pretend play skills and the child's theory of mind

25
Pretend Play Skills and the Child's Theory of Mind Angeline S. Lillard Stanford University and University of San Francisco LILLARD, ANGELINE S. Pretend Play Skills and the Child's Theory of Mind. CHILD DEVELOPMENT, 1993, 64, 348—371. Pretend play has recently been of great interest to researchers studying children's understanding of the mind. One reason for this interest is that pretense seems to require many of the same skills as mental state understanding, and these skills seem to emerge precociously in pretense. Pretend play might be a zone of proximal development, an activity in whieh children operate at a cognitive level higher than they operate at in nonpretense situations. Alternatively, pretend play might be fool's gold, in that it might appear to be more sophisticated than it really is. This paper first discusses what pretend play is. It then investigates whether pretend play is an area of advanced understanding with reference to 3 skills that are implicated in both pretend play and a theory of mind: the ability to represent one object as two things at once, the ability to see one object as representing another, and the ability to represent mental representations. There has recendy been a virtual explo- sion of research on children's understanding of the mind (e.g., Astington, Harris, & Olson, 1988; Butterworth, Harris, Leslie, & Well- man, 1991; Frye & Moore, 1991; Whiten, 1991). Researchers in this area are particu- larly interested in pretend play, in part be- cause pretense seems to require many of the same skills as mental state understanding, and these skills seem to emerge preco- ciously in pretense. One possible expla- nation for this apparent decalage is that pretend play is a "zone of proximal develop- ment" with reference to these skills. Vygot- sky (1978) described the zone of proximal development as a zone in which children, with the help of a more competent partner, show a level of competence above that which they show alone. The child's zone of proximal development at one time becomes her actual developmental level at a later time. Although zone of proximal develop- ment generally carries social connotations (in that someone provides a scaffold to sup- port the child at the higher level), it may be usefully applied to contexts in which a more advanced partner is not necessarily present. As Vygotsky (1978) suggested, pretend play might be one such context. In a pretense mode, children might generally be able to operate at a cognitive level higher than that at which they operate in nonpretense con- texts. A second possibility, however, is that' pretend play somehow makes children ap- pear to be more advanced than they really are, and hence pretend play is in a sense "fool's gold." There are logical reasons to suspect this is the case. Children use their cognitive structures to operate in both the pretend and the real world, and those struc- tures could be reasonably expected to sup- port the same level of competence in both. In addition, "The quality of fantasy an indi- vidual can produce ought to be related to the coherence and sophistication of his or her real-world knowledge" (Bretherton, 1984a, p. 37). Evidence indicating that children are more competent in pretend play scenar- ios might, upon investigation, prove inade- quate. It may be that pretend play pulls the wool over our eyes, and that the skills in- volved in it look more sophisticated than they really are. This paper investigates these two alter- Support for this work was provided in part by NIH Training Grant IT32MH19114-01, a Stanford University Fellowship, and NIMH Grant 40687 to John H. Flavell. The author thanks Judy DeLoaehe, John Flavell, Paul Harris, Ellen Markman, Amanda Woodward, and the anony- mous reviewers for their comments on earlier drafts of this manuscript. An earlier version of this manuscript was submitted in partial fulfillment of the requirements for the doctoral degree at Stanford University. Correspondence should be addressed to: Angeline Lillard, Department of Psychology, University of San Francisco, Ignatian Heights, San Francisco, GA 94117-1080. E-mail: [email protected] [Child Development, 1993, 64, 348-371. © 1993 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/93/6402-0010$01.00]

Upload: vuongkhuong

Post on 10-Feb-2017

224 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Pretend Play Skills and the Child's Theory of Mind

Pretend Play Skills and the Child'sTheory of Mind

Angeline S. LillardStanford University and University of San Francisco

LILLARD, ANGELINE S. Pretend Play Skills and the Child's Theory of Mind. CHILD DEVELOPMENT,1993, 64, 348—371. Pretend play has recently been of great interest to researchers studyingchildren's understanding of the mind. One reason for this interest is that pretense seems torequire many of the same skills as mental state understanding, and these skills seem to emergeprecociously in pretense. Pretend play might be a zone of proximal development, an activity inwhieh children operate at a cognitive level higher than they operate at in nonpretense situations.Alternatively, pretend play might be fool's gold, in that it might appear to be more sophisticatedthan it really is. This paper first discusses what pretend play is. It then investigates whetherpretend play is an area of advanced understanding with reference to 3 skills that are implicatedin both pretend play and a theory of mind: the ability to represent one object as two things atonce, the ability to see one object as representing another, and the ability to represent mentalrepresentations.

There has recendy been a virtual explo-sion of research on children's understandingof the mind (e.g., Astington, Harris, & Olson,1988; Butterworth, Harris, Leslie, & Well-man, 1991; Frye & Moore, 1991; Whiten,1991). Researchers in this area are particu-larly interested in pretend play, in part be-cause pretense seems to require many of thesame skills as mental state understanding,and these skills seem to emerge preco-ciously in pretense. One possible expla-nation for this apparent decalage is thatpretend play is a "zone of proximal develop-ment" with reference to these skills. Vygot-sky (1978) described the zone of proximaldevelopment as a zone in which children,with the help of a more competent partner,show a level of competence above thatwhich they show alone. The child's zone ofproximal development at one time becomesher actual developmental level at a latertime. Although zone of proximal develop-ment generally carries social connotations(in that someone provides a scaffold to sup-port the child at the higher level), it may beusefully applied to contexts in which a moreadvanced partner is not necessarily present.As Vygotsky (1978) suggested, pretend play

might be one such context. In a pretensemode, children might generally be able tooperate at a cognitive level higher than thatat which they operate in nonpretense con-texts.

A second possibility, however, is that'pretend play somehow makes children ap-pear to be more advanced than they reallyare, and hence pretend play is in a sense"fool's gold." There are logical reasons tosuspect this is the case. Children use theircognitive structures to operate in both thepretend and the real world, and those struc-tures could be reasonably expected to sup-port the same level of competence in both.In addition, "The quality of fantasy an indi-vidual can produce ought to be related to thecoherence and sophistication of his or herreal-world knowledge" (Bretherton, 1984a,p. 37). Evidence indicating that childrenare more competent in pretend play scenar-ios might, upon investigation, prove inade-quate. It may be that pretend play pulls thewool over our eyes, and that the skills in-volved in it look more sophisticated thanthey really are.

This paper investigates these two alter-

Support for this work was provided in part by NIH Training Grant IT32MH19114-01, aStanford University Fellowship, and NIMH Grant 40687 to John H. Flavell. The author thanksJudy DeLoaehe, John Flavell, Paul Harris, Ellen Markman, Amanda Woodward, and the anony-mous reviewers for their comments on earlier drafts of this manuscript. An earlier version of thismanuscript was submitted in partial fulfillment of the requirements for the doctoral degree atStanford University. Correspondence should be addressed to: Angeline Lillard, Departmentof Psychology, University of San Francisco, Ignatian Heights, San Francisco, GA 94117-1080.E-mail: [email protected]

[Child Development, 1993, 64, 348-371. © 1993 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/93/6402-0010$01.00]

Page 2: Pretend Play Skills and the Child's Theory of Mind

natives^ with reference to three skills thatseem to be involved in pretend play, but areapparently absent or at least greatly dimin-ished outside the play domain. Briefly, theseare (a) the ability to think of one object astwo things at once, (b) the ability to think ofone object as representing anotiier, and (c)the ability to represent mental representa-tions. Before discussing these abilities, how-ever, I will examine what pretend play isand discuss the first position further.Defining Pretense

Pretend play can be seen as the inter-section of two broader concepts: play andpretense. Although the defining features ofplay are difficult to pin down (see Rubin,Fein, & Vandenberg, 1983), for the presentpurposes it is usefully considered as any ac-tivity that is engaged in for the purpose offun, rather than survival. The other concept,pretense, involves stretching one "reality"over another, or holding "one thing in frontof another in order to protect or conceal ordisguise it" (Austin, 1979, p. 260). A usefulworking definition of pretend play, there-fore, is the projecting of a supposed situationonto an actual one, in the spirit of fun ratherthan for survival.

Pretend play is frequently manifested inbehavior, as in pretending to be asleep whenone is not. It often involves substitute ob-jects (pretending a cloth is a pillow), andimagined objects (pretending there's a pil-low where there's no object). Pretense canalso involve imagined roles and situations:pretending to be Sleeping Beauty, for exam-ple, awaiting a prince in the forest. In allthese scenarios, the real world is supplantedby the pretend world; the canvas of the pre-tend world is stretched over the frame of thereal (Bretherton, 1984a).

What exactly is involved in pretense, inthis stretching of a supposed situation overan actual one? There are five features that

Angeline S. Lillard 349could be considered necessary and suffi-cient:

1. A pretender2. A reality (obviously omnipresent)3. A mental representation that is differ-

ent from reality^4. A layering of the representation over

the reality, such that they exist within thesame space and time

5. Awareness on the part of the pre-tender of components 2, 3, and 4.

As an example, take the pretense that astick is a horse. First, (1) some mindful beinghas to do the pretending; pretending isstrictly a mental state and does not takeplace in the absence of a mind. (2) There isalways some reality with which the pretensecontrasts (or at least the pretender must be-lieve the pretense contrasts with reality; seefootnote 2). In this case reality is a stick, butin another case the reality could be thatthere is nothing there (imaginary entities).(3) The pretender represents a horse, ratherthan a stick (although she still represents astick at some level, see below). Pretendingpresupposes that the representation is atleast in some way different from (what thepretender perceives to be) reality, for exam-ple, I cannot pretend to be typing while Iactually am typing (assuming I know I amtyping). (4) T'he pretender represents thehorse right where the stick is, in fact pro-jecting the horse representation onto thestick. The top of the stick is treated as if itwere the horse's head, the bottom as if itwere the legs, and so on. (5) The pretenderis fully aware of what she is doing: sheknows the item is a stick, she knows what ahorse is, she knows they are different, andshe knows that she is pretending the stick isa horse. Pretending is done knowingly andintentionally.

A sixth element that frequently accom-panies pretense is activity. (Throughout this

' A third possibility, of course, is that theory-of-mind tasks underestimate children's realcapabilities. At this point there is a fairly large body of evidence suggesting that theory-of-minddeficits in young children are real, not artifactual (e.g., Astington et al., 1988; Butterworth et al.,1991; Frye & Moore, 1991; Whiten, 1991), and therefiDre this alternative is not discussed here.See Siegal and Beattie (1991) for a more skeptical stance.

^Leslie (1988) provides an example of a child pretending an empty cup is empty, defyingthis feature. However, this pretense is only in reference to the cup's prior state of being full; thechild could more accurately be described in this case as pretending that the cup was just emptied.However, there is one important specification for this feature: The representation need notactually be different from reality; technically speaking, what is essential for pretense is that thepretender believes that what is being pretended is different from reality. One might pretendthere is someone at the door when someone actually is at the door, but if one knows that theperson is there, one cannot simultaneously pretend she is there.

Page 3: Pretend Play Skills and the Child's Theory of Mind

350 Child Developmentdiscussion, activity is used to denote corpo-ral and not mental activity.) Informal inter-views suggest that adults vary in their esti-mation of whether pretending requiresactivity. Whereas some adults claim that pre-tending always involves bodily movement,others think that even immobile day-dreaming is a kind of pretense. In the pres-ent analysis, the action component of pre-tense is one of potentiality: One does nothave to move, but were one to move, onewould due so in accord with the pretense.However, there are two cases in whichbodily activity is central to pretense. First,pretending to perform some action requiressome sort of movement to mime the actionthat one is pretending to perform. It is forthis reason, as Austin (1979) points out, thatone cannot pretend to bend one's trunk: theonly way to mime such an action would beto do it. The second case is that of pre-tending for an audience (as in deception orplay-acting). Sueh pretense requires somesort of external manifestation of what is be-ing pretended. The content of the pretensemust be communicated, be it via actions,static appearances, or a Greek chorus.

Bateson (1955/1972), in writing aboutplay in animals, discussed the paradox in-herent in pretend play. "Expanded, thestatement 'This is play' looks something likethis: 'These actions in which we now engagedo not denote what those actions for whichthey stand would denote" (1955/1972, p.180). The word "not" is at the crux: the be-havior is not serious, the playful nip is not abite, and the real object is not the pretendobject. To be truly engaged in pretense, thepretender must be aware of this paradox. Forexample, if I am trying to eat a block becauseI believe it is a cookie, then I am not pre-tending; rather, I am mistaken. On the otherhand, if I enact the same eating behaviorknowing full well that what I am acting onis in fact a block, then I am pretending. Tobe truly pretending, one must grasp the situ-ation on two levels: both as the real situation(the block) and as the pretend situation (thecookie).

