees and hears and make new
predictions. He recently proclaimed, for ex-
ample, that the newly discovered species of
titanosaur on display at the American Mu-
seum of Natural History in New York City is
a plant eater, so that means it really isn’t
that scary.
Yet all that reaches Augie from his envi-
ronment is a stream of photons hitting his
retina and disturbances of air contacting
his eardrums. The neural computer that
sits behind his blue eyes manages some-
Alison Gopnik is a professor of psychology and an affiliate professor
of philosophy at the University of California, Berkeley. Her research
focuses on how young children learn about the world around them.
how to start with that limited information IN BRIEF
from his senses and to end up making pre-
• H
 ow do young children know what they know?
dictions about plant-eating titanosaurs. That question has long preoccupied philosophers
One lingering question is whether elec- and psychologists—and now computer scien-
tronic computers can do the same. tists.
During the past 15 years or so computer
scientists and psychologists have been try- • S pecialists in artificial intelligence are studying
ing to find an answer. Children acquire a the mental reasoning powers of preschoolers to
great deal of knowledge with little input develop ways to teach machines about the
world.
from teachers or parents. Despite enor-
mous strides in machine intelligence, even
• T wo rival machine-learning strategies—both
the most powerful computers still cannot halting attempts to mimic what children do
learn as well as a five-year-old does. naturally—have begun to transform AI as a
Figuring out how the child brain actual- discipline.
ly functions—and then creating a digital
version that will work as effectively—will
challenge computer scientists for decades
to come. But in the meantime, they are be-
ginning to develop artificial intelligence
that incorporates some of what we know
about how humans learn.
This Way Up
After the first burst of enthusiasm in the
1950s and 1960s, the quest for AI lan-
guished for decades. In the past few years,
though, there have been striking advances,
especially in the field of machine learning,
and AI has become one of the hottest de-
22
velopments in technology. Many utopian
or apocalyptic predictions have emerged
about what those advances mean. They
have, quite literally, been taken to presage
either immortality or the end of the world,
and a lot has been written about both these
possibilities.
I suspect that developments in AI lead
to such strong feelings because of our
deep-seated fear of the almost human. The
idea that creatures might bridge the gap
between the human and the artificial has
always been deeply disturbing, from the
medieval golem to Frankenstein’s monster
to Ava, the sexy robot fatale in the 2015
movie Ex Machina.
But do computers really learn as well as
humans? How much of the heated rhetoric
points to revolutionary change, and how
much is just hype? The details of how com-
puters learn to recognize, say, a cat, a spo-
ken word or a Japanese character can be
hard to follow. But on closer inspection, the
basic ideas behind machine learning are
not as baffling as they first seem.
One approach tries to solve the problem
by starting with the stream of photons and
air vibrations that Augie, and all of us, re-
ceives—and that reaches the computer as
pixels of a digital image and sound sam-
ples of an audio recording. It then tries to
extract a series of patterns in the digital
data that can detect and identify whole ob-
jects in the surrounding world. This so-
called bottom-up approach has roots in
the ideas of philosophers such as David
Hume and John Stuart Mill and psycholo-
gists such as Ivan Pavlov and B. F. Skinner,
among others.
In the 1980s scientists figured out a com-
pelling and ingenious way to apply bot-
tom-up methods to let computers hunt for
meaningful patterns in data. “Connection-
ist,” or “neural network,” systems take in-
spiration from the way that neurons con-
vert light patterns at your retina into rep-
resentations of the world around you. A
neural network does something similar. It
uses interconnected processing elements,
akin to biological cells, to transform pixels
at one layer of the network into increasing-
ly abstract representations—a nose or an
entire face—as data are crunched at pro-
gressively higher layers.
Neural-network ideas have gone through
a recent revival because of new techniques
called deep learning—technology now be-
ing commercialized by Google, Facebook
and other tech giants. The ever increasing
power of computers—the exponential in-
crease in computing capability that is cap-
tured by what is known as Moore’s law—
also has a part in the new success of these
systems. So does the development of enor-
mously large data sets. With better pro-
cessing capabilities and more data to
crunch, connectionist systems can learn far
more effectively than we might have once
thought.
Over the years the AI community has
seesawed between favoring these kinds of
bottom-up solutions to machine learning
and alternative top-down approaches. Top-
down approaches leverage what a system
already knows to help it learn something
new. Plato, as well as so-called rationalist
philosophers such as René Descartes, be-
lieved in a top-down approach to learn-
ing—and it played a big role in early AI. In
the 2000s such methods also experienced
their own rebirth in the form of probabilis-
tic, or Bayesian, modeling.
Like scientists, top-down systems start
out by formulating abstract and wide-rang-
ing hypotheses about the world. The sys-
tems then make predictions about what the
data should look like if those hypotheses
23
are correct. Also like scientists, the systems
then revise their hypotheses, depending on
the outcome of those predictions.
Nigeria, Viagra and Spam
Bottom-up methods are perhaps the most
readily understood, so let’s consider them
first. Imagine that you are trying to get your
computer to separate important messages
from the spam that arrives in your in-box.
You might notice that spam tends to have
certain distinguishing characteristics: a
long list of recipient addressees, an origi-
nating address in Nigeria or Bulgaria, refer-
ences to $1-million prizes or perhaps men-
tion of Viagra. But perfectly useful messag-
es might look the same. You don’t want to
miss the announcement that you have
earned a promotion or an academic award.
If you compare enough examples of
spam against other types of e-mails, you
might notice that only the spam tends to
have qualities that combine in certain tell-
tale ways—Nigeria, for instance, plus a
promise of a $1-million prize together spell
trouble. In fact, there might be some quite
subtle higher-level patterns that discrimi-
nate between the spam messages and the
useful ones—misspellings and IP addresses
that are not at all obvious, for example. If
you could detect them, you could accurate-
ly filter out the spam—without fear of miss-
ing a notice that your Viagra has shipped.
Bottom-up machine learning can ferret
out the relevant clues to solve this kind of
task. To do this, a neural network must go
through its own learning process. It evalu-
ates millions of examples from huge data-
bases, each labeled as spam or as an au-
thentic e-mail. The computer then extracts
a set of identifying features that separate
spam from everything else.
In a similar way, the network might
inspect Internet images labeled “cat,”
“house,” “stegosaurus,” and so on. By ex-
tracting the common features in each set of
images—the pattern that distinguishes all
the cats from all the dogs—it can identify
new images of a cat, even if it has never
seen those particular images before.
One bottom-up method, called unsu-
pervised learning, is still in its relative in-
fancy, but it can detect patterns in data that
have no labels at all. It simply looks for
clusters of features that identify an object—
noses and eyes, for example, always go to-
gether to form a face and differ from the
trees and mountains in the background.
Identifying an object in these advanced
deep-learning networks takes place through
a division of labor in which recognition
tasks are apportioned among different lay-
ers of the network.
An article in 2015 in Nature demonstrat-
ed just how far bottom-up methods have
come. Researchers at DeepMind, a compa-
ny owned by Google, used a combination of
two different bottom-up techniques—deep
learning and reinforcement learning—in a
way that enabled a computer to master At-
ari 2600 video games. The computer began
knowing nothing about how the games
worked. At first, it made random guesses
about the best moves while receiving con-
stant feedback about its performance. Deep
learning helped the system identify the
features on the screen, and reinforcement
learning rewarded it for a high score. The
computer achieved a high proficiency level
with several games; in some cases, it per-
formed better than expert human players.
That said, it also completely bombed on
other games that are just as easy for hu-
mans to master.
The ability to apply AI to learn from
large data sets—millions of Instagram im-
ages, e-mail messages or voice recordings—
24
allows solutions to problems that once
seemed daunting, such as image and speech Two Paths to AI´s Resurgence
recognition. Even so, it is worth remember-
ing that my grandson has no trouble at all Problems the average five-year-old solves readily can stump even the most powerful computers. AI has made a
recognizing an animal or responding to a spirited comeback in recent years by teaching computers to learn about the world somewhat like a child does.
spoken query even with much more limited The machine recognizes the letter “A” either from raw sensory information—a bottom-up approach—or by mak-
data and training. Problems that are easy ing a guess from preexisting knowledge—a top-down approach.
for a human five-year-old are still extreme-
ly perplexing to computers and much hard- Bottom Up Top Down
(Bayesian Methods)
er than learning to play chess. (Deep Learning)
A single example of the letter A suffices
Examples of the letter A teach a
Computers that learn to recognize a to recognize similar examples when
computer to distinguish patterns of
whiskered, furry face often need millions of Output: using Bayesian methods. The
light and dark pixels for various System is
Pixel by pixel, machine builds a model of the
examples to categorize objects that we can versions of the letter. Then, when this character primed with one
letter from its own internal
the machine receives a new input, resembles the example of a new concept,
classify with just a few. After extensive library of “parts,” assembling
it assesses whether the pixels training enough to support a range
training, the computer might be able to a figure made up of an acute
match the configuration from the raw data set; of output tasks
angle joined by a crossbar,
identify an image of a cat that it has never training set, confirming the letter therefore,
it is an A an A that can then be used to
is, in fact, an A. Deep learning
seen before. But it does so in ways that are identify slightly different
is a more complex
versions of the letter or to
quite different from human generaliza- version of this approach.
Input modify it in various ways.
tions. Because the computer software rea-
Input
sons differently, slipups occur. Some cat
images will not be labeled as cats. And the System is
computer may incorrectly say an image is a trained with raw data
(that is, pixels)
cat, although it is actually just a random
blur, one that would never fool a human
observer.
Output: Output: Output: Output:
Classifica- Generation Parsing Generation
All the Way Down tion of of object of
input new into parts new
The other approach to machine learning examples examples concepts
that has transformed AI in recent years
CREDIT: JEN CHRISTIANSEN; SOURCE: “HUMAN-LEVEL CONCEPT LEARNING THROUGH PROBABILISTIC PROGRAM INDUCTION,”
BY BRENDEN M. LAKE, RUSLAN SALAKHUTDINOV AND JOSHUA B. TENENBAUM,
IN SCIENCE , VOL. 350; DECEMBER 11, 2015 ( BAYESIAN EXAMPLE )

