The Impact of Pretend Play on Childrens Development

The Impact of Pretend Play on Childrens Development: A Review of the Evidence

Angeline S. Lillard, Matthew D. Lerner, Emily J. Hopkins, Rebecca A. Dore, Eric D. Smith, and Carolyn M. Palmquist

University of Virginia

Pretend play has been claimed to be crucial to children’s healthy development. Here we examine evidence for this position versus 2 alternatives: Pretend play is 1 of many routes to positive developments (equifinality), and pretend play is an epiphenomenon of other factors that drive development. Evidence from several domains is considered. For language, narrative, and emotion regulation, the research conducted to date is consistent with all 3 positions but insufficient to draw conclusions. For executive function and social skills, existing research leans against the crucial causal position but is insufficient to differentiate the other 2. For reasoning, equifinality is definitely supported, ruling out a crucially causal position but still leaving open the possibility that pretend play is epiphenomenal. For problem solving, there is no compelling evidence that pretend play helps or is even a correlate. For creativity, intelligence, conservation, and theory of mind, inconsistent correlational results from sound studies and nonreplication with masked experimenters are problematic for a causal position, and some good studies favor an epiphenomenon position in which child, adult, and environment characteristics that go along with play are the true causal agents. We end by considering epiphenomenalism more deeply and discussing implications for preschool settings and further research in this domain. Our take-away message is that existing evidence does not support strong causal claims about the unique importance of pretend play for development and that much more and better research is essential for clarifying its possible role.

Keywords: pretend play, preschool, cognitive development, social development

How does pretend play affect children’s development? Claims for its positive impact are resounding. The National Association for the Education of Young Children, the major preschool accred- iting body in the United States, stated in its recent position paper, “high-level dramatic play produces documented cognitive, social, and emotional benefits” (Copple & Bredekamp, 2009, p. 15). An article aimed at parents states that play “is a significant contributor to the child’s cognitive, physical, emotional, and social develop- ment”1 (Hurwitz, 2002, p. 101). Some even maintain that pretend play’s impact is unique: A clinical report on the subject for the American Association of Pediatrics opened, “play is essential to

development . . . so important . . . that it has been recognized by the United Nations High Commission for Human Rights as a right of every child”2 (Ginsburg, the Committee on Communications, & the Committee on Psychosocial Aspects of Child and Family Health, 2007, p. 182). P. K. Smith (2010, pp. 28–29) gave many other examples of the important and wide-reaching benefits attrib- uted to pretend play (see also Bredekamp, 2004; Brown & Vaughan, 2009; Elkind, 2007; Tullis, 2011). American parents concur (Roopnarine, 2011), and child development experts en- dorse pretend play even more strongly (K. R. Fisher, Hirsh-Pasek, Golinkoff, & Gryfe, 2008). Entire preschool curricula are designed around pretend play because of the “unequivocal evidence for [its] critical importance” to children’s development (Zigler & Bishop- Josef, 2004, p. 9). Master teachers’ discussions of why pretend play is so vital for children are convincing (e.g., Paley, 2005), and we agree: When we watch children in pretend play, it seems to us like a very important activity.

However, many non-Anglo cultures do not share this view of pretend play’s importance, and perhaps as a result, children grow- ing up in those cultures pretend much less (Gaskins & Goncu, 1992; Lancy, 2007). A recent survey found that in only five of 16

1 Elsewhere it is clear that pretend play is intended; for example, “in play, everything and anything can happen: a sheet over a table becomes a castle” (Hurwitz, 2002, p. 101).

2 Pretend play is Ginsburg et al.’s (2007) focus; for example, “play allows children to create and explore a world they can master, conquering their fears while practicing adult roles” (p. 183).

This article was published Online First August 20, 2012. Angeline S. Lillard, Matthew D. Lerner, Emily J. Hopkins, Rebecca A.

Dore, Eric D. Smith, and Carolyn M. Palmquist, Department of Psychol- ogy, University of Virginia.

Matthew D. Lerner is now at the Department of Psychiatry and Behav- ioral Neuroscience, University of Chicago Medical Center, Chicago, IL.

Preparation of this article was supported by National Science Founda- tion (NSF) Grant 1024293, a Brady Education Foundation grant, and a University of Virginia Sesqui award to Angeline S. Lillard; the American Psychological Foundation’s Elizabeth Munsterberg Koppitz Fellowship, the James H. and Elizabeth W. Wright Endowed Fellowship from the Jefferson Scholars Foundation, and grants from the American Psycholog- ical Association and the Association for Psychological Science to Matthew D. Lerner; and an NSF graduate fellowship to Eric D. Smith.

Correspondence concerning this article should be addressed to Angeline S. Lillard, Department of Psychology, University of Virginia, P.O. Box 400400, Charlottesville, VA 22904. E-mail: Lillard@virginia.edu

Psychological Bulletin © 2012 American Psychological Association 2013, Vol. 139, No. 1, 1–34 0033-2909/12/$12.00 DOI: 10.1037/a0029321

1

countries surveyed (the United States, the United Kingdom, Ire- land, Portugal, and Argentina) do the majority of mothers say their children (ages 1–12) often participate in imaginative play (D. G. Singer, Singer, D’Agostino, & Delong, 2009). Even within the United States, there is individual variation in how much children pretend (Fein, 1981). Should infrequent pretenders be pretending more? Would doing so help their development? Is the evidence strong enough to warrant designing curricula around pretend play and deriding preschools that do not encourage it? Here we examine evidence cited in support of pretend play’s importance to deter- mine whether there is a convincing case. The evidence concerns six domains of development, chosen because they are frequently claimed to be assisted by pretend play (e.g., see Ashiabi, 2007; Bergen, 2002; Ginsburg et al., 2007; Hirsh-Pasek, Golinkoff, Berk, & Singer, 2009; Isenberg & Quisenberry, 1988; Lillard, 2001a) and because we found at least a half dozen studies concerning each: nonsocial cognitive aptitudes (with five subdomains), social cognition, social skills, language, narrative skills, and self- regulation (with the subdomains of executive function and emotion regulation).

