osq216Gpnk3p.indd 4 3/9/16 5:37 PM RAISE GREAT Thirty years ago most psychologists, KIDS philosophers and psychiatrists thought that babies and young children were ir- rational, egocentric and amoral. They believed children were locked in the con- crete here and now—unable to under- Even the stand cause and effect, imagine the ex- periences of other people, or appreciate youngest the difference between reality and fanta- sy. People still often think of children as children know, defective adults. experience and But in the past three decades scien- tists have discovered that even the young- learn far more est children know more than we would ever have thought possible. Moreover, than scientists studies suggest that children learn about the world in much the same way that sci- Photographs by ever thought entists do—by conducting experiments, Timothy analyzing statistics, and forming intuitive Archibald possible theories of the physical, biological and psychological realms. Since about 2000, researchers have started to understand the underlying computational, evolutionary and neurological mechanisms that under- pin these remarkable early abilities. These revolutionary ndings not only change our ideas about babies, they give us a fresh perspective on human nature itself.

PHYSICS FOR BABIES Why were we so wrong about babies for so long? If you look cursorily at children who are four years old and younger (the age range I will discuss in this article), you might indeed conclude that not much is going on. Babies, after all, cannot talk. And even preschoolers are not good at re- porting what they think. Ask your aver- age three-year-old an open-ended ques- tion, and you are likely to get a beautiful but incomprehensible stream-of-con- sciousness monologue. Earlier research- ers, such as the pioneering Swiss psychol- ogist , concluded that chil- dren’s thought itself was irrational and illogical, egocentric and “precausal”— with no concept of cause and effect. By The new science that began in the late Alison Gopnik 1970s depends on techniques that look at what babies and young children do in- stead of just what they say. Babies look longer at novel or unexpected events than

MIND.SCIENTIFICAMERICAN.COM 5

osq216Gpnk3p.indd 5 3/9/16 5:38 PM H O W B A B I E S THINK

at more predictable ones, and experi- lieve that animals and plants have an “es- can explain, for instance, if someone is menters can use this behavior to gure out sence”—an invisible core that stays the acting oddly because he or she believes what babies expect to happen. The stron- same even if outside appearances change. something that is not true. gest results, however, come from studies For babies and young children, the By the end of the 20th century exper- that observe actions as well: Which ob- most important knowledge of all is iments had thus charted impressively ab- jects do babies reach for or crawl to? knowledge of other people. Andrew N. stract and sophisticated knowledge in How do babies and young children imi- Meltzoff of the University of Washing- babies and the equally impressive growth tate the actions of people around them? ton showed that newborns already un- of that knowledge as children get older. Although very young children have derstand that people are special and will Some scientists have argued that babies a hard time telling us what they think, imitate their facial expressions. must be born knowing much of what we can use language in more subtle ways In 1996 Betty Repacholi (now at the adults know about how objects and peo- to tease out what they know. For exam- University of Washington) and I found ple behave. Undoubtedly, newborns are ple, Henry Wellman of the University of that 18-month-olds can understand that far from being blank slates, but the Michigan has analyzed recordings of I might want one thing, whereas you changes in children’s knowledge also children’s spontaneous conversations want another. An experimenter showed suggest that they are learning about the for clues to their thinking. We can give 14- and 18-month-olds a bowl of raw world from their experiences. children very focused questions—for broccoli and a bowl of gold sh crackers One of the greatest mysteries of psy- instance, asking them to choose between and then tasted some of each, making chology and is how human just two alternatives rather than asking either a disgusted face or a happy face. beings learn about the world from a con- an open-ended question. In the mid-1980s and through the 1990s, scientists using these techniques discovered that babies already know a Babies look longer at novel or great deal about the world around them. unexpected events than at more That knowledge goes well beyond con- crete, here-and-now sensations. Re- predictable ones, and experimenters searchers such as Renee Baillargeon of can use this behavior to fi gure the University of Illinois and Elizabeth S. Spelke of Harvard University found that out what babies expect to happen. infants understand fundamental physi- cal relations such as movement trajecto- RAISE GREAT KIDS ries, gravity and containment. They look longer at a toy car appearing to pass Then she put her hand out and asked, fusing mess of sensory data. Over the through a solid wall than at events that “Could you give me some?” The past decade researchers have begun to t basic principles of everyday physics. 18-month-olds gave her broccoli when understand much more about how babies By the time they are three or four, she acted as if she liked it, even though and young children can learn so much so children have elementary ideas about they would not choose it for themselves. quickly and accurately. In particular, we biology and a first understanding of (The 14-month-olds always gave her have discovered that babies and young growth, inheritance and illness. This crackers.) So even at this very young age, children have an extraordinary ability to early biological understanding reveals children are not completely egocentric— learn from statistical patterns. that children go beyond super cial per- they can take the perspective of another ceptual appearances when they reason person, at least in a simple way. By age THE STATISTICS OF BLICKETS about objects. Susan A. Gelman, also at four, their understanding of everyday In 1996 Jenny R. Saffran, Richard N. Michigan, found that young children be- is even more re ned. They Aslin and Elissa L. Newport, all then at the University of Rochester, rst demon- strated this ability in studies of the sound FAST FACTS patterns of language. They played se- BABY BRAINS quences of syllables with statistical regu- Œ Babies’ and young children’s cognitive abilities far surpass those that psychologists long larities to some eight-month-old babies. attributed to them. They can, for instance, imagine another person’s experiences and grasp For example, “ro” might follow “bi” only cause and effect. one third of the time, whereas “da” might  Children learn about the world much as scientists do—in effect, conducting experiments, analyzing statistics and forming theories to account for their observations. always follow “bi.” Then they played the Ž The long helplessness of babies may be an evolutionary trade-off, a necessary consequence babies new strings of sounds that either of having brains wired for prodigious feats of learning and creativity. followed these patterns or broke them. Ba-