Furthermore, the real and pretend situa-tions are kept separate in pretense. Thereare two clearly separate worlds, one layeredover and projected onto the other. The pre-tend world is not expected to seep intothe real world (e.g., I do not expect the blockto become a real cookie in real life), nor isthe real world expected to adopt features ofthe pretend one (when I buy a bag of blocks,I do not expect it to contain cookies). Al-

though some reports of young children hav-ing difficulty with the real-pretend boundaryexist, they are rare and may be the exceptionrather than the rule. Many examples are an-ecdotal, or were collected in emotionallycharged situations, in which the child's be-havior may stem from the frightful emotionsrather than from uncertainty about reality(DeLoache & Plaetzer, 1985; Garvey &Berndt, 1975; Scarlett & Wolf, 1979). Otherexperimental examples might result fromthe pretense situation being "lab-manufac-tured" and not in line with the child's abili-ties and play patterns (DiLalia & Watson,1988). Younger children might not get as in-volved in such pretense, ancf this lack ofinvolvement would naturally be refiected inchildren's slipping out of the pretend worldinto the real one. Indeed, children must befairly good at maintaining a real-pretendboundary or they would be sorely confusedabout real-world relations (Leslie, 1987).Discussion now turns to evidence that pre-tense is an area in which children manifestadvanced abilities.

Pretend Play as an Environment forEarly Competence

Vygotsky (1978) wrote that "in play achild is always above his average age, abovehis daily behavior; in play it is as though hewere a head taller than himself" (p. 102). Ifthis were generally true, then it would makesense that children precociously manifest inpretend play certain skills that are also rele-vant to a theory of mind. This section briefiyreviews evidence for pretense being an areaof early competence for skills other than thethree with which this paper is mainly con-cerned. One such skill is deductive reason-ing. Dias and Harris (1988, 1990) found that4-6-year-olds are better able to reason aboutcounterfactual syllogisms when they con-cern a pretend planet than when they con-cern planet Earth. For example, when toldthat on another planet "All cats bark. Rex is acat. Does Rex bark?" children could answercorrectly that Rex the cat barks. Childrenalso did better on such syllogisms whenother means were used to encourage chil-dren to adopt a pretend attitude (such as thestoryteller adopting a "story intonation").These studies could be interpreted to meanthat the pretend world is an area in whichchildren manifest an early ability to free rep-resentations from their referents, and canallow those freed representations to be mod-ified (this is essentially Vygotsky's position[1978]). For example, "cat" is freed fromwhat children know about cats when one

Page 4: Pretend Play Skills and the Child's Theory of Mind

pre;tends that the cat is on another planet.In another study, Kuczaj (1981) found thatchildren were better able to reason hypo-theitically about fantasy characters thanabout their own parents. For example, chil-dien were better able to answer, "Whatwould have happened if Bingo [a toydragon] had played football yesterday?"than "What would have happened if yourmother had played football yesterday?"Children's thinking about the pretend worldseems to be more flexible than their think-ing about the real world, and this might indi-cate that pretend play is an environmentin which children are more competent attasks requiring flexible or divergent think-ing skills (see Rubin et al., 1983).

Other studies that seem to indicate pre-tend play is a facilitative environment arethose that investigate the relation betweenfrequency of pretend play and social-cognitive skills. Many theorists have sug-gested that fantasy play may even cause so-cial-cognitive development (Bruner, 1972;Flavell, Green, & Flavell, 1990; Mead, 1934;Singer, 1973; Smilansky, 1968; Vygotsky,1978), and that therefore children who pre-tend play more should be more advanced inthe social-cognitive realm. Indeed, childrenwith high fantasy predispositions are gen-erally better at tasks involving social-cogni-tive; skills, such as perspective-taking, role-taking, and social competence tasks(Connolly & Doyle, 1984; Rubin, 1976). (SeeMussen & Eisenberg-gerg, 1977, and Rubin& Schneider, 1973, for discussion ofthe rela-tion between such skills and the develop-ment of social cognition.) However, it isplausible that social-cognitive skills lead tosociodramatic play, rather than vice-versa,since sociodramatic play requires competentcommunication and negotiation (Brether-ton, 1989; Gottman, 1986; Rubin & Pepler,1980). Better evidence for play being a train-ing ground for social-cognitive skills wouldcome from studies that train children to en-gage in pretend play and then test whetherchildren's social-cognitive skills improverelative to those of control children. Severalstudies have engaged in this enterprise (e.g..Burns & Brainerd, 1979; Fink, 1976; Rosen,1974; Saltz, Dixon, & Johnson, 1977; Smi-lansky, 1968; Smith & Sydall, 1978; Spivak& Shure, 1974), and although the resultshave been somewhat: promising, the meth-ods used have some problems (Rubin et al.,1983). For example, outcome measuresmight not always have tapped the skills theywere intended to tap, pretreatment assess-

Angeline S. Lillard 351ments were not always taken, and controlgroups in some studies were not given anytreatment at all, leaving open the possibilitythat the results were caused by spendingtime in a special training group rather thanthe training itself (Hawthorne effects). In ad-dition, some of the data (Smilansky, 1968)are not quantitative; hence the signiflcanceof the flndings is uncertain. However, theflndings are generally positive over severalstudies, indicating that pretend play mightimprove social-cognitive skills. If it does,then it stands to reason that it is a facilitativeenvironment for such skills, and that the ear-liest appearance of such advanced skillscould therefore be in pretend play situa-tions. In sum, training studies and testsof children's reasoning ability each seemsomewhat suggestive that pretend play is anarena in which children show advanced so-cial-cognitive skills.

Discussion now turns to three succes-sive abilities that seem to be entailed in pre-tend play and that are also implicated in thechild's understanding ofthe mind. These arethe ability to apply multiple representationsto a single object, the ability to consider onething as representing another, and the abil-ity to represent another person's mental rep-resentation. These abilities each seem tobuild on each other. To see one object asrepresenting another, one must understandthat an object can be both a thing in itselfand a representation of something else. Tounderstand that the mind represents theworld, one must understand the generalpremise that one thing can represent an-other.

Considering Mntnal Identities—Exclusively in Pretense?

One characteristic of pretend play is thatduring such play children frequently pre-tend one thing is something else. However,during such episodes children also seem tobear both the pretend identity and the realidentity of the object in mind, and arethereby representing one objeet as beingtwo different things at once. A clear exampleof this occurs \vhen children pretend a non-food item is food. The child gives a convinc-ing display that she has identifled the ob-ject as food, but does not go so far as toactually eat it. For example, a child who ispretending a pile of sand is fantastic choco-late cake might call it cake, mimic eating it,say, "Yum-yum, what delicious cake!" andperhaps even mention the chocolate she goton her hands. But she does not actually eat

Page 5: Pretend Play Skills and the Child's Theory of Mind

352 Child Developmentthe sand. She is clearly aware of its real iden-tity all the while that she treats it as if it weresomething else. In this sense, one object isbeing simultaneously considered as havingtwo different identities.

However, outside of pretend situations,children seem to have a bias against viewingone object as two things at once (if not anoutright inability to do so). Flavell (1988) hasnoted that many other tasks with which 3-year-olds have difficulty seem to require thissame ability: appearance-reality, visual andconceptual perspective-taking, seriation, ref-erential communication, and learning multi-ple labels for one object, to name a few. Sev-eral of these tasks fall under the rubric of"theory of mind" tasks, since all involve un-derstanding that one object or situation canbe interpreted in many ways. For example,in order to entertain two different perspec-tives on a situation, the child must appreci-ate that one situation affords more than onepoint of view. Hence the ability to see oneobject as two things at once figures impor-tantly in developing theories of mind, aswell as in pretense. The fact that childrenappear to have the ability in pretense priorto age 2 but do not appear to have it outsideof pretense until at least 2 years later is puz-zling.

This section discusses children's objectsubstitutions in pretend play and then re-views the evidence for a single-representa-tion bias both in early word learning and inappearance-reality tasks."^ (Although learn-ing multiple object labels is certainly notconsidered a theory of mind task, it is dis-cussed here because the ability to representone object in two ways figures centrally init.) The focus of the discussion is whetherpretend play really requires this skill, and towhat extent children's early word learningand appearance-reality tasks deny their hav-ing it.Is the Pretend Cookie Also a Block?

Object substitution is the use of one ob-ject as if it were another, for example, usinga stick as if it were a horse. Pretend objectsubstitution usually emerges by 24 monthsof age (Rubin et al., 1983). For younger chil-dren, substitute objects must bear a strongsimilarity to the object for which they arebeing substituted (Elder & Pederson, 1978;Fein, 1979; Jackowitz & Watson, 1980). Atoothbrush can be mimed with a pencil but

not with a cup, for example. By 3 years ofage, however, quite different objects can besubstituted, including purely imaginary ob-jects (Overton & Jackson, 1973). In addi-tion to requiring less physical similarity insubstitute objects, older children can alsouse substitute objects whose real functionclearly contrasts with their pretend function.For example, 2-year-olds may be able to usea block but not a toy car as a telephone. Pre-sumably this is because the toy car clearlyhas another function, whereas the block'sfunction is ambiguous. A 3-year-old, how-ever, can use either the block or the car: hecan override the usual response of makingtoy cars behave like real cars (Pederson,Rook-Green, & Elder, 1981; Rubin et al.,1983; Ungerer, Zelazo, Kearsley, & O'Leary,1981). Thus, although object substitutionemerges in the flrst 2 years, over the next 2years children become increasingly able touse substitute objects whose function andappearance are clearly different from the ob-jects for which they are being substituted.In effect, they become able to entertain tworepresentations that are increasingly differ-ent in both appearance and function.

A second development in object substi-tution is an increase in the number of substi-tutions that can be made, where, for exam-ple, using a block for a horse is onesubstitution. Fein (1975) found that 94% of24-month-olds could pretend when no sub-stitutions were involved (i.e., could pretendto feed a toy horse with a toy cup); 70%could do so when one substitution was in-volved (i.e., a block served as the horse, butthe cup was used as a cup); and only 33%could do so when both objects were substi-tuted. In other words, with age, children be-come increasingly able to entertain multiplesets of dual representations.

Object substitution seems to entail con-ceiving of an object as two things at once.When a child pretends, for example, that ablock is a cookie, she seems to hold twoidentities for the block in mind at once. Onthe one hand, she treats the block as acookie, calling it a cookie, raising it to hermouth, and pretending to chew it. On theother hand, she treats it as a block, by notactually eating it. This could be thought ofas a linear rather than a dualistic capacity:The child might think of the object as acookie up to the point of mouth entry, at

^ Note that DeLoache (1989) claims that the ability to see one object as two things at onceis also what allows children to pass the scale model task. This is discussed in the next section.

Page 6: Pretend Play Skills and the Child's Theory of Mind

which point she switches to thinking aboutit as a block. But on inspection this possibil-ity is not tenable. Even at the moment of nottaking an actual bite, the child might referto the object as a cookie. She says "mmm"as she pretends to chew that which she isnot really chewing. The child clearly hasone foot in the pretend world and the otherin the real world, and is masterfully existingin both. The block is a cookie, but it is stilla block, too. In sum, pretend object sub-stitutions do seem to involve representingone object as two things at once, and withage children become better able to performthese substitutions with regard both to howdifferent the real and represented objectscan be, and to how many substitutions theycan entertain at once. The next section re-views two areas in which children do notseem to be able to represent one object astwo things at once: early word leaming andappearance-reality tasks.

The Mutual Exclusivity BiasOutside of pretend scenarios, conceiv-

ing of one thing as two things at once appar-ently ranges from undesirable to impossiblefor young children. For example, in leamingobject labels, children who know one labelfor an object tend to reject altemative labels.A 3-year-old who knows "dog" means a dogwill typically object to "pet" to refer to thesame type of item (Clark, 1987; Markman,1987; Merriman & Bowman, 1989). This biasmay explain why some classes of words, likesuperordinate category terms, are particu-larly difficult for children to learn (Markman,1989). As a further demonstration ofthe ten-dency to reject multiple labels, if a child ispresented with two objects, one of whichshe knows tiie label for, and is asked to givesomeone the "dax," children tend to choosethe object they do not know a label for(Markman, 1989). In addition, if shown justone object that they do know a label for, andgiven a new label, they tend to view the newlabel as referring to a part or property of theobject. Markman (1989) maintains that "mu-tual exclusivity" is a bias rather than a hard-and-fast rule in children's language learning.Children do learn multiple labels for ob-jects, presumably when they are given suf-ficient evidence that the new label applies(Markman, 1989). Flavell (1988) has sug-gested that this bias against assigning twolabels to an object stems from a propensityto view objects as having only one nature;the problem thus may be one of dual coding,or of representing one object as two thingsat once. Hence in language leaming we see

Angeline S. Lillard 353one manifestation of a reluctance to assigntwo different representations to one and thesame object.The Appearance-Reality Distinction

Another situation in which children'sproblem with dual coding arises is whenchildren are asked to describe both the realand apparent identities of illusory objects,for example, sponges that look like rocks,and candles that look like apples (Flavell,Green, & Flavell, 1986). In one study (Fla-vell, Flavell, & Green, 1987), for example,the experimenter took a candle that lookedlike an apple, carefully cupped her handsaround the candle so that only the wickcould be seen, and told the child that theobject was really a candle. Then the exper-imenter moved her hands, revealing thecandle's apple-like appearance. The exper-imenter proceeded to pretend to eat theapple-candle throughout the ensuing ques-tioning. The child was asked (by a secondexperimenter), "Right now, does that thinglook like an apple or look like a candle?"(emphasis theirs) and "For real, is that thingreally and truly an apple, or really and trulya candle?" In experiments of this sort, chil-dren younger than 4 years of age tend eitherto claim the object both looks like and isan apple (realism errors), or both looks likeand is a candle (phenomenism errors). Theproblem seems to be one of simultaneouslyseeing an object as both an apple in appear-ance and a candle in reality.