25
works in the opposite direction, from the
top down. It assumes that we can get ab-
stract knowledge from concrete data be-
cause we already know a lot and especially
because the brain is already capable of un-
derstanding basic abstract concepts. Like
scientists, we can use those concepts to for-
mulate hypotheses about the world and
make predictions about what data (events)
should look like if those hypotheses are
right—the reverse of trying to extract pat-
terns from the raw data themselves, as in
bottom-up AI.
This idea can best be illustrated by re-
visiting the spam plague through consider-
ing a real case in which I was involved. I re-
ceived an e-mail from the editor of a jour-
nal with a strange name, referring
specifically to one of my papers and pro-
posing that I write an article for the publi-
cation. No Nigeria, no Viagra, no million
dollars—the e-mail had none of the com-
mon indications of a spam message. But by
using what I already knew and thinking in
an abstract way about the process that pro-
duces spam, I could figure out that this
e-mail was suspicious.
To start, I knew that spammers try to ex-
tract money from people by appealing to
human greed—and academics can be as
greedy to publish as ordinary folks are for
$1-million prizes or better sexual perfor-
mance. I also knew that legitimate “open
access” journals have started covering their
costs by charging authors instead of sub-
scribers. Also, my work has nothing to do
with the journal title. Putting all that to-
gether, I produced a plausible hypothesis
that the e-mail was trying to sucker aca-
demics into paying to “publish” an article
in a fake journal. I could draw this conclu-
sion from just one example, and I could go
on to test my hypothesis further by check-
ing the editor’s bona fides through a
search-engine query.
A computer scientist would call my rea-
soning process a “generative model,” one
that is able to represent abstract concepts,
such as greed and deception. This same
model can also describe the process that is
used to come up with a hypothesis—the
reasoning that led to the conclusion that
the message might be an e-mail scam. The
model lets me explain how this form of
spam works, but it also lets me imagine
other kinds of spam or even a type that dif-
fers from any I have seen or heard about
before. When I receive the e-mail from the
journal, the model lets me work backward—
tracing step by step why it must be spam.
Generative models were essential in the
first wave of AI and cognitive science in the
1950s and 1960s. But they also had limita-
tions. First, most patterns of evidence
might, in principle, be explained by many
different hypotheses. In my case, it could
be that the e-mail really was legitimate,
even though it seemed unlikely. Thus, gen-
erative models have to incorporate ideas
about probability, one of the most import-
ant recent developments for these meth-
ods. Second, it is often unclear where the
basic concepts that make up generative
models come from. Thinkers such as Des-
cartes and Noam Chomsky suggested that
you are born with them firmly in place, but
do you really come into this world knowing
how greed and deception lead to cons?
Bayesian models—a prime example of a
recent top-down method—attempt to deal
with both issues. Named after 18th-centu-
ry statistician and philosopher Thomas
Bayes, they combine generative models
with probability theory using a technique
called Bayesian inference. A probabilistic
generative model can tell you how likely it
is that you will see a specific pattern of data
26
if a particular hypothesis is true. If the
e-mail is a scam, it probably appeals to the
greed of the reader. But of course, a mes-
sage could appeal to greed without being
spam. A Bayesian model combines the
knowledge you already have about poten-
tial hypotheses with the data you see to let
you calculate, quite precisely, just how like-
ly it is that an e-mail is legitimate or spam.
This top-down method fits better than
its bottom-up counterpart with what we
know about how children learn. That is
why, for the past 15 years, my colleagues
and I have used Bayesian models in our
work on child development. Our laboratory
and others have used these techniques to
understand how children learn about
cause-and-effect relationships, predicting
how and when youngsters will develop new
beliefs about the world and when they will
change the beliefs they already have.
Bayesian methods are also an excellent
way to teach machines to learn like people.
In 2015 Joshua B. Tenenbaum of the Mas-
sachusetts Institute of Technology, with
whom I sometimes collaborate, Brenden M.
Lake of New York University and their col-
leagues published a study in Science. They
designed an AI system that could recognize
unfamiliar handwritten characters, a job
that is simple for people but extremely tax-
ing for computers.
Think of your own recognition skills.
Even if you have never seen a character in a
Japanese scroll, you can probably tell if it is
the same or different from one on another
scroll. You can probably draw it and even
design a fake Japanese character—and un-
derstand as well that it looks quite differ-
ent from a Korean or Russian character.
That is just what Tenenbaum’s team mem-
bers got their software to do.
With a bottom-up method, the comput-
er would be presented with thousands of
examples and would use the patterns found
in those examples to identify new charac-
ters. Instead the Bayesian program gave
the machine a general model of how to
draw a character: for example, a stroke can
go right or left. And after the software fin-
ishes one character, it goes on to the next.
When the program saw a given charac-
ter, it could infer the sequence of strokes
that were needed to draw it, and it went on
to produce a similar set of strokes on its
own. It did so the same way that I inferred
the series of steps that led to my dubious
spam e-mail from the journal. Instead of
weighing whether a marketing scam was
likely to lead to that e-mail, Tenenbaum’s
model guessed whether a particular stroke
sequence was likely to produce the desired
character. This top-down program worked
much better than deep learning applied to
exactly the same data, and it closely mir-
rored the performance of human beings.
A Perfect Marriage
These two leading approaches to machine
learning—bottom up and top down—have
complementary strengths and weaknesses.
With a bottom-up method, the computer
does not need to understand anything
about cats to begin with, but it does need a
great deal of data.
The Bayesian system can learn from just
a few examples, and it can generalize more
widely. This top-down approach, though,
requires a lot of work up front to articulate
the right set of hypotheses. And designers
of both types of systems can run into simi-
lar hurdles. The two approaches work only
on relatively narrow and well-defined prob-
lems, such as recognizing written charac-
ters or cats or playing Atari games.
Children do not labor under the same
constraints. Developmental psychologists
27
have found that young children somehow
combine the best qualities of each ap-
proach—and then take them much further.
Augie can learn from just one or two exam-
ples, the way a top-down system does. But
he also somehow extracts new concepts
from the data themselves, like a bottom-up
system. These concepts were not there to
begin with.
Augie can actually do much more. He
immediately recognizes cats and tells let-
ters apart, but he can also make creative
and surprising new inferences that go far
beyond his experience or background
knowledge. He recently explained that if an
adult wants to become a child again he or
she should try not eating any healthy vege-
tables, because they make a child grow into
an adult. We have almost no idea how this
kind of creative reasoning emerges.
We should recall the still mysterious
powers of the human mind when we hear
claims that AI is an existential threat. Arti-
ficial intelligence and machine learning
sound scary. And in some ways, they are.
The military is researching ways to use
these systems to control weapons. Natural
stupidity can wreak far more havoc than
artificial intelligence, and we humans will
need to be much smarter than we have been
in the past to properly regulate the new
technologies. Moore’s law is an influential
force: even if advances in computing result
from quantitative increases in data and
computer power, rather than conceptual
revolutions in our understanding of the
mind, they can still have momentous, prac-
tical consequences. That said, we shouldn’t
think that a new technological golem is
about to be unleashed on the world.  M
MORE TO EXPLORE
Bayesian Networks, Bayesian Learning and Cognitive
Development. Alison Gopnik et al. in Developmental Science,
Vol. 10, No. 3, pages 281–287; May 2007.
Human-Level Concept Learning through Probabilistic Program
Induction. Brenden M. Lake et al. in Science, Vol. 350, pages
1332–1338; December 11, 2015.
The Gardener and the Carpenter: What the New Science of Child
Development Tells Us about the Relationship between Parents What makes some people more creative than others?
and Children. Alison Gopnik. Farrar, Straus and Giroux, 2016. For $9.99, this special edition explores
the intricacies of creativity from the rise
FROM OUR ARCHIVES of ingenuity in early humans to the
Machines Who Learn. Yoshua Bengio; June 2016. nurturing power of imaginative play to the
eccentricities of the unleashed mind, and more.
SCIENTIFIC AMERICAN ONLINE
Read more about how researchers use Bayesian methods at
BUY NOW
ScientificAmerican.com/jun2017/gopnik

 JAMIE GRILL GET T Y IMAGES
You Do Not Think Alone
A new book argues that thought and knowledge are community efforts
By Gareth Cook
T
he Thinker, Auguste Rodin’s
bronze sculpture, has be-
come a visual cliché, a com-
mon representation of deep
thought—a figure gazing
down, chin on hand, completely alone. This
is utterly misleading, according to the au-
thors of The Knowledge Illusion, which car-
ries the subtitle Why We Never Think Alone.
Steven Sloman, a professor at Brown Uni-
versity, and Philip Fernbach, a professor at
the University of Colorado Boulder’s Leeds
School of Business, argue that our intelli-
gence depends on the people and things
that surround us and to a degree we rarely
recognize. Knowledge, they say, is a com-
munity effort. Sloman answered questions
from Mind Matters editor Gareth Cook.
Scientific American: You argue that
we don’t know as much as we think
we do. Can you explain this?
Sloman: People overestimate how well they
understand how things work. Direct evi-
dence for this comes from the psychologi-
cal laboratory. The great Yale psychologist
Frank Keil and his students first demon-
strated the illusion of explanatory depth,
what we call the knowledge illusion. He
asked people how well they understand
how everyday objects (zippers, toilets, ball-
point pens) work. On average, people felt
they had a reasonable understanding (at
the middle of a 7-point scale). Then Keil
asked them to explain how they work. Peo-
ple failed miserably. For the most part, peo-
ple just can’t articulate the mechanisms
that drive even the simplest things. So
when he again asked them to rate their un-
derstanding, their ratings were lower. By
their own admission, the act of attempting
to explain had pierced their illusion of un-
derstanding. We have replicated this basic
finding many times, not only with everyday
objects but also with political policies. Psy-
chologist Matthew Fisher has shown that
people overestimate their ability to con-
struct logical justifications for their beliefs.
More indirect evidence comes from the
simple fact that people are surprisingly ig-
norant. [About] 50 percent of Americans
don’t know that antibiotics kill bacteria,
not viruses; only a minority can name even
a single Supreme Court justice. Cognitive
psychologist Rebecca Lawson has shown
that people can’t draw a bicycle, even with
substantial help. Yet people are surprised
when they discover what they can’t do.
That’s amazing that people have
trouble justifying their beliefs logically.
How can that be?
Human reasoning takes a couple of forms.
Most of the conclusions we come to are the
products of intuition. Intuitive processes
can be identified because we have no intro-
spective access to how they work; we are
only conscious of their output. For instance,
intuitive processes deliver stored conclu-
sions from memory. We can’t introspect to
see how memory retrieves information; it
just serves it up to consciousness.
To illustrate, most of us believe that
there was a great revolution in France in
the late 18th century. How do we justify
that belief? Most of us aren’t historians; we
just dredge up the fact from memory. We
can’t really justify it except by appealing to
our own memories, and we can’t even say
much about how we retrieved the memory.
It just comes to mind. Intuitive processes
are capable of more than just memory,
though. They are also capable of pretty so-
Gareth Cook is a Pulitzer Prize–winning journalist in Boston.
He edits Mind Matters, an online commentary blog at
www.ScientificAmerican.com/mind-matters
30
 phisticated pattern recognition. If you ask
me to reconstruct what I know about the
French Revolution, I can tell a story. The
story will be pretty superficial and miss a
lot of facts—really important facts—but
it’ll be largely coherent because my intui-
tive system is sophisticated enough to have
some sense of how the world works. For ex-
ample, I don’t remember the name of the
king, but I can tell you he was captured be-
fore his head was chopped off because you
can’t chop someone’s head off unless you’ve
captured the person. And, truth be told, I’m
just guessing his head was chopped off be-
cause my memory tells me that lots of peo-
ple had their heads chopped off in that pe-
riod. So intuition is pretty powerful—it can
THAD RUSSEL
tell a really good story. But it’s very limited
Steven Sloman in its factual basis. The late cognitive sci-
entist Thomas Landauer calculated that
humans can retain only about 1 gigabyte of
information, just a fraction of what a mod-
ern flash drive can hold.
Beyond intuition, we can also reason by
deliberating, thinking things through care-
fully. But we don’t do that very much, and
we’re not very good at it as individuals. We
need a lot of help. We often use things in
the world to help us, like whiteboards and
computers. But more than anything, we use
other people. Most thinking involves col-
laborating with other people. That’s why
scientists have lab meetings, why doctors
consult with specialists, and why it’s im-
portant to have someone to talk to when
you’re confused or upset. Individuals can’t
justify their beliefs, but groups are great at
justifying things (though not necessarily
justifications that would pass muster with
a philosopher). A little social support can
generate a lot of confidence.
Why is it important for
people to know this?
The main reason is mere curiosity. We are
told to “know thyself,” and what could be a
more important thing to know than what
we are capable of mentally? Moreover,
knowing that we’re more ignorant than we
think should make us more humble and
give us greater respect and gratitude for
others and the knowledge they bring to the
table. This is important in all our human
relationships, whether at home, at work or
elsewhere. And it’s also important to live
with others in a just and peaceful society.
31
Tell me more about this idea
that what we know is “social”?
People fail to distinguish the knowledge
that’s in their own heads from knowledge
elsewhere (in their bodies, in the world
and—especially—in others’ heads). And we
fail because whether or not knowledge is in
our heads usually doesn’t matter. What
matters is that we have access to the knowl-
edge. In other words, the knowledge we use
resides in the community. We participate
in a community of knowledge. Thinking
isn’t done by individuals; it is done by com-
munities. This is true at macro levels: fun-
damental values and beliefs that define our
social, political and spiritual identities are
determined by our cultural communities. It
is also true at the micro level: We are natu-
ral collaborators, cognitive team players.
We think in tandem with others using our
unique ability to share intentionality.
Individuals are rarely well described as
rational processors of information. Rather
we usually just channel our communities.
What do you mean that we are
“cognitive team players”?
The deliberative mind is designed to work
with other people. When we’re crossing
the street, we have to think about what on-
coming drivers are thinking, and we often
make eye contact with them to confirm
that we’re on the same page. This kind of
meshing of cognitive gears is even more
pronounced when we’re engaging in any
group activity: playing sports, sitting
around the dinner table telling jokes, fix-
ing our car or trying to crack the genetic
code. We think together. We feed one an-
other’s intuitions, we complete one anoth-
er’s thoughts, we hold knowledge that oth-
ers can make use of. There’s a division of
cognitive labor.
Some cognitive anthropologists have
made a strong argument that human be-
ings are the only animals—indeed, the only
EMILY SACCO
cognitive systems—capable of this kind of
collaboration. Humans are capable of shar-
ing intentionality. We don’t merely pursue Philip Fernbach
collective goals as many animals can do
(for instance, some species hunt in packs),
but we pursue a common goal together.
When, for example, a child and parent are
building a sand castle together, they are lit-
erally sharing thoughts: they are pursuing
a common end result and doing so with
knowledge that they hold in common,
about the weather, the tides (if they’re on
32
the beach), the availability of tools, et cet-
era. If one runs into a problem, the other
might help. This requires that they under-
stand that they share a goal. There are clev-
er experiments showing that this kind of
sharing of mental states is common and
easy for even young children but beyond
the capabilities of our closest genetic rela-
tive, chimpanzees.
How do you think about the way
technology fits into this?
Technology fits in many ways. First, it ex-
acerbates the knowledge illusion because it
is a powerful source of information. Studies
by psychologist Adrian Ward and others
have shown that we feel smarter around
Google. Of course, we should, as long as we
have access to it. It is a source of informa-
tion like no other. It is perhaps the most
significant member of our community of
knowledge. But its role is different from
that of other humans because it lacks the
critical ability that humans possess: It
doesn’t share our intentionality. It doesn’t
read our minds to figure out what we’re
looking for. It sometimes does a good imi-
tation of a human because some clever pro-
grammer has figured out what a good re-
sponse to a query would be. But the clever-
ness in that case is in the human
programmer, not the machine. As a result,
we have to be careful with technology. GPS
software has sent many a driver the wrong
way, sometimes causing them to drive into
lakes. And there have been disasters caused
by overreliance on technology because we
sometimes treat our technology as if it were
sharing our goals, as if it were human. So
we have to remain vigilant.
But the scariest role of technology is
what it has done to our social systems. It’s
so easy today to live in a bubble of
like-minded individuals on social media;
indeed, it’s hard not to. And the biggest
Web sites just tell us what we want to hear,
stuff that agrees with the view we already
have. As a result, we could be getting en-
tirely different news than people on other
The articles in this special edition ooer a host
sides of the political divide. We might be
of insights into raising children grounded solidly in
living in entirely different worlds with re-
scientific research. For $9.99, access
spect to the information we have about
what’s going on in the world. This is a rec- compelling articles on academic testing,
ipe for social tension. And we’re seeing unstructured play, the teen brain, and more!
precisely this kind of tension and the dam-
age it can cause.  M BUY NOW