First we define pretend play and review three theoretical posi- tions on whether and how it affects development generally. Next, we describe patterns of evidence that would support each position, then review and discuss the evidence domain by domain. Finally we consider one position more deeply, address the implications of our review for educational settings, and make suggestions for future research on this topic.

Defining Pretend Play

A preliminary issue is to define pretend play. Play itself is a notoriously difficult concept to pin down (Burghardt, 2011). For our purposes the four criteria of Krasnor and Pepler (1980) will define play: flexibility, positive affect, nonliterality, and intrinsic motivation (cf. Sutton-Smith & Kelly-Byrne, 1984). Flexibility denotes that play behaviors vary from real ones in form (they might be exaggerated, or truncated) and/or content (one might play at eating with a stick instead of a spoon). Positive affect touches on the idea that people look like they are having fun when they play. Nonliterality refers to the fact that, in play, behaviors lack their usual meaning while paradoxically retaining it; Bateson (1972) famously pointed out that, “the playful nip denotes the bite, but it does not denote what would be denoted by the bite” (p. 317). Intrinsic motivation suggests voluntariness: One engages in the activity by choice for its own sake.

Pretend play activities are the subset of play activities charac- terized by an “as-if” stance (Garvey, 1990). Beyond being simply nonliteral, in pretend play a “pretense” is layered over reality (Austin, 1979); specifically, a pretender knowingly and intention- ally projects some mentally represented alternative on to the pres- ent situation in the spirit of play (Lillard, 1993). Sometimes pretend play is social: A group of children share an alternative reality that they project, perhaps acting like they are different people in another place and time. Other times pretending is a solo activity. Pretend play can involve projecting imaginary objects and properties, or using one object as if it were another (Leslie, 1987). It is most prominent in early childhood, with ages 3 to 5 being declared its “high season” (D. G. Singer & Singer, 1992), although

it does continue into middle childhood and beyond (E. D. Smith & Lillard, in press).

There are several other forms of play besides pretend (see Pellegrini, 2009; P. K. Smith, 2010); in particular there is a small but important literature on physical play (such as hopscotch and rough-and-tumble play), which has been well reviewed elsewhere (Pellegrini & Bohn, 2005; Pellis & Pellis, 2009). Such forms of play assist sustained attention in conventional school situations (Pellegrini & Bohn, 2005); they also (in the case of rough-and- tumble play fighting) assist emotion regulation, social coordina- tion, and normal sexual behavior, at least in some rodents and nonhuman primates (Pellis & Pellis, 2009). Pretend play can overlap with these and other types of play. For example, physical play overlaps with pretend play when children pretend to be fighting warriors. Object play overlaps with pretending when a child animates those objects.

The literature is not always clear as to when pretend play specifically, versus play more generally, or some other specific type of play is at issue; this can be seen in the quotes with which we opened (but see footnotes 1–2), and probably arises because young children’s play is so often infused with pretense. Our aim here is not to resolve this ambiguity but rather to consider studies used to support claims that play is crucial to positive develop- ments, excluding the physical play literature just mentioned, and retaining focus on pretend play as much as possible. Our main exception to this is in two subdomains of nonsocial cognitive aptitudes, creativity and problem solving, because for those skills several studies concerning manipulative play with small objects (which might or might not involve pretending) are often cited as showing play’s cognitive benefits. When a study contrasted pre- tend play with some other form of play (like construction play, as in building with blocks) we focused on the pretend condition. Many studies strain the voluntary aspect of play in that children were told to play or were instructed in acting out a story, but because those studies have been cited as showing play’s benefits, they are reviewed here.

To locate studies, the first author began with references sup- porting claims of play’s benefits in articles like those in the opening paragraph, then back referenced those studies in a snow- ball fashion. Through this process she arrived at the six main topics and six subdomains of nonsocial cognitive skills; the subdomain of mathematics was subsequently eliminated due to an insufficient number of studies. From there a search engine (Google Scholar) was employed, searching by keywords (“social skills, pretend play”) and the “referenced by” and “related articles” features, as well as continuing to back reference from within articles. To avoid an unwieldy review, we passed over studies of atypical populations or cultural variation, and largely confined ourselves to published or in press peer-reviewed studies.3

Theoretical Background

P. K. Smith (2010) laid out three theoretically possible relation- ships between pretend play and positive developmental outcomes.

3 Exceptions were made for particularly important unpublished studies reported in other published work by the author or thesis advisor, and for one article under review.

2 LILLARD ET AL.

The first is that pretend play is crucial to optimal development. The second, which Smith supported, is equifinality: Pretending helps some developments, but it is only one possible route. Other activities can work as well or better. The third possibility is that pretending is an epiphenomenon or byproduct of some other selected-for capability, but in and of itself makes no contribution to development; rather, the other activity or condition to which it is sometimes attached is the actual contributor. Two major devel- opmental theorists, Vygotsky and Piaget, align with the first and third views, respectively.

For Vygotsky (1978), pretense has a crucial developmental role, because it is the activity by which children learn to separate referent from object. In play, children first understand that actions (and objects on which one might act) can be separated from reality and can be based on the meaning of a given situation rather than on the physical properties of objects (Vygotsky, 1967). In this way, for example, a banana could become a phone in a pretense situa- tion and the child could act on it as if it were a phone, inhibiting how he or she would act on it if it were a banana. The upshot of this is that children develop abstract thought through pretend play (Vygotsky, 1967). In addition, because reality must be inhibited, children also develop inhibitory control through pretending (Bodrova & Leong, 1996). Because of these features, “in play, it is as though [the child] were a head taller than himself” (Vygotsky, 1978, p. 102); play takes a child to the upper end of his or her “zone of proximal development” (p. 86).