6 MIND SPECIAL EDITION SUMMER 2016

osq216Gpnk3p.indd 6 3/9/16 5:39 PM bies listened longer to the statistically un- Statistician at Work Babies are skillful statistical analysts. Experiments usual strings. More recent studies show showed that eight-month-olds notice if an improbable number of red Ping-Pong that babies can detect statistical patterns balls are taken out of a collection that is mostly white. Variations of the experi- of musical tones and visual scenes, as well ments (such as swapping the role of red and white) control against alternative explanations (such as having a greater interest in red objects). Twenty-month-olds as more abstract grammatical patterns. tested with green and yellow toys inferred that a person taking an unusually large Babies can even understand the rela- number of the rare color would prefer to be given a toy of that color. Thus, babies tion between a statistical sample and a and young children learn about the world like scientists—by detecting statistical population. In a 2008 study my Universi- patterns and drawing conclusions from them. ty of California, Berkeley, colleague Fei Xu showed eight-month-old babies a box full of mixed-up Ping-Pong balls: say, these children could use probability to 80 percent white and 20 percent red. The discover brand-new and surprising facts experimenter would then take out  ve about the world. balls, seemingly at random. The babies In another experiment Laura Schulz, were more surprised (that is, they looked now at the Massachusetts Institute of longer and more intently at the scene) Technology, and I showed four-year-olds when Xu pulled four red balls and one a toy with a switch and two gears, one white one out of the box—an improbable blue and one yellow, on top. The gears outcome—than when she pulled out four turn when you  ip the switch. This simple white balls and one red one. toy can work in many ways. Perhaps the Detecting statistical patterns is just the switch makes both gears turn at once, or  rst step in scienti c discovery. Even more perhaps the switch turns the yellow gear, impressively, children (like scientists) use which turns the blue one, and so on. We those statistics to draw conclusions about ers can use probabilities to learn how the showed the children pictures illustrating the world. In a version of the Ping-Pong machine works. We repeatedly put one of each of these possibilities—the yellow ball study with 20-month-old babies us- two blocks on the machine. The machine gear would be depicted pushing the blue ing toy green frogs and yellow ducks, the lit up two out of three times with the yel- one, for instance. Then we showed them experimenter would take  ve toys from low block but only two out of six times for toys that worked in one or the other of the box and then ask the child to give her the blue one. Then we gave the children these ways and gave them rather complex a toy from some that were on the table. the blocks and asked them to light up the evidence about how each toy worked. For The children showed no preference be- machine. These children, who could not example, the children who got the “caus- tween the colors if the experimenter had yet add or subtract, were more likely to put al chain toy” saw that if you removed the taken mostly green frogs from the box of the high-probability yellow block on the blue gear and turned the switch, the yel- mostly green toys. Yet they speci cally machine. (More recently, Anna Waismey- low gear would still turn but that if you re- gave her a duck if she had taken mostly er of the University of Washington and moved the yellow gear and turned the ducks from the box—apparently the chil- I discovered that even 24-month-olds switch, nothing happened. dren thought her statistically unlikely se- could do this.) We asked the children to pick the pic- lection meant that she was not acting ran- They still chose correctly when we ture that matched how the toy worked. domly and that she must prefer ducks. waved the high-probability block over Four-year-olds were amazingly good at In my laboratory we have been inves- the machine, activating it without touch- ascertaining how the toy worked based tigating how young children use statisti- ing it. Although they thought this kind of on the pattern of evidence that we pre- cal evidence and experimentation to  g- “action at a distance” was unlikely at the sented to them. Moreover, when other ure out cause and effect, and we  nd start of the experiment (we asked them), children were just left alone with the ma- their thinking is far from being “pre- causal.” We introduce them to a device we call “the blicket detector,” a machine THE AUTHOR that lights up and plays music when you put some things on it but not others. ALISON GOPNIK is professor of psychology and af liate professor of philosophy at Then we can give children patterns of ev- the University of California, Berkeley. She has done groundbreaking research into idence about the detector and see what how children develop a “theory of mind,” the ability to understand that other people causal conclusions they draw. Which ob- have minds and may believe or want different things than they do. She helped to for- jects are the blickets? mulate the “theory theory,” the idea that children learn in the same way that scien- In 2007 Tamar Kushnir, now at Cor- tists do. Investigations of children’s minds, she argues, could help us resolve deep nell University, and I found that preschool- philosophical questions such as the mystery of consciousness.