 New Drugs for
Lou Gehrig´s Disease
Head for Clinical Trials
Newly discovered genetic mutations are
providing clues about how this disorder
relentlessly destroys motor neurons and
robs people of their mobility. The findings
may lead to drug therapies for a condition
that has long defied treatment
By Leonard Petrucelli
and Aaron D. Gitler
JERE
MY W
IL S O
N

34
A
myotrophic lateral sclero-
sis (ALS) strikes without
warning. The condition,
which strips nerve cells of
their ability to interact with
the body’s muscles, starts painlessly, with
subtle initial symptoms—such as stum-
bling, increased clumsiness and slurred
speech—that are often overlooked. The dis-
ease itself attracted little public attention
until legendary New York Yankees first
baseman Lou Gehrig began dropping balls
and collapsing on the field for no apparent
reason. Known as the Iron Horse for play-
ing in 2,130 consecutive games over 14
years, Gehrig was diagnosed with ALS in
June 1939 and delivered a poignant fare-
well at Yankee Stadium the next month.
Gehrig’s loss of muscle control progressed
so rapidly that by December he was too
weak to attend his National Baseball Hall
of Fame induction. Creeping paralysis
Leonard Petrucelli is a professor and chair of the department
of neuroscience at the Mayo Clinic in Jacksonville, Fla.
Aaron D. Gitler is an associate professor of genetics at the
Stanford University School of Medicine.
eventually left him bedridden. He died in IN BRIEF
June 1941 at the age of 37.
• A myotrophic lateral sclerosis (ALS), a neurode-
Today more than 6,000 people a year in generative disorder commonly known as Lou
the U.S. receive a diagnosis of ALS, now Gehrig’s disease, attacks nerve cells that lead
commonly known as Lou Gehrig’s disease from the brain and the spinal cord to muscles
in the States and as motor neuron disease throughout the body.
in Europe. It usually afflicts people between
the ages of 50 and 60 but can start much • S ophisticated gene-sequencing technologies
earlier or even as late as one’s 80s. At its on- have led to a flurry of discoveries revealing the
genetic underpinnings of ALS. Ongoing research
set, nerve cells in the brain and spinal cord
indicates that changes in any of many different
called motor neurons begin to die. Because
genes increase an individual’s susceptibility to
these cells send signals from the brain
the disease.
through the spinal cord to muscles, their
death causes a loss of mobility, dexterity, • G
 ene silencing using a synthetic molecule
speech and even swallowing. In most cases, called an antisense oligonucleotide has emerged
the higher functions of the brain remain as a potential treatment for some types of
undamaged: people stricken with ALS are ALS. Researchers are also seeking ways of
obliged to watch the demise of their own measuring the disease as it progresses to help
body as the disease advances unrelentingly. with early detection and the development
of drug therapies.
They soon become wheelchair-bound and,
eventually, bedridden. Left with no capacity
to communicate, eat or breathe on their
own, most die from respiratory failure with-
in three to five years. The sole Food and
Drug Administration–approved drug for
ALS is the glutamate blocker riluzole, which
prolongs survival by an average of three
months. There is no cure.
Pioneering French neurologist Jean-
35

Martin Charcot, who identified the disease
in 1869, encapsulated a description of it in
its name: “amyotrophic” means no muscle
nourishment; “lateral” refers to an area of
the spinal cord where portions of the dying
motor neuron cells are located; and “sclero-
sis” is the hardening or scarring that occurs
as the process of neural degeneration un-
folds. Despite Charcot’s straightforward
characterization, nearly a century and a half
later the complexity of ALS continues to
confound researchers. Although the disor-
der is invariably fatal, for unknown reasons
roughly 10 percent of patients survive for
more than 10 years, and some do so even
longer. That minority includes physicist
Stephen Hawking, who has famously lived
with ALS for more than five decades. Cur-
rent research suggests that environmental
factors play only a small role in triggering
ALS, probably by increasing the vulnerabil-
ity of individuals who are already genetical-
ly susceptible. Most puzzling is that the dis-
order occurs largely at random. Fewer than
10 percent of cases arise from genetic traits
passed down from one generation to the
next within a family. The remaining cases
are classified as uninherited, or sporadic.
During the past decade sophisticated
During the past decade sophisticated
sequencing technologies have
led to exponential growth in our
understanding of the disorder’s
underlying biology.
sequencing technologies have led to expo-
nential growth in our understanding of the
disorder’s underlying biology. Ongoing re-
search indicates that many different genes,
acting alone or in concert, can increase an
individual’s susceptibility. Specific muta-
tions have been tied to almost 70 percent
of familial cases and approximately 10 per-
cent of sporadic ALS. In turn, this wealth of
new genetic data is opening up many prom-
ising avenues for better therapy. Gene si-
lencing has emerged as a potential treat-
ment for some forms of ALS, with two drugs
that target separate rogue genes slated for
clinical trials this year. Meanwhile re-
searchers are identifying telltale biomark-
ers, including measurable substances in
bodily fluids or electrical activity in the
brain, that could help clinicians make early
diagnoses and better gauge the progress of
the disease. Such biomarkers may also be
useful in the development of other drug
treatments.
Early Genetic Clues
Although people with familial ALS, most of
whom have a 50 percent chance of passing
it to the next generation, make up a small
portion of ALS sufferers, they have played
an outsize role in helping to unravel the
36
 genetic underpinnings of the disease. The
A Rogue Gene Unmasked first genetic link to ALS came in 1993 from
studies that identified a mutation in a gene
Three Problematic Outcomes called SOD1 in approximately 20 percent of
The mutations occur in C9ORF72, SCENARIO 1: Recent studies have
the 72nd open reading frame, or protein- The excess repeats revealed that too many familial ALS cases. SOD1 codes for the an-
Output
encoding region, of the chromosome. The protein cause less RNA to be repeated “letters” in a tioxidant enzyme superoxide dismutase,
faulty messenger RNA transcribed from this DNA transcribed, leading to
might damage motor neurons in three ways, the production of too DNA sequence along which converts the highly reactive mole-
traveling to and from a cell’s ribosomes, where little of the protein coded chromosome 9 account
it is translated into proteins. by C9ORF72 and a loss for most inherited cases
cule superoxide—an oxygen free radical—
DNA of its normal, as yet
and some sporadic into less damaging forms.
unknown function.
C9ORF72 gene cases of ALS. While Researchers initially theorized that the
Ribosome
researchers unravel the
Mutation (more than mutation in SOD1 might weaken the en-
usual number of repeats) mystery of how these
mutations cause dis- zyme’s antioxidizing capabilities and thus
SCENARIO 2:
RNA containing extra ease, drug developers allow oxygen free radicals to wreak havoc
Misfolded
RNA
repeats is transcribed are testing a synthetic on motor neurons. A quarter of a century
from the gene’s double
variations molecule, called an
“sense” and “antisense” later we have learned with near certainty
strands of DNA, yielding antisense oligonucle-
RNA
misfolded RNA molecules otide, or ASO, to silence that that is not the case. Rather it seems
that can trap an array of them. that this mutation triggers what scientists
RNAs and proteins.
call a toxic gain of function, in which the
enzyme does something beyond what it is
SCENARIO 3: normally supposed to do.
Stopping the Mutation from Causing Trouble Instead of coding the
An ASO molecule is designed to enter a cell and seek typical protein output, In particular, the new function leads to
the extra repeats in the
out the RNA transcribed from the mutant C9ORF72 gene,
RNA are translated into
changes in the shape of certain proteins in
marking that RNA for destruction by one of the cell’s own
enzymes. Imminent trials will test this approach by infusing an assortment of useless neurons. Most autopsies of people with ALS
and toxic proteins that
the synthetic molecule into the nervous system of ALS patients
potentially damage brain reveal a typical pattern of brain pathology:
who have the C9ORF72 mutation via an intrathecal injection, Useless
a kind of reverse spinal tap. proteins and spinal cord neurons. clumps of proteins accumulated within mo-
ASO molecule
ASO molecules
tor neurons. For these neurons to function
introduced via optimally, the protein building blocks inside
spinal injection
the cells must be recycled efficiently; with
ALS, that recycling system breaks down. All
GRAPHIC BY JEN CHRISTIANSEN