In contrast, for Piaget (1962), pretend play is more an index than a promoter of development.4 Its appearance around 18 months indicates the development of the semiotic function, which also allows for deferred imitation and language. The semiotic function separates an idea from its referent, a memory from its context, and an object from its label, allowing one to entertain and elaborate on mental content that is separate from the physical, present reality.

Here we consider which of these views is best supported by the evidence. Each view is compatible with a particular pattern of evidence from correlational and experimental (short-term and training) studies, shown in Table 1 (cf. P. K. Smith, 2010, Table 9.2, p. 187). These patterns assume methodologically sound stud- ies including sufficient duration and sample sizes. First, if pretend play does crucially cause positive developments (Vygotsky’s po- sition), then strong positive correlations between pretend play and those developments should consistently be found; if a child pre- tends more, whether naturally (in a correlational study) or due to an experimental manipulation, the development should increase. If pretend play causes creativity, then children who pretend more will generally be more creative. Additional predictors, like intelligence, are also possible, but if pretending is truly the important causal factor, the unique and important relationship to pretend play should hold even when those other predictors are partialled out.

Conversely, if Smith’s equifinality position is correct, then one would generally expect positive relationships between play and the outcome but also correlations with other predictors that engender the outcome. For example, if social pretend play develops theory of mind and so does adult talk about mental states, then correla- tions should be found for both variables. Interventions increasing mental state talk and pretend play might have an additive effect when combined, which could lead to even larger effects (but not if there was a ceiling on development for that age). There could be cases when although equifinality is the best model, pretend play

fails to evince a significant effect. This might occur, for example, when there is substantial multicollinearity, or when an alternate predictor’s effect is much larger, masking pretend play’s effect. Thus equifinality does not insist on 100% consistent results, but it generally expects them.

The third, or epiphenomenon, position is supported if pretend play coincides with some other causal circumstance; in such cases pretend play might mistakenly be considered causal. For example, if social pretend play is related to theory of mind because adults who engage in a lot of mental state talk also happen to encourage pretend play, then perhaps what is actually leading to the increased theory of mind is not the pretend play, but the mental state talk; social pretend play is secondary or epiphenomenal to the mental state talk–theory of mind relationship. If pretend play is an epiphe- nomenon then one might find inconsistent correlations with outcomes (because pretend play does not always go along with the real predic- tors) but consistent correlations between real predictors and outcomes. Because different studies measure different possible predictors, the true predictors might not always be evident. Here we evaluate the patterns of evidence with an eye to each of these positions. Before beginning to do so, it is useful to note some recurring problems in this literature (see also Cheyne, 1982).

Common Methodological Problems

Several problems recur in the literature on whether pretend play helps development. Sometimes these problems occur because the

4 Others have claimed Piaget gives pretense a stronger role in develop- ment; for example, Singer and colleagues, citing Piaget (1962), claimed he “concluded that play was a vital component to children’s normal intellec- tual and social development” (D. G. Singer et al., 2009, p. 285). In our reading the closest Piaget (1962) comes to this is when he says it is undoubtedly “a preparation for imaginative aptitudes” (p. 155), where imagination (as in pretend play) is the assimilative pole of thought (in contrast to accommodation), and creative imagination arises only when one integrates the two. This is essentially the position taken by Harris (2000) and D. G. Singer and Singer (1992): Pretending assists imagination. But whereas for these modern authors this is a reason to centralize pretend play, our reading of Piaget’s text on play suggests that this role in imaginative development was a minor concern; pretend play was primarily an offshoot of the symbolic function. Perhaps confusion has arisen because elsewhere Piaget assigns manipulative activities (Piaget, 1929) and peer interaction (Piaget, 1932) as important to development, and pretend play often in- volves these other activities. But in considering manipulative activity, Piaget refers more to what is now referred to as embodied cognition (“manual work is essential to the child’s mental development”; Piaget, 1929, p. 383), and regarding peer involvement Piaget’s own focus on pretend play was particularly as a solitary activity. Piaget (1962) did think pretending served an egoistic function in that it allowed the child to fulfill wishes that he or she could not fulfill in reality. A child who wants to be a mother can simply pretend to be one. But Piaget was concerned with cognitive development, not personality development, and pretend play was pre-operational because it indicated what the child lacked. For Piaget, children outgrow pretending as they develop the ability to accommodate reality. Here he followed some major figures of his time in child psychol- ogy, such as Freud (1955, as discussed by Harris, 2000) and Montessori (1989). Aligning with our own reading, Sutton-Smith (1966) colorfully summarized Piaget’s view of pretending as “a buttress to an inadequate intelligence” (p. 108). For further discussion, see P. K. Smith (2010, pp. 31–37).

3PLAY DEVELOPMENT

research was conducted when experimental standards were not as high as they are today, pointing to the need to modernize the evidence base. In more recent studies, perhaps scholars did not apply more rigor because of a deep belief in the power of play (Elkind, 2007), what P. K. Smith (1988) dubbed the “play ethos” and Sutton-Smith (1995, p. 279) the rhetoric of “play as progress.” Here we strive to overcome the tendency to favor pretend play by holding all studies to a high standard.

One common problem in discussions of the impact of play on development is that correlational findings are often discussed as if they were causal. When children who play more do better on some other measure, of course it does not mean that the play definitely caused the outcome. Positive correlations between pretend play and a development are only a necessary precondition to pretend play being causal. Likewise prominent authors have described elaborate pretend worlds they constructed as children, and one might see the earlier behavior as causing their subsequent literary genius (Root-Bernstein, in press), but it is as plausible that their creativity led to conjuring up elaborate imaginary worlds at both time points.

