like father, like son: young children's understanding of how and why offspring resemble their...

Like Father, Like Son: Young Children'sUnderstanding of How and Why OffspringResemble Their Parents

Gregg E. A. Solomon and Susan C. JohnsonMassachusetts Institute of Technology

Deborah ZaitchikUniversity of Massachusetts, Boston

Susan CareyMassachusetts Institute of Technology

SOLOMON, GREGG E. A.; JOHNSON, SUSAN C ; ZAITCHIK, DEBORAH; and CAREY, SUSAN. Like Father,Like Son: Young Children's Understanding of How and Why Offspring Resemble Their Parents.CHILD DEVELOPMENT, 1996, 67,151-171.4 studies investigated the broad claim that preschoolersunderstand biological inheritance. In Study 1, 4-7-year-old children were told a story in whicha hoy was born to one man and adopted hy another. The biological father was described ashaving one set of features (e.g., green eyes) and the adoptive father as having another (e.g.,brown eyes). Subjects were asked which man the boy would resemble when he grew up. Pre-schoolers showed little understanding that selective chains of processes mediate resemblanceto parents. It was not until age 7 that children substantially associated the boy with his biologicalfather on physical features and his adoptive father on beliefs. That is, it was not until age 7 thatchildren demonstrated that they understood birth as part of a process selectively mediating theacquisition of physical traits and leaming or nurturance as mediating the acquisition of beliefs.In Study 2, subjects were asked whether, as a boy grew up, various of his features could change.Children generally shared our adult intuitions, indicating that their failure in Study 1 was notdue to their having a different sense of what features can change. Studies 3 and 4 replicatedStudy 1, with stories involving mothers instead of fathers and with lessened task demands. Takentogether, the results of the 4 studies refute the claim that preschoolers understand biologicalinheritance. The findings are discussed in terms of whether children understand biology as anautonomous cognitive domain.

Carey's (1985,1988) claim that an auton- logically distinct entities, and (3) uniqueomous domain of biology is not constructed causal mechanisms, which provide explana-until the end of the first decade of life has tions for the phenomena in the domain andcome under concerted criticism (e.g.. Gel- the properties of the entities in it (Carey,man & Wellman, 1991; Inagaki & Hatano, 1985). Wellman and Gelman (1992) further1993; Keil, 1989, 1994). At issue here are the point out that in an autonomous domain, do-explicit, accessible, conceptual representa- main-specific causal mechanisms operate intions—the framework theories—that are a manner that respects ontological distinc-fundamental to the organization of concep- tions within a coherent causal explanatorytual knowledge. We can credit children with framework. The notions of ontological dis-an autonomous cognitive domain only if tinctions and coherent causal explanatorythey can be shown to represent: (1) a set of frameworks are spelled out in related waysphenomena, involving (2) a domain of onto- by Carey (1985, in press), Keil (1979, 1989),

We wish to thank Adebisi Lipede, Adee Matan, Veronica Mendoza, David Millstein, KellyOlguin, Virginia Slaughter, Helen Tager-Flusberg, Linda Tickle Degnen, Mary Gale and theCambridge Montessori School, Lynn Stuart and the Cambridge School District, Linda Rings andthe Cambridge Compass Program, and the Solomon Schector Day School. This research wassupported by grant 91-20 from the James S. McDonnell Foundation to the flrst author and grant5-T32-MH18823 from NIMH to the second author. Please send correspondence concerning thisarticle to the authors at the Department of Brain and Cognitive Sciences, Massachusetts Instituteof Technology, 79 Amherst Street, Cambridge, MA 02139.

[Child Development, 1996, 67,151-171. © 1996 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/96/6701-0014$01.00]

152 Child Development

Murphy and Medin (1985), and Wellmanand Gelman (1992).

The above criteria for autonomous do-mains are not independent. For example,"ontological distinction" and "explanatoryframework" are interdefined; not all concep-tual distinctions are ontological distinctions(Carey, 1985; Keil, 1979). Cats are distin-guished from dogs, but this is not an ontolog-ical distinction. Carey (1985, in press) arguesthat ontological distinctions are those be-tween entities embedded in differentcausal/explanatory frameworks. Researcherswho claim that children understand an au-tonomous domain of biology have estab-lished that preschool children realize thatanimals (and perhaps also plants) are distin-guished from other entities, because theyundergo processes such as growth that otherkinds do not (e.g., Keil, 1994; Rosengren,Gelman, Kalish, & McCormick, 1991). Butin order to claim that these are understoodas ontological distinctions by children, onemust demonstrate that children have an ex-planatory framework that includes knowl-edge of biological causal mechanisms perti-nent to such animal-specific phenomena.

The question, then, is whether pre-school children know any causal explanatorymechanisms that are uniquely biological.Studies from several laboratories have beentaken as providing evidence that childrenknow at least one such mechanism, one thatexplains why blond parents tend to haveblond children. That is, it is claimed thatpreschool children have a biological under-standing of the inheritance of properties(Gelman & Weilman, 1991; Springer, 1992;Springer & Keil, 1989). However, a close re-view ofthese studies reveals them to be in-conclusive.

A biological understanding ofthe inher-itance of properties includes, at a minimum,two essential components: Resemblance toParents and Mediation by Reproduction.First, there must be some understandingthat offspring will resemble their parentswith respect to variation among species (e.g.,dogs have baby dogs not baby cats) and withrespect to variation among individuals ofthesame species (e.g., black parents tend tohave black children). And, second, theremust be some understanding of the ways bywhich children may come to resemble theirparents that entails a uniquely biological (asopposed to a psychological or mechanical)chain of causation. To be credited with a bio-logical concept of inheritance, children need

not understand anything like a geneticmechanism, but they must have some sensethat the processes resulting in Resemblanceto Parents differ from learning or other envi-ronmental mechanisms. Curly-haired par-ents may have curly-haired children becausethey give them permanents; prejudiced par-ents may have prejudiced children becausethey teach them to be so. These are not ex-amples of biological inheritance. Inheri-tance, the biological origin of an animal's bi-ological features, must be causally linked tobirth, the biological origin of the animal.And as the biological parent is causally asso-ciated with birth and birth is causally associ-ated with the origin of the body, so is thebiological parent causally associated withbodily properties. Thus, the second compo-nent: Children understand biological inheri-tance only to the extent that they understandthat, for certain characteristics, the chain ofprocesses underlying Resemblance to Par-ents crucially involves birth. These charac-teristics include those bodily traits, such asskin color, that do not change during a per-son's lifetime. Finally, children's under-standing of inheritance is part of a largerframework of biological causal explanationonly if birth is implicated in the origin ofbodily features and not in the origin of be-liefs and other properties that children knowto he learned. Inheritance judgments mustdistinguish among properties in a mannerthat is consistent with the finding that pre-school children know minds and bodies tobe ontologically distinct (Inagaki & Hatano,1993; Keil, 1989; Wellman & Gelman, 1992).

In this article we explore whether 4-7-year-olds understand biological inheritance.We grant that preschool children understandthe general notion of Resemblance to Par-ents. For example. Springer (1992) foundthat 4—8-year-olds understand that offspringresemble their parents. He told children thata pictured animal has an unusual property(e.g., "this horse has hair inside its ears").He probed for projection of this property toa physically similar animal, described as afriend unrelated by birth to the target, andto a physically dissimilar animal, describedas the target's baby. At all ages, the propertywas projected more to the baby than to thefriend. This important result confirms themounting evidence that preschool childrenare not appearance bound (Flavell, Flavell,& Green, 1983; Wellman & Gelman, 1988)and establishes the Resemblance to Parentscomponent of an understanding of biologicalinheritance. However, because the study

Solomon et al. 153

does not probe the mechanisms responsiblefor resemblance, it cannot bear on the sec-ond component.^ Springer (1992) success-fully distinguished what he considers a bio-logical relationship (parentage) from a socialrelationship (friendship), but, as Carey(1985, 1988) pointed out, parentage is alsoa social relationship. At the very least, onewould like to see biological parentage dis-tinguished from nonbiological parentagesuch as adoption.

The same issue arises with respect tothe data of Springer and Keil (1989). Adultsand 4-7-year-old children were told thatboth parents of an animal had a particularproperty atypical of the species (e.g., pinkrather liian tiie usual red hearts) and wereasked whether that animal would be bornwith the property. Springer and Keil manip-ulated further information about the abnor-mal property such as how the parents got it(whether they were born with it or acquiredit in an accident), whether the property wasinternal or external, and whether it wasdescribed as having "biological" func-tional consequences. Two important resultsemerged: First, only adults based their judg-ments solely on whether the parents wereborn with the property. That is, only adultsrelated birth to inheritance. The childrendid not appreciate that birth plays a mediat-ing role in the process by which offspringcome to resemble parents. Second, pre-school children did make systematic judg-ments; they were infiuenced by whether theproperty was described as having "biologi-cal functional consequences" or not. Fromthis result. Springer and Keil concluded thatpreschoolers have a biological concept of in-heritance. This conclusion must be tem-pered due to the lack of analysis of what isto be considered a biological functional con-sequence. For example. Springer and Keiltook "help them to see their enemies" as abiological functional consequence and"scared away all their friends" as a nonbio-logical functional consequence of "big,stretched-out eyes." But even if we grantthese designations, the result merely estab-lishes that preschool children expect Resem-blance to Parents for such properties. Again,

to probe further for an appreciation of birthas part of the causal process mediating theacquisition of certain properties, it would benecessary to compare natural parentage andadoptive parentage.