RNA
Enzyme 37
proteins, including enzymes, need to adopt
precise three-dimensional shapes as they
are synthesized in cells if they are to work
properly. Researchers eventually discovered
that mutations seem to cause individual
proteins to fold improperly and then clump
together. Cells tag these misshapen proteins
with ubiquitin, a molecular marker, which
signals that they need to be removed. When
this cellular disposal system becomes over-
whelmed, the trash builds up. In people with
certain types of familial ALS, motor neurons
are littered with clumps of aberrant SOD1
proteins tagged with ubiquitin.
A major breakthrough in ALS research
occurred in 2006, when scientists looked at
cases of ALS without SOD1 mutations. In
virtually every one, they discovered that an-
other protein, called TDP-43, also clumps
within motor neurons. TDP-43 belongs to a
class of proteins that regulate the activity
of messenger RNAs—mobile copies of DNA
that serve as templates for making the pro-
teins encoded by a gene’s DNA “letters.”
TDP-43 binds to a messenger RNA, guides
its processing in the nucleus, transports it
to where it needs to go in the cell and per-
forms other functions important for “trans-
lating” the RNA into a protein. Somehow in
ALS, the TDP-43 protein gets pulled out of
the nucleus and starts accumulating in the
surrounding cytoplasm. It might even act as
a kind of sink to pull additional copies of it-
self into that cytoplasm. Scientists have yet
to determine whether TDP-43 displays a
loss of function (because it is pulled from
the nucleus) or a toxic gain of function (be-
cause it builds up in the cytoplasm), or both.
Identification of TDP-43 as the key
clumping protein in most cases of ALS
helped geneticists home in on the specific
gene encoding it, TARDBP, and they found
rare mutations among some families with
an inherited form of the disease. The main
game changer in this work was the concep-
tual discovery that alterations in an
RNA-binding protein could cause ALS. Re-
searchers subsequently identified several
additional ALS-causing genes that give rise
to proteins involved in regulating RNA and
anticipate that there may be many more.
The late 2000s saw an explosion in ALS ge-
netics discoveries, with one or two new ALS
genes surfacing each year. But the most ex-
citing discovery was yet to come.
DNA Repeats Run Amok
The findings emerged from studies of sever-
al families with an inherited form of ALS. In
2011 two scientific teams independently re-
ported that they had found a peculiar type of
mutation in a gene with an equally peculiar
name—C9ORF72, which stands for the 72nd
open reading frame, or the part of a gene that
codes for a protein, on chromosome 9. In
healthy people, this gene includes a short se-
quence of DNA—GGGGCC—that is repeated
two to 23 times. In people with the C9ORF72
mutation, this segment is repeated hundreds
or sometimes thousands of times.
Subsequent research revealed that these
excessive repeats could explain 40 to 50 per-
cent of familial ALS cases and 5 to 10 per-
cent of seemingly sporadic cases. Intrigu-
ingly, the discovery of the mutations pro-
vided a genetic connection between ALS
and another disease, a form of dementia
called frontotemporal degeneration (FTD).
FTD is marked by changes in personality
and decision making. C9ORF72 mutations
can cause ALS or FTD, or even a combina-
tion of both called ALS-FTD. And clumps of
that ever present TDP-43 protein build up
in the neurons of people with C9ORF72 mu-
tations, providing yet another connection
between the two disorders. This association
implies that ALS and FTD might be part of a
38
spectrum of related conditions, although
how mutations in the same gene would lead
to such divergent symptoms is unclear.
Researchers are investigating three cel-
lular mechanisms that might explain how
the mutations in this mysterious gene
cause ALS. The repeating DNA segment
could interfere with the way the genetic
code is normally copied into messenger
RNA and then translated into C9ORF72
protein, decreasing the amount of protein
synthesized. This decrease could diminish
the protein’s effects, although its exact
function is still unknown. Alternatively,
there could be a toxic gain of function: per-
haps the repeating sequence causes the
RNA itself to form clumps that build up in
the nuclei of neurons and act like a sink,
trapping RNA-binding proteins and pre-
venting them from going about their usual
business. Or there could be a toxic gain of
function because of a bizarre twist of mo-
lecular biology in which the expanded re-
peat sequence gets translated into small
rogue proteins that are themselves prone
to clumping in the neurons of people with
C9ORF72 mutations.
So far the evidence suggests C9ORF72
mutations cause ALS through a toxic gain
of function, although the relative contribu-
tions of clumps of RNA and clumps of pro-
tein are still unclear. Ultimately the dis-
tinction may not matter, because thera-
peutic strategies are being developed that
could shut off the production of both RNAs
and proteins from the mutant gene in one
fell swoop.
Repeat Police to the Rescue?
Gene silencing using a synthetic molecule
called an antisense oligonucleotide (ASO)
represents one of the most exciting new
therapeutic advances in neurodegenera-
tive disease. An ASO molecule is designed
to locate and bind itself to the messenger
RNA molecule produced from a specific
gene, which in turn prompts an enzyme to
snap into action and attack the RNA-ASO
hybrid. ASOs can lead to the selective de-
struction of virtually any RNA produced
from a mutant gene. In the case of C9ORF72,
rodent studies indicate that antisense mol-
ecules engineered to destroy RNA clumps
in motor neurons can also destroy clumps
of aberrant repeat proteins and prevent
new protein clumps from forming.
Antisense drugs designed to target the
mutant C9ORF72 gene are expected to en-
ter clinical trials in humans this year. Mean-
while researchers have also designed an
antisense agent for the familial form of ALS
caused by SOD1, and results of an initial
clinical trial indicate it is safe to inject into
the fluid-filled space of the spinal column,
a site chosen to allow the drug to travel
through the cerebrospinal fluid that flows
around the brain and to find its way into
motor neurons.
The success of an ASO developed for an-
other neurodegenerative disease, called
spinal muscular atrophy, gives researchers
cause for cautious optimism. This genetic
motor neuron disease in infants is similar
to ALS. Very few children who suffer from
it live past their third birthday. In two re-
cent clinical trials of an antisense drug de-
signed to correct a gene defect that leads
to abnormal messenger RNA, children with
spinal muscular atrophy showed such dra-
matic improvement in their motor skills
that the FDA fast-tracked those trials and
gave formal approval for the drug in late
December 2016.
Solving Sporadic ALS
Studies of rare forms of ALS with a clear
familial inheritance pattern have paved
39
 the way for a better understanding of the
underlying biology of the disease. The big-
gest challenge going forward is to identify
mutations in the genomes of individuals
with sporadic ALS that make them suscep-
tible to the disease. Efforts are under way
around the world to collect DNA samples
from people with ALS and to scour their
genomes for data.
To expedite this task, geneticists have
developed a microchip that lets them con-
duct so-called genome-wide association
studies (GWASs) to readily compare the ge-
nomes of people with ALS with those of
healthy people. The chip focuses on ge-
nome regions known to have variants called
single-nucleotide polymorphisms—places
where a DNA letter, or nucleotide, can vary
from one person to another. GWASs are
correlational and thus cannot reveal wheth-
er something is causing ALS, but they can
identify suspect discrepancies that warrant
closer examination. Several recent interna-

MARK THOMPSON GET T Y IMAGES

ALS ice bucket challenge videos made by millions
of people, including Formula One driver Daniel
Ricciardo, helped to raise awareness and money
for research.

40
tional efforts to perform GWASs of more
than 10,000 people with ALS and more than
20,000 healthy people uncovered a number
of genomic differences that are now under
investigation. New technologies have also
simplified the process of collecting genetic
data, making it possible to sequence an in-
dividual’s entire genome in one day for less
than $1,000. It takes even less time and
money if you sequence only the exome, the
part of the genome that codes proteins.
Once researchers have assembled a com-
prehensive catalog of genetic variants as-
sociated with a predisposition for ALS, they
will attempt to decipher the complex ways
in which ALS-related genetic mutations in-
crease the risk of disease. That attempt will
include studying how various genes inter-
act and investigating whether multiple mu-
tant genes might be involved in some forms
of ALS, as well as considering how environ-
mental factors might help trigger the dis-
ease in some people. Some new studies
suggest that ALS may even result in part
from the reawakening of a dormant retro-
virus—a viral DNA sequence that long ago
inserted itself into the genome and nor-
mally would have sat quietly. It may be that
a retrovirus in some people with ALS jumps
from neuron to neuron in the brain, poten-
tially causing damage and initiating the
disorder in its wake.
Promising New Leads
A growing body of research suggests that
ALS is not merely a disease of dying motor
neurons. So-called glial cells, which are
even more abundant in the brain and the
central nervous system than neurons, may
also play an important role. Glial cells per-
form a variety of functions: some provide
physical support for neurons; others regu-
late the internal environment of the brain,
especially the fluid surrounding neurons
and their synapses. Recent studies of mice
with the SOD1 gene mutation produced a
surprise. Shutting off synthesis of the mu-
tant gene in glial cells prolonged life de-
spite the continued presence of toxic SOD1
protein in the animals’ motor neurons. It
appears that ALS may originate in the mo-
tor neurons but that communication with
glial cells helps to drive the progression of
the disease. Glial cells might also contrib-
ute to ALS by producing a toxic factor, al-
though scientists are not exactly sure of
what that factor is or how it works. Once
the factor (or factors) is identified, ways to
block its production or hinder its ability to
transmit its bad signal to motor neurons
could be developed to slow or halt ALS.
Amid the quest to unravel the myriad
causes of ALS, researchers have also been
scrambling to identify biomarkers that can
help doctors assess the progress of the dis-
ease. For example, ongoing efforts aim to
detect the abnormal repeat proteins made
from that C9ORF72 DNA expansion in eas-
ily accessible body fluids, such as the blood
or spinal fluid. In March one of us (Petru-
celli) reported that he had detected these
proteins in the cerebrospinal fluid of peo-
ple with ALS and ALS-FTD—as well as in
asymptomatic carriers of the mutated gene.
Such measurements could potentially aid
in early diagnosis. Other biomarker re-
search is focusing on developing imaging
techniques to help detect the TDP-43 pro-
tein clumps that build up in the brains of
people with ALS before these aggregates
start to kill motor neurons. All these bio-
markers could also serve as useful bench-
marks to judge the success of possible ther-
apies in clinical trials.
The rapid advances taking place in ge-
netics and genomics, as well as the devel-
opment of new and improved biomarkers,
41
will usher in an era of precision medicine
for ALS. In the near future, patients will be
grouped together based on the type of ALS
they have and will then receive a treatment
or preventive tailored to them.
The Power of Social Media
Much of the progress in ALS research during
the past decade can be attributed to the will-
ingness of large numbers of individuals af-
flicted with the disease to volunteer both
their time and their DNA to participate in
large-scale genomics studies. People with
ALS and their families have also helped in-
crease public awareness and canvas funds
to support ongoing research and patient
services through the power of social media.
The “ALS Ice Bucket Challenge” took the
Internet by storm in 2014. Pete Frates, a for-
mer captain of Boston College’s baseball
team who was diagnosed with ALS two years
earlier, at age 27, helped to get things rolling
when he posted a video on Facebook chal-
lenging his friends to dump buckets of ice
water over their heads to raise money for
the ALS Association. The campaign quickly
went viral as a host of celebrities, including
Mark Zuckerberg, Bill Gates, Oprah Winfrey,
Leonardo DiCaprio and LeBron James, took
the challenge. During an eight-week period, MORE TO EXPLORE
Facebook users posted more than 17 million State of Play in Amyotrophic Lateral Sclerosis Genetics. Alan E.
videos of themselves getting drenched for Renton, Adriano Chiò and Bryan J. Traynor in Nature Neuroscience,
the cause. Supporters ended up raising more Vol. 17, No. 1, pages 17–23; January 2014.
than $115 million, of which 67 percent went Decoding ALS: From Genes to Mechanism. J. Paul Taylor,
to research, 20 percent went to patient and Robert H. Brown, Jr., and Don W. Cleveland in Nature, Vol. 539,
community services, and 9 percent went to pages 197–206; November 10, 2016.
public and professional education.
ALS is a relentlessly cruel disease. Be- FROM OUR ARCHIVES
fore Gehrig’s stirring retirement speech at Playing Defense against Lou Gehrig's Disease. Patrick Aebischer
Yankee Stadium—in which he famously re- and Ann C. Kato; November 2007.
ferred to himself as “the luckiest man on
the face of the earth”—and news of his di- SCIENTIFIC AMERICAN ONLINE
agnosis spread, most people who contract- Read about how Stephen Hawking lives with ALS at
ed the disease suffered in silence. But now ScientificAmerican.com/jun2017/als-genes
public awareness continues to grow, in part
because of people like Frates. The social
media campaign he helped to spark revital-
ized the ALS Association, which has since
tripled its annual budget for research. Sci-
entists are optimistic that the explosive
growth in our understanding of ALS biolo-
gy will continue and that casting an ever
widening dragnet for rogue genes will lead
to better therapies for holding this stealth
killer at bay.  M
 42
Post-Truth: The Dark Side of the Brain
A growing number of politicians are talking nonsense
with impunity. False information is proliferating. What’s
worse, the human brain loves it
By Sebastian Dieguez