One might ask to what extent the objectsstill look like the objects that they are meantto look like once their real identity is known.Perhaps once the child knows it is a candle,the real identity may become so salient thatit really looks like a candle to her. Alter-natively, perhaps children conceptualize"looks like" as a question concerning whatthey know about the object. However, con-vergent evidence from several studiesspeaks against such interpretations. For ex-ample, if children really do understand theappearance-reality distinction but are con-fused by the questions used to tap theirknowledge, then concerted attempts to trainthe children on the meaning ofthe questionsshould lead to improved performance. Threedifferent research groups have used differ-ent methods to try to train children on thesetasks, all without success (Braine & Shanks,1965; Flavell et al., 1986; Taylor & Hort,1990). In addition, very simplified word-ing and memory cues do not significantlyimprove children's performance (Flavell,Green, Wahl, & Flavell, 1987). Finally, chil-

Page 7: Pretend Play Skills and the Child's Theory of Mind

354 Child Developmentdren apparently have the same difficultywith assigning multiple representations in avariety of tasks using widely differing sortsof questions (e.g., Astington & Gopnik, 1988;Beal, 1988; Taylor, 1988; Wimmer & Perner,1983), providing convergent validity for Fla-vell et al.'s explanation. Children's difficultyseems to lie in ascribing two identities to anobject at once.

To summarize, in leaming object labels,children are biased against using two labelsto refer to one and the same object. In mak-ing the appearance-reality distinction, thisbias seems to extend to an outright inability:most 3-year-olds appear to be incapable ofconsidering an object as having divergentreal and apparent identities. And yet in pre-tend play, children willingly and withouthesitation claim that blocks are cookies, sandpiles are cakes, and bananas are telephones.Furthermore, they seem to keep the realidentity in mind all the while: they do nottry to eat block cookies or sand cakes. Howcould children be capable of this feat in pre-tend play but not in real life? If they havewhat it takes (perhaps cognitive structuresor processing capacity) to assign two identi-ties in pretense, and they do so regularly andwillingly, why do they not use the capabilityin understanding illusions or learning objectlabels? This is a skill that truly seems ad-vanced in pretend play relative to other ar-eas. In the next section I discuss three possi-ble explanations for this apparent decalage.Note that these explanations need not be ex-clusive; two or even all three might be im-plicated.

Possible ExplanationsDirection of fit.—Seaile (1983) ob-

served that different mental states seem tojibe with reality in different ways. Mentalstates like desire have a mind-to-world di-rection of fit, meaning that the mental stateserves to change the world. Our desires mo-tivate us to take actions that in turn fulfillthose desires. Other mental states, like be-lief, have a world-to-mind direction of fit.Our beliefs change to match the worldaround us. Gopnik and Slaughter (1991)have raised the possibility that children maylearn about desires before beliefs becausethe mind-to-world direction of fit may beeasier to learn about than the world-to-minddirection of fit. Indeed, children do seem tounderstand desires before beliefs (Gopnik &Slaughter, 1991; Lillard & Flavell, 1992;Wellman & Woolley, 1990) and difficulty inunderstanding mental states that have a

world-to-mind direction of fit may partly ex-plain why.

Gopnik and Slaughter (1991) note thatpretense does not require consideration ofthe relation between a representation andthe world, implying that it does not haveclear direction of fit. However, although pre-tense does not change reality in the samesense that desires do (e.g., one's desire forice cream may result in the ice cream beingeaten), pretense does generally emanatefrom the mind. The world does not make themind pretend in the same sense that theworld makes the mind believe. If I see atruck, I believe the truck is there, but I donot pretend anything unless my mind im-poses a pretense, for example, imposes thepretense that the truck has a driver andneeds gas. Hence early understanding ofpretense also satisfies a direction-of-flt inter-pretation of why some mental states are eas-ier for children to understand than othermental states.

Appearance-reality tasks, on the otherhand, have a world-to-mind direction of fit:understanding "It looks like an apple" isclearly a matter of understanding the world'seffect on the mind, not vice-versa. Hencechildren's difficulty in understanding ap-pearance-reality tasks also fits with the di-rection-of-fit explanation. Object labels are amore complicated case in this regard. Al-though children occasionally make up objectlabels (e.g., Horgan, 1980), implying thatthey may see labels as having a mind-to-world direction of fit, in general object labelsare given to them by more competent oth-ers—in other words, by the world. In gen-eral, then, it seems that children shouldview object labels and appearance-realitytasks as involving a world-to-mind directionof fit, and pretense as involving a mind-to-world direction of fit. This may in part ex-plain why children can more easily applymultiple identities to an object in pretensethan in other scenarios. The direction-of-fitanalysis claims that children can apply mu-tual identities in pretend play but not inother tasks because pretend play is an easycase, or a facilitative environment.

Implicit knowledge.—Another possibil-ity for why pretend play appears to be anarea of advanced skill with regard to consid-ering multiple representations is that knowl-edge about the real and pretend identitiesmay operate at different levels. An analogymight be made to the highest stage of Fitts

Page 8: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 355and Posner's (1967) model of skill acquisi-tion. In the autonomous stage of skill devel-opment, knowledge need not be consciouslyinvoked. One can drive a car, for instance,vi'ithout thinking consciously about pressurecm the gas pedal, or using the turn signal.Meanwhile, on a conscious level, one canhold a conversation with a passenger in thecar on a subject totally removed from drivingthe car. Knowledge about how to drive is"implicit" in that it does not have to be con-sciously invoked. (For discussion of implicitand explicit knowledge, see Dretske, 1988;Karmiloff-Smith, 1991; Mandler, 1988.) Theconversation, however, does need to be keptat a conscious level. Perhaps the substituteobject's real identity is kept in mind at thesame level as the knowledge of how to drivetlie car, whereas the pretend identity is keptin mind at the level of the conversation. Incomputer terminology, facts concerning realidentity could be "backgrounded" with ref-erence to facts concerning the pretend iden-tity, but both sets of facts could contribute totlie behavior. The implicit knowledge couldcome into play whenever unpleasant results,like eating sand, would result from the pre-tend behavior. Exactly how this would workis a matter for further speculation, but on theface of it does seem plausible that childrenmanage to deal with objects as both theirpretend and real identities simultaneouslyby retaining the real identity at the level ofimplicit knowledge, backgrounded with ref-erence to the pretend identity. Under thisanalysis, children can apply mutual identi-ties in pretend play in part because pretensecreates a special situation, in which infor-mation can be backgrounded. On the otherhand, because the information is back-grounded, it is not being considered in thesame conscious way that is required in mak-ing the appearance-reality distinction or inla.beling objects. Hence pretend play couldalso be seen as a sort of fool's gold by thisexplanation.

Labels versus usage.—A related possi-bility for how children manage to deal withtv;'O identities for one object in pretense con-cems the different demands of labels andusage in the different tasks. In appearance-reality tasks and in learning object labels,the child has to accept two different labelsfor the same object. Conversely, in pretense,the child does not need to apply two labelssimultaneously. The real identity may sur-face only in limitations on how the objectis used; in other words, the child may bear

in mind only one aspect of the concept"block," namely, "inedible." Although nostudies have specifically addressed this is-sue, it may be that children do not routinelyuse the real labels for substitute objects dur-ing their play. They may say "pretend this isa telephone" more often than "pretend thisbanana is a telephone." In addition, the pre-tend identity surfaces in part through usage:the banana is being used as if it were a tele-phone. Whereas in language learning andin appearance-reality tasks children are re-quired to label an object as two things atonce, in pretend play the child uses an ob-ject as though it were two things at once, insome ways a telephone, and in other ways abanana. Perhaps in appearance-reality tasks,children would do better if the task con-cerned use rather than identity. For exam-ple, if a child were instructed to select anitem to put in a fruit bowl, perhaps shewould choose the apple-candle, despite hav-ing claimed that it looks like a candle. Whenpressed to explain her behavior, perhaps shecould then even acknowledge that it lookslike an apple. In sum, it may be that childrencan use an item differently than its real iden-tity would dictate, but still cannot label it astwo different things at once.

Flavell, Flavell, and Green (1987) in ef-fect tested this possibility in a series of stud-ies comparing children's ability to make theappearance-reality distinction with theirability to make the pretend-reality distinc-tion. A test of children's ability to makean appearance-reality distinction was de-scribed earlier, using the apple-candle as anexample. In the pretend-real condition, thechild was told, "Ellie is pretending" (em-phasis theirs). As in the appearance-realityversion ofthe task, the second experimenterpretended to take bites from the applethroughout questioning. "Is she pretendingthat thing is a candle or pretending it's anapple?" or "That thing she's holding, is thata real apple or a pretend apple?" (The ap-pearance-reality equivalent of this task was"Look at what she's holding. Right now,does that thing look like an apple or looklike a candle?") Children were also given astandard reality question: "For real, is thatthing really and truly an apple or really andtruly a candle?" Children performed sig-nificantly better on the pretend-reality ques-tions than on the appearance-reality ques-tion, despite the fact that both tasks seem torequire bringing two discrepant labels for anobject—a real label, and either a pretend or

Page 9: Pretend Play Skills and the Child's Theory of Mind

356 Child Developmentan appearance label—to the foreground. De-scribing a scenario as pretense thus seemedto bring out a competence that was lackingin the nonpretend version.

However, three factors should be notedwith regard to this study. First, the pretendquestion ("Is she pretending that thing is acandle or pretending it's an apple?") mighthave been easier than the analogous appear-ance-reality question not because childrenunderstand pretense more easily, but be-cause pretending something is a candlewould involve some different action, likelighting it, rather than miming eating it asthe experimenter was doing. In addition, thesecond question ("That thing she's holding,is that a real apple or a pretend apple?") usesthe label "apple" for both choices; the childonly has to choose the appropriate modifier(real or pretend). Choosing a modifier ratherthan a discrepant identity label may be eas-ier, and thereby facilitate children's perfor-mance. Third, although these studies doseem to involve commenting on an object'sidentity, they could actually only involve la-beling object usage. In this study, for exam-ple, the experimenter acted like she waseating the apple-candle while the questionswere being asked. In other words, she wasusing the object as if it were an apple. Chil-dren might have responded correctly by fo-cusing on the action rather than on the ob-ject's pretend identity. The authors raisedthis possibility, but dismissed it as implausi-ble, since the experimenter's activity andthe questions focused on the object in bothconditions. However, it is possible that anobject's identity is tied to activity or usage inpretend tasks but not in appearance-realitytasks. Perhaps the word "pretend" cues chil-dren to look at activity. Supporting this pos-sibility, Woolley and Wellman's (1990) studyof children's ability to make pretend-realitycontrasts used a method in which children'sattention was conceivably drawn to the ac-tivity even more than in the Flavell et al.study. For example, in one condition, the ex-perimenter pretended to brush her teethwith a real toothbrush and the child wasasked, "For real, am I really and truly brush-ing my teeth or am I pretending to brush myteeth?" In this study, children performednear ceiling, as opposed to the 58%—67%correct range found by Flavell et al.; atten-tion to activity might be one explanation forthis (Green, personal communication, July,1990).

In sum, young children may have suc-ceeded on this test by conceiving of pre-

tending as something someone is doing,rather than as something someone is repre-senting. Such tasks seem to require consid-ering mutual identities, but may in fact onlyrequire considering an activity that is dis-crepant with an identity. Three approachescould be taken to testing this possibility.First, to test whether children are con-sciously holding two identities in their ownminds during pretense, one could stop chil-dren in the middle of play and ask, "What'sthis really?" and "What are you pretendingthis is?" An especially relevant case wouldbe one in which the pretend object was notbeing acted upon at the time when the ques-tions were asked, to ensure that childrenwere referring to the pretend identity ratherthan to how it was being used or acted upon.(A potential problem with this, however, isthat children might refer to what they havebeen pretending the object is, or what theywill pretend it is in the near future.) Second,while someone was acting on an object, forexample pretending to eat an apple-candle,one could ask questions that address usageversus representation directly. For example,one might ask, "What's she doing with thatthing?" or "What is that thing, to her, rightnow?" as opposed to "For real, what shouldyou do with that thing?" Third, one couldtest whether in nonpretend scenarios it iseasier to consider discrepant usage than di-vergent identities. For example, childrenknow that toothbrushes are for brushingteeth. If they could say that someone wasusing a toothbrush as if it were a hairbrush,and still identify the object as really andtruly a toothbrush, this would support thepossibility that usage or activity might ac-count for children's apparent ability to applymutual identities in pretend play.

SummaryYoung children seem to have a bias

against applying two identity labels to thesame object and seem unable to describean object as being one thing in reality anda different thing in appearance. However,when they engage in object substitutions inpretend play, young children must be keep-ing two identities in mind at once or elsethey would, for example, really try to eat pre-tend food. Therefore applying two identitiesto an object at once is a skill that childrenhave in pretend play but not outside of it,indicating that pretend play is an area of spe-cial competence with respect to this skill.But it may be that the way children managethis feat requires a much simpler under-standing than is required in appearance-

Page 10: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 357reality tasks and learning multiple labels.For example, they might background knowl-edge of the real identity, or might refer toactivity rather than representation in ascrib-ing the identities in the pretend world. Intiiis sense pretend play might not be as so-phisticated as it seems. Additional researchis needed to determine how children con-ceive of the real and pretend identities inobject substitution.