Gelman and Wellman (1991) specifi-cally contrasted nature and nurture. Theyasked, for example, whether a cow, Edith,who had been separated from other cows atbirth and raised with pigs, would moo oroink and whether she would have a straightor a curly tail. Even 4-year-oIds judged thatEdith would moo and have a straight tail.But the result does not bear directly on in-heritance (nor was it intended to do so): Thestory asserts that Edith is a cow, in spite ofhaving been raised in the company of pigs,and so it prejudges the question of interest.There is a wealth of evidence, most of it fromGelman herself, that preschoolers take cate-gory membership as predictive of category-relevant properties in the face of conflictinginformation (Gelman, 1988; Gelman &Markman, 1987). Furthermore, Gelman andWellman's stories did not stress that thebaby cow was raised in a pig family, an off-spring among otiier offspring who were pigs.Thus, it did not cleanly contrast adoptiveand biological family relationships.^

To fill the gap in the literature, the studiespresented here contrast adoptive and biologi-cal family relationships in determining Resem-blance to Parents. Study 1 uses both direct andindirect methods of examining what mecha-nisms children believe mediate similarity be-tween offspring and parents. It indirectly as-sesses children's understanding by examiningwhether their pattern of judgment dependsupon what kind of property is at issue. A pattemof judgment that offspring will resemble theirbiological parent on physical properties andtheir adoptive parent on clearly nonphysicalproperties would provide indirect evidence fora differentiation of two classes of mediatingprocesses by which offspring come to resembleparents. Study 1 also directly examines chil-dren's understanding ofthe mechanisms medi-ating resemblance between parents and off-spring by asking them for explanations of theirjudgments.

' Springer and Keil (1991) did explore children's understanding of various mechanisms, butthose data are inconclusive (see Ceneral Discussion).

^ Gelman and Wellman did attempt to test whether category information was driving theinference. They posed stories about a seed described as coming from one plant (e.g., an apple)and grown in tiie environment of another (e.g., in a flower pot, in the company of flowers). Byage 5, children judged it would come up an apple rather than a flower. But here they werecontrasting/oiriiiy with environment (like Springer, 1992), not adoptive versus biological family.Thus, they again established that children know that offspring resemble parents.


Study 1: Similarity to Adoptive andBiologieal Parents

MethodForty-eight children from Boston area

schools and day-care centers participated inthe study, 16 each in the 7-year-old group(M = 7-7, range = 7-0 to 7-11), the 6-year-old group (M = 6-5, range = 6-0 to 6-11),and the preschooler group (M = 5-1, range= 3-11 to 5-10). Sixteen adults also partici-pated in the study. The subjects includedblack, white, and Asian males and females ofdifferent economic backgrounds. The adultsranged from high school graduates to thosewith advanced degrees. There was no effectof level of education on adult performance.

The subjects were interviewed individ-ually. Each interview took approximately 10to 15 min. The subjects were read a fairy talein which a little boy is born to a king andadopted by a shepherd or one in which alittle boy is bom to a shepherd and adoptedby a king (see Appendix A). The experiment-ers pointed to schematic line drawings oftheking, the shepherd, the boy, the king's cas-tle, and the shepherd's hut in order to main-tain the children's interest and make thestory clearer. None of the features subse-quently probed in the task is represented inthe pictures. Before proceeding with thetesting, the experimenters questioned thesubjects to make sure that they understoodthe story. They asked, "Where was the littleboy born?" and "Where did he grow up?"Two children did not respond correctly andwere dropped from the study. When the ex-perimenters were confident that the subjectsunderstood the story, they asked a series of18 inheritance questions in a two-alternative, forced-choice paradigm. The bi-ological parent was described as having oneof a pair of features, the adoptive parent ashaving the other, and the subjects wereasked what they thought the little boy wouldbe like when he grew up. For example, "Theking has green eyes and the shepherd hasbrown eyes. When the little boy is all grownup and is a young man, do you think that hewill have green eyes like the king or browneyes like the shepherd?" If subjects did notanswer they were encouraged to do so, andif they chose something other than one ofthe two alternatives, that response was notedand the question was repeated. Childrenwere encouraged to provide explanations oftheir judgments; the adults were not askedfor explanations.

The 18 feature pairs were dividedamong five types of traits (see Table 1). The

TABLE 1

LIST OF STUDY 1 FEATURE PAIRS, BY TRAIT TYPE

Physical traits:has green eyes/has brown eyeshas a high voice/has a low voicehas curly hair/has straight hairis tall/is shorthas liver on right/has liver on lefthas pink heart/has red heart

Beliefs:believes that skunks can see in the dark/

believes that skunks cannot see in the darkbelieves that tin cans rust/believes that tin

cans do not rustbelieves that oil floats in water/believes that

oil sinks in waterPreferences:

likes candy more than pickles/likes picklesmore than candy

likes cats more than dogs/likes dogs morethan cats

likes parties more than movies/likes moviesmore than parties

Temperaments:is shy/likes to meet new peopledoesn't share things/shares thingslaughs all the time/angry all the time

Skills:is better at reading than at counting/is better

at counting than at readingis better at football than at baseball/is better

at baseball than at footballis better at cooking than at sewing/is better

at sewing than at cooking

contrast of greatest theoretical interest is be-tween the six physical traits (e.g., liver onthe right/liver on the left) and the three be-liefs (e.g., believes that skunks can see inthe dark/believes that skunks cannot see inthe dark), for a differentiation of these traitslies at the heart of the distinction betweenresemblance to family due to biologicalcauses and that due to teaching and leamingcauses. The nine other traits were dividedamong three preferences (e.g., likes candymore than pickles/likes pickles more thancandy), three skills (e.g., is better at footballthan baseball/is better at basebcJl than foot-ball), and three temperaments (e.g., laughsall the time/is angry all the time). Thesetraits were included mainly for exploratoryreasons. Experts who study heritability ofsuch traits are divided as to the degree ofbiological determination of them, but it wasour intuition that adults in our culture con-sider them to be mainly under environmen-tal infiuence, their acquisition mediated byteaching and learning.

The experiment was balanced acrosssubjects to control for the following possible

Solomon et al. 155

confounds: whether the king or the shep-herd was the biological father; whether theking or the shepherd was mentioned firstduring the forced-choice questioning(thereby avoiding the confound of having,for example, the king always described asbeing tall and the shepherd always short);and which combinations of features weretrue of the same father (two different shuf-flings of features were used so that, for exam-ple, having brown eyes sometimes wentwith being tall and sometimes with beingshort). The features were interspersed tocontrol potential order effects. There were,all told, eight different experimental ver-sions. A preliminary analysis revealed nosignificant effects related to the differentversions.

ResultsThe goal of the present analysis was to

identify individual subjects who showed ev-idence of having reasoned differentiallyabout the origins of Physical traits and Be-liefs. The Preferences, Temperaments, andSkills had been included in the study in or-der to explore whether children might con-sider them to be unique in some way. Infact, the subjects judged them quite simi-larly and much like the Beliefs (they had notbeen predicted to vary). The greatest differ-ence shown among these four types of traits,in percent of features judged to resemble thebiological father, at any age, was only 10 per-centage points. Therefore, for the purpose ofsmoother exposition, the following analysesfocus on the theoretically critical contrast ofthe Physical traits aiid Beliefs.

Pattern of responding over traittypes.—The subjects were characterized ac-cording to their individual judgment pat-terns. The most important of these initialpattern groups was the Differentiated pat-tern group, defined to include those subjectswho did distinguish between the physicaltraits and beliefs. Subjects were consideredto have shown a Differentiated pattern ifthey judged the boy to resemble his biologi-cal father on at least five of the six Physicalfeatures and on none of the three Beliefs oron all six ofthe Physical features and on oneofthe Beliefs. The probability (binomial the-

orem) of an individual subject's showingsuch a pattern by chance is .0195. Of thosesubjects who did not meet these criteria forinclusion into the Differentiated patterngroup, three further categories capture theirresponse patterns. A number of childrenshowed a bias to judge the boy to resembleone of the parents on virtually all features,regardless of trait type. Subjects were con-sidered to have shown an Adoptive Parentbias pattern if they judged the boy to resem-ble his adoptive parent on more than 13 ofthe 18 features (significantly more thanwould be expected by chance).^ Conversely,subjects were considered to have shown aBiological Parent bias if they had judged theboy to resemble his biological parent onmore than 13 of the features (significantlymore than chance). Finally, any pattem notfalling into the other categories was consid-ered to be a Mixed pattern.

Subjects of different ages gave markedlydifferent patterns of judgments. As can beseen in Table 2, between the ages of 4 and7, increasing numbers of children begin toappreciate that offspring resemble theirbirth parents with respect to some traits andtheir adoptive parents with respect to others.The association between age and the num-ber of subjects in the Differentiated patternwas significant according to a chi-squareanalysis, crossing the four levels of age withthe two levels of pattern (Differentiated vs.Not Differentiated), x^(3. A? = 64) = 28.44,p < .001.

The finding of greatest theoretical im-portance here is that most of the childrenunder the age of 7 did not show a Differen-tiated pattern. As this is the only pattemproviding clear evidence that subjects dis-tinguished two mechanisms underlyingresemblance to parent, these results under-mine the claim that preschoolers as a groupunderstand biological inheritance. Note,however, that it is entirely possible that indi-vidual preschoolers might have such an un-derstanding. A second-order application ofthe binomial theorem, based on a probabil-ity of .02 that a subject would show a Differ-entiated pattem by chance, indicates thatthe 15 adults {p < .001), nine 7-year-olds

^ It is, of course, statistically possible for subjects with Differentiated judgment patterns tomeet these criteria (this was, in fact, true of only one cbild in our study). Note, however, thatour hypothesis was that young children do not selectively associate the origin of physical traitswith the biological father and beliefs with the adoptive father. Therefore, we flrst determinedwhich subjects could be included in tbe Differentiated pattern group, tbe group most directlyrelevant to testing our hypothesis. The bias patterns therefore apply only to those subjects notincluded in the Differentiated group. This method of analysis credits children with the mostadvanced understanding possible.