MIND.SCIENTIFICAMERICAN.COM 7

osq216Gpnk3p.indd 7 3/9/16 5:39 PM H O W B A B I E S THINK

chine, they played with the gears in ways that helped them learn how it worked— as if they were experimenting. Another study by Schulz used a toy that had two levers and a duck and a puppet that popped up. One group of preschoolers was shown that the duck appeared when you pressed one lever and that the puppet popped up when you pressed the other one. The second group saw that when you pressed both levers at once, both toys popped up, but they nev- er got a chance to see what the levers did separately. Then the experimenter had the children play with the toy. Children from the rst group played with the toy much less than those from the second group. They already knew how it worked and were less interested in exploring it. The second group faced a mystery, and they spontaneously played with the toy, soon uncovering which lever did what. These studies suggested that when children play spontaneously (“getting into everything”) they are also exploring basic ideas. First, they use mathematics adults at grasping this unusual causal cause and effect and doing experiments— to describe the hypotheses that children structure. The adults seemed to rely more the most effective way to discover how might have about things, people or words. on their prior knowledge that things usu- the world works. For example, we can represent a child’s ally do not work that way, even though causal knowledge as a map of the causal the evidence implied otherwise for the THE BABY COMPUTER relations between events. An arrow could machine in front of them. Obviously children are not doing exper- point from “press blue lever” to “duck In other recent research my group RAISE GREAT KIDS iments or analyzing statistics in the self- pops up” to represent that hypothesis. found that young children who think they conscious way that adult scientists do. Second, the programs systematically are being instructed modify their statisti- The children’s brains, however, must be relate the hypotheses to the probability cal analysis and may become less creative unconsciously processing information in of different patterns of events—the kind as a result. The experimenter showed a way that parallels the methods of sci- of patterns that emerge from experimen- four-year-olds a toy that would play mu- enti c discovery. The central idea of cog- tation and statistical analysis in science. sic if you performed the right sequence of nitive science is that the brain is a kind of Hypotheses that better t the data be- actions on it, such as pulling a handle and computer designed by evolution and pro- come more likely. I have argued that chil- then squeezing a bulb. For some children, grammed by experience. dren’s brains may relate hypotheses the experimenter said, “I don’t know Computer scientists and philosophers about the world to patterns of probabili- how this toy works—let’s gure it out.” have begun to use mathematical ideas ty in a similar way. Children reason in She proceeded to try out various longer- about probability to understand the pow- complex and subtle ways that cannot be action sequences for the children, some erful learning abilities of scientists—and explained by simple associations or rules. that ended with the short sequence and children. A whole new approach to devel- Furthermore, when children uncon- made music and some that did not. When oping computer programs for machine sciously use this Bayesian statistical anal- she asked the children to make the toy learning uses what are called probabilistic ysis, they may actually be better than work, many of them tried the correct models, also known as Bayesian models adults at considering unusual possibili- short sequence, astutely omitting actions or Bayes nets. The programs can unravel ties. In a study published in 2014 in Cog- that were probably super uous based on complex gene expression problems or help nition, my colleagues and I showed four- the statistics of what they had seen. understand climate change. The approach year-olds and adults a blicket detector With other children, the experiment- has also led to new ideas about how the that worked in an odd way, requiring er said that she would teach them how computers in children’s heads might work. two blocks on it together to make it go. the toy worked by showing them se- Probabilistic models combine two The four-year-olds were better than the quences that did and did not produce

