Professional Documents
Culture Documents
M
y nine-month-old daughter, Ellie, is a statistics genius. arena that’s been investigated, babies and children are remarkably
This may sound like typical new mom bragging, but adept at learning. But while they may be excellent at figuring out
it’s not; it’s scientific fact. According to research from the world around them, it’s still unclear exactly how much they
Professor Fei Xu’s Infant Cognition and Language know, and when, and what mechanisms are in place to allow this
Lab in the Department of Psychology at UC Berkeley, the average rapid learning.
six-month-old is pretty good at making basic estimates of probability, Armed with colored ping pong balls, light-up lollipops, stuffed
and by the time they learn to walk—around a year old—most babies animals, and invented words, researchers in the Xu lab are making
are experts. Children are also masters of language acquisition, pat- strides toward answering these questions. The answers they find may
tern recognition, and inductive reasoning. In fact, in almost every have applications in fields from parenting to computer programming.
Features Baby lab
walking
rolling over crawling running
Months 6 9 12 24 36 48 60
What’s in a name? but may not always come in the same color “daxes,” while animals with a different set
It’s the oldest debate in developmental or size. of marks on these two appendages might be
research: do we learn to learn, or are we These biases have long been thought to “blickets.” Here, the usual biases aren’t help-
simply biologically programmed to soak be innate, since they arise so early in devel- ful, so children have to learn not only what
up information from our surroundings and opment and are so universal. Intriguingly, each animal is called (the hypothesis), but at
experiences? The obvious answer, of course, though, children seem to weigh information the same time figure out a weird new rule
is that it’s probably a little bit of both, but the differently depending on the type of object that governs how these animals are named
precise location of the boundary between that’s being defined (color, for example, is (the overhypothesis).
nature and nurture is a matter of intense more important when learning the names By manipulating this basic experimental
debate. of foods, while texture becomes important setup, researchers can ask what variables
Postdoctoral researcher Sylvia Yuan when learning the names of animals), sug- affect how children form overhypotheses.
is investigating this boundary by studying gesting that experience might play a role in One key finding has been that the number
word learning in toddlers. Previous research bias formation. of categories presented seems to be more
has suggested that by the age of two or so, This process of bias-building is referred important than the number of examples per
children have a number of cognitive biases to as overhypothesis formation. “As they’re category. Preschoolers shown eight animals
that help them solve the nearly impossible learning about each word,” Yuan explains, are more easily able to classify them if there
logic problem of what words mean. “Even “they might be testing in their head: is it the are four categories with two animals each
if we explicitly label something, like, ‘this texture, is it the color, is it the shape? And (two daxes, two blickets, two faps, two zoogs)
is a Lego’,” Yuan says, holding up a yellow when they see another example, they might than if there are two categories with four
block, “there are lots of logical possibilities be thinking, ‘okay, it doesn’t seem like it’s the animals each (four daxes and four blick-
as to what the word could be referring to. It texture, it seems like it’s more the shape.’” As ets). This suggests that each new category
could be referring to the color yellow, it could children form a hypothesis about what each a child sees either strengthens or changes
be referring to something hard, it could be object is called, they’re also forming a more her overhypothesis about how categories
referring to something on my hand.” abstract rule, or overhypothesis, that defines are defined in general, indicating that it’s a
But kids don’t run through all of these how object names are assigned in general. dynamic process.
possibilities every time they learn a new word. Yuan and others in the lab are trying to Yuan plans to use these initial studies as
Instead, they have a number of biases that determine what factors affect overhypothesis a launching point to investigate how other
help them narrow in on the right definition formation when children are acquiring new factors, like adding noise by varying the
fairly quickly. For example, children tend to vocabulary. Instead of examining established sizes or shapes of the items, or introducing
assume that a word refers to an entire object, biases, like shape, the lab introduces artificial exceptions to the rules, affect overhypothesis
rather than just part of it—“car” describes categories so they can study the overhy- formation. Increasing the number of non-
the whole vehicle, not just its hood. Similarly, pothesis formation process as it happens. informative marks, for example, could go
they assume that labels apply to the shape of In one typical experiment, preschoolers must either way—it might help children focus in on
the object, rather than another character- figure out that markings on the tail and left the actually useful information more quickly,
istic, like color or texture. This is a useful foot of otherwise identical stuffed animals or it might just confuse them. “We’re trying
Marek jakubowski
assumption, because objects are likely to determine their identity—for example, ones to figure out what the environmental inputs
have a stereotypical shape (a “ball” is usually with a question mark on the tail and an are that make it easier or harder for them to
ball-shaped, a “cup” is usually cup-shaped) exclamation point on the left foot might be achieve an overhypothesis,” Yuan says.