TABLE 2

PATTERNS OF JUDGMENTS FROM STUDY 1, BY AGE

PATTERN

DifferentiatedBiological biasAdoptive biasMixed

Adults

15001

7

9313

AGE

6

4309

Preschoolers

1168

ip < .001), and four 6-year-oIds (p < .001)who showed this pattern are in each casesignificantly more than would be expectedby chance out of a group of 16. However, thesame is not true ofthe single preschooler (inthe group of 16) who showed a differentiatedpattem. It is therefore probable that some ofthe 6-year-old children understood biologi-cal inheritance, though most did not,whereas such a claim cannot be made re-garding preschoolers.

Ofthe children who did not show a Dif-ferentiated pattern, a substantial number ofpreschoolers showed an Adoptive Parentbias. It is possible that these children maythink learning or environmental factors me-diate the acquisition of all properties. Chil-dren with a Biological Parent bias (shownalmost exclusively by 6- and 7-year-olds)may think that birth mediates the acquisitionof all properties. Or it may be that childrenin both groups responded solely on the basisof resemblance to parents, but had differentcriteria for determining what constitutes"parent." The modal pattern shown by boththe 6-year-olds and the preschoolers was theMixed pattem. This pattern may reflect ran-dom or inconsistent judgments or it may re-flect judgments based on a different set ofcriteria than is reflected in the distinctionbetween adoptive and biological parents.

The Mixed group likely also includesthose subjects ^vho have begun to distin-guish two mechanisms underlying resem-blance to parents, but who, for whatever rea-son, do not distinguish between the Physicalfeatures and Beliefs quite sharply enough toshow a Differentiated pattem. Let us definethe Partially Differentiated pattern group asincluding those subjects in the Mixed pat-tem group who judged the boy to resemblehis father on at least half (three) ofthe Physi-cal features and on fewer than half (one ornone) ofthe Beliefs. A subject would have achance probability of .33 of meeting these

more liberal criteria, as determined by thebinomial theorem. One adult, one 7-year-old, two 6-year-oIds, and one preschoolershowed this pattem. Note that even whenwe combine these subjects with those in theDifferentiated group, the results still weighagainst the claim that preschoolers overallunderstand biological inheritance. A sec-ond-order application of the binomial theo-rem indicates that the two preschoolers andthe six 6-year-olds showing a Differentiatedor Partially Differentiated pattem are notsignificantly more than would be expectedby chance at each age, whereas the 10 7-year-olds (p < .02) and 16 adults (p < .001)are significantly more than would be ex-pected.

Explanations.—The children were en-couraged to explain their judgments (adultswere not asked for explanations). The expla-nations were coded to reveal evidence ofwhether some children explicitly knowcausal mechanisms that selectively underlieresemblance to parents on physical traitsand those that underlie resemblance on be-liefs. Therefore, the explanations werecoded as Biological Parent, Nurture, orOther (described below). Each response wasassigned to the most sophisticated categorypossible. Two experimenters coded the jus-tifications, independently of the judgmentsto which they applied, with 95% agreement;disagreements were resolved by fisticuffs.

The Biological Parent justificationswere those responses that invoked the dif-ference between the biological and adoptiveparents in explaining the feature judgments.This category included a wide variety of pos-sible responses all making reference to thefundamental difference between the fathers("because he was born with the shepherd,""because the king was his real father," etc.).There were children who used "he's his fa-ther" to justify both biological and adoptivefather judgments without distinguishing in

any other way between the fathers. Theseresponses were coded as Other.

Nurture justifications were those re-sponses that referred to the father's role inraising the boy or to a specific mechanismsuch as teaching, leaming, or self-discovery.The explanation "because he got used to it"was also considered a nurture explanationbecause it implied a change due to experi-ence. Other examples of nurture explana-tions are "because he was with him moretime, so he could have leamed," "becausehe was with cats longer," and "because theking took him out [to the movies] a lot."

The Other category simply includedthose responses not characterized by theprevious categories, even when loosely in-terpreted. The responses "I don't know" aswell as simple failures to respond were in-cluded in this category. Many ofthese expla-nations suggest that some children judgedthe boy to resemble his father on particularfeatures based on reasoning that ignoredparentage in favor of external considerationssuch as the children's own beliefs about thedesirability or truth of a particular propertyin the world (e.g., "because cats are nicer,""because it's better to share," and "becausehearts are red"). Explanations of this typelikely reflect the difficulty some childrenhad in seeing the relevance of the distinc-tion between biological and adoptive par-ents to the questions at hand, leading themto reason from the trait alone.

Children varied in the number of spon-taneous justifications that they gave, and sixchildren gave no explanations at all (three ofthem fell in the Adoptive Parent bias andthree in the Mixed pattern group). Twelveof the 14 children in the Differentiatedgroup ever gave a Biological Parent explana-tion for any trait, as did 20 ofthe 34 childrenin the Nondifferentiated groups. Interest-ingly, all seven ofthe children in the Biolog-ical Bias group gave at least one explanationof this type, whereas only one of the sevenchildren in the Adoptive Bias group did so.Thirteen ofthe 14 children in the Differenti-ated group and 13 of the 34 children in theNondifferentiated groups ever gave a Nur-ture explanation. That only one ofthe sevenchildren in the Adoptive Bias group did sosuggests that these children did not thinkthat some psychological mechanism medi-ated resemblance on all traits, but, rather,that they were at a loss to explain their judg-ments. Finally, eight of the 14 children inthe Differentiated group and 22 of the 34

Solomon et al. 157

children in the Nondifferentiated groupsever gave an Other explanation (and manyof these children, three of the children inthe Differentiated group and 16 of those inthe Nondifferentiated groups, gave explana-tions that referred to the desirability or prob-ability of the features).

Children who understand two distincttypes of mechanisms underlying resem-blance to parents should give Biological Par-ent explanations for their judgments thatphysical traits, but not beliefs, are inheritedfrom birth parents and they should give Nur-ture explanations for their judgments thatchildren acquire beliefs, but not physicaltraits, from the parents that raise them. Fur-thermore, when children judge the boy toresemble his birth parent on a Preference,Temperament, or Skill, they should give aBirth Parent explanation, and when theyjudge him to resemble his adoptive parent,they should use a Nurture explanation.(Note that our argument concerning thePreferences, Temperaments, and Skills doesnot hinge on which parent we a priori thinkthe boy will resemble.) Following this rea-soning, children were considered to haveshown such a Selective pattem to their ex-planations if (1) they gave at least one Bio-logical Parent explanation for the Physicaltraits or for a Preference, Temperament, orSkill they judged to be like that of the birthparent, but never did so for the Beliefs or forthe Preferences, Temperaments, and Skillsthey judged to be like that of the adoptiveparent; and (2) they gave at least one Nur-ture explanation for Beliefs or for a Prefer-ence, Temperament, or Skill they judged tobe like that ofthe adoptive parent, but neverdid so for the Physical traits or for the Prefer-ences, Temperaments, or Skills they judgedto be like that of the birth parent. The NotSelective pattem included all of the otherchildren.

Eleven of the 14 children in the Differ-entiated group showed the Selective pat-tem, but only two of the 34 children in theNondifferentiated groups did so. A chi-square analysis, crossing the two levels ofexplanation pattem (Selective and Not Se-lective) with the two levels of inheritancejudgment pattem (Differentiated and Non-differentiated), was significant, x^(l, N = 48)= 26.53, p < .001. This analysis supports theinference from the judgment patterns thatthe children in the Differentiated group arelikely to be explicitly aware that distinct pro-cesses mediate the acquisition of physicalfeatures and beliefs. Note that the results do

158 Child Developmient

not necessarily indicate that the children inthe Nondifferentiated groups believe thatpsychological mechanisms mediate resem-blance on Physical traits. Only six ofthe 34children in the Nondifferentiated groupsever gave Nurture explanations of their judg-ments of Physical resemblance and onlynine children ever gave Birth Parent expla-nations of their judgments of resemblanceon the Beliefs. Rather, most ofthe childrenin the Nondifferentiated groups producedOther explanations or no explanation at all.

DiscussionChildren's judgment pattems and expla-

nations tell a consistent story and appear toundermine the broad claim that preschoolchildren have a biological understanding ofinheritance. Only 56% of the 7-year-olds,25% of the 6-year-olds, and 6% of the pre-schoolers selectively differentiated betweenthe origins of physical traits and beliefs.Rather, most ofthe preschoolers and 6-year-olds displayed a Mixed pattern to their judg-ments. This pattern group may have in-cluded those who simply respondedrandomly as well as those who based theirjudgments on another set of criteria such asthe desirability or probability ofthe individ-ual features rather than upon the nature ofthe fathers. The children who judged thatthe boy would resemble the adoptive fatherin almost all respects may merely havebased their judgments on resemblance toparent alone, considering the parent thatraised the boy to be the relevant parent.Their explanations suggest that it is unlikelythat they believed that teaching and learningcan affect physical traits, for very few gaveeither Nurture or Biological Parent explana-tions.

We speculate that children with a Bio-logical Parent bias may be a transitionalgroup, both in age (only one preschoolershowed this pattem) and in conceptual de-velopment. Their explanations suggest thatthey understand that birth plays an impor-tant role in resemblance to parent, but theydo not yet appear to have worked out its rolein mediating resemblance on only certaintraits. These children judged that virtuallyall traits would resemble those of the birthfather. They may realize the importance ofbirth in determining who the "real" fatheris, and then globally judge the boy to resem-ble his "real father" without differentiatingamong types of traits. It is unlikely that thesechildren truly considered birth as part oftheprocess that fixes all of the boy's traits, forsuch judgments are baldly inconsistent with

the awareness children of this age have dem-onstrated previously that processes such asteaching and learning fix beliefs (e.g., Ho-grefe, Wimmer, & Perner, 1986; Sullivan &Winner, 1991; Zaitchik, 1991).

