8/7/2019 science starts early

http://slidepdf.com/reader/full/science-starts-early 1/3

DOI: 10.1126/science.1195221, 1022 (2011);331 Science

Frank C. KeilScience Starts Early

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by, yo u can order high-quality copies for yourIf you wish to distribute this articl e to others

here.following the guidelines

can be obtained byPermission to republis h or r epurpose articles or portions of articles

): February 27, 2011 www.sciencemag.org (this infomation is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/331/6020/1022.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/331/6020/1022.full.html#ref-list-1, 3 of which can be accessed free:cites 12 articlesThis article

http://www.sciencemag.org/cgi/collection/psychologyPsychology

subject collections:This article appears in the following

registered trademark of AAAS.is aScience 2011 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

8/7/2019 science starts early

http://slidepdf.com/reader/full/science-starts-early 2/325 FEBRUARY 2011 VOL 331 SCIENCE www.sciencemag.org022

PERSPECTIVES

Science Starts EarlyPSYCHOLOGY

Frank C. Keil

Infants and children grasp surprisinglysophisticated correlational and causal patterns.

signals may be found in widely divergentprotein localization pathways, not just duringprotein secretion.

Additional work will reveal whether mRNA signals represent a general pro-tein localization strategy or if they remainexceptional examples. For example, a recentreport revealed that several transcripts in E.

coli and Caulobacter crescentus also local-ize speci cally, but that they remained veryclose to the chromosomal site where theywere synthesized ( 9 ). In any case, the obser-

vations that mRNA molecules localize spe-ci cally within a bacterium predict the exis-tence of pathways that mediate this localiza-tion, and these components will need to beidenti ed. It appears that mRNA can con-tribute not simply to encoding proteins, butto delivering them as well.

References and Notes

1. L. Shapiro, H. H. McAdams, R. Losick,Science 326 , 1225(2009).

2. G. Blobel,Proc. Natl. Acad. Sci. U.S.A. 77 , 1496 (1980).3. K. Nevo-Dinur, A. Nussbaum-Shochat, S. Ben-Yehuda,

O. Amster-Choder,Science 331 , 1081 (2011).4. M. A. Lande, M. Adesnik, M. Sumida, Y. Tashiro, D

Sabatini, J. Cell Biol. 65 , 513 (1975).5. G. Blobel, B. Dobberstein,J. Cell Biol. 67 , 852 (19756. G. Blobel, B. Dobberstein,J. Cell Biol. 67 , 835 (19757. A. F. Palazzoet al ., PLoS Biol. 5 , e322 (2007).8. D. M. Anderson, O. Schneewind,Science 278 , 1140

(1997).9. P. Montero Llopiset al ., Nature 466 , 77 (2010).

10. This work was funded by the Intramural Research Pgram of the NIH National Cancer Institute Center foCancer Research.

10.1126/science.1201

Infants and young children can exhibitstriking confusion about how the world

works, from failing to grasp that wind causes waves, to being mysti ed about howbabies are created. Indeed, some researchershave characterized a child’s knowledge of theworld as a bundle of misconceptions awaitingreplacement with correct concepts througheducation ( 1 ).

Evidence is mounting, however, thatyoung children are often quite adept at uncov-ering statistical and causal patterns and thatmany foundations of scienti c thought arebuilt impressively early in our lives. Thisgrowing understanding of how children

acquire many of the thinking skills used inscience has implications not only for educa-tion but also for understanding how all of usmake scienti c progress in the face of igno-rance and incomplete knowledge.

For cognitive psychologists, scientistshave long presented an intriguing puzzle.Whether a biologist or a geologist, scientistsroutinely, and with seeming ease, call upona diverse set of cognitive skills to do their jobs. They detect correlations, often betweenseemingly unrelated phenomena. They infer causation from these correlations. If all goeswell, they uncover the mechanisms thatexplain it all and then share their knowl-edge and build upon it by acquiring knowl-edge from others.

Each of these abilities has early origins.Consider, for example, how children respond to the challenge of noticing correlations asthey encounter them in the ow of experi-ence. For instance, an infant learning lan-guage, upon hearing streams of syllables, not

only has to notice how often certain syllablesoccur but also needs to infer higher-order

patterns arising from those syllables. Onestudy ( 2 ) showed that 5-month-old infantscan handle this challenge by rapidly trackingnot only the sounds of the syllables but alsovisual patterns associated with each syllable.In the experiment, infants looking at a com-puter screen were repeatedly presented withabstract patterns of syllables and shapes. An“ABB” pattern, for instance, could be repre-sented by certain shapes corresponding tothe syllables “di ga ga.” When presented witha new pattern (ABA) with new syllables such as “le ko le” the infants looked longer

at the shapes on the screen than if the newsyllables were in the old ABB pattern. Thissuggests that they recognized it as a new,unfamiliar correlation.