Another day
Behind door number three... is used to researchers’ advantage Infant Cognit at the
ion and Lang
This sort of “smart” mechanism that allows in the classic “looking time/viola- We get the be uage Lab...
st one-liners fro
children to draw up broad, organizing prin- tion of expectation” measure, a I had this pres m preschoolers
chooler today, .
ciples based on a small number of examples well-established test for deter- So, why do yo actually. I said
u think that th ...
is crucial for learning. Without an efficient mining what babies are able to Oh. I know LO at ’s how this w
TS of things. orks?
way to generalize the knowledge gained from predict. Because babies spend Oh, you do?
one experience and apply it to another, it a longer time looking at things I do. I really kn
ow a lot of thin
She’s barely fo gs.
would simply take too long for kids to figure that are novel or surprising, ur...
out how the world works. (And as anyone an infant’s looking behavior -Stephanie Den
ison, PhD Studen
who has ever watched a child repeatedly test can be measured to provide a t
gravity with the food items on her high chair metric of whether he finds an
tray will tell you, it takes long enough as it is.) event expected or unexpected.
But until recently, it hasn’t been clear whether Measuring looking time was crucial for These experiments make it clear that
overhypothesis formation is limited to word the Xu lab’s studies of overhypothesis forma- infants can recognize patterns very quickly
learning, or when this skill first arises. tion in infants (as opposed to the toddlers in and use them to make generalizations at a very
“Our working hypothesis is that there is the object naming study). In these experi- early age. According to Xu, that suggests the
a set of learning mechanisms in children that ments, nine-month-olds watched while a presence of a powerful learning mechanism
support rapid learning,” says principal inves- researcher removed objects from various that might underlie many different biases
tigator Fei Xu. She and others have predicted boxes. The first few boxes contained objects that were previously thought to be innate.
that even babies less than one year old might of the same shape, but of different colors Of course, it remains to be seen whether
be able to form overhypotheses, but probing and sizes. Then, surprise! The final box con- this mechanism is itself learned—are there
infant psychology can be difficult. Simply tained, say, a star and a circle. If the babies over-overhypotheses to be discovered? In the
working with babies can be a challenge in and had formed an overhypothesis based on future, comparisons between overhypothesis
of itself. After all, there aren’t many fields their previous experience—“boxes contain formation in infants and toddlers may also
in which papers routinely include lines like, items with the same shape”—they should help illuminate how this process changes
“An additional four subjects were tested but have looked longer at this unexpected event. with age. If pattern recognition is something
excluded due to fussiness.” Entertainingly, And, indeed, this was the case. Importantly, that improves with practice, it’s possible that
researchers report that fussiness isn’t as babies formed overhypotheses equally well young babies will have a harder time with
much of a problem as bodily functions. when the items in each box were all of the confusing cases than more experienced
Stephanie Denison, a graduate student in same color but different shapes, showing that toddlers and children; on the other hand,
the lab, puts it delicately: “Occasionally they this learning mechanism is general and not, it’s also possible that the younger subjects
get distracted by... digestion during the trial.” say, the manifestation of an innate shape bias. may actually have an easier time because
Yuan elaborates, “We would have observ-
ers write down, for example, ‘face is all red
and squinty’... the kids sort of stop looking
at what’s going on on the stage and in the
trial.” Distractability can also be problematic.
“One little one just pulled off her socks in the
middle of it. There’s a foot flying over there,
a foot flying over here,” lab manager Christie
Reed recounts. And the occasional baby will
fall asleep during a study, too.
It’s also tough to find experimental
methodologies that can truly illuminate
infant cognition. “Smart as infants are, it
is hard to work with them, since they do
not yet talk or follow instructions,” Xu says.
Researchers can’t just ask very young babies
what they’re thinking—they have to figure
it out in some other way. “We often capital-
Jacqueline Chretien
a
ink your game is
w ill le t yo u kn ow that they th
Kids say...
ey’ll sigh, they’ll me.
little boring. Th ? Th is ga m e’s too easy for
al m os t do ne
Are we t they say...
lly bombing, bu
And they’re tota just so you know.