These results undermine Springer andKeil's (1989) claim that young children un-derstand biological inheritance, eventhough they eure actually in accord with theirresults. Springer and Keil found that most4—6-year-old children do not consider the or-igin of a feature (whether the parents hadacquired the feature through birth orthrough life experience) to be a factor intheir judgments of whether offspring wouldbe born with that feature. In other words,their children, like most of ours, failed toappreciate that birth plays a central and se-lective role in the chain of causation thatfixes physical traits. And, like us, they foundthat it is only at about age 7 that most chil-dren begin to show such an understanding.

Study 2: What Can Change and How

It is possible that children understandthat some physical features are initially de-termined at birth, by the same process bywhich the animal comes into being, but alsobelieve that these features can later changeso as to match those of an adoptive parent.For example, one child in Study 1 judgedthat the prince started out with brown eyes(like his biological father), but that theyturned green when he was adopted, becausebeing adopted is like "being reborn." Inother words, children may understand thebiological inheritance of physical features,but allow for changes that adults do not. ifso, then Study 1 may have underestimatedchildren's understanding of biological inher-itance. The principal goal of Study 2 was toassess whether children have intuitionsabout feature change that are radically dif-ferent from those of adults. Subjects wereasked directly whether a given set of fea-tures could change over the course of an in-dividual's life. For those features judgedchangeable, subjects were then asked to ex-plain their judgments.

There is evidence that young childrendo share many of our adult intuitions aboutfeature change. For example, young chil-dren know that animals, unlike inanimateobjects, grow in size over time, and thatsome mental features (e.g., having a quicktemper) as well as some physical features(e.g., weight) can change, but they also knowthat some bodily features (e.g., skin color)

Solomon et al. 159

do not change (Hirschfeld, in press; Inagaki& Hatano, 1993; Rosengren etal., 1991). Ina-gaki and Hatano (1993) have further shownthat children understand that some bodilyproperties are not subject to a person's inten-tional control. They use this to claim thatchildren distinguish two different explana-tory frameworks for understanding distincttypes of change in features, but this claimrequires closer scrutiny. Inagaki and Hatanoanalyzed their subjects' explanations in or-der to determine whether children under-stand that the processes mediating thechange of physical features are differentfrom those mediating the change of mentalfeatures. They claimed that children weremore apt to attribute changes in physicaltraits to physical practice (e.g., practice run-ning to improve running speed) and to attri-bute change in mental traits to mental prac-tice, volition, or effort (e.g., practice sayinga phone number to improve memory). Thisclaim is problematic, for though it is clearthat researchers differentiate the two sortsof practice, it is not clear that children do.

A second goal of Study 2 was to relatechildren's understandings of feature changeto their understandings of inheritance (thesame subjects who took part in this studyalso took part in Study 3, an inheritance taskvery much like that of'Study 1). This relationbetween these understandings is important,for biologically inherited features are thosethat are present at birth and do not changeduring the life cycle, or those that emergethrough some maturational bodily processesand not through teaching. Of course, oneprerequisite for constructing an understand-ing of inheritance is accepting the explana-tory goal of accounting for the properties ofindividual people. Study 2 focuses childrenon this explanatory goal by asking them toreason about which features change andwhat causes these changes. Notice that ifchildren have accepted this explanatory goaland understand the relevance of changeabil-ity to determining feature resemblance, thensuccess on our inheritance task requires onlyinferences of the following sort: skin colordoes not change over development; babiesresemble their parents in skin color; there-fore, children's skin color will be the sameas their birth parents'.

A final goal of Study 2 was to investigatewhether children's judgments about featurechange were influenced by the desirabilityof the features in question, as was suggestedby many of the children's Study 1 explana-tions. The specific mental features that chil-

dren had judged to be modifiable in Inagakiand Hatano's (1993) experiment were bothundesirable temperaments (quick tem-peredness and forgetfulness), and the twobodily features judged modifiable (runningspeed and weight) also involved a changefrom less desirable (slow and skinny) tomore desirable states. It is not clear that thejudgments of Inagaki and Hatano's subjectswould have been quite so adult-like had theproposed changes been from the more to theless desirable states.

MethodSixteen new subjects in each of four age

groups, adults, 7-year-olds (M = 7-7, range= 7-0 to 8-4), 6-year-olds, (M = 6-6, range= 6-0 to 6-11), and preschoolers (M = 5-2,range = 4-3 to 5-11), took part in Studies2 and 3. They were drawn from the samepopulation as those in Study 1.

The 10 feature pairs were like the 18used in Study 1. In order to shorten the task,fewer trait pairs of each type were used: fourPhysical traits, two Beliefs, two Preferences,and two Temperaments (see Table 3). TheSkills traits were eliminated entirely be-cause they provided no additional informa-tion about children's reasoning; the childrenin Study 1 had judged and given explana-tions for the skills very much as they hadfor the preferences and temperaments. Theindividual beliefs and temperaments werechosen in order to test systematicallywhether desirability or probability played asignificant role in children's judgments.Both of the temperaments were chosen be-cause children were assumed to know whichvalue was more socially or psychologically

TABLE 3

LIST OF STUDY 2 FEATURE PAIRS, BY TRAIT TYPE

Physical traits:is short/is tallis black/is whitehas liver on right/has liver on lefthas baby teeth/has grown-up teeth

Beliefs:believes that a red light means stop/believes

that a red light means gobelieves that a lion has 32 teeth/believes

that a lion has 36 teethPreferences:

likes cats more than dogs/likes dogs morethan cats

likes coconut/does not like coconutTemperaments:

doesn't share things/shares thingslaughs a lot/cries a lot


desirable. One of the beliefs (whether a redlight means stop or go) was chosen becauseit was assumed that children would knowwhat was true. The other belief (how manyteeth a lion has) was chosen as a contrastbecause it was assumed that children wouldnot know what was true. Two ofthe physicaltraits do change in the normal course ofgrowth (the Mutable Physical traits) and twodo not (the Immutable Physical traits). Bothtypes were presented to avoid confoundingthe modifiability of a trait with its beingphysical. The features were presented inblocks by trait type. Half the subjects werepresented with the Physical traits first, fol-lowed by Beliefs, Preferences, and Temper-aments. The other half received the Beliefsfirst, followed by the Physical traits. Prefer-ences, and Temperaments. The order of pre-sentation of feature values within pairs wascounterbalanced.

The subjects were asked whether thefeatures of a child named Fred could changeas he grew up (e.g., "Fred is a little boyabout the same age you are. Fred's liver ison his left side. Could Fred change as hegrows up so that his liver is on his right?").The subjects gave a yes/no response. If theyresponded yes, they were asked to explainin their own words how this change mightoccur. After these free explanations, theywere asked to judge the acceptability (yes/no) of both a body-related explanation ("Isthere something about Fred's body thatmight make him change like that?") and ateaching/learning explanation ("Couldsomebody teach Fred to become taller?").In order to facilitate comparisons betweenStudies 2 and 3, subjects were given corre-sponding versions. For example, those sub-

jects who were told in Study 2 that Fred waswhite and were asked whether he couldchange to black as he grew up were told inStudy 3 that the biological mother was whiteand the adoptive mother black.

ResultsYes/no responses.—All 16 of the adults,

7-year-olds, and 6-year-olds and 11 of thepreschoolers judged that neither of the Im-mutable Physical traits could change. Theprobability of showing such a pattern bychance responding is .25. A second-order ap-plication ofthe binomial theorem indicatesthat even the 11 preschoolers are signifi-cantly more than would be expected bychance (p < .001). The results weigh againstthe strong claim that children have a radi-cally different sense that certain physicaltraits can change. But this is not to say thattheir judgments of feature change were en-tirely adult-like.

Desirability I probability Response Bias.—It was assumed that subjects might basetheir judgments on the desirability of thefeatures and not on any principled reasoningabout feature acquisition and change. There-fore, for each of the comparisons in whichone state is more desirable than the other(the red light belief and both tempera-ments), half of the subjects were asked aboutchange from the more desirable to the lessdesirable feature, and half were asked aboutchange in the reverse direction.

Table 4 shows that the adults judged allof these features to be changeable, regard-less of desirability. The children showedquite a different pattem. That is, the chil-dren were disproportionately more likely tosay that Fred could go from believing that a

TABLE 4

DESIRABILITY JUDGMENTS FROM STUDY 2

DIRECTION OF CHANGE

CHILDREN'SJUDGMENT

Yes

233

2214

196

No

121

210

518

ADULTSJUDGMENTS

Yes

88

88

88

No

00

00

00

Belief: trafRc light:Undesirable to desirableDesirable to undesirable

Temperament: sharing:Undesirable to desirableDesirable to undesirable

Temperament: mood:Undesirable to desirableDesirable to undesirable

Solomon et al. 161

red light means go to believing that it meansstop than from believing that it means stopto believing that it means go, x (̂l> ^ = 48)= 33.57, p < .001; they were disproportion-ately more likely to say that he could go fromnot sharing to sharing than the reverse, x (̂l>N = 48) = 7.11, p < .008; and that he couldgo from being angry a lot to laughing a lotthan the reverse, x^(l, N = 48) = 14.11.p < .001.

These results bear directly on Inagakiand Hatano's (1993) finding that childrenjudged mental and physical features to bemodifiable much as adults might. The partic-ular modifiable features they used were un-desirable temperaments or bodily featuresthat the children then judged could changeinto more desirable features. In light of ourfinding that desirability is a large factor inchildren's reasoning about feature change,the children in Inagaki and Hatano's studymay not have been as adult-like in their rea-soning about feature change as they wouldappear.