YUJI SAKAI GET T Y IMAGES

43
I
t may seem surprising that after be-
ing elected president, Donald Trump
continued to insist that the elec-
tions were rigged. Or that he accused
his predecessor of having tapped his
phone—without any proof. Or that one of
his advisers claimed that the inauguration
ceremony had shattered the record for at-
tendance, which clearly it had not. But that
would underestimate the new and baffling
phenomenon of “post-truth,” of which
Trump is the most striking example.
Post-truth triumphed the day after the
U.K. voted in favor of Brexit. That’s when its
defenders acknowledged that they had mis-
led the public about the health care benefits
of leaving Europe. It triumphed again when
François Fillon’s center-right campaign for
the French presidency rushed to exaggerate
the number of supporters who turned out
for a campaign rally in the Place du Tro-
cadéro in Paris. And again when the Spanish
government expressed pride in having stra-
tegically placed many researchers abroad in
Sebastian Dieguez is a postdoctoral researcher in cognitive
neuroscience in the Laboratory of Cognitive and Neurological
Sciences at the University of Fribourg in Switzerland. This article
originally appeared in Cerveau & Psycho.
scientific collaborations, when in fact they
had left as a result of relentless budget cuts.
Further examples include the Turkish gov-
ernment, which complains of being cen-
sored in Europe—while imprisoning its jour-
nalists. And Russia, which supplies the
whole world with dubious information
through clever use of media propaganda.
Leaving aside politicians, “alternative
facts” are created and disseminated at all
levels. It is this planetary tsunami of false
information that has driven many experts
to refer to the era of “post-truth.” But what
exactly is it? Derived from the title of a book
published by journalist Ralph Keyes in 2004,
the term achieved its 15 minutes of fame
after being selected in 2016 as the Oxford
Dictionaries Word of the Year. The Oxford
Dictionaries defines post-truth as an adjec-
tive “relating to or denoting circumstances
in which objective facts are less influential
in shaping public opinion than appeals to
emotion and personal belief.” This defini-
tion reflects a profound lack of confidence
in the “legitimate” dispensers of facts (par-
ticularly the media and experts), making
any truth, or any claim to the truth, sus-
pect. Falsehood, in all its forms, thus takes
on a routine character, becomes omnipres-
IN BRIEF
• M
 any political leaders have recently taken to
making fanciful statements. These assertions
combine with the mass of false information
circulating on the Internet and influence public
opinion. People no longer seem to care about
objective reality. We have gotten to the point
where experts now refer to the era of post-truth.
• A lthough “truth” is discussed all the time, it
has never had as little currency as it does now.
Researchers explain this development as the
product of our ancestral brain encountering
modern individualism and new technologies.
Studies show that what we find interesting
about information is not so much its veracity
as the social prestige it brings, especially on
social networks.
• T he phenomenon is difficult to control because
it is armed with formidable mechanisms of
self-defense. But fact-checking and education
in critical thinking initiatives must continue to
develop. We may also need to rehabilitate
fiction, emphasizing its capacity to give us plea-
sure without needing to make it into reality.
44
ent and overwhelming, and enjoys virtually
complete immunity. The sciences, unfortu-
nately, are not spared, as shown by the cli-
mate controversy and the categorical rejec-
tion of vaccines by a segment of the popu-
lation. Some people will even assert (as
former NBA player Shaquille O’Neal did re-
cently) that Earth is flat and defend their
right to say it.
The term “post-truth,” however, is not
without its detractors. Some observers point
out that the phenomenon of disinforma-
tion has always existed and that, conse-
quently, there is nothing new or “post” un-
der the sun. Because when have we ever re-
ally known an “era of truth”? Or been
perfectly and impartially informed? Or
bothered to listen to points of view contrary
to ours and argued solely on the basis of
facts? Others dispute the idea that “truth”
is even a relevant concept, given that it most
often is left to certain dominant elites to
determine what is false and what is true, ac-
cording to their own interests.
An Unholy Alliance?
The lack of consensus notwithstanding,
there does seem to be something new about
the current phenomenon. Many experts see
it as the monstrous fruit of the encounter
between our ancestral psychological pro-
pensities and technological progress.
In fact, our good old Homo sapiens brain
is not as concerned with objectivity as one
might think. Its priority is rather to safe-
guard its own regime of truth. The “argu-
mentative theory of reasoning” of French
researchers Hugo Mercier and Dan Sperber
postulates that our very ability to reason is
constrained and shaped by our need to be
right and to convince. This explains our nu-
merous errors of reasoning, which far from
being anarchic or random, often stroke our
ego. Thus, we more readily accept and retain
information that suits us and that reinforces
our beliefs—a tendency known as confirma-
tion bias. According to a recent study by Jo-
nas Kaplan of the University of Southern
California and his colleagues, an entire ce-
rebral network is involved in ego represen-
tation (called the default mode network).
This network is activated when we receive
information contrary to our political ideas,
as if our very identity were being attacked.
In another study, Micah Edelson of the
Weizmann Institute in Rehovot, Israel, and
his colleagues showed that changing one’s
mind depends on increased activity in a
small region called the anterior lateral pre-
frontal cortex. Yet this activity is inhibited
by the combined action of the amygdala
and the hippocampus, or the centers of
emotion and memory, respectively. The
memories and beliefs recorded in our brain
appear to have the power to prevent us
from changing our mind, especially if they
are emotionally laden.
Rise of Individualism
Yet another factor—this one sociopoliti-
cal—is involved in maintaining false be-
liefs: economic progress goes hand in hand
with a rise in individualism. This link was
recently shown by Henri Santos of the Uni-
versity of Waterloo in Ontario and his col-
leagues. The researchers looked at data
from 77 countries over 51 years, using be-
havioral criteria (such as the percentage of
people living alone) and values associated
with individualism (for example, using data
from surveys assessing the importance of
independence). In an individualist society,
personal expression and forming opinions
are highly valued. Truth and memory are
thus perceived less as a common, shared
legacy than in traditional societies and
more as strictly sacrosanct, private goods.
45
 Consequently, when partisans of François
Fillon or Donald Trump believed they saw
larger crowds than were truly present, their
brain is blocking both the information that
contradicts their beliefs and the brain re-
gions that would allow them to change their
mind. The sociopolitical context then rein-
forces their right to maintain their be-
liefs—a perverse outcome of what begins as
a positive impulse, namely, to form an indi-
vidual opinion.
But individual blindness is hardly the
whole story. Another key characteristic of
the era of post-truth is the way in which
false information spreads. Here is where
new technologies, and in particular social
networks, come into play: the ability to
share and disseminate makes it possible for
DAVID STEPHENSON GET T Y IMAGES (LEF T ); LUCAS JACKSON GET T Y IMAGES (RIGHT )

any belief to turn into “information,” espe-

cially as the distinction between them
tends to disappear. Walter Quattrociocchi
and his team at the IMT School for Ad-
vanced Studies in Lucca, Italy, showed how
Facebook amplifies the confirmation bias
through its “personalized” algorithms. In
fact, these algorithms lead to the creation
of isolated communities that tend to be-
come polarized, ones whose beliefs are re- Inauguration of Inauguration of
inforced and become ever more extreme. Barack Obama, 2009 Donald Trump, 2017

46
Buzz Neurons
A study done in 2017 shows how much our
lack of ancestral concern for objectivity is
reflected on social networks. Using func-
tional MRI, researchers at the University of
Pennsylvania analyzed the brain activity of
subjects reading New York Times articles
and then asked them to rate their likeli-
hood of sharing the articles later. The ques-
tion the researchers wanted to answer was:
Does this brain activity predict whether in-
formation will become viral on social net-
works? By comparing the number of actual
sharing generated by the article with the
participants’ fMRI scans, the researchers
concluded that this was indeed the case.
The key was the activity of a “judgment sys-
tem” comprising the ventral striatum and
the ventromedial prefrontal cortex: the
more these regions were activated during
reading, the more success the article had
on the Internet. The ventral striatum is in-
volved in motivation and pleasure, and the
ventromedial prefrontal cortex in self-rep-
resentation and social cognition.
These results suggest that the choice of
sharing an article is based on anticipating
the reactions of others (“Will my social net-
work like it and respond?”) and on the hope
for increased personal prestige (being
picked up and “liked” is a reward in itself
that undoubtedly contributes to the suc-
cess of social networks). It is not, however,
based on the veracity of the information.
Consequently, even information that is
false spreads. Conditions seem ripe for a
threshold effect: drowning in a mess of
falsehoods and vagaries, the truth loses all
prescriptive power to lay out a course of ac-
tion. This is an unprecedented situation
that leaves observers perplexed. How can
we fight against post-truth when there is
no recognition of the difference between
objective reality and personal opinion?
Fortunately, remedies are at hand: never
before have we seen so much analysis and
fact-checking. Even sales of 1984 by George
Orwell, a forerunner in the denunciation of
disinformation, are surging.
The Hydra Head of Post-Truth
But to what effect? Unfortunately, post-
truth possesses formidable self-defense
mechanisms. Disseminating corrective up-
dates, however factual they may be, often
reinforces false information, simply because
it is thus repeated and propagated. Attack-
ing an alternative fact also gives it weight,
making it more credible and memorable
than it deserves. Another difficulty arises
from the fact that “official” sources evoke a
certain degree of mistrust. Vaccination cam-
paigns therefore tend to reinforce the hos-
tility of people who oppose vaccines—in
other words, precisely those whose behavior
should be changed (the so-called boomer-
ang effect). Why? Because they feel that
their convictions are being attacked, one of
which is that an unnamed “they” seeks to si-
lence them by any means.
Worse, a recent study shows that even if
we succeed in changing false beliefs, this
does not guarantee a change in behavior.
Briony Swire of the Massachusetts Insti-
tute of Technology and her collaborators
interviewed subjects of all political stripes
during the presidential campaign of 2016,
before Trump was elected. They were asked
to rate whether the billionaire’s statements
were true or false. Not surprisingly, Trump’s
supporters were more likely to believe
them. After they were presented with cor-
rections of objectively false statements,
however, an amazing thing happened:
participants reduced their belief in the
falsehoods, regardless of the source of the
explanation—whether from a pro- or an-
47
ti-Trump expert—but they did not change
their voting preferences.
In short, all signs indicate that we ex-
pect to be lied to and that deep down many
of us find it normal or hardly realize it. As
the authors concluded, “Something other
than the truth [of his statements] account-
ed for his success.”
End of Cognitive Dissonance
In principle, being directly contradicted by
the facts should create a form of dissonance,
an uncomfortable cognitive state that must
be resolved one way or another, at the very
least by acknowledging bad faith—basic
postulates of a theory put forward by psy-
chologist Leon Festinger in the mid-1950s.
Here, too, we may be looking at an unprec-
edented phenomenon deserving of study:
post-truth could be killing cognitive disso-
nance, which had the virtue at least of sig-
naling some kind of incoherence.
How, then, are we to resist? Tried-and-
true methods are obviously of paramount
importance: to reestablish truth in all cir-
cumstances, to gain confidence by being
rigorous and impartial, to teach critical
thinking at school. Yet perhaps the truth is
ill equipped to win this battle alone. Per-
haps it is also worth rehabilitating fiction by
invoking its distinctiveness: after all, “post- ..................................................
truth” could well imply a “post-fiction”
world. Today the boundaries are blurred: Comprehensive Coverage
at Your Fingertips
Trump was, after all, a champion of reality
TV, the genre that introduced the idea that
one could film unscripted day-to-day life
and live a fairy tale. In this context, Françoise
Lavocat, professor of comparative literature BUY NOW
at the New Sorbonne University, Paris, ad-
vocates reestablishing a clear distinction
between truth and fiction. As a species, we
are particularly fond of stories. Of that there
is no doubt. But it only increases the urgen-
cy of resolving to defend truth and facts. We
are at a critical juncture. It may never be
more important to make the most of our ex-
traordinary ability to invent alternative
worlds and to learn to enjoy them without
confusing them with reality. M

In Baby Mode
The birth of a first child alters parents’ lives
suddenly and forever: sleepless nights,
afternoons in pajamas and hardly a
moment’s respite. Parents are able to make
this transition because of changes that take
place in the brain
By Anna von Hopffgarten

ANNIE ENGEL GET T Y IMAGES

49
A
few weeks after the birth of
my first son, I caught my-
self rocking a stick of but-
ter to sleep while in line at
the supermarket checkout.
I gently moved the shopping cart back and
forth, until I became aware of what I was
doing and stopped in the hope that no one
had observed me.
My son was one of those kids who often
got jolted out of his sleep for no apparent
reason. His baby carriage stopping at a cor-
ner was reason enough. Because of this, I
soon learned to push his wagon back and
forth when waiting for the light to change
or when standing on line. This movement
had become second nature to me.
During the first weeks and months after
birth, most parents find themselves in a
kind of baby mode: their daily schedule is
determined by their infant, and their
thoughts revolve around his or her needs.
Mommy and Daddy do whatever is neces-
sary to keep the baby fed and happy.
But newborns do not develop a day-
Anna von Hopffgarten is a biologist and editor at Gehirn&Geist,
where this article originally appeared.
night rhythm until later, and all of them IN BRIEF
want to nurse every two to three hours.
• A fter birth, a mother’s brain reorients complete-
Postponements are not accepted gracious- ly to care of her infant. The hormone oxytocin
ly. Even the most exhausted parents wake sensitizes her motivation and emotion centers
up with a start when they hear their baby’s to her child.
shrill cries.
But for some strange reason, many par- • T he neural networks in her brain are also
ents seem not to mind the frequent night- transformed structurally. For example, during
time wake-up calls. They nurse or feed their pregnancy the gray matter in the regions
baby and change diapers because that is of the frontal and temporal lobes shrinks.
Superfluous synapses there may also be re-
just what you do. Only the next morning,
moved at this time.
when they see their child sleeping peace-
fully, do they forget the rigors of the night • W
 e know from animal studies that the number
before. What makes that possible? of nerve fibers that respond to oxytocin and
Apparently the brain simply makes in- vasopressin increases in the brain of male par-
ternal adjustments to these new challeng- ents, which makes fathers especially nurturing.
es—an evolutionary adaptation without
which mammals would be unequipped to
nurture offspring. During the first few
years of life, infants are completely help-
less and would not survive for long with-
out engaged long-term parental support.
No matter how sleep-deprived newly mint-
ed parents may be, they prioritize the
well-being of their children.
The transformation to a life dedicated
to motherhood occurs during pregnancy
and the initial postpartum weeks. During
this time the mother’s body is flooded with
50
a cocktail of hormones, among them the unhappy. Fatigue? Hunger? Perhaps she to interpret their baby’s facial expressions
attachment hormone oxytocin as well as wants some attention—or to be left alone. correctly and know intuitively how to po-
estrogen and prolactin, which stimulates Some mothers have an uncanny feel for sition themselves so that the infant,
lactation. Some regions of the brain have what the problem might be. They are able who cannot yet focus her eyes, can
receptors for these hormones—the areas
that deal with stress, for instance. One no-
table region is the hypothalamus, which
controls important functions such as sleep
and hunger. A subregion, the medial pre-
optic area (mPOA), probably plays a criti-
cal role in parental nurturing. If this is ex-
cised from the brain of mother rats, they
suddenly neglect their offspring.