Other research ( 3 ) has found that6-month-olds can take the next step and infer causation from certain kinds of correlations.In these experiments, researchers measured how long infants looked at animations show-ing “collisions” of shapes. In some anima-tions, one object “launched” a second one,causing it to move, as when two billiard ballscollide. When shown animations in whichthe balls reversed roles, infants looked lon-ger at the new pattern than at the originalone. They did not react as strongly, however,when the original and reversed animationscontained half-second gaps between themoment when the rst object stopped mov-ing and the second one started to move. Thissuggests that the infants recognized theseevents as noncausal, or unrelated.

Later studies showed that infants makecausal interpretations by integrating infor-mation in ways that closely mirror adults.For instance, they will form causal interpre-

tations based on information that is collectedin brief time windows after the occurrence of

the critical event (such as a possible collision)(4 ); these post-event decision-making win-dows are similar to those measured in adults.Thus, it appears that certain sequences ofevents automatically elicit thoughts of causa-tion at all ages.

Older infants expand on these inferencesof causality to sense more abstract and sub-tle causal relationships. In one recent study(5 ), 11-month-olds were shown animationsof two sets of blocks: one initially orderedinto a neat array, the other scattered into dis-order. Then, a screen obscured the blocks and

either a lifelike “animate” agent appeared,such as an object with a face, or an “inani-mate” agent, such as a ball. Finally, the screenwas removed, revealing that either an orderlystack of blocks had become disordered or theopposite. By measuring how long the infantslooked at various combinations, the research-ers concluded that the infants learned thatonly the animate object could cause disor-dered blocks to become orderly but that boththe animate and inanimate agents could scat-ter an orderly pile.

Once out of infancy, children becomeable to examine more complex networksof correlations to infer causal patternsincluding hidden ones, and they readily dothis using sample sizes too small for tradi-tional statistical tests of signi cance. Theyare particularly sensitive to the usefulnessof “intervening on a system” or manipu-lating conditions to separate causal linksfrom those that are just correlational. Forexample, when confronted with a novel boxconsisting of gears and a switch, preschoolchildren are easily able to gure out cause-and-effect relations, and rule out mere cor-

Department of Psychology, Yale University, New Haven, CT06520–8205, USA. E-mail: frank.keil@yale.edu

Published by AAAS

8/7/2019 science starts early

http://slidepdf.com/reader/full/science-starts-early 3/3www.sciencemag.org SCIENCE VOL 331 25 FEBRUARY 2011

PERSPECTIV

relations, by manipulating key componentsand observing the consequences ( 6 ). Thus,they anticipate a key motivation of experi-mental design, in which some variables aremanipulated while others are held constant.

In addition to guring out the causal rela-tions underlying novel devices, children arealso sensitive to highly abstract causal pat-terns associated with speci c “domains” thatcorrespond roughly to formal areas of sci-ence, such as biology, physical mechanics,and psychology. For example, while beingcompletely ignorant about the biologicaldetails, most preschoolers do know that food gets transformed after it enters the body and that the transformed version is critical for helping the body to grow and to move ( 7 ).

Some of these “core knowledge” domainsmay have origins in infancy and then becomecombined into larger conceptual systems inchildhood ( 8 ).

A child’s understanding of the world isnot driven simply by assembling correla-tions in a bottom-up manner. Instead, evenvery young children bring to most learningsituations broad intuitions and expectationsabout plausible and implausible patterns.This kind of top-down analysis enables themto rule out or narrow an overwhelming rangeof possible correlations ( 9 ). For example, inthinking about biological phenomena suchas disease or inheritance, children may makedifferent inferences from patterns of covari-ation than they do for physical phenomenasuch as collisions or rotating gears ( 10 ).

Another top-down expectation that childrenbring to living things, but not to artifacts, isan “essentialist bias”: the idea that some-thing you can’t see (e.g., “microstructuralstuff ”) causes what you can see (“surfacephenomena” such as feathers or fur) and isthe essence of the thing being observed. Thisis a guiding principle in much of formal sci-ence, even as it can also lead to false infer-ences, such as that species are de ned by xed essences ( 11 ).