The Study 2 results are consistent withthe suggestion from Study 1 that children arelikely to weigh a feature's desirability toostrongly in reasoning about feature acquisi-tion. This tendency cannot result from chil-dren's ignorance ofthe mechanisms of belieffixation, for we know from the theory ofmind literature that 4-year-olds are able toattribute false beliefs on the basis of anactor's exposure to information (e.g., Ho-grefe et al., 1986; Sullivan & Winner, 1991;Zaitchik, 1991). It may be that the childrenunderstood the task not so much as testingthe possibility of change but the probabilityof change, and it is indeed highly improba-ble that someone in our culture could be-lieve that, for example, a red light means go.Perhaps the young children lent such prag-matic factors greater weight than did theadults who, with their metacognitive under-standing that these questions are about theinfluence of nature and nurture, focusedonly on the mechanism relevant to such aconflict and ignored the feature's desirabil-ity. Many of the children, by contrast, mayhave failed to see the relevance of mecha-nisms of change and so resorted to consider-ations of probability or desirability.

Free explanations.—Children's free ex-planations for those features they consideredchangeable were coded and analyzed. Ofparticular interest is whether liiey un-derstood that bodily processes mediatechange in physical features and teaching

and learning mediate change in beliefs.Two researchers coded the explanationsinto the following three categories with 92%agreement.

Teaching explanations included all ex-plicit appeals to teaching and learning aswell as tiiose justifications that referred tosome particular piece of evidence that Fredwould acquire (e.g., Fred might change hisbelief about the number of teeth a lion hasbecause "he will go to a zoo, not a zoo, to a,yeah, a gypsy carnival and see them doingtricks and the lion might open his mouth andhe could count very quickly").

The Body-related explanations in-cluded those that referred to some bodilyprocess. Almost all were of the specific sort"you get taller because you eat" or "you getgrown-up teeth because your baby teeth fallout." Appealing to Fred's growing up as anexplanation for how features change as hegroics up is tantamount to restating thegiven infonnation. Such explanations de-scribe no processes and so were consideredto be Other explanations.

The Other explanations did not appealto mechanisms at all. They included thoseexplanations that merely restated the child'sjudgment that the feature will change, ap-peals to a canonical direction of change (e.g.,"because everyone gets taller"), confirminginstances (e.g., "because 1 changed likethat"), and mere restatements ofthe informa-tion provided (e.g., "he'll get taller becausehe'll grow"). These explanations also in-cluded appeals to the feature's desirability(e.g., "because it's better to be tall") or truth(e.g., "because red lights really do meanstop").

Nine of the preschoolers, 10 of the 6-year-olds, and 12 of the 7-year-olds evergave a Teaching explanation for any trait.Eight ofthe preschoolers, nine ofthe 6-year-olds, and 11 of the 7-year-olds ever gave aBody-related explanation for any trait. Al-most all of the children (81%) gave at leastone Other explanation. Of these, five of thepreschoolers, nine ofthe 6-year-olds, and sixof the 7-year-olds explained the change interms of the desirability of the features. Fi-nally, two ofthe preschoolers, two ofthe 6-year-olds, and one of the 7-year-oIds nevergave any explanations.

Recall that we were particularly inter-ested in whether children would appeal toteaching explanations for changes in beliefsand bodily explanations for physical


changes. We considered children to haveshown a Selective explanation pattern if theygave a Body-related explanation for thechange of at least one Physical trait, butnever for a Belief, and if they gave at leastone Teaching explanation for at least oneBelief, but never for a Physical trait. Anyother pattern was considered a Not Selectiveexplanation pattern. Five of the preschool-ers, four of the 6-year-olds, and eight of the7-year-olds showed the Selective explana-tion pattern. A chi-square analysis crossingthe two levels of explanation pattem typewith the three levels of children's age wasnot significant.

Consistent with Inagaki and Hatano's(1993) findings, teaching was never used toexplain physical changes, but fewer thanhalf of the children showed a Selective ex-planation pattern. Recall that after givingtheir free explanations, children were alsoasked to judge whether two presented expla-nation types could explain the changeabilityofthe trait ("Is there something about Fred'sbody that might make him change like that?"and "Could somebody teach Fred to become[feature]?"). Unfortunately, the question isflawed, so it is not clear that children under-stood the body explanation that was beingproposed.'' Therefore, we have dropped thePresented Explanations from our analyses.It is striking that only about half of the pre-schoolers referred to bodily mechanisms atall in their free explanations given that forevery trait they judged mutable they wereasked whether there "was something aboutFred's body that might make him change."

DiscussionThe results of Study 2 demonstrate that

children's failure in Study 1 cannot be attrib-uted simply to a belief that immutable physi-cal traits can change. Even preschoolersdemonstrated, in their judgments of possiblechange of physical traits and in many of theirexplanations, that they generally share ouradult intuitions. Their explanations furtherreveal that they understand that teachingand learning will not affect even those phys-

ical traits that can change. These results areconsistent with the finding that children dis-tinguish between mental and physical ob-jects (Wellman & Gelman, 1992) and real-ize that there are some bodily phenomenaover which we do not have intentional con-trol (Inagaki & Hatano, 1993). Study 3 allowsus to examine how children's awareness ofthe constrained processes involved in fea-ture change relates to their awareness oftheprocesses involved in inheritance.

Study 3: Similarity to Adoptive andBiological Parents, Redux

The results of Study 1 undermine theclaims that preschool children understandbiological inheritance. This may be because,as we have claimed, previous studies failedto examine the necessary contrast betweenbiological and adoptive parentage. The prin-cipal goal of Study 3 was to relate children'sjudgments of resemblance to parents to theirStudy 2 judgments of feature change; thisconjunction of tasks addresses the possibleobjection that children performed as theydid in Study 1 because they have a differentunderstanding of what physical traits canchange over a lifetime. Study 3 also ad-dressed the possibility that certain method-ological aspects of Study 1 masked chil-dren's understanding. For example, thesechildren may have been less aware of thebiological relations between children andtheir fathers than of that between childrenand their mothers. Furthermore, the sheernumber and variety of features probed mayhave simply overwhelmed the youngestchildren.

MethodSubjects took part in Study 3 immedi-

ately after Study 2. The stimuli and proce-dure were nearly identical to those of Study1, but with changes designed to lessen thetask demands. First, the task was shortenedto include nine instead of 18 feature pairs:three physical traits, two beliefs, two prefer-ences, and two temperaments (see Table 3).The features were identical to those used in

•* In accepting the body explanation for the change of Beliefs, the children might havereasoned that because thinking occurs in the brain, and because the brain is part of the body,there is something about Fred's body that could cause him to change beliefs. Such childrenwould then have seemed not to distinguish bodily and psychological mechanisms. Moreover, itis not even clear what we are to infer from children's acceptances of the body explanation forthe change of physical traits. It verges on the tautological to say that something about the bodyexplains changes in the body. The children may simply be answering that the physical traits areofthe body. Finally, because the explanations are for those features that were judged changeable,we get a skewed sample, for we do not represent the thinking of the children on features theyerroneously judged not changeable.

Solomon et al. 163

Study 2, except that the teeth feature pair(forcing children to choose between a parentwith grown-up teeth and one with babyteeth) was dropped. Second, because chil-dren understand some role of the mother inbirth and creation long before they under-stand the involvement of the father (Bern-stein & Cowan, 1975), the story was changedto involve a queen and a shepherdess ratherthan a king and shepherd (see Appendix B).Third, the procedure no longer involved thepictures ofthe parents and the baby becauseof the possibility that those pictures maysomehow have influenced the siibjects'judgments in Study 1. The pictures of thecastle and the hut were still used. The traitswere grouped by type to facilitate the chil-dren's abilities to access and apply whateverknowledge they think relevant. Finally, thestudy was balanced across subjects to controlfor the order in which the blocks of traitswere presented (i.e.. Physical, Belief, Pref-erence, then Temperament as opposed toBelief, Physical, Preference, Temperament),for whether the queen or the shepherdesswas associated with particular features (e.g.,half of the subjects were told that the queenis short and half were told that the shepherd-ess is short), and for whether the queen orthe shepherdess is the biological mother. Asin Study 1, a preliminary analysis revealedno significant order effects.

ResultsPattem of responding over trait types.—

The subjects' overall judgment patternswere sorted according to the patterns of re-sponding introduced in Study 1. As in Study1, we note that subjects at all ages generallyjudged the acquisition of the Preference-Temperament traits and the Beliefs simi-larly. Therefore, we focus on the contrast ofthe physical traits and beliefs. The Differen-tiated pattern group was defined to includethose children who had judged the girl toresemble her biological mother on all threeof the physical traits but on none of the be-

liefs. The binomial theorem indicates that asubject has a chance probability of .03125 ofshowing this pattem. Of those subjects whodid not meet these criteria for inclusion intothe Differentiated pattern group, three fur-ther categories were defined. Subjects wereconsidered to have shown an Adoptive Par-ent bias pattem if they judged the girl toresemble her adoptive parent on more thanseven of the nine features (significantlymore than would be expected by chance).Conversely, subjects were considered tohave shown a Biological Parent bias if theyhad judged the girl to resemble her biologi-cal parent on more than seven ofthe features(significantly more than chance). Finally,those patterns not described by the abovecategories were considered Mixed patterns.The results are presented in Table 5.

The number of subjects showing theDifferentiated pattern as opposed to anyother pattern varied significantly by age,X (̂3, iV = 64) = 23.90, p < .001. A second-order application of the Binomial theorem,based on a chance probability of .03 of show-ing this pattern, indicates that the 16 adults(p < .001), 10 7-year-olds (p < .001), six 6-year-olds (p < .001), and three preschoolers(p < .05) showing this pattem were all moreat each age than one would expect bychance. Again, it is likely that some pre-schoolers command the understanding of bi-ological inheritance tapped by this task. Weemphasize, rather, that there is insufficientground for the claim that preschoolers as agroup understand biological inheritance.Most of the children under 7 years of ageshowed a Nondifferentiated pattern.