The Neural Engine of

the Maternal Instinct
During pregnancy the neurons, or nerve
cells, in the mPOA undergo change. Re-
searchers have observed that in pregnant
rats these cells grow, and the number and
length of the dendrites that conduct sig-
nals from other neurons also increase. The
researchers achieved the same effect by ad-
ministering a mixture of pregnancy hor-
mones to virgin animals. In all probability,

ALFRED PASIEKA GET T Y IMAGES

these hormones prepare this region for
birth and motherhood.
At a practical level, parents must strug-
Oxytocin hormone crystals
gle to determine what makes a crying baby

51
nonetheless recognize who is there for her.
This behavior is the result of numerous
adjustments and changes that have taken
place within the brain. These modifications
affect the regions that enable us to direct
our attention and the networks of the re-
ward system. Two important protagonists
in this process are oxytocin and dopamine;
the latter enables us to take action and in-
creases drive. The mere interaction with
the child stimulates the release of both
these chemical messengers, thereby inten-
sifying caring behavior.
Studies have shown that women who
have high blood levels of oxytocin during
pregnancy and the first postpartum months
respond empathetically to their children.
Duration of gaze, tone of voice, smiles—ev-
erything is just right. In contrast, mothers
who have low oxytocin levels more fre-
quently exhibit more intrusive and less em-
pathetic behavior.
As experiments have shown, even the
most subtle signals, such as a baby’s gur-
gling or facial expression, have a rewarding
and therefore motivating effect on the ma-
ternal brain. The power of this feeling of
elation was demonstrated by neuroscien-
tist Joan Morrell and her co-workers at Rut-
gers University. In a thought-provoking ex-
periment, they gave mother rats a choice
shortly before giving birth: Would they pre-
fer the presence of three infant animals or
a dose of cocaine? Eight days after giving
birth most of the mothers preferred the ba-
bies. Later, however, that pattern was re-
versed; after 16 days almost all the animals
succumbed to the drug. The rewarding ef-
fect of the babies apparently decreases over
time—at least in rats.
Nursing stabilizes attachment between
a mother and her child because the moth-
er’s body releases large quantities of oxyto-
cin during nursing. Physiologist Craig Fer-
ris of Northeastern University and his col-
leagues sought to understand exactly what
this messenger substance does in the brain.
They placed mother rats in a brain scanner
while they were nursing their young. They
also administered a dose of oxytocin to an-
other group of mother rats.
The brain activity of these rodents was
astonishingly similar. The olfactory system
and the emotional and reward centers, such
as the nucleus accumbens, the insula, the
amygdala and various core regions of the
hypothalamus, were all stimulated by nurs-
ing and by administered oxytocin. If, how-
ever the researchers administered a sub-
stance that blocks the binding sites for oxy-
tocin, the activity in these areas was
considerably weaker.
And in humans? In fact, studies using
functional MRI indicate that in women
oxytocin sensitizes brain circuits that pro-
cess emotions and increase motivation and
empathy, especially in the limbic system.
It is well established that women release
large quantities of oxytocin during vaginal
delivery. The hormone causes the contrac-
tions in the muscles of the womb that trig-
ger labor. This gave psychiatrist James Swain,
now at Stony Brook University, an idea. He
and his team placed mothers in a brain scan-
ner and played them audio and videotapes
of the crying of their own baby and that of a
strange baby. Half of the women had under-
gone natural delivery a few weeks earlier;
the other half had delivered by cesarean sec-
tion. The result was that the emotion and
motivation networks of mothers who had
given birth to their children vaginally re-
sponded to their own baby’s cries with far
greater neural activity than to the cries of a
strange baby. Also, their responses were
greater than those of the cesarean mothers.
52
Postcesarean Baby Blues
As researchers have suspected, a lack of
oxytocin is also a predisposing factor in
postpartum depression after a cesarean
section. Mothers who have delivered by this
procedure often feel depressed and irrita-
ble and demonstrate comparatively less in-
terest in their newborn.
This observation does not, however, mean
that a cesarean section permanently dam-
ages the mother-child relationship. Swain
repeated his crying experiment three to four
months later with the same mothers—and
the effect that he saw earlier had disap-
peared. In both groups of mothers, the brain
responded equally to the sound of their baby
MICHELLEGIBSON GET T Y IMAGES
crying. It seems that the initial difference in
oxytocin release eventually decreases.
A similar experiment was conducted by
Pilyoung Kim, now at the University of Den-
ver, and her team. All his test subjects had
had vaginal deliveries, but they differed in
how they fed their newborns. One half of
the women nursed their babies, whereas
the other half bottle-fed them. As expect-
ed, the oxytocin released during nursing
sensitized the brain’s motivation and emo-
tion centers. The nursing mothers respond-
ed more strongly to their child’s cries.
The neural reaction to infant crying
changes over time, however. Immediately
after the birth of their first child, some
mothers begin to sweat in response to the
slightest whimper. Did I do something
wrong? Is my baby sick? How can I calm
him? Of course, such urgent questions are
not exactly a rarity among mothers. This
initial sense of insecurity is also reflected
in brain activity. As Swain noted in his stud-
ies, the amygdala and insula respond espe-
cially strongly to the baby’s crying during
the first two to four weeks postpartum. And
these regions of the brain are important
centers for the processing of fear.
After three to four months, however,
53
this activity migrates to regions in the hy-
pothalamus and the medial prefrontal cor-
tex. The latter regulates negative emo-
tions by suppressing the amygdala’s reac-
tions to unpleasant stimuli. This change
can be seen as mothers become more
self-assured and respond more flexibly to
their child’s needs.
But neural activity is not all that changes
as a result of pregnancy and birth. The “hard-
ware” of the maternal brain adapts to the
new challenges. In 2010 psychologist Bene-
detta Leuner, now at Ohio State University,
counted the dendritic spines in a segment

SALLY ANSCOMBE GET T Y IMAGES

of the prefrontal cortex of rats. These
mushroomlike protrusions, or spines, cover
neurons’ branching structure, the dendrite,
where signals are received from other cells.
In fact, young mother rats had considerably
more spines than virgin animals did, al-
though it is not yet clear how this finding af-
fects behavior. Nevertheless, behavioral ex-
periments seem to indicate that mother an-
imals are cognitively more flexible and less
distractible—possibly because their pre-
frontal cortex is more efficient.
How do researchers define efficiency,
though? It is not possible to count directly
individual cells or dendritic spines in the
living human brain. Researchers must thus
fall back on methods such as MRI, where
the resolution is not as fine. Elseline
Hoekzema of Leiden University in the
Netherlands and her colleagues did just
that in a study published in 2017.
The researchers determined the volume
of gray matter of childless women–which
is made up largely of cell bodies of neurons.
When 25 of the test subjects became moth-
ers during the following year, their brain
was rescanned. To the researchers’ great
surprise, the gray matter in some regions of
the frontal and temporal lobes actually
shrank during pregnancy. This shrinkage
affected those regions that enable us to de-
code the emotions, expectations and inten-
tions of others. These changes were not
seen either in the fathers or in the childless
control group. Apparently the changes re-

54
sult from the pregnancy itself and not from
an adaptation to the parenting role.
According to Hoekzema, this decrease in
volume does not necessarily imply a reduc-
tion in the number and size of brain cells.
What is more probable is that the neural
connections undergo restructuring. She
compares this process with so-called prun-
ing in puberty, in which excess synapses
are removed. This change ensures that neu-
ronal communication grows more efficient
among the synapses that remain.
So much for the mothers. Does the pa-
ternal brain also adapt to the new role? The
answer is a resounding yes. Changes begin
during his partner’s pregnancy. As is seen
in women, the quantity of blood prolactin
and estrogen increases in men, albeit of
course at a much lower level. The testoster-
one level of fathers-to-be, on the other
hand, is somewhat lower than in childless
men. Testosterone levels remain somewhat
lower after the birth, especially in men who
spend a great deal of time with their chil-
dren. At this point, researchers can only
speculate as to the reasons. It could well be
that lower testosterone levels make fathers
calmer and more averse to risky behavior.
As studies on mice have shown, nerve fi-
bers in the brain that respond to oxytocin
and vasopressin are more common in
mouse fathers. Vasopressin promotes nur-
turing behavior in female rodents. If re-
searchers block the vasopressin system, the
animals exhibit less nurturing behavior to-
ward their offspring. Among other things,
these nerve fibers also receive touch sig-
nals from the skin. Researchers suspect
that a round of cuddling with children may
activate a caregiving response.
In 2014 Ruth Feldman of Bar-Ilan Uni-
versity in Israel and her co-workers under-
took a study of what exactly takes place in
the brain of a father after childbirth. In a
brain scanner, the neuroscientists showed
test subjects—heterosexual fathers, their
wives and fathers from a same-sex rela-
tionship—video sequences of their infant.
The brain of the mothers exhibited typical
excitation of the emotion centers, such as
the amygdala and insula. In contrast, brain
regions in the cortex of heterosexual fa-
thers that engage in planning and empa-
thizing with others became active.
What was astonishing was that when both
parents were men, the “maternal” brain re-
gions of the fathers who had assumed the
greater responsibility for child care were also
activated, among others the amygdala and
insula. With regard to brain activity, they
were simultaneously father and mother.
The more time that heterosexual fa-
thers spent alone with their children, that
is, without their partner, in the months be-
fore measurement, the more their brain
recruited the involvement of the amygda-
la, the typically female component of the
neural caregiving network. According to
the researchers, the father’s emotional
center switches to standby when the moth-
er is a steady presence. In effect, he lets her
bear the brunt of the worry. But when he
alone is responsible, his brain increasingly
takes on this function. And what is the up-
shot? Dads, let your wives get out of the
house more often—chances are you’ll do
just fine.  M
 55
 OPINION
Cracking the
Popularity Code
Do you know the two types
of popularity—and which is
better for you?