A second recurring problem is failure to replicate. For example, one study shows increases in empathy associated with pretense training (Saltz & Johnson, 1974) and another does not (Iannotti, 1978), and typically only the positive finding is cited. If other key experimental factors are essentially equal, either the reported pos- itive result reflects a Type I error or the failure reflects a Type II error. Inconsistent findings in correlational studies contradict the causal view but would be expected with either the equifinality or epiphenomenalism. For equifinality, nonreplications would occur when the alternate route was stronger in one study, and including it masked the effect of pretend play; for the epiphenomenon position, nonreplications would occur because the underlying cause sometimes accompanied pretend play and sometimes did not. In the literature extolling play’s benefits, failures to replicate are often ignored.

A third problem concerns experimenter bias. Every under- graduate research methods course should impart the importance

of experimenters being “masked” insofar as possible: that is, unaware of (a) the hypotheses being tested and (b) participants’ conditions. Yet cognitive development research rarely uses masked experimenters. This might usually be fine: Child de- velopment researchers and the kinds of tests they give might not be vulnerable to experimenter bias under the usual circum- stances. For example, we know of no research suggesting that false belief or conservation errors occur at certain ages only when experimenters are unmasked. However, for research on the benefits of pretend play there are several cases where results obtained with knowledgeable experimenters went away when masked ones were employed (Christie, 1983; Guthrie & Hud- son, 1979; Pepler & Ross, 1981; Simon & Smith, 1983, 1985; P. K. Smith, Simon, & Emberton, 1985; P. K. Smith & Whit- ney, 1987). Nonreplications with masked experimenters make a strong case for being cautious about pretend play results ob- tained with knowledgeable experimenters.

Besides correlational data, nonreplication, and unmasked ex- perimenters, other recurrent problems are very small sample sizes, nonrandom assignment, confounding implementer with intervention (particularly concerning when there is only one implementer per condition and interventions last for several months), control conditions that differ beyond pretend play, confounding content with pretend play, and unsound statistical practices like using subsets of data and one-tailed tests without prior rationale.

Methodological problems are so prevalent in this literature that meta-analysis is precluded. E. P. Fisher (1992) did a meta-analysis of the impact of play (generally) on development, despite aware- ness of these limitations (see “Shortcomings of the Studies,” pp. 164–168), but he also did not have a consistent even-handed approach to which statistics he included, and further, he used some wrong statistics that inflated his result. As a particularly egregious illustration of this, from Christie (1983) he used the statistic pertaining to a variable named variable (F � 257.67), reflecting the overall sample scores on five variables, when the far smaller Variable � Time � Condition statistic (F � 0.49) is what should

Table 1 Three Possible Relationships Between Pretend Play and Development

Expected pattern of results Crucial (Vygotsky) Equifinality (P. K. Smith) Epiphenomenon (Piaget)

Correlational studies Strong, unique, and consistent. Generally consistent but not unique. Including other causal variables could mask pretend play’s effects, so correlations could be inconsistent.

Inconsistent, but consistent with other variables that are causal. For example, if presence of certain toys increases pretending in children who are more creative, but other objects have no impact, then correlations between pretending and creativity will be seen only in environments with those toys.

Experimental (short-term and training) studies

Strong, unique, and consistent. Strong and consistent but not unique, so other conditions could also affect development. For example, skills training and pretend play training could both increase the development.

Effects found only if the crucial underlying factor(s) is (are) influenced by the intervention. For example, suppose pretend play only assists development when intensive adult interaction is part of that training; when children pretend but there is no intensive adult interaction, the pretend play does not increase the development; in addition, another condition might show that intensive adult interaction alone increases the development, even in the absence of play.

4 LILLARD ET AL.

have been used. Careful reading reveals many more problems, yet this article is often cited (126 times, Google Scholar, as of May 28, 2012) as evidence that play helps development (e.g., Bergen, 2009; Ginsburg et al., 2007; Wyver & Spence, 1999).

Because so many studies in this area are methodologically unsound, the current literature base is best suited to a descriptive review, on which we now embark. In each section, we begin with theoretical and construct issues, then review studies. A series of 10 tables compiles the studies pertinent to each domain or subdomain of development.

After reviewing the studies, each section concludes by dis- cussing the evidence with respect to the three views (summa- rized in Table 12). In these discussions, we sometimes rely on the absence of evidence to support a position. We do this with caution, since one can never prove that a relationship does not exist (Altman & Bland, 1995). However, inconsistent correla- tion patterns across studies with similar samples and methods and reliable coders are against a causal view. Likewise, when sound experimental methods yield null effects or even effects showing play is less positive than the alternative, this is also relevant. Finally, doubt is also cast on a causal view when masking experimenters or equalizing other aspects of interven- tions nullifies previous findings.

A final note before treading into the evidence concerns the “straw person” element of the crucial causal view. When put on the stand, perhaps few would endorse the position that pretend play is crucial (in the sense of essential or vital) for various aspects of development. Yet the quotes with which we opened and additional references throughout this review show that this stance is taken in the literature, so we consider it here.

Nonsocial Cognitive Aptitudes

As was seen in our opening paragraph, many scholars have asserted that pretend play produces cognitive benefits. One way pretend play could help cognition is by predisposing children to a generally playful attitude (Dansky & Silverman, 1973) that could lead to production of unusual ideas, creative problem solving (Vandenberg, 1980), and then to other cognitive aptitudes. This view is compatible with Fredrickson’s (2001) broaden-and-build theory of positive emotions, with play eliciting joy, which in turn leads to a broadening of individuals’ thought–action repertoires. Vygotsky’s ideas on symbolic and abstract thought, just reviewed, also suggest how pretend play could assist cognitive abilities. Here we discuss evidence that pretend play assists development in five subdomains: creativity, problem solving, intelligence, conserva- tion, and reasoning.