As in Study 1, the Adoptive Parent biaswas shown more by the preschoolers thanany other age group, the Biological Parentbias was shown most by the 6-year-olds, andrelatively fewer ofthe 7-year-olds than oftheyounger children fell into the Mixed cate-gory. Again, the Mixed pattern group was

TABLE 5

PATTERNS OF JUDGMENTS FROM STUDY 3, BY AGE

AGE

PATTERN

Differentiated ..Biological bias .Adoptive bias ..Mixed

Adults

16000

7

10015

6

6208

Preschoolers

3139

164 ChUd Development

further analyzed to identify the subgroup ofsubjects who showed a Partially Differenti-ated pattern. This pattern was liberally de-fined to include those subjects who judgedthe girl to resemble her biological mother onmore than half (two or three) of the Physicalfeatures and on at most half (one or none) ofthe Beliefs. Even when we combine the four7-year-olds, two 6-year-olds, and three pre-schoolers who showed this pattern withthose in the Differentiated group, we do notfind that most of the 6-year-olds and pre-schoolers meet these more liberal criteria.A second-order application of the binomialtheorem (based on a chance probability of.38 of showing such a pattern) indicates thatthe six preschoolers and eight 6-year-oldsshowing a Differentiated or Partially Differ-entiated pattern are not significantly morethan would be expected by chance at eachage, whereas the 14 7-year-olds (p < .001)and 16 adults (p < .001) are significantlymore than would be expected. These resultssupport the Study 1 findings that most chil-dren under the age of 7 did not demonstratean understanding of biological inheritance.

More children showed the Differenti-ated pattern and the Partially Differentiatedpattern in Study 3 than in Study 1, but com-parisons between children's performancesin Studies 1 and 3 must be made cautiously.The binomial theorem indicates that thechance probability of showing a Differenti-ated pattern is slightly greater in Study 3(.03) than it is in Study 1 (.02), as is the prob-ability of showing a Partially Differentiatedpattern (.38 as compared to .33). Nonethe-less, the distribution of patterns by ages inthe two studies is quite similar.

Explanations.—The subjects wereasked to justify all of their judgrnents andtheir responses were coded according to theStudy 1 classifications {Biological Parent,Nurture, and Other). Two researchers codedthe explanations with 99% agreement.Again, we looked for the association of Bio-logical Parent explanations with acquisitionof Physical traits and Nurture explanationswith the acquisition of Beliefs. As in Study1, we considered children to have shown aSelective explanation pattern if they gave atleast one Biological Parent explanation forthe physical traits and never for the beliefs,and gave at least one Nurture explanationfor the beliefs, but never for the physicaltraits. The Not Selective pattern included allof the other children.

Fourteen of the 19 children in the Dif-ferentiated group showed the Selective pat-tern, but only three of the 29 children in theNondifferentiated groups did so. A chi-square analysis, again crossing the two lev-els of explanation pattern (Selective and NotSelective) with two levels of inheritancejudgment pattern (Differentiated and Non-differentiated) showed a significant associa-tion between explanation pattern and inheri-tance judgment pattern, x^(l. ^ = 48) =20.13, p < .001. These analyses of the chil-dren's explanations were consistent with theinheritance judgment results: The childrenin the Differentiated inheritance patterngroup were far more likely than the otherchildren to make explanations differentiat-ing between those processes mediating theacquisition of physical features and thosemediating the acquisition of beliefs.

Comparison to Study 2 changeabilityjudgments.—A comparison of Studies 2 and3 allows us to reject the possibility that thosechildren who judged offspring to resembletheir adoptive parents on physical featureshad a radically different sense from adults ofwhat features can change. Children as agroup made adult-like changeability judg-ments (at least on those feature pairs forwhich one value was not clearly preferable),but still failed to make adult-like inheritancejudgments. Let us look at the judgments inthe two studies on a within-subject basis.The Immutable Physical traits of race andliver position bear most directly on the rela-tion of children's understanding of biologi-cal inheritance to their understanding of fea-ture change. Twenty-one times childrenjudged the girl to resemble her adoptivemother on these traits, but only four timesdid they also judge those traits changeableafter birth.

Many children appeared unaware thatchangeability had any relevance whatsoeverto judgments about inheritance. For exam-ple, one child judged the girl to be whitelike her adoptive mother, though he hadminutes earlier judged that a person couldnot change from being black to being white.After the experiment, when asked again forhis judgments on race, he repeated his ear-lier responses. When asked what color thelittle girl would be when she was bom, hesaid "white [the color of the adoptive par-ent], because race cannot change." Clearly,children's notions of the changeability of aphysical feature does not adequately ac-

count for why their resemblance judgmentsdiverge from those of adults.

Comparison of Studies 2 and 3 allowsus to pursue the question of whether thosechildren who understand changeability inStudy 2 to be mediated by selective mecha-nisms will be more likely to understand re-semblance to parents in Study 3 to be medi-ated by selective mechanisms (though notnecessarily the same mechanism). Nine ofthe 17 (53%) children who showed a selec-tive pattern of explanation of changeabilityin Study 2 also showed a selective patternof explanation of resemblance to parent inStudy 3, whereas only eight of the 31 (26%)children who did not show a selective pat-tern of explanation in Study 2 did show aselective pattern of explanation in Study 3.A chi-square analysis, crossing two levels ofStudy 2 explanation pattern (Selective vs.Not Selective) with two levels of Study 3explanation pattern (Selective vs. Not Selec-tive) shows that this association falls justshort of significance, x^(l, IV = 48) = 3.54p = .06.

DiscussionThe results of Study 3 were very much

in keeping with those of Study 1: It is only atage 7 that most children make differentiatedjudgments and give explanations that revealan understanding of birth as part of a causalchain mediating the fixing of immutablephysical features. Study 2 shows that chil-dren do understand that such features asrace and liver position cannot change; whatmost do not understand is the role of thebiological parent in determining them in thefirst place. In Study 2, explanations that re-ferred to explicit mechanisms of change didnot account for more than about a third of allexplanations given by children at any age.But when the children are grouped by Study3 inheritance pattern, we see a coherence totheir judgments. Children in Study 3 whowere more likely to make explanations thatimplied that they were aware of distinctcausal chains mediating resemblance to par-ents (those in the Differentiated groups)were more likely to make Study 2 explana-tions that implied an awareness of distinctcausal chains mediating feature change.These tasks may well tap the same system ofunderstanding. There may be a conceptualcoherence to children's understandings ofthese seemingly separate phenomena; un-derstanding which features can and cannotchange is related to an understanding ofwhat mechanisms determine what those fea-

Solomon et al. 165

tures are likely to change into. It is this kindof coherence of understanding that definesa domain of biology.

Study 4: Similarity to Adoptive andBiological Parents, Redux Redone

Study 4 addressed the possibility thataspects of the methodologies of Studies 1and 3 prevented young children from dem-onstrating an understanding of inheritancethey in fact possess. Perhaps our placing thestory in the context of a fairy tale led chil-dren to accept magical transformations. Andperhaps our including traits that differed indesirability and our including preferencesand temperaments distracted children fromthe contrast of interest. Study 4 differed fromStudies 1 and 3 in each of these respects andalso in its simplified wording of the generalinstructions. At issue is whether preschool-ers would now succeed at the task.

MethodSixteen preschoolers (M = 5-0, range =

4-2 to 5-10), drawn from the same populationas those in the previous three studies, tookpart in the study. Subjects were told anadoption story in which a woman in a hospi-tal dies immediately after giving birth to ababy girl and another woman then adoptsthe baby girl and takes her back to live withher (see Appendix C). In addition toavoiding the fairy tale nature of the previousstories, the hospital story also eliminates thepossibility that the biological mother couldhave visited the adopted child and so influ-enced her. Furthermore, the subjects wereno longer shown pictures of the homes of themothers, but rather were shown schematicpictures of the mothers in order to lessen thefocus on environmental factors. The wordingof the instructions was simplified and morerepetition was included in order to make itless likely that subjects simply did not un-derstand what judgments they were beingasked to make. The list of features no longerincluded preferences or temperaments, butonly five physical features and five beliefs,the contrast of most interest (see Table 6).Moreover, subjects were asked to judge onlyimmutable physical features so as to focustheir attention on how those features wereinitially determined. Finally, the featureswere chosen so that one feature of each pairwas not inherently more desirable than theother.

As in Study 3, the Beliefs and Physicaltraits were presented in blocks to highlight


TABLE 6

LIST OF STUDY 4 FEATUKE PAIRS, BY TRAIT TYPE

Physical traits:Brown eyesLiver on right [something

inside you]Turned up nose [like tbis]

Blonde hairFlat appendix [something

inside you]

Beliefs:Lion has 40 teethAardvarks eat only meatTo bake a cake put oven

at 300°Mr. Rogers lives in a city

called ChicagoRock called feldspar is

gray

Green eyesLiver on left

Turned down nose [likethis]

Red hairRound appendix

Lion has 30 teethAardvarks eat only plantsTo bake a cake put oven at

400°Mr. Rogers lives in a city

called SeattleRock called feldspar is

black

the contrast of trait types. The study was alsobalanced across subjects to control for theorder in which the blocks of traits were pre-sented (i.e.. Physical then Beliefs as op-posed to Beliefs then Physical), for whichparent was described first (the biological orthe adoptive parent), and for the order inwhich the five feature pairs within each traittype were presented.

Results and DiscussionPattern of responding over trait

types.—The children's judgments were ana-lyzed according to the Study 1 and 3 inheri-tance pattern groups. The Differentiatedpattem group was defined to include thosechildren who had judged the girl to resem-ble her biological mother on at least four ofthe five Physical traits and on at most one ofthe Beliefs. The binomial theorem indicatesthat a subject has a chance probability of.03516 of showing this pattern. Of those sub-jects who did not meet these criteria for in-clusion into the Differentiated patterngroup, three further categories were defined.Subjects were considered to have shown anAdoptive Parent bias pattern if they judgedthe girl to resemble her adoptive parent onmore than seven of the 10 features (signifi-cantly more than would be expected bychance). Conversely, subjects were consid-ered to have shown a Biological Parent biasif they had judged the girl to resemble herbiological parent on more than seven of thefeatures (significantly more than chance). Fi-nally, those patterns not described by theabove categories was considered Mixed pat-terns.