Science education should build uponthese early-emerging cognitive founda-tions. Like the vast majority of adults, chil-dren need instruction about detailed mecha-nisms, but they also bring to the classrooma rich repertoire of skills that enable them

to track patterns and infer causality in waysthat constrain and support learning aboutspecific mechanisms. Educational strate-gies that leverage those skills, such as con-necting mechanism information (e.g., natu-ral selection) to a higher-level causal pat-tern children already know (e.g., adapta-tion), will have a huge head start over thosethat do not ( 12 ).

The study of formal science is alsoinformed by studies of children. Recentresearch reveals how all of us are able toaccumulate knowledge with only partialunderstanding of mechanisms outside our own narrow areas of expertise ( 13 ). Scientistsuse their senses of more abstract causal and correlational patterns to navigate and rely onthe divisions of cognitive labor that are essen-

tial to scienti c progress. Thus, all scientistsmust outsource some understanding to other experts by grasping the coarse causal and correlational patterns associated with distinctareas of expertise. We are only beginning tounderstand how they nd the right depth of analysis in their own areas for optimal prog-ress and then link that work to research doneby others ( 14 ).

Future studies must explore how chil-dren, and adults, can build upon these foun-dations. How, for instance, can we preventthe essentialist biases and other narrowingstrategies that children use to understand the world from lingering into adulthood and impairing scientific reasoning ( 15 )Imbuing organisms with xed essences, for example, can impair understanding of natu-ral selection ( 16 ), and a child’s assumptionthat order is created by intentional agents ( 5

can ultimately impair recognition of other order-creating forces such as those at work

in evolution. Finally, we need to examinethe relationship between implicitly trackingpatterns and explicitly characterizing their natures. Taken together, these future studieswill continue to transform how we all think about the scienti c skills of children. Rather than bumbling babies, they are individualswho right from the start are deeply inter-ested in and can learn surprisingly fast aboutthe patterns of nature.

References and Notes1. D. Gil-Perez, J. Carrascosa,Sci. Educ. 74 , 531 (1990).2. M. C. Frank, J. A. Slemmer, G. F. Marcus, S. P. Johnson

Dev. Sci. 12 , 504 (2009).3. A. M. Leslie, S. Keeble,Cognition 25 , 265 (1987).4. G. E. Newman, H. Choi, K. Wynn, B. J. Scholl,Cognit.

Psychol. 57 , 262 (2008).5. G. E. Newman, F. C. Keil, V. A. Kuhlmeier, K. Wynn,Proc

Natl. Acad. Sci. U.S.A. 107 , 17140 (2010).6. L. E. Schulz, A. Gopnik, C. Glymour,Dev. Sci. 10 , 322

(2007).7. K. Inagaki, G. Hatano,Curr. Dir. Psychol. Sci. 15 , 177

(2006).8. S. Carey,The Origin of Concepts(Oxford Univ. Press, N

York, 2009).9. W. K. Ahn, J. K. Marsh, C. C. Luhmann, K. Lee,Mem. Cog

nit. 30 , 107 (2002).10. T. L. Grif ths, J. B. Tenenbaum,Psychol. Rev. 116 , 661

(2009).11. S. A. Gelman,The Essential Child: Origins of Essential-

ism in Everyday Thought (Oxford Univ. Press, New York

2009).12. R. A. Duschl, H. A. Schweingruber, A. E. Shouse, Eds.Ta

ing Science to School: Learning and Teaching Science inGrades K-8 (National Academies Press, Washington, DC2007).

13. D. M. Sobel, K. H. Corriveau,Child Dev. 81 , 669 (2010).14. M. Strevens,Depth: An Account of Scienti c Explanation

(Harvard Univ. Press, Cambridge, MA, 2009).15. P. Bloom, D. S. Weisberg,Science 316 , 996 (2007).16. A. Shtulman, L. Schulz,Cogn. Sci. 32 , 1049 (2008).17. Preparation of this essay was supported by NIH grant

R37-R37-HD023922 to F.C.K. Thanks to K. Lockhart fcomments.

CREDIT: HEMERA/THINKSTOCK

Budding scientist. Many aspects of scienti c understanding appear early.

10.1126/science.1195221

Published by AAAS