ar t for this,
I’m WAY too sm n, PhD Student
-Stephanie Deniso
of pink to yellow ping pong balls. (Other is blindfolded the 11-month olds know to to probe more deeply into how infants arrive
versions of this study have used red, white or disregard the researcher’s preference and at these estimates. Humans are notorious for
green balls—colors selected “entirely based expect a representative sample. “This tells failing to evaluate probabilities accurately,
depending on a variety of external factors task, and the baby should be so engaged, and to create models of reasoning, learning, and
(the 10 percent of American homeowners they’re like, ‘hmm, I think I’m just going language acquisition that inform artificial
who are underwater on their mortgages can to go see what’s over there on the door,’” intelligence and natural language process-
tell you about the perils of optimism bias), Denison says. Short attention spans have also ing. There are also applications in clinical
but not much is known about why we make complicated the experiment in cuter ways. psychology. Infants’ performance in basic
the mistakes we do. Studying whether babies Some babies, when asked to select a lollipop, cognitive tasks like these is increasingly
are susceptible to the same kinds of errors as choose to hug the experimenter instead. understood to be correlated with their abili-
adults may help us solve this cognitive puzzle. So, the experimenters try to make the ties later in life, so more detailed knowledge
single preference trial really count. After of typical development may make it easier to
Come on over, baby the selection is made, the researchers add identify atypical development at very early
While looking time is a well-respected and some positive reinforcement, clapping and stages, when interventions would be the most
frequently used experimental measure, it generally encouraging the baby to feel that effective.
can be tricky in practice. Looking behavior she’s made a truly excellent choice. (This Normally developing children can ben-
can be affected by many different factors, all technique will also be familiar to anyone efit from new insights, too. Knowing when
of which need to be controlled. Even then, who has ever tried to convince a skeptical specific cognitive skills are emerging can
the difference between the reactions to an baby that she likes the new vegetable she help parents and educators engage with these
expected and unexpected condition may only just tried.) And at the end of the experiment, processes and give children richer learning
be a few seconds. And of course, no matter babies shown jars containing all environments. “I feel that if you’re aware
how good the assay, it’s always nice to have pink or all black lollipops usu-
a complementary experiment, particularly ally head for the preferred color,
one that’s very different in approach. suggesting that the preference Another day
at the ICL Lab
With this in mind, the Xu lab has devel- is consistent throughout the The cutest thin ...
g is seeing th
th ey kind of squi ei r lit tle faces when
oped a novel, active, and frankly just darn experiment. nt, or they’ll be
see they’re thin perplexed. You
cool measure of babies’ thought processes: The researchers also aren’t king, they’re lo can
tell that in thei oking.... you ca
crawling toward a hidden lollipop. For this above using some tricks in an r little minds, th n just
e gears are go
assay, babies are first offered a black lollipop effort to achieve uniform color - Christie Reed, ing.
Lab manager
and a pink lollipop; whichever one they reach preference. “The pink one
for or crawl to is established as the preferred lights up now, which has made
color. Once a preference is determined, the it much, much, much easier to get basically
babies are shown two boxes containing oppo- all the babies to prefer pink,” Denison says. of this sort of thing, that could make you
site ratios of pink to black, and the researcher So far, this new method has been used to interact with the baby differently, or maybe
removes one lollipop from each container show that babies not only understand which provide different kinds of stimulation,”
in such a way that the baby can only see the bin is more likely to yield a pink pop, but Yuan says. Personal experience bears this
stick. If this were a looking time experiment, that they can apply this understanding to out—now when I’m browsing at the toy store,
the lollipop’s color would be revealed and the guide their physical actions. Now, variations I’m on the lookout for games that will chal-
baby’s reaction would be monitored. In this on this setup can be used to pick apart any lenge Ellie’s probabilistic reasoning skills.
new measure, however, the baby is allowed to number of cognitive processes, including And when she approaches the age at which
crawl or walk to either cup to show that she overhypothesis formation. It’s also much language acquisition explodes on the scene,
knows which one is more likely to contain easier to apply an active measure like this I’ll be sure to rein in my use of expletives at
the preferred color. Eleven-month-olds pick to non-human animals (in fact, Denison just the right time.
the right cup about 70 to 80 percent of the and Xu originally came up with the idea as Though major questions still remain,
time, showing that they have a reasonably something that could be used with rhesus and the nature versus nurture debate is
firm grasp of single-event probability. macaques), and future comparative experi- increasingly thought of as something of a
As with any research with babies, how- ments are planned in monkeys and even straw man, work from the Xu lab and others
ever, the crawling measure has its fair share squirrels. These comparisons may help us in the field has certainly shown that babies
of difficulties. First, it’s difficult to be sure understand what makes human cognition are—as Denison puts it—“really, really
that the baby has a true preference for one so unique. smart.” As a doting mom, this just confirms
color. After all, babies can be fickle. The what I already believed, but as a scientist? It’s
initial experimental design called for four Lab to life nice to have some peer-reviewed citations to
preference trials, but the babies lost inter- The Xu lab’s insights into baby cognition are back me up.
est by the time the test trials rolled around, fascinating in their own right, but there are
making the results difficult to interpret. “It’s also practical applications for this research.
always funny when as a researcher you think Increasingly, computer scientists are col- Jacqueline Chretien is a graduate student in
you’re doing this really intelligent, wonderful laborating with developmental psychologists molecular and cell biology.