Creativity

Although creativity has been operationalized in a number of ways, in the studies on play it has typically been defined as the ability to produce original content relevant to a particular task (Wallach & Kogan, 1965). The most commonly used measure of creativity in this literature is the alternate uses task (R. C. Wilson, Guilford, & Christensen, 1953), in which participants give possible uses for common objects, like a paper towel or a paperclip (Dan- sky, 1980a). Task responses are typically coded for fluency (num- ber of uses named) and originality (number of uses not given by

any other participant). Another common task is the Torrance Thinking Creatively in Action and Movement Test or TCAM (Torrance, 1981), which includes several subtests, including hav- ing children move like trees in the wind, and also alternate uses.

Correlational studies. Several studies have addressed the claim that pretending makes children more creative (Ginsburg et al., 2007; J. L. Singer, 1973) by looking for correlations between naturalistic play and creativity, since they should exist if more frequent pretenders have become more creative via their pretend activity. Of course correlations are not evidence of causation, but if causation exists, correlations should be consistently found.

Naturalistic classroom play has been categorized differently in different studies, but a combination of Smilansky (1968) and Parten’s (1932) schemes has been used most often (Rubin, 2001). Smilansky’s scheme (derived from Piaget) divides play into cog- nitive categories (functional play when a child repeats motor actions on objects, construction play when a child builds things, dramatic or symbolic play when the child substitutes an imagined world for reality, and games-with-rules like Checkers). Parten’s scheme is focused on the social aspect of children’s play: coop- erative when children are actively interacting in a common group endeavor; associative when they interact but not toward a single common endeavor; parallel when they play similarly but side by side, with little interaction; solitary independent when they play alone at their own games; onlooker when they watch others play; and unoccupied.

We found eight studies correlating pretend play and creativity (see Table 2). Seven concerned preschoolers. Typically children’s play in preschool was coded for 1–5 min per day for a period of 20 days or more, and then alternate uses with two to four objects (or in some cases the TCAM or another test) was administered. Results were inconsistent.

Johnson (1976) found that, controlling for IQ, amount of social but not solitary fantasy play was related to fluency. This would suggest that something about the social element, rather than pre- tending in and of itself, was related to creativity. However, John- son (1978) later failed to replicate this finding in a very similar study, showing no relationship between pretend play (social or solo) and alternate uses. A different study with the same age range also found no relationship between creativity and social pretend play (L. Dunn & Herwig, 1992) but found a negative relationship between originality of responses and solo pretend play that disap- peared when IQ was partialled out. Pellegrini and Gustafson (2005) also did not find children’s frequency of pretend play to be related to creativity (r � –.18) yet did find frequency of construc- tion play was related. On the other hand, a different study found all pretend play was significantly related to creativity in a sample of 15 high-IQ preschoolers (Moran, Sawyers, Fu, & Milgram, 1984). Wyver and Spence (1999) found that particularly fantastical pre- tend play was related to semantic creativity (naming all the objects one could think of) but not figural creativity (making objects from shapes); two other categories of pretend play that they coded were not related to either. The remaining preschool study, Lloyd and Howe (2003), was not useful regarding pretend play because they combined it with functional play, which was twice as common as pretend play, but it is worth noting that in contrast to Pellegrini and Gustafson (2005) they found construction play (which was coded separately) was not related to creativity. In the eighth and final correlational study, looking at somewhat older children, Russ and

5PLAY DEVELOPMENT

Grossman-McKee (1990) examined play with Russ’s Affect in Play Scale (APS), which uses 10 min of play with puppets and blocks coding generativity. Alternate uses also taps generativity, so it is not surprising that positive correlations were found, both concurrently and over time (Russ, Robins, & Christiano, 1999).

The eight correlational studies show an inconsistent pattern of relationships that does not support the causal model. The other models seek alternate routes or third variables that could underlie relationships when they are found. With this particular set of studies, no obvious other variable emerges. IQ is sometimes re- lated to creativity (Johnson, 1976; Wyver & Spence, 1999) but sometimes not (L. Dunn & Herwig, 1992; Lloyd & Howe, 2003; Moran et al., 1984; Wyver & Spence, 1999); Johnson (1978) did not test IQ. Johnson worked with low-socioeconomic status chil- dren, whereas others involved middle-class ones, but this cannot explain the discrepant results across Johnson’s own studies. An- other third variable that fits either view is environment (i.e., types of toys supplied), which can drive the types of play children engage in (McLoyd, 1983). Because environment was not mea- sured, we cannot evaluate if this could underlie the inconsistencies. Regardless, distinguishing the second and third views requires experimental studies.

Experimental studies. Four experimental studies found higher associative fluency when children first played with objects

for which they later named uses (Dansky, 1980b; Dansky & Silverman, 1973, 1975; Li, 1978). This fit with Sutton-Smith’s (1968) quasi-experimental study in which boys came up with more alternate uses for traditional boy toys than did girls (although for girl toys, they were equal). To check whether experience with the objects, rather than playing with them per se, was important, Dansky and Silverman (1973) included an imitation group, in which children gained experience by imitating the experimenter handling the objects. The imitators gave no more uses than the controls, suggesting that experience with the objects did not ex- plain the first results (see also Hughes, 1981, as cited in Hutt, Tyler, Hutt, & Christopherson, 1989).

Pretend play’s effect on creativity could be limited to the objects at hand. To test this, Dansky and Silverman (1975) used a different set of objects in the test phase, and here again the play group produced more uses, suggesting play’s hypothesized effect on creativity generalizes.