External RepresentationsUnderstanding external representations

figures importantly both in theory of mindand in pretend play. In pretend play, chil-dren using substitute objects are often as-sumed to think of the substitute object asrepresenting the referent. Children begin touse substitute objects during the latter halfof the second year. Understanding externalrepresentations is important to theory ofmind because it is a likely building blockfor understanding mental representations, orthat internal representations refer to the ex-temal world. In order to understand a men-tal representation, one must first understandthe basic fact that one thing can representanother. This might be flrst understood ina realm that allows one to see both the rep-resentation and the represented, in othervv ords, when both the representation and therepresented are external entities. This sec-tion discusses how external representationsfigure in pretend play, then reviews litera-ture concerning a general understanding ofextemal representation: studies of children'sunderstanding of scale models and photo-graphs.

Representation in PretendPretend play has often been assumed

to be representational (Bretherton, 1984a,1984b; Fein, 1979; Leslie, 1987; McCune-Nicolich, 1981; Piaget, 1962). There are twov/ays one might take this claim: childrenmight use their own intemal representationsof the world in their pretend play, or theymight conceive of one extemal object as rep-resenting some other object. As to the formerinterpretation, children do seem to use men-tiil representations of objects in their pre-tense. For example, in pretending a block isa cookie, a child uses her representation ofa cookie. Indeed, having a mental represen-tjition of some alternate reality is one of thedeflning features of pretense. The second in-terpretation ofthe claim that pretend play isrepresentational is more central to the issueof understanding external representations.Although surely the child uses her own in-

temeJ representation in creating the pretendobject, does she also think of the pretendobject as representing its referent? Dis-cussing representations and symbols raisesa host of complex issues that are well beyondthe scope of this discussion (but see Good-man, 1968; Huttenlocher & Higgins, 1978;Olson & Campbell, in press; Perner, 1991;Pierce, 1940; Potter, 1979). Sufflce it to saythat, under a broad definition of "represent,"there are many classes of things which couldbe said to represent other things: words rep-resent their referents, a picture of a bear rep-resents a bear, and a whistle can represent atrain. Children can produce and interpretsuch representations from a very young age(DeLoache, Strauss, & Maynard, 1979). Pic-ture interpretation may be similar to inter-preting models (like model cars and horses),although pictures are two-dimensional rep-resentations of referents, and models arethree-dimensional ones. In both cases, how-ever, the child demonstrates awareness ofthe similarity between the representationand the referent when she uses the sameword for the representation and the referent,or makes a model car move like a real car (asin functional play). The question of interesthere is whether recognition of similarity insuch cases is tantamount to understandingrepresentation.

With regard to object substitutions inpretend play, the child could conceive of amodel car in either of two ways. She mightthink of the model car as representing a bigcar, or she might think ofthe model car sim-ply as a very small car, which she can makemove in the manner that big cars move.Goodman (1968) makes a similar point v 'ithregard to portraits: a painting of a man canbe interpreted as designating some particu-lar man or as a generic "man-painting." Thisdistinction is of the essence in trying to de-termine whether children's pretend playdemonstrates an understanding that one ob-ject can represent another.

Pemer (1991) further clarifies this dis-tinction. He calls the case of thinking of themodel car as representing a big car symbolicsubstitution. In symbolic substitution, oneobject serves as a symbol for another. Forexample, a general can use sticks in a sand-box as symbols of his soldiers. "Symboliza-tion is the representing of an object or eventby something other than itself" (Potter,1979, p. 4). Perner calls the second case—using the model car as if it were a big car—hypothetical substitution; his example ofthis form of substitution is pretending on

Page 11: Pretend Play Skills and the Child's Theory of Mind

358 Child DevelopmentMonday that it is Sunday (a fun prospect!).In this case, Monday does not representSunday; one merely instantiates Sunday'spattern of activity on Monday. Similarly, if Istir my coffee with a pen, I do not necessar-ily think ofthe pen as representing a spoon;I am simply using it as if it were a spoon. Toact as if X were Y (as in hypothetical substi-tution) is not the same as using X to repre-sent Y (symbolic substitution). This is per-haps reminiscent of the usage versus labelissue that was raised in the previous section:to use X as if it were Y is to act as if X wereY, whereas to label X as Y is (arguably) anal-ogous to representing X as Y.

The question, then, is whether a childwho pretends that a banana is a telephone isengaging in symbolic or hypothetical substi-tution. Huttenlocher and Higgins (1978) dis-cuss the difficulty in determining whether aprocess is symbolic, but conclude that whena child explicitly says what she is pre-tending, then one can be certain of symbolicsubstitution. Leslie (1987) discounts thiscriterion, because the verbalization couldbe a learned response. The child may sim-ply know, for example, that when you haveteacups, you pretend to sip from them andsay, "Mmm, delicious tea." Verbalization istherefore not a clear criterion for whether asubstitution is hypothetical or symbolic. Per-ner (1991) also points out that this criterionis insufficient, since the child would makethe same statement to comment on a hypo-thetical substitution: "I'm pretending thisbanana is a telephone" could either mean"I'm using it as a telephone" or "I see thisas a symbol for a telephone."

Exactly how one does determinewhether a child's object substitution is hypo-thetical (acting as if) or symbolic is not clear.Pretend play might not tell us whether chil-dren understand external representation,because perhaps when a child engages inobject substitution there is no way of de-termining if the child is thinking ofthe sub-stitute object as representing the referent oris just acting as if one thing were somethingelse. As far as can be determined from thepresent evidence, therefore, pretend objectsubstitutions are representational in thesense that children must use their own men-tal representations of referents when theypretend that some other object is that refer-ent, but it is unclear whether, to the child,substitute objects represent their referents.

Some research often discussed in con-junction with children's theories of mind

speaks to the issue of when children under-stand external representation, or that onething can stand for another (symbolic substi-tution). The sticks in the general's sandboxare symbolic because they stand for the sol-diers. One can use the sticks as indicatorsabout the actual soldiers: for example, onemight learn about how many soldiers thereare, and where they are located, if one couldcorrectly interpret the sticks. Conversely,one can look at the real-life situation, andmake some interpretations about the sticksin the sandbox. These kinds of situationshave been studied with regard to children'sunderstanding of representation. DeLoache(1989) has investigated whether childrencan use a scale model or a photograph ofa room to detennine where an object islocated in another room, and Liben andDowns (1989) have looked at children's un-derstanding of maps. Several investigators(Charman & Baron-Cohen, in press; De-Loache, 1989, 1991a; Lillard & Cho, 1991;Pemer & Leekam, 1990; Zaitchik, 1990)have taken the opposite approach, studyingwhether children who know a situation(analogous to the room) can determine whata representation of that situation (analogousto the model) looks like. An important dis-tinction in the studies is that in DeLoache'swork, the situation and the representationare concordant, whereas in the other studies,the situation has been changed since therepresentation was made. This will be dis-cussed below.

Scale Models and MapsDeLoache's (1989, 1990, 1991a, 1991b)

scale model studies focus on an abrupt shiftin childen's use of scale models between 2y2and 3 years of age. The basic paradigm usesa full-sized, fumished room, and a smallscale model ofthe room. The child is shownthe big room and the model, and their corre-spondence is pointed out. Then, while thechild watches, a small Snoopy is hidden inthe model, for example, behind the couch,and the child is told that a big Snoopy ishidden in the same place in his big room.The child is led to the big room, and is askedto find big Snoopy. Children of 2V2 years ofage are at a loss as to where to find bigSnoopy, whereas 3-year-olds go straight tothe corresponding hiding place, in this case,the big couch.

In the scale model task, the child mustunderstand that she can use the little roomas a guide to the big room. The question ofinterest here is whether this implies that thechild understands the model as representing

Page 12: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 359ihe big room. Perner (1991) has argued thatto succeed on DeLoache's task, a child needonly understand correspondence. Perneruses the example of duplicated fioor plans.For instance, two fiats in the same buildingtend to have the same fioor plan. Someonevi'ho lives in the upper fiat can enter thelower fiat for the first time and know exactlyvi'here the kitchen is. This is not because theupper flat represents the lower flat; it is sim-ply because they have the same fioor plan.An understanding that two spaces can corre-spond may be all the child needs to passscale model tasks.

DeLoache's (1989, 1991a) analysis ofthe scale model task is that it requires morethan understanding correspondence. Shesuggests that it requires thinking of an objectas both a thing in itself and as a representa-tion of something else. DeLoache suppliesseveral kinds of evidence supporting this in-terpretation. One kind of evidence is thatchildren's pretend play object substitutionsbecome much less dependent on perceptualsupport around 3 years of age (Elder & Ped-erson, 1978; Ungerer et al., 1981), the sameage at which the scale model is first used asa guide to the larger room. In other words,children become able to use a cup as if itv/ere a toothbrush, whereas previously onlyother long thin objects like pencils couldserve as substitutes for a toothbrush. On theone hand, it seems that children's emerg-ing ability to use less similar substitute ob-jects is not really analogous to using a scalemodel, since children's object substitutionsbecome less dependent on perceptual sup-port in pretend play after age 3, and yet thescale model is very much like the largeroom. It seems that to use the scale modelas if it were the big room should be easy for a2-year-old if pretend play object substitutionrequired the same skill. The model is liketlie big room, just as an egg-shaped containeris like a cup (Elder & Pederson, 1978). Ontlie other hand, both cases do involve over-looking what one object is or looks like inorder to use or see that object with referenceto a different object.

The second line of evidence De Loacheuses to support her claim is 2y2-year-olds'improved performance on photograph andon drawing conditions of the task, in whichphotographs or drawings of the larger roomreplace the scale model (DeLoache, 1991a).Photographs have no role other than as sym-bolic representations, DeLoache argues, sothe child does not have to see them as twothings at once. It could be, however, that to

interpret a photo in this way is analogous tointerpreting- language: the photo of thecouch might in essence say to the child "golook behind the couch." Why doesn't themodel say this as well? DeLoache arguesthat children are so distracted by the modelas a toy that they cannot get past that in orderto use information from it to guide theirsearch of the big room. This makes goodsense, but it does not indicate that childrenuse the photo as a guide because they under-stand representation. Rather, children mightuse the photo as a guide only because theynote the correspondence between the photoand the big room.

The third line of evidence is that chil-dren do better on the scale model task whenthe salience of the model as an object is de-creased (by being placed behind a windowin a puppet theater) and worse when its sa-lience is increased (by allowing children toplay with it) (DeLoache, 1991b). Althoughthis is again intriguing evidence for the posi-tion that children's difficulty is in seeing themodel both as a thing in itself and as relatingto something else, it still does not show thatthe nature of that relating is one of represen-tation. It could be that playing with themodel makes it hard to see that it structurallycorresponds to something else as well. In-deed, it is difficult to see how one might cir-cumvent this lower-level interpretation ofwhat allows children to succeed on the scalemodel task. At this point it is therefore un-clear whether children really understandthat the model represents the big room, orwhether they simply understand that thetwo spaces match, such that one can findSnoopy in the matching location in the otherspace. The attractiveness of the model as aplay toy does appear to get in the way ofchildren seeing that the model provides use-ful information about the larger room. Itwould be interesting to see if children haveless trouble with less attractive models. Forexample, one could fill the two spaces withgeometric objects instead of furniture, andhide an uninteresting object like a block in-stead of a Snoopy. Although this changecould result in decreased motivation forflnding the object, it is also possible that un-der conditions that provided less temptationto play, younger children would better beable to see that all the objects in the smallspace match the objects in the big space.However, to reiterate, realizing that tworooms match is not the same as seeing oneroom as representing another.

Because it appears that younger chil-

Page 13: Pretend Play Skills and the Child's Theory of Mind

360 Child Developmentdren fail the model task because they areoverly intrigued by the model as a toy, De-Loache's studies can be discussed in thecontext of children understanding that oneobject can be two different things at once:both a toy and a model, for example. How-ever, nothing in the tasks asks children toexplicitly declare such a dichotomy: there isnothing analogous to asking children, "Whatis this really, and what does this look like toyour eyes right now?" It could be the casethat children who pass the task are onlythinking of the small room as one thing—amodel of the larger room. Hence the scalemodel tasks do not provide solid evidencethat children can see one object as twothings at once—they might instead see themodel only as a model. Perhaps childrenbackground the identity of the model as toyrelative to the identity of the model asmodel.

To see something as something else isto understand hypothetical substitution: thattwo flats have the same floor plan, or that apen can be used like a spoon. This does notimply understanding symbolic substitution.Understanding maps is a more certain caseof understanding symbolic substitution. Onthe one hand, the roads on the map reflectthe real roads precisely, curving at the samejunctures, and so on; in a sense, then, under-standing correspondence could get one byin interpreting maps. But the structure of themap is so entirely different from the struc-ture of the referent space that more than cor-respondence must be at issue. A map is notjust a photograph or a drawing or a miniatureof a larger space. Instead, red lines representroads, blue areas represent water, small dotsrepresent towns, and larger dots representcities. Children's understanding of mapstherefore might provide evidence for theirunderstanding of external representation.