None of the preschoolers showed theDifferentiated pattem. Eight preschoolersshowed the Adoptive Parent bias, fourshowed the Biological Parent bias, and fourshowed a Mixed pattern. Again, the Mixedpattem group was analyzed to identify thesubgroup of subjects showing a PartiallyDifferentiated pattern. This pattern was de-fined to include those subjects who judgedthe girl to resemble her biological mother onmore than half (at least three) ofthe Physicalfeatures and on at fewer than half (at mosttwo) of the Beliefs. The Binomial theoremindicates that subjects have a chance proba-bility of .25 of showing this pattern. A sec-ond-order application of the binomial theo-rem indicates that the two preschoolers whoshowed this pattern aire not significantlymore than would expected by chance.Therefore, despite methodological alter-ations, these results support the Study 1 and3 findings that most preschoolers did notdemonstrate an understanding of biologicalinheritance.

General Discussion

Callanan and Oakes (1992, pp. 221-222)provide a striking example of a 4-year-old'sunadult-like understanding of resemblanceto family:

Child: Why does Daddy, James [big brother],and me have blue eyes and you have green eyes?

Parent: [Told ber she got her eyes fromDaddy. Then said goodnight and left the room.]

Child: [Child calls mother back 5 minuteslater.] I like Pee Wee Herman and I have blue

eyes. Daddy likes Pee Wee Herman and he hasblue eyes. James likes Pee Wee Herman and hehas blue eyes. If you liked Pee Wee Herman youcould get hlue eyes too.

. Parent: [I told her it would take more thanmy liking Pee Wee Herman to make my eyes blue.I realized that she didn't understand me, so I ex-plained that God gave me tbis color and that theycouldn't be changed.]

Child: Could you try to like Pee Wee Her-man so we could see if your eyes turn blue?

Consistent with this anecdote, the results ofour studies refute the broad claim that pre-school children understand biological inher-itance. This child expected resemblance tofamily for eye color, but had no ideas aboutthe biological mechanisms underlying i t Asargued above, a child's understanding of in-heritance is uniquely biological only if it isembedded within a conceptual frameworkthat explains resemblance to parent on phys-ical traits in a manner distinct from that ex-plaining resemblance on beliefs. With ouradult notions of biology and psychology, weunderstand that children adopted at birthwill tend to resemble their biological par-ents on bodily traits because of one chain ofcausal processes (of which birth is a crucialelement) and their adoptive parents on be-liefs because of another chain of processes(typically involving teaching or learning).For the most part, the preschoolers and 6-year-olds in Studies 1, 3, and 4 neither dif-ferentiated among features by trait type indeciding which parent the offspring wouldresemble, nor differentiated among traittypes in their explanations. They failed ourtask.

What might the children's failure mean?Many of them demonstrated a bias to judgethe little boy or girl to resemble one parenton virtually all traits. These children maynot have been thinking about causal pro-cesses at all, but may simply have basedtheir judgments upon their knowledge thatoffspring resemble their parents, decidedwhich parent was the "real" parent, andthen judged the offspring to resemble thatparent on all traits. It is possible that some ofthe children who showed a biological parentbias may have begun to understand thatbirth plays a role in determining who is the"real" parent, but they have evidently failedto understand how birth might be part of aseries of causal processes that selectivelyfixes bodily traits and not beliefs. The chil-dren who showed a Mixed pattern alsolacked this understanding. Many of thesechildren may simply have judged the fea-

Solomon et al. 167

tures randomly. But in Studies 1 and 3, oth-ers may, in a nonrandom fashion, have con-sidered the desirability or probability of afeature to be of determining importance, re-gardless of parentage. In either case, thesechildren did not differentiate among fea-tures in a systematic enough fashion to indi-cate that they understood resemblance toparents to be mediated by distinct chains ofcausal processes.

The failure of the young children wasnot due to their lack of some domain-generalability to reason' causally, for there is con-vincing evidence that they understand cau-sation within the cognitive domains of psy-chology and mechanical physics (e.g., Carey& Spelke, 1994; Leslie, 1994; Wellman &Celman, 1992). And Study 2 indicated thatthis failure was not due to their having a rad-ically different sense from adults of whatphysical features can change. Instead, thechildren's failure is likely due to their igno-rance of uniquely biological mechanismsrelevant to the origin and determination ofbodily features.

Previous studies directly asking pre-school children about origins reveal thatthey know that people are involved in theorigin of artifacts and their properties, butnot in the origin of animals and plants andtheir properties (Celman & Kremer, 1991;Keil, 1989; Springer & Keil, 1991). Butknowing that people are not involved insome process is not the same as knowing thenature of the process. When directly askedhow it is that rabbits come to hop or to havelong ears, 4-7-year-oIds have very little tosay.

Springer and Keil (1991) attempted tocircumvent young children's lack of knowl-edge of specific processes by asking them tochoose from a set of explanations for how,for example, a fiower gets its color. They de-picted a flower growing inside anotherfiower and asked how the baby fiower cameto be blue. Children accepted explanationsin terms of sun and rain falling on the fioweror those in which the mother gave the flowersomething that made it turn blue. They fur-ther accepted a mediating process in whichthe sun and rain melted some of the bluefrom the mother flower onto the baby. Thisstudy is very important, for it addresses howchildren constrain the explanations they findplausible for phenomena they as yet cannotexplain, but it does not establish that theyhave constructed any specifically biologicalexplanations for inheritance before age 7.


The study does confirm Celman andKremer's (1991) finding that children do notthink that human agency plays a role inflower color. But beyond that, the children'schoices of explanations are simply con-strained by what they know about flowersand color. For example, the children ac-cepted the explanation that "the flower turnsblue because the sun and rain fell on themother when the little flower was growinginside it." This, by itself, provides no mech-anism at all; it is something children knowabout flowers—that sun and rain are neces-sary for growth. Furthermore, the explana-tions simply reflect general knowledgeabout color transfer: children only accept ex-planations wherein something becomesblue because it is covered with somethingblue. Finally, the study does not establishthat children understand the role of repro-duction in inheritance, for though they weretold that the flowers were babies growinginside mothers, children of this age think itis a category mistake to talk about motherplants at all (Keil, 1979).

By the age of 7, most children suc-ceeded at our tasks; they associated thephysical traits with the biological parent andthe beliefs with the adoptive parent and alsomade a sharp distinction between trait typesin giving the nurture explanations and thebiological parent explanations. What doestheir success indicate about their under-standing? Certainly, the nurture explana-tions explicitly describe genuine causal me-diating mechanisms (learning, being taught,getting used to it), but the birth parent expla-nations are ambiguous. It is not completelyclear how deep an understanding of biologi-cal inheritance these children had.

We adults understand birth not so muchas the single direct causal mechanism of in-heritance, but as part of a causal chain. Ulti-mately, it is the biological parents who causeoffspring to have certain of their biologicalproperties (as we understand it, from thetransmission of their genetic code). Onemight conceive of a number of alternativecausal paths by which biological parentscould influence their children, but onlythose causal chains that normally involvebirth can be considered as part of biologicalinheritance. It may be that the children'sbirth parent explanations merely reflectedtheir belief that the birth parent is the "real"parent, with no sense at all of causal pro-cesses involved. Yet, the fact that these chil-dren generally gave such explanations selec-tively for physical traits and not beliefs is

indirect evidence that many of them wereaware of birth as part of a causal process.The birth parent explanations, then, mayhave been a way of implicating birth as partof the chain of processes by which a brown-eyed father comes to have a brown-eyed son.Even if these children do not know whatthose processes are (and it is exceedinglyunlikely that they do), they may know thatsuch processes exist. Birth, in this sense,may function as something of a conceptualplaceholder and as a constraint on later the-ories.

It may be that those children who suc-ceeded on the inheritance task were in theprocess of constructing an autonomous cog-nitive domain of biology. Recall that thechildren in Study 2 who selectively referredto some body-related process in explainingthe change of physical traits, but not thechange of beliefs were also more likely torefer to birth in explaining resemblance onphysical traits, but not resemblance on be-liefs. These children may know that theseostensibly distinct phenomena within thedomain are indeed explanatorily related bybodily processes, if processes more inferredthan understood. One hallmark of domain-specific thinking is just this kind of theoreti-cal coherence.

Of course, most of the preschoolers didnot demonstrate an understanding of thegreater relation of biological inheritance toother phenomena within the domain. Howthey eventually come to such an understand-ing is a question of critical importance, andthe focus of ongoing research. Hirschfeld (inpress) has argued that children first under-stand that certain essential properties (racein particular) are fixed at birth and that laterchildren come to understand how otherproperties are fixed. Partially in response toa previous version of this paper, Hirschfeldconducted a study in which about three-quarters of the 4- and 5-year-old subjects ina "switched-at-birth" study judged anadopted girl to have the skin color of herbirth parents rather than that of her adoptiveparents. Hirschfeld may be correct in claim-ing that preschoolers have an essentialist un-derstanding of race, but our Study 3 resultsdirectly contradict his findings. They indi-cate that preschoolers were not significantlymore likely than chance to judge that the girlwould resemble her birth parent on skincolor. One possible explanation of this dis-crepancy is that we placed a greater empha-sis than did Hirschfeld on the adoptive fam-ily as a family. Unlike Hirschfeld, we

Solomon et al. 169

explicitly told our subjects that the little girlcalled her adoptive parent "Mother" and theparent called tiie little girl "Daughter." Afterall, we grant that preschoolers know that off-spring are more likely to resemble familymembers than nonfamily members (e.g..Springer, 1992). Studies in our lab are cur-rently exploring whether children have anessentialist understanding of species-kindmuch as Hirschfeld claims of race.