In the Dansky studies just described, although the theoretical rationale concerned pretend play, it is unclear whether children pretended with the objects or just manipulated them. Li (1978) tested whether pretend play might improve creativity above and beyond play generally. In the pretend play condition, the experi- menter told a short fantasy story, then showed children the stim- ulus objects and said, “Let’s make-believe or imagine that these

Table 2 Studies Examining the Effect of Play on Creativity

Type Citation � � � Masked Int Masked Exp Notes

C Johnson (1976) SPP Solo C Johnson (1978) SPP/Solo C L. Dunn & Herwig (1992) SPP Solo Solo disappears when partial IQ C Pellegrini & Gustafson (2005) Const. PP C Moran et al. (1984) PP Holds when partial IQ C/T Wyver & Spence (1999) PP PP/SPP No No C: One type of pretense related to one

task; T: effects work both ways; fantastical themes; finding for one of two measures for associative PP

C Lloyd & Howe (2003) Const. C Russ & Grossman-McKee (1990) Solo Creativity/play measures are redundant E Dansky & Silverman (1973) Play No No E Dansky & Silverman (1975) Play No No Extended to different objects E Li (1978) PP No No Two of four objects only E Dansky (1980b) Play No No Pretenders only E P. K. Smith & Whitney (1987) Play No Yes E Pepler & Ross (1981) Play No Yes See footnote 5 in text E Pellegrini & Greene (1980) Play No No Structured questioning better E Pellegrini (1981) Play No No Structured questioning better E Russ & Kaugars (2001) PP No No Scorer but not administrator masked E Howard-Jones et al. (2002) Play Irrel. Yes Control task very dull T Dansky (1980a) SPP Yes Yes Adult contact not controlled; “natural”

pretenders only T Feitelson & Ross (1973) PP PP No No Only on one measure; adult contact

not controlled T Christie (1983) PP � Skills No Yes Adult contact controlled T P. K. Smith & Syddall (1978) PP � Skills Yes No Adult contact controlled T P. K. Smith et al. (1981) PP � Skills No Yes Adult contact controlled T Moore & Russ (2008) Solo No Yes Adult contact controlled

Note. Type of study: C � correlational; E � experimental; T � training. Type of play: Play � pretense status unspecified; Const. � construction play (e.g., blocks); Solo � pretend play alone; SPP � social pretend play; PP � pretend play (social unspecified). �: positive relationship to play; �: no correlation or play � nonplay condition; � � negative relationship to play. Masking: Intervention (Int) or posttest experimenters (Exp). If masking status was not specified, we assume experimenters were not masked, since that is the unmarked case. Masking for correlational studies is omitted because it is rarely mentioned, even when it is likely (because play observations occurred several years earlier than testing, for example). Irrel. � irrelevant.

6 LILLARD ET AL.

objects could become anything you want them to be. Play with all of these things” (p. 33). Free play, imitation, and control condi- tions were similar to those used in Dansky’s studies. After 10 min, the same experimenter administered the alternative uses test with the three objects used in the intervention and a new fourth object. Significant differences were found for one of three old objects (a paperclip), for the make-believe group and the free play group, and for the new object (a screwdriver) only for the make-believe group versus the control. Overall, Li’s children came up with far fewer uses than children had in the previous studies, perhaps reflecting population variance.

Taking a different tack, Dansky (1980b) examined whether children who naturally engage in more pretend play would benefit more from a play intervention. Children were classified as pre- tenders if they engaged in pretense more than 25% of the time, or nonpretenders if they engaged in it less than 5%. They were then assigned to free play or control conditions. Suggesting the class- room classifications had validity, in the free play condition 88% of the pretenders but only 6% of the nonpretenders pretended with the objects. The alternate uses test was given with a different set of objects, and only the pretenders in the free play condition had higher fluency. Dansky concluded that play induces creativity only for those who are predisposed to pretend.

These experimental studies suggest that play might have a causal effect on creativity, at least for children who frequently pretend (see also Sutton-Smith, 1967). However, in these studies the experimenters administering the creativity test knew which condition each child was in, and perhaps their expectations influ- enced children’s responses. With alternate uses, the experimenter elicits answers until they think a child has run out of possibilities. More coaxing might inadvertently occur when children are ex- pected to produce more uses. In an attempted replication of Dan- sky and Silverman (1973), P. K. Smith and Whitney (1987) used different experimenters for the intervention phase and the posttest, with the latter masked to condition. Results showed no significant differences—in fact the control condition obtained the highest mean score. This finding is consistent with another study using a masked experimenter: The number of uses given was not signifi- cantly different for children in a play condition (Pepler & Ross, 1981, Study 2).5 A less direct test of possible experimenter effects occurs when a different hypothesis is being tested. Two studies tested the hypothesis that focused questioning would elicit more uses than would playing with objects (Pellegrini, 1981; Pellegrini & Greene, 1980). The hypothesis was upheld: Focused questioning led to more uses, and free play was not different from the control. Another way to check for experimenter biasing is to videotape task delivery, which could encourage standardization; Russ and Kaugars (2001) did so, and children who engaged in pretend play generated no more uses than children who did puzzles. All this suggests that the alternate uses task might be particularly vulner- able to experimenter bias.

A solution is to use a different creativity task, like making a collage, the creativity of which is assessed by masked judges (Howard-Jones, Taylor, & Sutton, 2002). Unfortunately the study that did this did not use a good control condition. Six- and 7-year-olds played with salt dough for 25 min or copied words from the board (and if they finished early, were told to start over). Afterward, they made collages, and the play group’s were more creative. However, perhaps forced copying had a deleterious effect

on creativity, rather than play having a positive one. Further research with a more neutral control condition is needed.

Training studies. Experimental studies (as defined here) examine short-term change; perhaps play does influence creativity but requires longer incubation periods. We located seven longitu- dinal play training studies. The first three produced effects that were in the expected direction, but four others controlled for adult contact and found play training itself had no effect.

Dansky (1980a) compared 36 low-income preschoolers in so- ciodramatic play, free play, and object exploration interventions over 3 weeks, with three 30-min sessions per week. The sociodra- matic play training involved enacting pretend play themes like going on a picnic. Children in the free play group could play as they wished, and they rarely engaged in pretend play. The explo- ration training group explored and discussed objects. All experi- menters were masked. The sociodramatic play group outperformed both other groups on alternate uses.