Downs and Liben (1987) found that 5 -7-year-olds did not understand the represen-tational nature of maps. For example, theyclaimed that an airplane on a map indicatedthat an airplane was in that location in thereferent space. If that plane flew away, therewould be no plane there. A road representedby a red line was described as really beingred. Finally, the compass symbol was de-scribed as being "a basketball stadium,feathers, and the place where the lifeguardsits" (p. 213). Hence, even at an age whentheir pretend play is very sophisticated, chil-dren's interpretation of map symbols is veryliteral. This might indicate that even olderchildren do not understand external repre-

sentations, but it more certainly indicatesthat such children have simply not beengiven the opportunity to learn map conven-tions. To test whether they are capable oflearning about maps as representations, onewould need to train children on the variousmap symbols and see if they could transferthat training to other maps and their referentspaces—in effect, if they could understandmap conventions as symbols for things in thereal world. If children could, it would sug-gest that they were able, at least with train-ing, to understand that one thing can repre-sent another, because the symbols used onmaps are so different from their referentsthat understanding simple correspondencewould not suffice. For example, blue standsfor water, but that water might be green orbrown rather than blue.

False Photographs and DrawingsThe tasks described in the previous sec-

tion involved noncontradictory representa-tions: for example, the scale model wasaligned with the big room. Such tasks mightbe easier than tasks requiring children toreason about contradictory representations:ones that are not aligned with a current situ-ation. In fact, just as many theorists have ar-gued that one cannot determine whetherchildren understand beliefs until they dem-onstrate understanding of false beliefs (falsebelief tasks are discussed in the last section),one can argue that the true test for under-standing representation is understandingfalse representation. Understanding that fea-tures of one object can correspond to thoseof another can get the child by in noncontra-dictory representation tasks. However, toreason about contradictory representationsrequires a more fundamental understandingof what a representation is.

Zaitchik (1990) devised a method to testchildren's understanding of "false photo-graphs": photographs that represent a pastrather than a current situation. After pre-training on the basic features of a Polaroidcamera, children were introduced to Bertand Ernie. Bert was placed on a mat to sun-bathe, and Ernie took a photo of him. Thephoto was placed face down on the table.Ernie and Bert left the scene, and Big Birdwandered in and lay down on the mat. Thechild was asked, "In the picture, vi ho is onthe mat?" Three-and-a-half-year-olds tendedto fail this task, claiming that Big Bird ratherthan Bert was on the mat in the photo. Four-year-olds, on the other hand, succeeded.The younger children's failure has been

Page 14: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 361taken as evidence that children do not un-derstand representation generally, corrobo-rating the claims of many theorists that suchchildren fail some theory of mind tasks be-cause they do not understand a specific typeof representation, namely, mental represen-tation.

However, to accept the conclusion thatchildren do not understand false photo-graphs (and therefore external representa-tions), one must discount the possibility thatthey are simply confused by the camera it-self. Otherwise, such studies might only bediagnosing children's misunderstanding ofhow Polaroid cameras work. For example,children might think photos are of scenes atthe moment they develop, rather than at themoment when taken. Pemer and Leekam(1990) attempted to control for this problemby adding a "color transmission" conditionto Zaitchik's basic paradigm. This conditioninvolved a nonrepresentational analog of thestandard photo condition. For example, thecamera took a picture of a yellow cloth. Itvras explained to the child that when thecamera clicked "it makes the paper insidetbe same color" as the yellow cloth. In therepresentational version, the photo was of adoll named Judy, who was wearing a greendress. The child was told that when the cam-era clicks "it makes a picture of Judy wear-ing this color." In both versions the coloreditem was changed after the photo was taken,and the child was asked, "What color isJudy/the screen now? What color was Judy/tlie screen when you pressed the button?"and either "What color is Judy in the pic-ture?" (Judy-photo condition) or "What coloris the paper in my hand?" (color transmis-sion condition). Children's performance ontliie color transmission condition was sig-nificantly better than their performance oneither the Judy-photo condition or on a falsebelief task that was administered as part ofthe study. Because both conditions involvedcameras, Perner and Leekam concluded thatchildren's problem lies with understandingrepresentations generally, rather than anyadded difficulty with cameras. There areseveral merits to their method, especiallytide attempt to tightly control across photoand color transmission conditions. However,there is a fundamental difference betweenthe two conditions that might account for thecolor transmission task being easier; namely,tlie form of their questions differed across

conditions. The question in the color trans-mission task ("What color is the paper in myhand?") could draw children's attention tothe photo, whereas the question in the Judy-photo task ("What color is Judy in the pic-ture?") could draw their attention to the doll,who is currently wearing a different colordress than the doll in the picture. Becauseof this, one cannot rule out the possibilitythat the cause of children's errors is questionfocus rather than failure to understand whatis in the photo. A truly analogous task stillneeds to be done.*

Some researchers have attempted to cir-cumvent the camera technology issue byasking children about "false drawings": infull view of the child, a drawing was madeof a scene, the drawing was placed upsidedown on the table, the scene was changed,and children were asked what was in thedrawing (Charman & Baron-Cohen, in press;Lillard & Cho, 1991). However, it is unclearwhether such false drawing studies even tapchildren's understanding of representation,since children need only remember whatwas drawn to pass the task. On the otherhand, Polaroid camera studies also seem un-satisfactory because they use complex cam-era technologies that children may not un-derstand. One might get around both ofthese problems by using rubber stamps in-stead of a Polaroid camera to test children'sunderstanding of what is in a picture. An ob-ject could be placed on a tray, and a rubberstamp of that object could be used to makean image of the object on a piece of paperhidden inside a boxlike apparatus. Theworkings of such a system would thereforebe quite similar to those of a camera butwould not be technologically complex. Theitem on the tray could then be switched, andthe child could then be asked what was onthe original piece of paper. Such a studywould provide a better test of whether chil-dren understand pictorial representations.

SummaryChildren's understanding of representa-

tions outside of the pretend domain appearsto be limited. When a representation doesnot differ from its referent, as in a scalemodel task, 3-year-olds can use it to guidetheir behavior toward the referent, but thiscan be explained by appeal to understandingcorrespondence rather than representation.Although further work along the lines of

* Pemer (personal eommunication, July, 1992) and his colleagues are currently working onsuch a study.

Page 15: Pretend Play Skills and the Child's Theory of Mind

362 Child Development

false photographs is needed, evidence avail-able thus far indicates that children do notunderstand representations when they differfrom referent situations. Nor is there evi-dence that children understand representa-tion within pretense, although how exactlyone would marshal evidence of true sym-bolic understanding here is unclear. In sum,there is no substantive evidence that youngchildren understand external representationeither within or outside of the pretensedomain.

The next section discusses children'sability to represent mental representations.Although this could include representingone's own past or future mental representa-tions, for simplicity discussion is limited tothe representation of others' mental repre-sentations. Understanding nonmental (ex-ternal) representations might be easier thanunderstanding mental representation. Onereason for this is because to understand thatthe mind has internal symbolic contents thatrepresent the world, one would first needto understand that one object can representanother. On the other hand, learning in so-cial domains might precede learning in non-social domains (see Cosmides, 1989), andtherefore representation of other people'smental representations might precede un-derstanding external representations.

Representing MentalRepresentations

"Pretend you thought I was alive, but I wasdead." [Child recorded by Giffin, 1984]The third skill that children appear to

use in pretend play but apparently lack inother domains is the capacity to representothers' mental representations. This sectionfirst discusses character play, in which chil-dren, as part of their pretense, appear toenact another person's view or mental rep-resentation of the world. Next it looks at rep-resentation within pretense in a differentlight, investigating whether when childrenobserve others pretending, they understandthat those others are mentally representingthe pretend situation. Finally, it looks brieflyoutside the pretend domain at tasks that in-vestigate whether children under 4 can rep-resent others' false beliefs.Character Play

The term character play is used here todenote activities that involve the positing ofother characters. To pretend to be anotherperson—or to pretend that a doll is somespecific person—seems to involve repre-

senting the internal life of that person aswell as the person's external qualities. Toplay another character well, one must repre-sent the world as that person represents it.Some studies have found significant positivecorrelations between frequency of dramaticplay and such skills as perspective-taking,cooperativeness, and social competence(Connolly & Doyle, 1984; Cragg, 1977, citedin Rubin & Pepler, 1980; Rubin, 1976; Ru-bin & Maioni, 1975). Representation of oth-ers' mental representations (metarepresenta-tion) is at the core of perspective-taking. Itmay be a component of cooperativeness andsocial competence as well, since being ableto see the world through another's eyes canallow one to better determine how best tocooperate with that person.

One might take the position that thesecorrelations result from pretend play servingas a training ground for the other skills, per-haps training in metarepresentation. How-ever, correlations between frequency ofthose skills and sociodramatic play mightalso be due to a different underlying fac-tor rather than metarepresentation; namely,children with better social skills are proba-bly better able to negotiate play. Pretendplay training studies might answer this ques-tion, but existing studies (Burns & Brainerd,1979; Fink, 1976; Rosen, 1974; Saltz et al.,1977; Smilansky, 1968; Smith & Sydall,1978; Spivak & Shure, 1974) do not controlfor the effects of training in general, and theyoften use measures that may not be valid.Hence the relation between social-cognitiveskills and pretend play is unclear. Further-more, the relation between metarepresenta-tional skills and both pretend play and othersocial-cognitive assessments is unclear (al-though performance on false belief tasks iscorrelated with performance on perspective-taking tasks; see Astington & Gopnik, 1988;Flavell, 1988). We do not know, for exam-ple, if children who are precocious in theirpretend play pass false belief tasks at an ear-lier age than other children. In addition, nostudies have examined children's pretendplay with respect to the level of metarepre-sentational knowledge children show intheir character portrayals.

This section focuses on metarepresenta-tion in two different types of character play:sociodramatic play and play with dolls. Al-though both seem to involve different chal-lenges and aids to metarepresentation, itis unclear which would show the mostadvanced skill levels. Some studies haveshown that children apply internal states to

Page 16: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 363dolls later than they apply them to people(Wolf, Rygh, & Altshuler, 1984), hence theirplay with real live partners may look moreadvanced (but see Dolgin & Behrend, 1984;Gelman, Spelke, & Meek, 1983). Real part-ners might scaffold children to higher lev-els of social understanding than they canachieve alone (Dunn, 1988, 1991; Vygotsky,1978), also making play with live partnerslook more advanced, at least when thosepartners are themselves more advanced. Onthe other hand, sociodramatic play involvescoordinating one's actions with live playerswho have their own ideas about what to do.This might bring to sociodramatic play alevel of complexity that would interfere withrepresenting the mental representations ofcharacters within the play. In playing withdolls, the child is in complete control ofvi'hat each character does and thinks.

Sociodramatic play.—Sociodramaticplay emerges around 3 years of age (Fein,1981; Garvey, 1990; Rubin et al., 1983) (al-tliough it may emerge as early as 18 monthsfor younger siblings in the family setting;see Dunn, 1988). At first, such play doesnot appear to involve metarepresentation.Piather, it appears to be scripted, ritualizedbehavior that proceeds in a turn-taking pat-tern. For example, one child will say some-thing that the other child will then repeat,possibly with a slight alteration. Severalpieces of evidence support the notion thatearly sociodramatic play revolves aroundscripts. For example, roles are sometimesadopted but are easily abandoned when theturn-taking structure is not adhered to (Gar-vey, 1990; Rubin et al., 1983). Not untilaround 5 years of age do children genuinelyintercoordinate their actions and easily in-corporate other players' unexpected inven-tions (Rubin et al., 1983), as would be ex-pected if they were not relying heavily onscripted routines. Garvey (1990; Garvey &Berndt, 1975) outlined three types of rolestliat children adopt in play: functional roles,which revolve around activities, like cook;character roles, based on stereotypes, likefireman; and relational roles, like motherand child. These are all roles about whichchildren would seem to have abundantscripted knowledge. At younger ages, themost frequently adopted roles are relationalones, particularly those involving domes-tic themes (see Forys & McCune-Nicolich,1984). For younger children, the adoptedcharacter conforms very closely to stereo-typed images of those characters; by 5 yearsof age, children add their own variations

(Garvey & Berndt, 1975; Mathews, 1977). Inaddition, 3-year-olds tend to develop theirroles independently of their play partners,whereas 5-year-olds develop their roles verymuch in reference to each other (Iw^anaga,1973). These factors all support the impres-sion that early sociodramatic play involvesthe carrying out of scripted routines, ratherthan imagining others' mental representa-tions of the world and acting on them.

Carrying out scripts certainly does notentail metarepresentation. When one is trulymetarepresenting, one is free to vary from ascript because one is "in the character" thatone is playing. Taken together, therefore,these studies give the impression that notuntil around 5 years of age does sociodra-matic play involve the careful intercoordina-tion that one would expect if children weremetarepresenting. Although children en-gage in sociodramatic pretend play earlier,it appears to be largely scripted and to de-pend on staying within those scripts. Suchbehavior is consistent with (or even delayedrespective to) the level of metarepresenta-tional knowledge children are thought tohave outside of the play domain, for exam-ple, in visual perspective-taking, appear-ance-reality, and false belief tasks. But socio-dramatic play is complex, since it requiresintercoordination. The child must be con-structing and maintaining the same fantasyframework that the other child is working in(Dunn 6s: Dale, 1984). Children might ex-hibit metarepresentational skills earlier insolitary play with dolls, since dolls do nothave their own ideas about what to play.