8 SCIENTIFIC AMERICAN MIND SPECIAL EDITION SUMMER 2016

osq216Gpnk3p.indd 8 3/9/16 5:39 PM Natural Experimenters ities we see in very young children. Re- cient, but over time they prune out un- Four-year-olds are adept at interpreting cent biological thinking is in close accord used connections and strengthen useful evidence to learn about cause and with what we see in the psychology lab. ones. Baby brains also have a high level of e ect, such as determining if one cog From an evolutionary perspective, the chemicals that make brains change on a machine is turning another (oppo- one of the most striking things about hu- connections easily. site page). Some even carried out the man beings is our long period of imma- The brain region called the prefrontal correct experiments (and drew the right conclusion) while freely “playing” with turity. We have a much longer childhood cortex is distinctive to humans and takes the toy. Research involving a “blicket than any other species. Why make babies an especially long time to mature. The detector” (below ), which is more likely so helpless for so long and thus require adult capacities for focus, planning and to light up for some combinations of adults to put so much work and care into ef cient action that are governed by this blocks than for others, found that four- keeping their babies alive? brain area depend on the long learning that year-olds could use sta tistics to learn how the machine worked, even when Across the animal kingdom, the in- occurs in childhood. This area’s wiring it showed new, unexpected behavior. telligence and  exibility of adults are may not be complete until the mid-20s. Indeed, they were more open-minded correlated with the immaturity of babies. The lack of prefrontal control in than adults when faced with evidence “Precocial” species such as chickens rely young children naturally seems like a that the machine responded to blocks on highly specific innate capacities huge handicap, but it may actually be tre- in an unusual way. adapted to one particular environmental mendously helpful for learning. The pre- niche, and so they mature quickly. “Al- frontal area inhibits irrelevant thoughts tricial” species (those whose offspring or actions. But being uninhibited may need care and feeding by parents) rely on help babies and young children to explore learning instead. Crows, for instance, freely. There is a trade-off between the can take a new object, such as a piece of ability to explore creatively and learn  ex- wire, and work out how to turn it into a ibly, like a child, and the ability to plan tool, but young crows depend on their and act effectively, like an adult. The very parents for much longer than chickens. qualities needed to act ef ciently—such as A learning strategy has many advan- swift automatic processing and a highly tages, but until learning takes place, you pruned brain network—may be intrinsi- are helpless. Evolution solves this prob- cally antithetical to the qualities that are lem with a division of labor between ba- useful for learning, such as  exibility. bies and adults. Babies get a protected A new picture of childhood and hu- time to learn about their environment, man nature emerges from the research of without having to actually do anything. the past decade. Far from being mere un- When they grow up, they can use what nished adults, babies and young children they have learned to be better at surviv- are exquisitely designed by evolution to ing and reproducing—and taking care of change and create, to learn and explore. music, and then she acted on the toy in the next generation. Fundamentally, ba- Those capacities, so intrinsic to what it exactly the same way. When asked to bies are designed to learn. means to be human, appear in their pur- make the toy work, these children never Neuroscientists have started to un- est forms in the earliest years of our lives. tried a shortcut. Instead they mimicked derstand some of the brain mechanisms Our most valuable human accomplish- the entire sequence of actions. Were these that allow all this learning to occur. Baby ments are possible because we were once children ignoring the statistics of what brains are more  exible than adult brains. helpless dependent children and not in they saw? Perhaps not—their behavior is They have far more connections between spite of it. Childhood, and caregiving, is accurately described by a Bayesian mod- neurons, none of them particularly ef - fundamental to our humanity. M el in which the “teacher” is expected to choose the most instructive sequences. In simple terms: if she knew shorter se- MORE TO EXPLORE quences worked, she would not have ■ The Scientist in the Crib: Minds, Brains, and How Children Learn. Alison Gopnik, Andrew N. shown them the unnecessary actions. Meltzoff and Patricia K. Kuhl. William Morrow and Company, 1999. ■ The Philosophical Baby: What Children’s Minds Tell Us about Truth, Love, and the Meaning EVOLUTION AND NEUROLOGY of Life. Alison Gopnik. Farrar, Straus and Giroux, 2009. If the brain is a computer designed by ■ When Younger Learners Can Be Better (or at Least More Open-Minded) Than Older Ones. A. Gopnik, T. L. Grif ths and C. G. Lucas in Current Directions in Psychological Science, evolution, we can also ask about the evo- Vol. 24, No. 2, pages 87–92; April 2015. lutionary justi cation and neurological ■ The Gardener and the Carpenter: What the New Science of Child Development Tells Us about the basis for the extraordinary learning abil- Relationship between Parents and Children. Alison Gopnik. Farrar, Straus and Giroux (in press).

MIND.SCIENTIFICAMERICAN.COM 9

osq216Gpnk3p.indd 9 3/9/16 5:39 PM