In a similar study, Feitelson and Ross (1973) compared 24 kindergarteners in play tutoring, music tutoring, toy play without tutoring, and control groups, with ten 30-min interventions over 5 weeks. Creativity was measured with the picture completion sub- test of Thinking Creatively With Pictures (Torrance, 1966), in which children complete up to 10 pictures in a way that “no one else will think of” and give each picture a title, and Dog and Bone (Banta, 1970), in which children make up different routes from a dog to a bone. The play tutoring group’s scores increased the most on the number of unusual titles given (but not number of pictures or picture elements) and trended toward better performance on Dog and Bone.

Wyver and Spence (1999) trained 38 children in three types of play based on Parten’s social categories and play type (associative fantastical, cooperative fantastical, or cooperative constructive/ everyday sociodramatic) for 4 hr over 4 weeks, with pre- and posttests of figural and semantic creativity; they included a no- intervention control. The former two groups increased their fan- tastical pretend play, and the third did not increase in any type of pretend play. However, only the third group showed increases in both semantic and figural creativity. Associative fantastical play training improved semantic but not figural creativity, and cooper- ative fantastical play did not improve in either. These results are quite mild then for the hypothesis that pretend play increases creativity, since it did not reliably do so for the fantasy groups and the sociodramatic group also engaged in construction play. Other experiments in this study tested the reverse direction by training children in creativity and then observing play; results suggested complex and bidirectional relationships.

The results of these studies suggest that the play training can increase children’s creativity. However, it is unclear whether ex-

5 Their first experiment, using a knowledgeable experimenter, yielded some significant differences

: A Review of the Evidence

Angeline S. Lillard, Matthew D. Lerner, Emily J. Hopkins, Rebecca A. Dore, Eric D. Smith, and Carolyn M. Palmquist

University of Virginia

Pretend play has been claimed to be crucial to children’s healthy development. Here we examine evidence for this position versus 2 alternatives: Pretend play is 1 of many routes to positive developments (equifinality), and pretend play is an epiphenomenon of other factors that drive development. Evidence from several domains is considered. For language, narrative, and emotion regulation, the research conducted to date is consistent with all 3 positions but insufficient to draw conclusions. For executive function and social skills, existing research leans against the crucial causal position but is insufficient to differentiate the other 2. For reasoning, equifinality is definitely supported, ruling out a crucially causal position but still leaving open the possibility that pretend play is epiphenomenal. For problem solving, there is no compelling evidence that pretend play helps or is even a correlate. For creativity, intelligence, conservation, and theory of mind, inconsistent correlational results from sound studies and nonreplication with masked experimenters are problematic for a causal position, and some good studies favor an epiphenomenon position in which child, adult, and environment characteristics that go along with play are the true causal agents. We end by considering epiphenomenalism more deeply and discussing implications for preschool settings and further research in this domain. Our take-away message is that existing evidence does not support strong causal claims about the unique importance of pretend play for development and that much more and better research is essential for clarifying its possible role.

Keywords: pretend play, preschool, cognitive development, social development

How does pretend play affect children’s development? Claims for its positive impact are resounding. The National Association for the Education of Young Children, the major preschool accred- iting body in the United States, stated in its recent position paper, “high-level dramatic play produces documented cognitive, social, and emotional benefits” (Copple & Bredekamp, 2009, p. 15). An article aimed at parents states that play “is a significant contributor to the child’s cognitive, physical, emotional, and social develop- ment”1 (Hurwitz, 2002, p. 101). Some even maintain that pretend play’s impact is unique: A clinical report on the subject for the American Association of Pediatrics opened, “play is essential to

development . . . so important . . . that it has been recognized by the United Nations High Commission for Human Rights as a right of every child”2 (Ginsburg, the Committee on Communications, & the Committee on Psychosocial Aspects of Child and Family Health, 2007, p. 182). P. K. Smith (2010, pp. 28–29) gave many other examples of the important and wide-reaching benefits attrib- uted to pretend play (see also Bredekamp, 2004; Brown & Vaughan, 2009; Elkind, 2007; Tullis, 2011). American parents concur (Roopnarine, 2011), and child development experts en- dorse pretend play even more strongly (K. R. Fisher, Hirsh-Pasek, Golinkoff, & Gryfe, 2008). Entire preschool curricula are designed around pretend play because of the “unequivocal evidence for [its] critical importance” to children’s development (Zigler & Bishop- Josef, 2004, p. 9). Master teachers’ discussions of why pretend play is so vital for children are convincing (e.g., Paley, 2005), and we agree: When we watch children in pretend play, it seems to us like a very important activity.

However, many non-Anglo cultures do not share this view of pretend play’s importance, and perhaps as a result, children grow- ing up in those cultures pretend much less (Gaskins & Goncu, 1992; Lancy, 2007). A recent survey found that in only five of 16

1 Elsewhere it is clear that pretend play is intended; for example, “in play, everything and anything can happen: a sheet over a table becomes a castle” (Hurwitz, 2002, p. 101).

2 Pretend play is Ginsburg et al.’s (2007) focus; for example, “play allows children to create and explore a world they can master, conquering their fears while practicing adult roles” (p. 183).

This article was published Online First August 20, 2012. Angeline S. Lillard, Matthew D. Lerner, Emily J. Hopkins, Rebecca A.

Dore, Eric D. Smith, and Carolyn M. Palmquist, Department of Psychol- ogy, University of Virginia.

Matthew D. Lerner is now at the Department of Psychiatry and Behav- ioral Neuroscience, University of Chicago Medical Center, Chicago, IL.

Preparation of this article was supported by National Science Founda- tion (NSF) Grant 1024293, a Brady Education Foundation grant, and a University of Virginia Sesqui award to Angeline S. Lillard; the American Psychological Foundation’s Elizabeth Munsterberg Koppitz Fellowship, the James H. and Elizabeth W. Wright Endowed Fellowship from the Jefferson Scholars Foundation, and grants from the American Psycholog- ical Association and the Association for Psychological Science to Matthew D. Lerner; and an NSF graduate fellowship to Eric D. Smith.