Doll ploy.—Wolf et al. (1984) describea three-stage model of the development ofreplica play. At about 2 years of age, childrenattribute independent agency to dolls (seealso Watson & Fischer, 1977, 1980; Wolf,1982). About 6 months later, they attributeperceptual, sensory, and emotive experi-ences to them. Finally, just before 3V2 yearsof age, they attribute cognitive experi-ences to dolls. (These are about the sameages when children apply these differentstates to others in real life as well—seeBretherton & Beeghly, 1982; Dunn, Breth-erton, & Munn, 1987; Shatz, Wellman, &Silber, 1983). Sensory and emotional expe-riences are not intentional states, henceapplying them to others does not constituterepresenting representational states (Den-nett, 1987). Hence only the third stage cantruly be considered metarepresentational,and it was first seen to occur in Wolf et al.'ssubjects at an average age of 3 years, 4

Page 17: Pretend Play Skills and the Child's Theory of Mind

364 Child Developmentmonths. This is 6 to 8 months earlier thanmost children pass tests like those of falsebelief and perspective-taking that are as-sumed to require the ability to metarepre-sent. Two possibilities arise: Since the Wolfet al. study describes the first noted appear-ance of the skill, it may be that children havefledgling metarepresentational skills thatthey occasionally use botli within and out-side of pretense. Under this interpretation,because Wolfe et al. observed children overmany hours, they chanced to observe earlymanifestations of the skill. A second possi-bility is that even in applying cognitivestates to dolls, children are simply reen-acting scripts. Alternatively, it may be thatmetarepresentational skills are manifestedin pretend play earlier than in nonpretendscenarios. In this case, pretense would be anarea of advanced development of metarepre-sentational skills.

In sum, although sociodramatic playseems to involve representing others' men-tal representations, it may only involve en-acted scripted routines. Further data exam-ining this issue are needed. The next sectionlooks at whether children understand pre-tending itself to involve mental represen-tations.Metarepresenting the PretendRepresentation

Leslie's (1987, 1988) computational ac-count of pretend play is frequently inter-preted as viewing pretend play as an activityin which children show advanced develop-ment with regard to representing others'mental representations. According to Leslie,the infant initially has only primary repre-sentations of the world: she sees the worlddirectly, and represents it as it is. In pretendplay, however, such a representational sys-tem would not suffice. A child's primary rep-resentation of a banana would be greatly dis-turbed by watching someone pretend abanana was a telephone: the child wouldstart to think of a banana as something youtalk into. Therefore it must be the case thatthe child develops another type of represen-tational system for use in pretend play. Thissame system, according to Leslie, is what isused to understand mental states. The prod-ucts of this system are secondary representa-tions, or representations of representations(he calls these metarepresentations, al-though many others reserve the term forcases in which the representational relationis also represented, e.g., Forguson & Gop-nik, 1988; Olson, 1988; Perner, 1991; Pyly-shyn, 1978; Whiten & Perner, 1991). Sec-ondary representations are freed up from

their usual meanings, so that one object cansubstitute for a different object without thechild confusing actual semantic relations.Leslie posits a close relation between pre-tense and developing a theory of mind. Heclaims that the child uses the same mecha-nism to understand that someone is pre-tending a banana is a telephone that she usesto understand that someone mistakenly be-lieves a banana is a telephone; in eithercase, the primary representation of a bananais freed from the secondary one (see Pemer,1991, for some problems with this view). Al-though Leslie does not think that the child'suse of these secondary representations inpretend play causes development of a theoryof mind, he does think that in pretend playwe see a "nascent ability to understand cog-nition" (1988, p. 31). "Pretending oneself isthus a special case of the ability to under-stand pretense in others (someone else's atti-tude to information). In short, pretense is anearly manifestation of what has been calledtheory of mind (Premack & Woodruff,1978)" (Leslie, 1987, p. 416).

What exactly Leslie means by an "atti-tude to information" is not specified. Theissue is whether children understand the in-formation to be mentally represented by theother person or not, and there are three rea-sons why Leslie's work is often interpretedas claiming that they do (e.g., Forguson,1989; Siegler, 1991; Whiten & Byrne, 1991).The first is terminology: metarepresentationtypically implies a representation of a repre-sentation, and even appreciation of the rep-resentational relation. For example, whenLeslie (1991) writes "the child makes an in-ference from the others' perceived behaviorto a metarepresentation of the other's in-tended message" (p. 75), it is easily parsedas the infant metarepresenting the other'srepresentation of the situation. Second, thevery structure of Leslie's (1987) argumentinvites the interpretation that children un-derstand pretense as involving representa-tions. His argument begins by stating thatpretense involves two simultaneous repre-sentations of a situation (the real and the pre-tend), and then states that children under-stand pretense as soon as they begin topretend: "Understanding pretense is simplypart and parcel of being able to pretend one-self" (1987, p. 416). Because pretense surelyinvolves sophisticated use of representa-tions, Leslie's statement seems to mean thatthe child understands representation in a so-phisticated manner. The third source of theinterpretation is the association Leslie(1987, 1988) draws between understanding

Page 18: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 365false beliefs and understanding pretense.He argues that the cognitive structures un-derlying false belief and pretense under-standing are the same, and that because chil-dren have those structures in place by age2, there must be a nonstructural reason whythey fail false belief tasks until age 4. (Hisinitial claim was that false belief tasks wereharder than pretense because they requiredinferential ability. This claim has been re-tracted [Leslie, 1991] as others have devel-oped false belief tasks that do not requirechildren to make inferences [Flavell, Fla-vell, Green, & Moses, 1990; Lillard & Fla-vell, 1992].) Other investigators took the tackthat children fail false belief tasks becausethey do not understand representation (e.g.,Flavell, 1988; Forguson & Gopnik, 1988;Pemer, 1988); Leslie denied that could bethe problem since the requisite cognitivestructures for understanding representationmust be in place. Leslie's posing a nonstruc-tural reason for false belief failures also im-plied that in his view children do under-stand representation, and understandpretense as a representational mental state.Many have read Leslie's work to mean this,and have accepted that pretense is under-stood representationally. Indeed, evidencefrom autism fits the view that pretend playand a theory of mind are linked by one un-derlying factor, since autistic children dis-play an absence of both (Baron-Cohen, 1987;Leslie, 1991). In sum, Leslie's argumentstrongly implies that pretense is an area inwhich children show early competence forunderstanding mental representation.

However, no experimental evidence hassupported that claim, and Lillard (1991, and1993, in this issue) has recently gathered ev-idence that contradicts it. As noted pre-viously, pretense always entails a mentalrepresentational component, and sometimesalso is accompanied by an action compo-nent. To understand the representationalcomponent of pretense requires metarepre-sentation: one must represent oneself or an-other person representing a block as acookie, and must understand that represen-tation as a representation. To understand theaction component, on the other hand, oneneed only know that someone is acting onthe block as if it were a cookie. Althoughboth action and representation are usuallyinvolved in pretense, the representationalcomponent is the more crucial one. For ex-iimple, a block may be a cookie in a pretendgame, but not be acted upon at any givenmoment. Even when it is not acted upon,however, it is mentally representefd by the

pretender as a cookie. What evidence do wehave that children understand the represen-tational component of pretense? Most stud-ies of pretense involve pretense with actions(Flavell, Flavell, & Green, 1987; Woolley &Wellman, 1990). In these studies, childrencould perform correctly by referring to theaction, rather than by metarepresenting. Infact, as was discussed above, children per-formed better in Woolley and Wellman'sstudy, in which actions provided a solid cueto the pretense, than in Flavell et al.'s study.Lillard (1991, and 1993, in this issue) testedthe hypothesis that children first understandpretending only as an action, and only latercome to see it as involving mental represen-tations. For example, in one experiment atroll doll named Moe, described as beingfrom a mythical "Land of the Trolls," wasintroduced, and was shown to be hoppingaround. The experimenter said, "He's hop-ping around, kind of like a rabbit hops. Moedoesn't know that rabbits hop like that—hedoesn't know anything about rabbits—buthe is hopping like a rabbit" (Lillard, 1993,Experiment 3, in this issue). Children wereasked three questions: (a) "Is he hoppinglike a rabbit?" (b) "Does he know that rab-bits hop like that?" and (c) "Would you sayhe is pretending to be a rabbit, or he is notpretending to be a rabbit?" Although theycorrectly answered yes to the first questionand no to the second, over 60% of 4-year-olds claimed on at least three of four suchtrials that Moe was pretending to be a rabbit.In other words, they seemed to interpretpretense as the action alone, and failed toconsider it as a mental representationalstate. Interestingly, these same childrentended to pass a false belief task (Astington& Gopnik, 1988), indicating that childrenmay understand that different people canrepresent the same object differently beforethey understand the representational com-ponent of pretense.

In an additional experiment, 4-year-oldswere shown pictures of other children whowere variously described as either hoppinglike a rabbit or not hopping like a rabbit, andas either thinking about being a rabbit or notthinking about being a rabbit. They werethen asked, for each child, whether she waspretending to be a rabbit. Altliough all chil-dren passed four control trials in which thepremises were both positive (hopping likeand thinking about being a rabbit) or bothnegative, over 60% claimed on three of fourtrials that the girl was pretending to be arabbit when she was performing the actionof hopping but failed to have the necessary

Page 19: Pretend Play Skills and the Child's Theory of Mind

366 Child Developmentmental representation of thinking about be-ing a rabbit (Lillard, 1993, Experiment 4, inthis issue).

These studies indicate that, contrary towhat many have supposed, young childrenmay not mentally represent others' mentalrepresentations even in pretend situations.They may instead simply think of pre-tending as acting in a certain way. However,it could be that in naturalistic situations chil-dren have some understandings they do nothave in experimental situations. For exam-ple, they may know not to ask someone elseto pretend to be Big Bird if they know thatthe person does not know who Big Birdis. Further work is needed to determinewhether this is in fact the case.

Metarepresentation Outside thePretend World

Recent years have seen a surge of theo-retical and experimental work on children'sunderstanding of the mind. Much of thiswork addresses children's understanding ofand ability to represent others' mental repre-sentations of situations. This section dis-cusses children's performance on tasks in-volving representation of others' beliefs.Beliefs are representations of the world: in-temal mental symbols that are supposed toreflect the way the world really is (whenthey fail to do so, they are false beliefs). Tofully understand a belief is also to representwhat a belief is—what sorts of things causebeliefs, and what beliefs in turn cause. Al-though some belief research has focused onchildren's understanding of true beliefs,most has focused on false beliefs. This is duelargely to a problem interpreting successfulperfonnance on true belief tasks. In theprevious section it was argued that correctperformance on tasks concerning externalrepresentations could be achieved via an un-derstanding of correspondence betweenspaces, and that contradictory representa-tions provided a truer test of whether chil-dren understand representations. The sameis true of metarepresentation; one can bemore sure that children are metarepresent-ing when the tasks could not be solved bysimpler understandings like correspon-dence. For example, when the belief to beunderstood is false and therefore does notcorrespond to reality, children cannot arriveat correct answers by assuming correspon-dence between beliefs and reality.

In 1983, Wimmer and Perner reported aseries of studies investigating children's un-derstanding of false beliefs. In these studies.

a character named Maxi put some chocolatein a cupboard, and then left the scene. InMaxi's absence, his mother moved the choc-olate to a new location, a drawer. Adults, andeven 4-year-olds, understand that Maxi's be-lief that his chocolate is in the cupboard isnow false, since it has been moved withouthis knowledge. Three-year-olds, however,do not seem to understand that people canhave false beliefs. For example, they claimthat Maxi thinks the chocolate is in its newlocation, the drawer, even though he has noaccess to that fact. The false belief task hasbeen replicated many times under many con-ditions, some of which were designed to di-minish or even remove potential inferentialdifficulties about the origins of beliefs andto take them out of their causal-explanatoryframework, and the consensual evidence isthat the 3-year-olds' deficit is real (e.g., Fla-vell, Flavell, Green, & Moses, 1990; Gopnik& Slaughter, 1991; Lillard & Flavell, 1992;Moses & Flavell, 1990; Russell, Mauthner,Sharpe, & Tidswell, 1991). Although somehave found improved performance undersome conditions (Hala, Chandler, & Fritz,1991; Lewis & Osborne, 1990; Mitchell &Locohee, 1991; Siegal & Beattie, 1991), thefact remains that in the majority of studiesmost 3-year-olds have real and abiding dif-ficulties with the task. Interestingly, chil-dren do significantly better on the same taskwhen it involves mental states other than be-lief (Lillard & Flavell, 1992). The problemfor 3-year-oIds seems to arise only when thefocal mental state involves representing theworld in some way. Outside of the pretenddomain, therefore, children do not appear tounderstand how people can misrepresent asituation, or have a false belief about an ob-ject or situation, until around 4 years of age.