The present studies challenge the claimthat preschoolers' understanding of inheri-tance is biological. And these data also chal-lenge the claim that preschoolers have con-structed an autonomous cognitive domain ofbiology insofar as that claim rests on theirunderstanding biological inheritance. It maybe that a different way of characterizing bio-logical reasoning will better capture chil-dren's causal explanatory frameworks (e.g.,Inagaki & Hatano, 1993; Keil, 1994), but theburden is upon those researchers who makesuch claims to demonstrate that preschoolersreason about a range of phenomena in a the-oretically coherent manner. Suffice it to saythat, in this nascent domain of research, sucha claim is premature.

Appendix AThe King/Shepherd StoriesVersion 1Once upon a time, there was a king. Here is apicture of him. [Show king.] He lived in this bigpalace. [Show palace.] The king could not havechildren, but he wanted a child very much. So hewent out into his kingdom where he met a shep-herd who had many children. Here is a picture ofthe shepherd. [Show shepherd.] This is where theshepherd lived. [Show hut] The shepherd lovedall of his children and they loved him. The kingtold the shepherd that he wanted to adopt theshepherd's baby boy and raise him as his own sonand that the baby would then grow up to be theprince. This is the shepherd's baby boy. [Showbaby, place by hut, by shepherd.] The shepherdagreed that this was a good idea so the kingadopted the baby boy and took him to the palace.[Place king and baby by palace.] The king lovedthe baby and the baby loved the king. The babygrew up in the palace with the king and becamea prinee.

Version 2Once upon a time, there was a king. Here is apicture of him. [Show king.] He lived in this bigpalace. [Show palace.] The king had a baby sonwho he loved very mueh and who loved him too.[Show baby.] Here is a picture of a shepherd wholoved children but didn't have any of his own.This is where the shepherd lived. [Show hut.] Un-fortunately, there was a wicked witch who wantedto eat the king's baby. The king wanted to hide

his baby son from the witch so he let the shepherdadopt the baby and take him back to live with himat the hut. [Place shepherd and baby by hut.] Theshepherd loved the baby very much and the babyloved the shepherd. The baby grew up with theshepherd and the witch never found him.

Appendix BThe Queen/Shepherdess StoriesVersion 1Once upon a time, there was a queen. She livedin this big palace. [Show palace.] The queeneould not have children, but she wanted a childvery much. So she went out into her kingdomwhere she met a shepherdess who had many chil-dren. This is where the shepherdess lived. [Showhut.] The shepherdess loved all of her childrenand they loved her. The queen told the shepherd-ess that she wanted to adopt the shepherdess'slittle girl and raise her as her own child and thatthe little girl would then grow up to be the prin-cess. The shepherdess agreed that this was a goodidea so the queen adopted the little girl and tookher to the palace. The queen loved the little girland the little girl loved the queen. The queencalled her "Daughter" and she called the queen"Mother." The little girl grew up in the palacewith the queen and became a princess. She livedher whole life with the queen in the palace. Nowthe little girl is a young woman and we have somequestions about what you think she's like now.

Version 2Once upon a time, there was a queen. She livedin this big palace. [Show palace.] The queen hada little daughter who she loved very much andwho loved her too. In the queen's kingdom therewas also a shepherdess who loved children butdidn't have any of her own. This is where theshepherdess lived. [Show hut.] Unfortunately,there was a wicked witch who wanted to eat thequeen's little girl. The queen wanted to hide herlittle girl from the witch so she let the shepherdessadopt her daughter and take her back to live withher in the hut. The shepherdess loved the girlvery much and she loved the shepherdess. Thelittle girl lived with the shepherdess all of her lifeand the shepherdess called her "Daughter" andshe called the shepherdess "Mother." The shep-herdess raised the girl and when she grew up shebecame a shepherdess too. And the witch neverfound her. Now the little girl is a young womanand we have some questions about what you thinkshe's like now.

Appendix CThe Study 4 StoryThere was a woman named Mrs. Smith; here's adrawing of her [point to picture]. Mrs. Smith wentinto a hospital and gave birth to a baby girl. Here'sa drawing of the baby just after she was born[point to picture]. Unfortunately, Mrs. Smith diedright after having the baby and she never even gotto see the baby. Fortunately, there was a really


nice woman named Mrs. Jones who was visitingthe hospital. See, here's a drawing of Mrs. Jones[point to Mrs. Jones]. Mrs. Jones saw what hap-pened. She saw that a baby girl was born to Mrs.Smith, but that Mrs. Smith died right away [point-ing to pictures]. Mrs. Jones always wanted to havea little girl, and she saw that the little baby girlwas all alone, so she adopted the little baby girland brought her home to live with her. Mrs. Jonesloved the little girl very much and called her"Daughter" and the little girl loved Mrs. Jonesvery much and called her "Mommy." The littlegirl spent her whole life with Mrs. Jones.

Do you remember who the little baby girlwas bom to? And who did the little baby girl livewith?

Now I'm going to ask you some questionsabout what you think the little girl is going to belike when she grows up into a young woman. Mrs.Smith has [feature X] and Mrs. Jones has [featureY]. The little girl is grown up now. Does she have[X] like Mrs. Smith or does she have [Y] like Mrs.Jones? [pointing all the time].

References

Bernstein, A., & Cowan, P. (1975). Children's con-cepts of how people get babies. Child Devel-opment, 46, 77-91.

Callanan, M. A., & Oakes, L. M. (1992). Preschool-ers' questions and parents' explanations:Causal thinking in everyday activity. Cogni-tive Development, 7, 213-233.

Carey, S. (1985). Conceptual change in childhood.Cambridge, MA: Bradford Books.

Carey, S. (1988). Conceptual differences betweenchildren and adults. Mind ir language, 3,167-181.

Carey, S. (in press). On the origin of causal/ex-planatory notions. In A. J. Premack, D. Pre-mack, & D. Sperber (Eds.), Causal cognition:A multidisciplinary debate. Oxford: Claren-don.

Carey, S., & Spelke, E. S. (1994). Domain-specificknowledge and conceptual change. In L.Hirschfeld & S. Gelman (Eds.), Mapping themind: Domain specificity in cognition andculture. New York: Cambridge UniversityPress.

Flavell, J., Flavell, E., & Green, F. (1983). Devel-opment of the appearance-reality distinction.Cognitive Psychology, 15, 95-120.

Gelman, S. A. (1988). The development of induc-tion within natural kind and artifact catego-ries. Cognitive Psychology, 20, 65-90.

Gelman, S. A., & Kremer, K. E. (1991). Under-standing natural cause: Children's explana-tions of how objects and their properties origi-nate. Child Development, 62, 396-414.

Gelman, S. A., & Markman, E. M. (1987). Young

children's inductions from natural kinds; Therole of categories and appearances. Child De-velopment, 58, 1532-1541.

Gelman, S. A., & Wellman, H. M. (1991). Insidesand essences: Early understandings of thenonobvlous. Cognition, 38, 213-244.

Hirschfeld, L. (in press). The child's representa-tion of human groups. In D. Medin (Ed.), Thepsychology of learning and motivation: Ad-vances in research and theory (Vol. 30). SanDiego: Academic Press.

Hogrefe. J., Wimmer, H., & Pemer, J. (1986). Ig-norance versus false belief: A developmentallag in attribution of epistemic states. ChildDevelopment, 57, 567-582.

Inagaki, K., & Hatano, G. (1993). Young children'sunderstanding of the mind-body distinction.Child Development, 64, 1534-1549.

Keil, F. C. (1979). Semantics and conceptual de-velopment: An ontological perspective. Cam-bridge, MA: Harvard University Press.

Keil, F. C. (1989). Concepts, kinds, and cognitivedevelopment. Cambridge, MA: BradfordBooks.

Keil, F. C. (1994). The birth and nurturance ofconcepts by domains: The origins of conceptsof living things. In L. Hirschfeld & S. Gelman(Eds.), Mapping the mind: Domain specificityin cognition and culture. New York: Cam-bridge University Press.

Leslie, A. M. (1994). ToMM, ToBY, and Agency:Core architecture and domain specificity. InL. Hirschfeld & S. Gelman (Eds.), Mappingthe mind: Domain specificity in cognitionand culture. New York: Cambridge Univer-sity Press.

Murphy, G. L., & Medin, D. L. (1985). The roleof theories in conceptual coherence. Psycho-logical Review, 92, 289-316.

Rosengren, K. S., Gelman, S. A., Kalish, C. W.,& McCormick, M. (1991). As time goes by:Children's early understanding of growth inanimals. Child Development, 62, 1302-1320.

Springer, K. (1992). Children's beliefs about thebiological Implications of kinship. Child De-velopment, 63, 950-959.

Springer, K, & Keil, F. C. (1989). On the develop-ment of biologically specific beliefs: The caseof inheritance. Child Development, 60,637-648.

Springer, K., & Kell, F. C. (1991). Early differenti-ation of causal mechanisms appropriate to bi-ological and nonbiological kinds. Child De-velopment, 62, 767-781.

Sullivan, K., & Winner, E. (1991). When 3-year-olds understand ignorance, false beliefs, andrepresentational change. British Journal ofDevelopmental Psychology, 9, 159-171.

Wellman, H. M., & Gelman, S. A. (1988). Ghil-

Solomon et al. 171

dren's understanding of the nonohvlous. In core domains. Annual Review of Psychology,R. J. Sternberg (Ed.), Advances in the psy- 43, 337-375.chology of human intelligence (Vol. 4). Hills- Zaitchik, D. (1991). Is only seeing really he-dale, NJ: Erlbaum. lieving? Sources ofthe true helief in the false

Wellman, H. M., & Gelman, S. A. (1992). Cogni- helief tasks. Cognitive Development, 6,tive development: Foundational theories of 91-103.

like father, like son: young children's understanding of how and why offspring resemble their...

Documents