BETSIE VAN DER MEER GET T Y IMAGES

By Gareth Cook
56
W
e live in an age ob-
sessed with populari-
ty. Adults spend more
and more of their
time thinking, and
behaving, like high school students. In a
new book—called, yes, Popular—psycholo-
gist Mitch Prinstein explores popularity
with a scientist’s eye. Prinstein, a professor
at the University of North Carolina at Chap-
el Hill, argues that there are, in fact, two
types of popularity and that we as a culture
have settled on the more dysfunctional
type. There is, he says, a better way. Prin-
stein answered questions from Mind Mat-
ters editor Gareth Cook.
Scientific American: Does our popularity
in high school affect us later in life?
Prinstein: It does. Quite remarkably, in fact.
Research findings suggest that even 40 years
later, we can predict who will graduate from
high school or college, who will succeed at
work, who will apply for welfare or social
Gareth Cook is a Pulitzer Prize–winning journalist in Boston.
He edits Mind Matters, an online commentary blog at
www.ScientificAmerican.com/mind-matters
services, and who may suffer from debilitat-
ing mental health difficulties or addictions
all by knowing how popular folks were in
high school. Our popularity even predicts
our physical health—those who were least
popular in childhood are more likely to have
cardiovascular and metabolic illnesses de-
cades later than those who were well liked.
One analysis suggested that the risks of un-
popularity on our mortality are as strong as
the risks that come from smoking!
SOMER HADLEY REVOLUTION STUDIOS
What may be most surprising, however, is
that our popularity plays a role that cannot
be accounted for by our socioeconomic sta-
tus, IQ, family background, prior mental
health difficulties or appearance. There’s
something about the way we are regarded Psychologist Mitch Prinstein
by others that changes our life trajectories
quite meaningfully and substantially.
What do you mean by “popularity”?
This is an important question because most
do not realize that scientists have identi-
fied two different types of popularity, each
associated with very different outcomes. In
childhood, our popularity is defined by how
much we are liked by others. The most pop-
ular kids are those who lead quietly, help
others and cooperate, and this type of pop-
57
ularity predicts many desirable long-term
outcomes.
A second type of popularity emerges in
adolescence, however, reflecting changes
in our neural circuitry that are triggered by
pubertal hormones. This is the period when
popularity begins to reflect our “status”
more than our “likability.” The markers of
status—visibility, influence, dominance
and power—all activate the social reward
centers in our brain and change our rela-
tionship with popularity forever. Through-
out adulthood, we have a choice to pursue
greater likability or greater status—a deci-
sion made so much more difficult by the
growing number of platforms (reality TV,
social media, et cetera) designed to help us
gain status. In fact, our focus on easily ob-
tained status now is perhaps stronger than
at any other point in human history. That’s
a problem, however. Because unlike the
positive outcomes associated with high lik-
ability, research findings indicate that hav-
ing high status leads to later aggression,
addiction, hatred and despair.
Can you tell me more about the prob-
lems you see with pursuing status?
There are problems for individuals who
pursue status and problems for us as a soci-
ety that has continued to invest in this type
of popularity. Research findings, such as
those published by University of Virginia
professor Joe Allen, indicate that those who
care the most about their social standing
grow up to have difficulties with their in-
terpersonal relationships years later. They
remain fixated on their status and even on
others’ popularity rather than on attributes
that may lead to fulfilling human connec-
tion. Other research suggests that those
who wish the most for status are most like-
ly to later report anxiety, depression and
problems with substance use.
As a society, there already are signs that
our desire for status has taken a toll. Com-
pared with just a few decades ago, research
suggests that our life goals now reflect
wishes to own more possessions, acquire
more power, and feel more visible and in-
fluential than others. This is in stark con-
trast to our desire to foster community and
cooperation just a couple of decades ago.
Even our children are getting the message
that the number of their social media fol-
lowers is an accomplishment worth striv-
ing for. Yet ironically, the more we seek
these online markers of status—retweets,
“likes,” shares—the more we feel segment-
ed and disconnected from one another.
Were you popular as a kid?
Not really. In fact, at 4’7” in 10th grade,
with a perfect attendance record since kin-
dergarten, I would have easily served as the
poster boy for the nerds! If we think of the
most popular as those who had the highest
“status,” I wasn’t popular at all.
But I think I was likable. And it turns out
that this was more important.
How so? How has this shaped your life?
Studies show that people who are likable
are afforded privileges that become rein-
forcing and self-perpetuating. As children,
those who are liked are invited to join oth-
ers more often, and each of these interac-
tions offers extra opportunities to learn
skills that were denied to unlikable, exclud-
ed peers. Over time these skills lead to even
greater likability, additional learning occa-
sions, and so on, creating a cycle that leads
likable people to not only enjoy more posi-
tive reputations but truly advanced skills
and competencies.
The same is true for adults. Two equally
qualified employees, for instance, will get
58
further in their career if they differ in lik- however. For some, simply reading the book Like us on Facebook
ability—not solely because of favoritism and recognizing the distinct outcomes as-
but because the more likable person will sociated with likability versus status will
actually become the better employee. offer a sufficient opportunity to reflect on
I was lucky. A good sense of humor about past behavior and think about instincts
my physical differences in adolescence more critically. Once people realize how
helped me to avoid most egregious instanc- clearly science has linked status with nega- facebook.com/ScientificAmerican
es of bullying and form friendships with tive outcomes, convincing folks to return
peers that guided me toward good deci- to a focus on likability may be easy. For oth-
sions and new opportunities. I wasn’t ers, it may take a bit more, however. With-
“cool,” but I was able to enter whatever out knowing who would be elected last No-
context I needed and learn skills that I still vember, the book now seems to offer a pro-
draw on to this day. phetic message to our president—perhaps
the best illustration of an insatiable appe-
Are there ways to convince people tite for status that has become distracting,
to switch from pursuing status to the unsatisfying and even dangerous.  M
“likable” kind of popularity?
I hope this book will offer folks an opportu-
nity to acknowledge that we all have hu-
man instincts to seek popularity. It may
feel juvenile, reflecting the time in our lives
when popularity seemed paramount. But
there’s remarkable evidence that populari-
ty still matters today, and it guides our be-
havior and our desires as much as it did
back then. Our longing for popularity is lit-
erally a part of our DNA.
As adults, we have the option of choos-
ing the type of popularity we strive for,


 OPINION
Is the U.S.
Education System
Producing a
Society of
“Smart Fools”?
One distinguished psychologist
explains why he believes this
is so and how to reverse course

MICHAELQUIRK GET T Y IMAGES

By Claudia Wallis
60
A
t this year’s annual meeting
of the Association for Psy-
chological Science (APS) in
Boston, Cornell University
psychologist Robert Stern-
berg sounded an alarm about the influence
of standardized tests on American society.
Sternberg, who has studied intelligence
and intelligence testing for decades, is well
known for his “triarchic theory of intelli-
gence,” which identifies three kinds of
smarts: the analytic type reflected in IQ
scores; practical intelligence, which is more
relevant for real-life problem solving; and
creativity. He offered his views in a lecture
associated with receiving a William James
Fellow Award from the APS for his lifetime
contributions to psychology. He explained
his concerns to Scientific American. Edited
excerpts follow.
Scientific American: In your talk, you
said that IQ tests and college entrance
exams such as the SAT and ACT are
essentially selecting and rewarding
“smart fools”—people who have a
certain kind of intelligence but not the
kind that can help our society make
progress against our biggest challenges.
What are these tests getting wrong?
Sternberg: Tests like the SAT, ACT, the
GRE—what I call the alphabet tests—are
reasonably good measures of academic
kinds of knowledge, plus general intelli-
gence and related skills. They are highly
correlated with IQ tests, and they predict a
lot of things in life: academic performance
to some extent, salary, level of job you will
reach to a minor extent—but they are very
limited. What I suggested in my talk today
is that they may actually be hurting us. Our
overemphasis on narrow academic skills—
the kinds that get you high grades in
school—can be a bad thing for several rea-
sons. You end up with people who are good
at taking tests and fiddling with phones
and computers, and those are good skills,
but they are not tantamount to the skills
we need to make the world a better place.
What evidence do you see of this harm?
IQ rose 30 points in the 20th century around
the world, and in the U.S., that increase is
continuing. That’s huge; that’s two standard
deviations, which is like the difference be-
tween an average IQ of 100 and a gifted IQ
of 130. We should be happy about this, but
the question I ask is, If you look at the prob-
lems we have in the world today—climate
change, income disparities in this country
that probably rival or exceed those of the
gilded age, pollution, violence, a political
situation that many of us never could have
imagined—one wonders, What about all
those IQ points? Why aren’t they helping?
What I argue is that intelligence that’s
not modulated and moderated by creativi-
ty, common sense and wisdom is not such
a positive thing to have. What it leads to is
people who are very good at advancing
themselves, often at other people’s ex-
pense. We may not just be selecting the
wrong people; we may be developing an
incomplete set of skills—and we need to
look at things that will make the world a
better place.
Do we know how to cultivate wisdom?
Yes, we do. A whole bunch of my colleagues
and I study wisdom. Wisdom is about using
your abilities and knowledge not just for
Claudia Wallis is an award-winning science journalist whose work
has appeared in Time magazine, Fortune, the New York Times and
other national publications. She is the former managing editor of
Scientific American Mind and former science editor of Time.
61
your own selfish ends and for people like
you. It’s about using them to help achieve a
common good by balancing your own in-
terests with other people’s and with
high-order interests through the infusion
of positive ethical values.
You know, it’s easy to think of smart peo-
ple, but it’s really hard to think of wise peo-
ple. I think a reason is that we don’t try to
develop wisdom in our schools. And we
don’t test for it, so there’s no incentive for
schools to pay attention.
Can we test for wisdom,
and can we teach it?
You learn wisdom through role modeling.
You can start learning that when you are
six or seven. But if you start learning what
our schools are teaching, which is how to
prepare for the next statewide mastery
tests, it crowds out of the curriculum the
things that used to be essential. If you look
at the old McGuffey Readers, they were as
much about teaching good values and good
ethics and good citizenship as about teach-
ing reading. It’s not so much about teach-
ing what to do but how to reason ethically;
to go through an ethical problem and ask,
“How do I arrive at the right solution?”
I don’t always think about putting
ethics and reasoning together. What
do you mean by that?
Basically, ethical reasoning involves eight
steps: seeing that there’s a problem to deal
with (say, you see your roommate cheat on
an assignment); identifying it as an ethical
problem; seeing it as a large enough prob-
lem to be worth your attention (it’s not like
he’s just one mile over the speed limit);
seeing it as personally relevant; thinking
about what ethical rules apply; thinking
about how to apply them; thinking, “What
are the consequences of acting ethical-
ly?”—because people who act ethically usu-
ally don’t get rewarded; and, finally, acting.
What I’ve argued is that ethical reasoning
is really hard. Most people don’t make it
through all eight steps.
If ethical reasoning is inherently
hard, is there really less of it and less
wisdom now than in the past?
We have a guy [Greg Gianforte of Montana]
who allegedly assaulted a reporter and was
then elected to the U.S. House of Represen-
tatives—and that’s after a 30-point average
increase in IQ. We had violence in cam-
paign rallies. Not only do we not encourage
creativity, common sense and wisdom, I
think a lot of us don’t even value them any-
more. They’re so distant from what’s being
taught in schools. Even in a lot of religious
institutions, we’ve seen a lot of ethical and
legal problems arise. So if you’re not learn-
ing these skills in school or through reli-
gion or your parents, where are you going
to learn them? We get people who view the
world as being about people like them-
selves. We get this kind of tribalism.
So where do you see the
possibility of pushing back?
If we start testing for these broader kinds
of skills, schools will start to teach to them
because they teach to the test. My col-
leagues and I developed assessments for
creativity, common sense and wisdom. We
did this with the Rainbow Project, which
was sort of experimental when I was at
Yale University. And then at Tufts Univer-
sity, when I was dean of arts and sciences,
we started Kaleidoscope, which has been
used with tens of thousands of kids for ad-
mission to Tufts. They are still using it.
But it’s very hard to get institutions to
change. It’s not a quick fix. Once you have
a system in place, the people who benefit
62
from it rise to the top, and then they work
very hard to keep it.
Looking at the broader types of admis-
sion tests you helped to implement—
such as Kaleidoscope at Tufts, the
Rainbow Project at Yale or Panorama
at Oklahoma State University, is there
any evidence that kids selected for
having these broader skills are in any
way different from those who just
score high on the SAT?
The newly selected kids were different. I
think the folks in admissions would say so,
at least when we started. We admitted kids
who would not have gotten in under the old
system—maybe they didn’t quite have the
test scores or grades. When I talk about
this, I give examples, such as those who
wrote really creative essays.
Has there been any longitudinal
follow-up of these kids?
We followed them through the first year of
college. With Rainbow, we doubled predic-
tion [accuracy] for academic performance,
and with Kaleidoscope, we could predict
the quality of extracurricular performance,
which the SAT doesn’t do.
Do you think the emphasis on narrow
Follow us on Instagram
measures like the SAT or GRE is
hurting the STEM fields in particular?
I think it is. I think it’s hurting everything.
We get scientists who are very good for- @scientific_american
ward incrementers—they are good at doing instagram.com/scientific_american
the next step, but they are not the people
who change the field. They are not redirec-
tors or reinitiators, who start a field over.
And those are the people we need.
Are you hopeful about change?
If one could convince even a few universi-
ties and schools to try to follow a different
direction, others might follow. If you start
encouraging a creative attitude, to defy the
crowd and to defy the zeitgeist and if you
teach people to think for themselves and
how what they do affects others, I think it’s
a no-lose proposition. And these things can
be taught, and they can be tested.  M

 OPINION
Can Stimulating a
Nerve in the Ear
Make You a Whiz in
Mandarin Class?
A technology program
tries to enhance the brain´s
learning ability