Sum,maryGhildren's performance on tasks con-

cerning beliefs and pretense does not pro-vide evidence of metarepresentational skillsprior to age four. Ghildren's sociodramaticplay does not indicate that they understandmetarepresentation at 3 years of age either,although sociodramatic play is admittedlycomplex, given the need to intercoordinatewith others. Solitary doll play removes theproblem of others' wills, and children doseem to apply mental states to dolls at anearlier age than they succeed on metarepre-sentational tasks. However, exactly what thisshows is unclear without further research.Perhaps what seems to be metarepresenta-tional in applying cognitive states to dolls isin fact only scripted knowledge (Nelson &Seidman, 1984). A careful investigation of

Page 20: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 367children's replica play with this question inmind would be useful.

ConclusionEvidence reviewed here indicates that

in some scenarios 3-year-olds really do seemto show understandings in pretense that arebeyond those that they show in nonpretenddomains. In particular, they seem to applytwo identities to an object at once in pretendplay, but not in other domains like learningobject labels or understanding illusions. Per-haps children background the real objectidentity while dealing explicitly with thepretend identity. Alternatively, they mayhave achieved better understanding in pre-tense because pretense emanates more fromthe mind than the world, or because theysee it as usage or activity rather than as amental state.

No studies directly address the issue ofsymbolic versus hypothetical substitutionvi' ith regard to children's own object substi-tutions. Research from nonpretend scenariosindicates that although children may under-stand correspondence prior to 4 years of age,they might not understand representations,since they fail to accurately report the con-tents of contradictory representations. How-ever, there are problems with the researchon contradictory external representationsthat must be addressed before firm conclu-sions can be drawn.

It has been suggested that even in natu-ralistic pretense children are not represent-ing others' mental representations untilsometime after 3 years of age, consistentv ith several studies of children's metarep-resentational abilities in nonpretend situa-tions. However, they do attribute cognitionto dolls earlier than they pass metarepresen-tiitional tasks. On the other hand, recent evi-dence suggests that even at 4 years of age,when children seem to understand false be-liefs, they do not appreciate the mental rep-resentational underpinnings of pretense it-self In sum, although children's behaviorin pretense often seems metarepresenta-tional, it might not actually be.

What might make children's pretendplay seem more sophisticated than it reallyis? Several theorists have suggested thatchildren understand pretense as acting-as-if(Harris, 1991; Lillard, 1991, and 1993, in thisissue; Lillard & Flavell, 1992; Perner, 1991;Wellman, in press). It may be that the capac-ity to participate in as-if worlds is the impor-tant development in pretense (Bretherton,1989). This capacity is exercised early in

pretend play, and in children's love of sto-ries (Walton, 1990). It may also be mani-fested in adults' general ability to share inthe imaginations of others: novels, operas,paintings, and music, to name a few. In addi-tion, it probably contributes to our ability toplan ahead, to intend, and to decide be-tween altemative actions. However, in earlypretend play, children might not yet under-stand that the alternative worlds of pretenseand stories exist foremost in minds; instead,they may see them as existing in attitudes,behaviors, and books. Perhaps it is only withthe onset of an understanding of the repre-sentational mind around age 4 that childrenbegin to see alternative worlds as mentallybased. Pretend play may be a zone of proxi-mal development in the same sense as en-joying stories may be one: it may free chil-dren to participate in other realities. Butthey may not yet understand these realitiesto be mentally based.

ReferencesAstington, J. W., & Gopnik, A. (1988). Knowing

you've changed your mind: Children's under-standing of representational change. In J. W.Astington, P. L. Harris, & D. R. Olson (Eds.),Developing theories of mind (pp. 193-206).Cambridge: Cambridge University Press.

Astington, J. W., Harris, P. L., & Olson, D. R.(1988). Developing theories of mind. NewYork: Cambridge University Press.

Austin, J. L. (1979). Pretending. In J. O. Urmson& G. J. Warnock (Eds.), Philosophical papers.Oxford: Oxford University Press.

Baron-Cohen, S. (1987). Autism and symbolicplay. British journal of Developmental Psy-chology, 5, 139-148.

Bateson, G. A. (1972). A theory of play and fantasy.In G. A. Bateson (Ed.), Steps to an ecology ofmind (pp. 39-51). New York: Chandler (Orig-inal work published 1955)

Beal, C. R. (1988). Children's knowledge aboutrepresentations of intended meaning. In J. W.Astington, P. L. Harris, & D. R. Olson (Eds.),Developing theories of mind (pp. 315-325).Cambridge: Cambridge University Press.

Braine, M. D. S., & Shanks, B. L. (1965). The con-servation of a shape property and a proposalabout the origin of conservations. Canadianjournal of Psychology, 19, 197-207.

Bretherton, I. (1984a). Representing the socialworld in symbolic play: Reality and fantasy.In I. Bretherton (Ed.), Symbolic play (pp. 3—41). New York: Academic Press.

Bretherton, I. (1984b). Symbolic play: The de-velopment of social understanding. Orlando,FL: Academic Press.

Bretherton, I. (1989). Pretense: The form and

Page 21: Pretend Play Skills and the Child's Theory of Mind

368 Child Developmentfunction of make-believe play. Develop-mental Review, 9, 383-401.

Bretherton, I., & Beeghly, M. (1982). Talkingabout internal states: The acquisition of anexplicit theory of mind. Developmental Psy-chology, 18, 906-921.

Bruner, J. S. (1972). Nature and uses of immatu-rity. Am,erican Psychologist, 27, 1-28.

Burns, S. M., & Brainerd, C. J. (1979). Effects ofconstructive and dramatic play on perspec-tive-taking in very young children. Develop-mental Psychology, 15, 512-521.

Butterworth, G., Harris, P., Leslie, A., & Wellman,H. (1991). Perspectives on the child's theoryof mind [Special issue] British Journal of De-velopmental Psychology, 9.

Charman, T., & Baron-Cohen, S. (in press). Re-search Note: Understanding drawings andbeliefs: A further test of the metarepresen-tational theory of autism, journal of ChildPsychology and Psychiatry.

Clark, E. V. (1987). The principle of contrast: Aconstraint on acquisition. In B. MacWhinney(Ed.), Mechanisms of language acquisition:The 20th annual Carnegie Symposium onCognition. Hillsdale, NJ: Erlbaum.

Connolly, J. A., & Doyle, A. (1984). Relation ofsocial fantasy play to social competence inpreschoolers. Developmental Psychology, 20,797-807.

Cosmides, L. (1989). The language of social ex-change: Has natural selection shaped how hu-mans reason? Studies with the Wason Selec-tion Task. Cognition, 31, 187-276.

Cragg, S. (1977). The effects of empathy, role-taking, and play on altruism. Unpublishedmanuscript. University of Waterloo.

DeLoache, J. S. (1989). Young children's under-standing of the correspondence between ascale model and a larger space. Cognitive De-velopment, 4, 121-139.

DeLoache, J. S. (1990). Young children's under-standing of models. In R. Eivush & J. A. Hud-son (Eds.), Knowing and remem.bering inyoung children (pp. 94-127). Cambridge:Cambridge University Press.

DeLoache, J. S. (1991a). Symbolic functioning invery young children: Understanding of pic-tures and models. Child Development, 62,736-752.

DeLoache, J. S. (1991b). When a picture is notworth a thousand words: Understanding thesymbolic function of pictures and models. Pa-per presented at the biennial meeting of theSociety for Research in Child Development,Seattle.

DeLoache, J. S., & Plaetzer, B. (1985). Tea fortwo: joint mother-child symbolic play. Paperpresented at the biennial meeting of theSociety for Research in Child Development,Toronto.

DeLoache, J. S., Strauss, M. S., & Maynard, J.(1979). Picture perception in infancy. InfantBehavior and Development, 2, 77-89

Dennett, D. (1987). The intentional stance. Cam-bridge, MA: MIT Press.

Dias, M. G., & Harris, P. L. (1988). The effectof make-believe play on deductive reasoning.British Journal of Developmental Psychol-ogy, 6, 207-221.

Dias, M. G., & Harris, P. L. (1990). The inBuenceof the imagination on reasoning by youngchildren. British journal of DevelopmentalPsychology, 8, 305-318.

DiLalla, L. F., & Watson, M. W. (1988). Differenti-ation of fantasy and reality: Preschoolers' re-actions to interruptions in their play. Devel-opmental Psychology, 24, 286-291.

Dolgin, K. G., & Behrend, D. A. (1984). Children'sknowledge about animates and inanimates.Child Development, 55, 1646-1650.

Downs, R. M., & Liben, L. S. (1987). Children'sunderstanding of maps. In P. Ellen & C.Thinus-Blanc (Eds.), Cognitive processesand spatial orientation in animal and m,an(pp. 202-219). Dordrecht, Holland: MartinusNijhoff

Dretske, F. (1988). Explaining behavior. Cam-bridge, MA: Bradford/MIT Press.

Dunn, J. (1988). The beginnings of social under-standing. Cambridge, MA: Harvard Univer-sity Press.

Dunn, J. (1991). Young children's understandingof other people: Evidence fronfi observationswithin the family. In C. Moore & D. Frye(Eds.), Children's theories of mind (pp. 97-114). Hillsdale, NJ: Erlbaum.

Dunn, J., Bretherton, I., & Munn, P. (1987). Con-versations about feeling states between moth-ers and their young children. DevelopmentalPsychology, 23, 132-139.

Dunn, J., & Dale, N. (1984). I a daddy: 2-year-olds'collaboration in joint pretend play with sib-ling and with mother. In I. Bretherton (Ed.),Sym-bolic play (pp. 131-158). London: Aca-demic Press.

Elder, J. L., & Pederson, D. R. (1978). Preschoolchildren's use of objects in symbolic play.Child Development, 49, 500-504.

Fein, G. G. (1975). A transformational analysis ofpretending. Developm.ental Psychology, 11,291-296.

Fein, G. G. (1979). Play and the acquisition ofsymbols. In L. G. Katz (Ed.), Current topicsin early childhood education (pp. 195—225).Norwood, Nj: Ablex.

Fein, G. G. (1981). Pretend play in childhood: Anintegrative review. Child Development, 52,1095-1118.

Fink, R. S. (1976). Hole of imaginative play in cog-nitive development. Psychological Reports,39, 895-906.

Page 22: Pretend Play Skills and the Child's Theory of Mind

Angeline S. Lillard 369Fitts, P. M., & Posner, M. I. (1967). Human perfor-

mance. Belmont, CA: Brooks-Cole.Flavell, J. H. (1988). The development of chil-

dren's knowledge about the mind: From cog-nitive connections to mental representations.In J. W. Astington, P. L. Harris, & D. R. Olson(Eds.), Developing theories of mind (pp. 244-271). Cambridge: Cambridge UniversityPress.

Flavell, J. H., Flavell, E. R., & Green, F. L. (1987).Young children's knowledge about the appar-ent-real and pretend-real distinctions. Devel-opmental Psychology, 23, 816-822.

Flavell, J. H., Flavell, E. R., Green, F. L., &Moses, L. J. (1990). Young children's under-standing of fact beliefs versus value beliefs.Child Development, 61, 915-928.

Flavell, J. H., Green, F. L., & Flavell, E. R.,(1986). Development of knowledge about theappearance-reality distinction. Monographsofthe Society for Research in Child Develop-ment, 51(1, Serial No. 212).

Flavell, J. H., Green, F L., & Flavell, E. R. (1990).Developmental changes in young children'sknowledge about the mind. Cognitive Devel-opment, 5, 1—27.

Flavell, J. H., Green, F. L., Wahl, K. E., & Flavell,E. R. (1987). The effects of question clarifica-tion and memory aids on young children'sperformance on appearance-reality tasks.Cognitive Development, 2, 127-144.

Forguson, L. (1989). Common sense. London:Koutledge.

I^orguson, L., & Gopnik, A. (1988). The ontogenyof common sense. In J. W. Astington, P. L.Harris, & D. R. Olson (Eds.), Developing the-ories of mind (pp. 226-243). New York: Cam-bridge University Press.

Forys, S. K. S., & McCune-Nicolich, L. (1984).Shared pretend: Sociodramatic play at 3 yearsof age. In I. Bretherton (Ed.), Symbolic play(pp. 159-194). London: Academic Press.

Frye, D., & Moore, C: (1991). Children's theoriesof mind. Hillsdale, NJ: Erlbaum.

Garvey, G. (1990). Play. Gambridge, MA: HarvardUniversity Press.

Garvey, G., & Berndt, R. (1975). Organization inpretend play. Paper presented at the meetingof the American Psychological Association,Chicago.

Gelman, R., Spelke, E. S., & Meek, E. (1983).What preschoolers know about animate andinanimate objects. In D. Rogers & J. Sloboda(Eds.), The acquisition of symbolic skills (pp.297-326). New York: Plenum.

Giffin, H. (1984). The coordination of meaning inthe creation of shared make-believe play. In I.Bretherton (Ed.), Symbolic play (pp. 73-100).Orlando, FL: Academic Press.

Goodman, N. (1968). Languages of art. Indianapo-lis: Bobbs-Merrill.

Gopnik, A., & Slaughter, V. (1991). Young chil-dren's understanding of changes in their men-tal states. Child Development, 62, 98-110.

Gottinan, J. M. (1986). The world of coordinatedplay: Same- and cross-sex friendships in chil-dren. In J. M. Gottman & J. G. Parker (Eds.),Conversations of friends (pp. 139-191). NewYork: Cambridge University Press.

Hala, S., Chandler, M., & Fritz, S. A. (1991).Fledgling theories of mind: Deception as amarker of three-year-olds' understanding offalse belief. Child Development, 62, 83-97.

Harris, P. L. (1991). The work ofthe imagination.In A. Whiten (Ed.), Natural theories of mind(pp. 283-304). Oxford: Basil Blaekwell.

Horgan, D. (1980). Nouns: Love 'em or Leave 'em.In V. Teller & S. White (Eds.), Studies inchild language and multilingualism. (pp. 5—26). New York: New York Academy of Sci-ences.

Huttenlocher, J., & Higgins, E. T. (1978). Issuesin the study of symbolic development. InW. A. Collins (Ed.), Minnesota symposium onchild psychology (Vol. 11, pp. 98-140). Hills-dale, NJ: Erlbaum.

Iwanaga, M. (1973). Development of interper-sonal play structures in 3, 4, and 5-year-oldchildren, journal of Research and Develop-ment in Education, 44, 196-200.

Jackowitz, E. R., & Watson, M. W. (1980). Thedevelopment of object transformations inearly pretend play. Developmental Psychol-ogy, 16, 543-549.

Karmiloff-Smith, A. (1991). Beyond modular-ity: Innate constraints and developmentalchange. In S. Carey & R. Gelman (Eds.), Epi-genesis of the m,ind: Essays in biology andknowledge (pp. 171-198). Hillsdale, NJ:Erlbaum.

Kuczaj, S. A. (1981). Factors influencing children'shypothetical reference./ournai of Child Lan-guage, 8, 131-137.

Leslie, A. M. (1987). Pretense and representation:The origins of "theory of mind." Psychologi-cal Review, 94, 412-426.

Leslie, A. M. (1988). Some implications of pre-tense for mechanisms underlying the child'stheory of mind. In J. W. Astington, P. L. Har-ris, & D. R. Olson (Eds.), Developing theoriesof mind (pp. 19—46). New York: CambridgeUniversity Press.

Leslie, A. M. (1991). The theory of mind impair-ment in autism: Evidence for a modularmechanism of development? In A. Whiten(Ed.), Natural theories of mind (pp. 63-78).Oxford: Basil Blackweli.

Lewis, C, & Osborne, A. (1990). Three-year-olds'problems with false belief: Conceptual deficitor linguistic artifact? Child Development, 61,1514-1519.

Liben, L. S., & Downs, R. M. (1989). Understand-

Page 23: Pretend Play Skills and the Child's Theory of Mind

370 Child Developmenting maps as symbols: The development ofmap concepts in children. In W. H. Reese(Ed.), Advances in child development and be-havior (pp. 145-201). San Diego, CA: Aca-demic Press.

Lillard, A. S. (1991). Young children's conceptual-ization of pretend. Unpublished doctoral dis-sertation, Stanford University.

Lillard, A. S. (1993). Young children's conceptual-ization of pretense: Action or mental repre-sentational state? Child Development, 64,372-386.

Lillard, A. S., & Cho, V. (1991, August). Youngchildren's understanding of representations.Paper presented at the American Psychologi-cal Association, San Francisco.

Lillard, A. S., & Flavell, J. H. (1992). Young chil-dren's understanding of different mentalverbs. Developm.ental Psychology, 28, 626-634.

Mandler, J. M. (1988). How to build a baby: Onthe development of an accessible represen-tation system. Cognitive Development, 3,113-136.

Markman, E. M. (1987). How children constrainthe possible meanings of words. In U. Neisser(Ed.), Concepts and conceptual development(pp. 255-287). Cambridge: Cambridge Uni-versity Press.

Markman, E. M. (1989). Categorization and nam-ing in children. Cambridge, MA: Bradford/MIT Press.

Mathews, W. S. (1977). Modes of transformation inthe initiation of fantasy play. DevelopmentalPsychology, 13, 212-216.

McCune-Nicolich, L. (1981). Toward symbolicfunctioning: Structure of early pretend gamesand potential parallels with language. ChildDevelopment, 52, 785-797.

Mead, G. H. (1934). Mind, self, and society. Chi-cago: University of Chicago Press.

Merriman, W. E., & Bowman, L. L. (1989). Themutual exclusivity bias in young children.Monographs of the Society for Research inChild Development, 54(3-4, Serial No. 220).

Mitchell, P., & Lacohee, H. (1991). Children'searly understanding of false belief. Cogni-tion, 39, 107-127.

Moses, L. J., & Flavell, J. H. (1990). Inferring falsebeliefs from actions and reactions. Child De-velopment, 61, 929-945.

Mussen, P., & Eisenberg-Berg, N. (1977). Roots ofcaring, sharing and helping. San Francisco:Freeman.

Nelson, K., & Seidman, S. (1984). Playing withscripts. In I. Bretherton (Ed.), Sym.bolic play(pp. 45-72). London: Academic Press.

Olson, D. R. (1988). Representation and misrep-resentation: On the beginnings of symbol-ization in young children. Unpublished

manuscript, Ontario Institute for Studies inEducation.

Olson, D., & Campbell, R. (in press). Construct-ing representations. In C. Pratt & A. Garten(Eds.), Representational development. NewYork: Wiley.

Overton, W. F., & Jackson, J. P. (1973). The repre-sentation of imagined objects in action se-quences: A developmental study. Child De-velopment, 44, 309-314.

Pederson, D. R., Rook-Green, A., & Elder, J. L.(1981). The role of action in the developmentof pretend play in young children. Develop-mental Psychology, 17, 756-759.

Pemer, J. (1988). Developing semantics for theo-ries of mind: From propositional attitudes tomental representation. In J. W. Astington,P. L. Harris, & D. R. Olson (Eds.), Developingtheories of m.ind (pp. 141-172). Cambridge:Cambridge University Press.

Perner, J. (1991). Understanding the representa-tional mind. Cambridge, MA: MIT Press.

Perner, J., & Leekam, S. (1990). Photographyversus color transmission: Zooming in onchildren's problem, with "representation."Unpublished manuscript. University of Sus-sex.

Piaget, J. (1962). Play, dreams, and imitation inchildhood. New York: Norton.

Pierce, C. S. (1940). Logic as semiotic: The lan-guage of signs. In J. Buchler (Ed.), The philos-ophy of Pierce: Selected writings. London:Routledge & Kegan Paul.

Potter, M. C. (1979). Mundane symbolism: Therelations among objects, names, and ideas. InN. R. Smith & M. B. Franklin (Eds.), Symbolicfunctioning in childhood (pp. 41—65). Hills-dale, NJ: Erlbaum.

Premack, D., & Woodruff G. (1978). Does thechimpanzee have a theory of mind? Behav-ioral and Brain Sciences, 1, 515-526.

Pylyshyn, Z. W. (1978). When is attribution of be-liefs justified? Behavioral and Brain Sci-ences, 4, 592-593.

Rosen, C. (1974). The effects of sociodramatic playon problem-solving behavior among cultur-ally disadvantaged preschool children. ChildDevelopment, 45, 920-927.

Rubin, K. H. (1976). The relationship of socialplay preference to role taking skills in pre-school children. Psychological Reports, 39,823-826.

Rubin, K. H., Fein, G. G., & Vandenberg, B.(1983). Play. In E. M. Hetherington (Ed.),P. H. Mussen (Series Ed.), Handbook of childpsychology: Vol 4. Socialization, personality,and social development (pp. 693-774). NewYork: Wiley.

Rubin, K. H., & Maioni, T. L. (1975). Play prefer-ence and its relationship to egoeentrism, pop-

Page 24: Pretend Play Skills and the Child's Theory of Mind

ularity, and classification skills in preschool-ers. Merrill-Palmer Quarterly, 21, 171-179.

Rubin, K. H., & Pepler, D. J. (1980). The relation-ship of child's play to social-cognitive growthand development. In H. G. Foot, A. J. Chap-man, & J. R. Smith (Eds.), Friendship and so-cial relations in children (pp. 209-233). NewYork: Wiley.

Rubin, K. H., & Schneider, F. W. (1973). The rela-tionship between moral judgment, egoeen-trism, and altruistic behavior. Child Develop-ment, 44, 661-665.

Russell, J., Mauthner, N., Sharpe, S., & Tidswell,T. (1991). The "windows" task as a measureof strategic deception in preschoolers and au-tistic subjects. British Journal of Develop-mental Psychology, 9, 331-350.

Saltz, E., Dixon, D., & Johnson, J. (1977). Trainingdisadvantaged preschoolers on various fan-tasy activities: Effects on cognitive functionand impulse control. Child Development, 48,367-380.

Scarlett, W. G., & Wolf, D. (1979). New directionsfor child development: When it's only make-believe: The construction of a boundary be-tween fantasy and reality. In E. Winner & H.Gardner (Eds.), New directions for child de-velopm.ent: Fact,fiction, and fantasy in child-hood (Vol. 6, pp. 29-40). San Francisco: Jos-sey-Bass.

Searle, J. R. (1983). Intentionality. Cambridge:Cambridge University Press.

Shatz, M., Wellman, H., & Silber, S. (1983). Theacquisition of mental verbs: A systematic in-vestigation of the first reference to mentalstates. Cognition, 14, 301-321.

Siegal, M., & Beattie, K. (1991). Where to lookfirst for children's knowledge of false beliefs.Cognition, 31, 1-12.

Siegler, R. S. (1991). Children's thinking. Engle-wood Cliffs, NJ: Prentice-Hall.

Singer, J. L. (1973). The child's world of make-believe. New York: Academic Press.

Smilansky, S. (1968). The effects of sociodramaticplay on disadvantaged children: Preschoolchildren. New York: Wiley.

Smith, P. K., & Sydall, S. (1978). Play and non-playtutoring in preschool children: Is it play ortutoring that matters? British Journal of Edu-cational Psychology, 48, 315-325.

Spivak, G., & Shure, M. B. (1974). Social adjust-ment of young children. San Francisco: Jos-sey-Bass.

Taylor, M. (1988). The development of children'sunderstanding of the seeing-knowing distinc-tion. In J. W. Astington, P. L. Harris, & D. R.Olson (Eds.), Developing theories of mind(pp. 207-225). Cambridge: Cambridge Uni-versity Press.

Taylor, M., & Hort, B. (1990). Can children be

Angeline S. Lillard 371trained in making the distinction between ap-pearance and reality? Cognitive Develop-ment, 5, 89-99.

Ungerer, J., Zelazo, P. R., Kearsley, R. B., &O'Leary, K. (1981). Developmental changesin the representation of objects in symbolicplay from 18 to 34 months of age. Child Devel-opment, 52, 186-195.

Vygotsky, L. S. (1978). Mind in society. Cam-bridge, MA: Harvard University Press.

Walton, K. L. (1990). Mimesis as make-believe.Cambridge, MA: Harvard University Press.

Watson, M. W., & Fischer, K. W. (1977). A devel-opmental sequence of agent use in late in-fancy. Child Development, 48, 828-836.

Watson, M. W., & Fischer, K. W. (1980). Develop-ment of social roles in elicited and spontane-ous behaviour during the preschool years. De-velopmental Psychology, 16, 483-494.

Wellman, H. M. (in press). Early understanding ofmind. The normal case. In S. Baron-Cohen,H. Tager-Flusberg, & D. Cohen (Eds.), Un-derstanding other m.inds: Perspectives fromautism. Oxford: Oxford University Press.

Wellman, H. M., & Wooley, J. D. (1990). From de-sires to beliefs: The early development of ev-eryday psychology. Cognition, 35,245-275.

Whiten, A. (1991). Natural theories of mind. Ox-ford: Basil Blackweli.

Whiten, A., & Byrne, R. W. (1991). The emergenceof metarepresentation in human ontogenyand primate phylogeny. In A. Whiten (Ed.),Natural theories of mind (pp. 267-283). Ox-ford: Basil Blackweli.

Whiten, A., & Pemer, J. (1991). Fundamental is-sues in the multidisciplinary study of mind-reading. In A. Whiten (Ed.), Natural theoriesof mind (pp. 1-18). Oxford: Basil Blackweli.

Wimmer, H., & Perner, J. (1983). Beliefs aboutbeliefs: Representation and constrainingfunction of wrong beliefs in young children'sunderstanding of deception. Cognition, 13,103-128.

Wolf, D. (1982). Understanding others: A longitu-dinal case study of the concept of indepen-dent agency. In G. E. Forman (Ed.), Actionand thought (pp. 297-327). New York: Aca-demic Press.

Wolf, D. P., Rygh, J., & Altshuler, J. (1984).Agency and experience: Actions and states inplay narratives. In I. Bretherton (Ed.), Sym-bolic play. London: Academic Press.

Woolley, J. D., & Wellman, H. M. (1990). Youngchildren's understanding of realities, non-realities, and appearances. Child Develop-ment, 61, 946-961.

Zaitchik, D. (1990). When representations conflictwith reality: The preschooler's problem withfalse beliefs and "false" photographs. Cogni-tion, 35, 41-68.

Page 25: Pretend Play Skills and the Child's Theory of Mind