ALFRED PASIEKA GET T Y IMAGES

By Gary Stix
64
I
n April the Defense Advanced Re-
search Projects Agency (DARPA)
announced contracts for a pro-
gram to develop practical methods
to help someone learn more quick-
ly. In the ensuing press coverage, the en-
deavor drew immediate comparisons to The
Matrix—in which Neo, Keanu Reeves’s char-
acter, has his brain reprogrammed so that
he instantly masters kung fu.
DARPA is known for setting ambitious
goals for its technology development pro-
grams. But it is not requiring contractors
for the $60-million, four-year effort to find
a way to let a special-forces soldier upload
neural codes to instantaneously execute a
flawless wushu butterfly kick.
The agency did award contracts, though,
to find some means of zapping nerves in
the peripheral nervous system outside the
brain to speed the rate at which a foreign
language can be learned by as much as 30
percent, a still not too shabby goal.
Sending an electric current into the va-
gus nerve in the neck from a surgically im-
planted device is already approved for treat-
ing epilepsy and depression. The DARPA
program, in tacit acknowledgment of the
fact that mandatory surgery might be un-
acceptable for students contemplating an
accelerated Mandarin class, wants to devel-
op a noninvasive device to stimulate a pe-
ripheral nerve, perhaps in the ear. The goal
is not just to hasten the learning of foreign
languages but also to facilitate pattern-rec-
ognition tasks, such as combing through
surveillance imagery.
Learning, of course, has nothing to do
with laying down bits on a formatted mag-
netic disk. Sensory inputs to the eyes or
ears set off a cascade of events that relays
signals to various regions of the vast neural
substrate of the brain’s some 86 billion
neurons.
The first part of DARPA’s Targeted Neu-
roplasticity Training (TNT) program will
consist of basic research. It will try to con-
firm that coursing an electric current
through a peripheral nerve outside the cen-
tral nervous system will eventually enable
someone to perform a task such as distin-
guishing more rapidly whether the word
ma in Mandarin has several meanings, such
as mother (mā) or horse (mă), depending
on the tonal inflections.
DARPA decided learning enhancement
was worth pursuing because of published
research that has shown that peripheral
nerve stimulation can result in the produc-
tion of signaling chemicals—neuromodu-
lators such as norepinephrine—that aid in
the rewiring that takes place during learn-
ing—“synaptic plasticity” to invoke the te-
chie term of art.
I asked Doug Weber, who heads the TNT
program at DARPA, about the difference
between using a nerve-zapping device and
taking modafinil, Ritalin or any other phar-
maceutical compound that purportedly ex-
hibits cognitive-enhancing properties.
“The thing you can’t do easily with a pill
that you can do very easily with neurostim-
ulation is control timing and dosing,” We-
ber says. Swallowing the pill might ratchet
up levels of acetylcholine, norepinephrine
or other signaling molecules after they go
through your system, but ingestion cannot
be paired to the precise moment that a ma
tone sounds.
Besides a cultural nod to The Matrix,
TNT also evokes the current tech fad for
neurohacking—downing a pill or donning
headgear to improve cognition. If the on-
line do-it-yourself crowd is to be believed,
zapping the brain to help language learn-
ing has already gotten off to a desultory
start. Try searching for the keywords “tDCS”
65
and “language learning.” Then look for re-
sults such as a Reddit post entitled “Accel-
erating Foreign Language Learning with
TDS.” Also, don’t miss “Did tDCS Give Me
Dyslexia?”
Purchased or do-it-yourself tDCS (trans­
cranial direct-current stimulation) is the
fad du jour for neurohackers. It channels a
small electric current into the brain through
two electrodes placed on the scalp. Many of
the posts about tDCS are devoted more to
the proper electrode placement on the
scalp than the subtle differences of of mā
versus mă.
Skepticism also pervades the published
literature. A study published online on May
16 in Psychological Science showed little
benefit from adding tDCS to a cogni-
tive-training program for older adults. We-
ber says the TNT program intentionally
avoided tDCS:
You’re applying this very small cur-
rent to the scalp, and if you look at it from
the standpoint of the path of least resis-
tance, it means that most of it is going to
be shunted through the scalp, and so
your skin is going to absorb most of the
charge. Very little, if any, will get through
the skull because the skull is thick, and
it’s a high-impedance barrier. And what-
ever current does trickle through the
skull, which again is very little, will again
be shunted through by the CSF [cerebro-
spinal fluid] and other fluid that sits be-
tween the neurons of the brain and the
skull. So just from a pure physics stand-
point, it’s magical to me that there’s any
effect, whereas with peripheral nerves,
we know exactly what the direct effects of
stimulation are.
Even so, the researchers must still figure
out what happens during the remapping of
the neural connections that occur when we
learn something—no small task. Detecting
when signaling molecule levels increase—a
sign that nerve stimulation may be having
the desired effect—must be done by exam-
ining changes in pupil size and other indi-
rect measures.
The closer one looks at what happens
during the learning process, the more hu-
mility is in store. In May, I posted an inter-
view with Falk Huettig of the Max Planck
Institute for Psycholinguistics, a research-
er who had just published a study about the
enormous changes that occur in the brain
when an illiterate adult learns to read. In
addition to the questions Scientific Ameri-
can published online, I asked Huettig
whether he thought a simple pulsing of
electric current through a peripheral nerve
would suffice to speed learning. “The hy-
pothesis that vagal and trigeminal nerve
stimulation allows people to learn foreign
languages faster is certainly interesting but
very much a long shot,” he told me.
Weber says the decision was made by
DARPA to focus on a few areas, such as
learning foreign languages, because pe-
ripheral nerve stimulation is thought to be
a good candidate for enhancing the ability
to discriminate sounds—again, the má ver-
sus mà challenge. He readily acknowledges
that much is still unknown. It might be nec-
essary, for instance, to take steps to ensure
that the memory of a new skill is not quick-
ly forgotten, perhaps by enhancing sleep in
some way. But DARPA’s ambitions could
still yield payoffs, Weber notes: “Any par-
tial success in this program will have a high
impact if we learn something new about
how the brain adapts, acquires and refines
cognitive skills.”  M
 66
 OPINION
Why You Shouldn´t
Tell People about
Your Dreams
They are really meaningful to
you but not to anybody else

ORLA GET T Y IMAGES

By Jim Davies
67
I
had a friend who tried hard to remem-
ber more of her dreams. She would
write them down and then tell people
about them. She stopped, though, be-
cause it started interfering with her
social life. She would start talking about her
dreams, and people would leave the room.
There are several major theories about
why we dream. One is the activation-syn-
thesis theory, which holds that dreams are
interpretations by our forebrain of essen-
tially random activity from the spinal cord
and cerebellum during sleep, especially
rapid eye movement (REM) sleep.
Part of the explanation for why dreams
can be so weird is that they are interpreted
from chaotic information. The evolution-
arily older parts of our brain are also the
seat of our basic emotions. According to
this theory, the emotion comes first, and
dreams are made to make sense of that
emotion. Evidence for this position comes
from scene changes that happen: when we
have anxiety dreams, for example, they of-
ten switch from one anxious situation to a
different one—so rather than us feeling
anxious because of the content of our
dream, it could be that our feeling is caus-
ing an anxious narrative in the dream!
Another major theory of dreaming is
threat-simulation theory, which holds that
the evolutionary function of dreaming is
for us to practice how to behave in threat-
ening situations. There is a lot of evidence
for this theory, too.
First, most dream emotion is negative.
Also, people tend to dream of ancestral
threats: falling, being chased, natural di-
sasters, and so on. These frightening ele-
ments are overrepresented in dreams—that
is, we see them in dreams much more than
our experience in our day-to-day world
would predict. Many people dream of being
chased by animals, but how often does this
actually happen to people? The overrepre-
sentation of animals chasing us in dreams,
especially for children, suggests that we
have some innate fear of them. In contrast,
we do not dream of modern threats, such as
heart attacks, as much as we would expect
if dreams were based on the problems we
actually face in today’s world.
These two theories of dreaming are of-
ten presented as competing, but as far as I
can tell, they are compatible—that is, even
if dreams are interpretations of chaotic in-
put from the spinal cord, there is still a the-
ory needed to describe how that chaotic in-
put is elaborated into narratives that we
experience as dreams, and it is quite possi-
ble that the mind takes advantage of this
opportunity to practice dealing with dan-
gerous things.
Why do we feel the urge to talk about
our dreams? A suggested ramification of
threat-simulation theory relates to the idea
that “two heads are better than one”: dis-
cussion of dreams might be adaptive if they
help us mentally prepare for threats. We
like to talk about dreams to help us prepare
for how to act in dangerous situations in
the future.
Which leads us to why we find our own
dreams so interesting. There are three rea-
sons, based on known psychological ef-
fects, although all are speculative, in terms
of my application of them to dreams.
The first is negativity bias, which makes
us pay attention to dangerous things.
Because most dreams are negative (sup-
port for the threat-simulation theory), our
bias in favor of negative information makes
them feel important.
The second reason has to do with the
emotional primacy of dreaming—because
so many dreams are so emotional, they feel
important in a way that people hearing
68
about them, not feeling that emotion, might
find hard to relate to. Once I dreamed of a Digital Matter
About Your Gray Matter
terrifying staircase. When I told my girl-
friend about it, she laughed at me for being
scared of such a harmless thing. In the
dream, it was scary, but clearly my audi-
ence could not appreciate that.
We tend to think of dreams as being re-
ally weird, but in truth, about 80 percent of

eBooks
dreams depict ordinary situations. We’re
just more likely to remember and talk about
the strange ones. Information we do not
understand can often rouse our curiosity,
particularly in the presence of strong emo- In-depth Coverage on
tion. Just like someone having a psychotic Neurology, Consciousness,
Behavior and More
experience, the emotional pull of dreams
makes even the strangest incongruities
seem meaningful and worthy of discussion
and interpretation.
These reasons are why most of your BUY NOW
dreams are going to seem pretty boring to
most people. But if you’re going to talk
about some of your dreams, pick the ones
in which you deal with a problem in some
new way. The negativity bias would make
them more interesting than your happy
dreams, and if you feel that you learned
something about how to deal with a threat,
maybe your audience will, too.  M



Editor in Chief and Senior Vice President: Mariette DiChristina
Digital Content Manager: Curtis Brainard
Collections Editor: Andrea Gawrylewski
Chief Features Editor: Seth Fletcher
Chief News Editor: Dean Visser
Chief Opinions Editor: Michael D. Lemonick
Senior Editor: Gary Stix
Creative Director: Michael Mrak
Art Directors: Marc Grove, Ryan Reid
Photography Editor: Monica Bradley
Assistant Photo Editor: Liz Tormes
Copy Director: Maria-Christina Keller
Senior Copy Editors: Michael Battaglia, Daniel C. Schlenoff
Copy Editor: Aaron Shattuck
Prepress and Quality Manager: Silvia De Santis
Senior Editorial Product Manager: Angela Cesaro
Technical Lead: Nicholas Sollecito
Senior Web Producer: Ian Kelly
Web Production Associate: Eli Rosenberg
Editorial Administrator: Ericka Skirpan
Senior Secretary: Maya Harty
President: Dean Sanderson
Executive Vice President: Michael Florek
Head, Marketing and Product Management: Richard Zinken
Associate Consumer Marketing Director: Catherine Bussey
Senior Consumer Marketing Manager: Lou Simone
Marketing Manager: Marie Cummings
Marketing and Customer Service Coordinator: Christine Kaelin
Rights and Permissions Manager: Felicia Ruocco
Head of Communications, USA: Rachel Scheer
Some of the articles in Scientific American Mind originally appeared in our
German edition Gehirn&Geist (Editor in Chief: Carsten Könneker).
Get the most out of your
Mind digital subscription.
Keep up with the latest discoveries
LETTERS TO THE EDITOR: related to the brain and human behavior
anytime, anywhere
Scientific American Mind, 1 New York Plaza, Suite 4500, New York, NY
10004-1562, 212-451-8200 or editors@sciam.com. Letters may be
edited for length and clarity. We regret that we cannot answer each one.
HOW TO CONTACT US: EXPLORE NOW
For Advertising Inquiries: Scientific American Mind, 1 New York Plaza, Suite
4500, New York, NY 10004-1562, 212-451-8893, fax: 212-754-1138
For Subscription Inquiries: U.S and Canada: 888-262-5144, Outside North
America:, Scientific American Mind, PO Box 5715, Harlan IA 51593,
515-248-7684, www.ScientificAmerican.com/Mind
To Order Reprints: Reprint Department, Scientific American Mind, 1 New
York Plaza, Suite 4500, New York, NY 10004-1562, 212-451-8877, fax:
212-451-8252, reprints@sciam.com
For Permission to Copy or Reuse Material From SciAmMIND: Permissions
Department, Scientific American Mind, 1 New York Plaza, Suite 4500,
New York, NY 10004-1562, 212-451-8546, www.ScientificAmerican.com/
permissions. Please allow three to six weeks for processing.
Scientific American Mind (ISSN 1555-2284), Volume 28, Number 5,
September/October 2017, published bimonthly by Scientific American, a
divisionof Nature America, Inc., 1 New York Plaza, Suite 4500, New York,
N.Y. 10004-1562. Subscription rates: one year (six issues), U.S. $29.99;
Canada, $34.99 USD; elsewhere, $39.99 USD. For subscription inquiries,
call (888) 262-5144. To purchase back issues, call (800) 925-0788.
Printed in U.S.A. Copyright © 2017 by Scientific American, a division of
Nature America, Inc. All rights reserved.
Scientific American is part of Springer Nature, which owns or has
commercial relations with thousands of scientific publications
(many of them can be found at www.springernature.com/us).
Scientific American Mind maintains a strict policy of editorial
independencein reporting developments in science to our readers.
Springer Nature remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.