Correspondence concerning this article should be addressed to Angeline S. Lillard, Department of Psychology, University of Virginia, P.O. Box 400400, Charlottesville, VA 22904. E-mail: Lillard@virginia.edu

Psychological Bulletin © 2012 American Psychological Association 2013, Vol. 139, No. 1, 1–34 0033-2909/12/$12.00 DOI: 10.1037/a0029321

1

countries surveyed (the United States, the United Kingdom, Ire- land, Portugal, and Argentina) do the majority of mothers say their children (ages 1–12) often participate in imaginative play (D. G. Singer, Singer, D’Agostino, & Delong, 2009). Even within the United States, there is individual variation in how much children pretend (Fein, 1981). Should infrequent pretenders be pretending more? Would doing so help their development? Is the evidence strong enough to warrant designing curricula around pretend play and deriding preschools that do not encourage it? Here we examine evidence cited in support of pretend play’s importance to deter- mine whether there is a convincing case. The evidence concerns six domains of development, chosen because they are frequently claimed to be assisted by pretend play (e.g., see Ashiabi, 2007; Bergen, 2002; Ginsburg et al., 2007; Hirsh-Pasek, Golinkoff, Berk, & Singer, 2009; Isenberg & Quisenberry, 1988; Lillard, 2001a) and because we found at least a half dozen studies concerning each: nonsocial cognitive aptitudes (with five subdomains), social cognition, social skills, language, narrative skills, and self- regulation (with the subdomains of executive function and emotion regulation).

First we define pretend play and review three theoretical posi- tions on whether and how it affects development generally. Next, we describe patterns of evidence that would support each position, then review and discuss the evidence domain by domain. Finally we consider one position more deeply, address the implications of our review for educational settings, and make suggestions for future research on this topic.

Defining Pretend Play

A preliminary issue is to define pretend play. Play itself is a notoriously difficult concept to pin down (Burghardt, 2011). For our purposes the four criteria of Krasnor and Pepler (1980) will define play: flexibility, positive affect, nonliterality, and intrinsic motivation (cf. Sutton-Smith & Kelly-Byrne, 1984). Flexibility denotes that play behaviors vary from real ones in form (they might be exaggerated, or truncated) and/or content (one might play at eating with a stick instead of a spoon). Positive affect touches on the idea that people look like they are having fun when they play. Nonliterality refers to the fact that, in play, behaviors lack their usual meaning while paradoxically retaining it; Bateson (1972) famously pointed out that, “the playful nip denotes the bite, but it does not denote what would be denoted by the bite” (p. 317). Intrinsic motivation suggests voluntariness: One engages in the activity by choice for its own sake.

Pretend play activities are the subset of play activities charac- terized by an “as-if” stance (Garvey, 1990). Beyond being simply nonliteral, in pretend play a “pretense” is layered over reality (Austin, 1979); specifically, a pretender knowingly and intention- ally projects some mentally represented alternative on to the pres- ent situation in the spirit of play (Lillard, 1993). Sometimes pretend play is social: A group of children share an alternative reality that they project, perhaps acting like they are different people in another place and time. Other times pretending is a solo activity. Pretend play can involve projecting imaginary objects and properties, or using one object as if it were another (Leslie, 1987). It is most prominent in early childhood, with ages 3 to 5 being declared its “high season” (D. G. Singer & Singer, 1992), although

it does continue into middle childhood and beyond (E. D. Smith & Lillard, in press).

There are several other forms of play besides pretend (see Pellegrini, 2009; P. K. Smith, 2010); in particular there is a small but important literature on physical play (such as hopscotch and rough-and-tumble play), which has been well reviewed elsewhere (Pellegrini & Bohn, 2005; Pellis & Pellis, 2009). Such forms of play assist sustained attention in conventional school situations (Pellegrini & Bohn, 2005); they also (in the case of rough-and- tumble play fighting) assist emotion regulation, social coordina- tion, and normal sexual behavior, at least in some rodents and nonhuman primates (Pellis & Pellis, 2009). Pretend play can overlap with these and other types of play. For example, physical play overlaps with pretend play when children pretend to be fighting warriors. Object play overlaps with pretending when a child animates those objects.

The literature is not always clear as to when pretend play specifically, versus play more generally, or some other specific type of play is at issue; this can be seen in the quotes with which we opened (but see footnotes 1–2), and probably arises because young children’s play is so often infused with pretense. Our aim here is not to resolve this ambiguity but rather to consider studies used to support claims that play is crucial to positive develop- ments, excluding the physical play literature just mentioned, and retaining focus on pretend play as much as possible. Our main exception to this is in two subdomains of nonsocial cognitive aptitudes, creativity and problem solving, because for those skills several studies concerning manipulative play with small objects (which might or might not involve pretending) are often cited as showing play’s cognitive benefits. When a study contrasted pre- tend play with some other form of play (like construction play, as in building with blocks) we focused on the pretend condition. Many studies strain the voluntary aspect of play in that children were told to play or were instructed in acting out a story, but because those studies have been cited as showing play’s benefits, they are reviewed here.

To locate studies, the first author began with references sup- porting claims of play’s benefits in articles like those in the opening paragraph, then back referenced those studies in a snow- ball fashion. Through this process she arrived at the six main topics and six subdomains of nonsocial cognitive skills; the subdomain of mathematics was subsequently eliminated due to an insufficient number of studies. From there a search engine (Google Scholar) was employed, searching by keywords (“social skills, pretend play”) and the “referenced by” and “related articles” features, as well as continuing to back reference from within articles. To avoid an unwieldy review, we passed over studies of atypical populations or cultural variation, and largely confined ourselves to published or in press peer-reviewed studies.3

Theoretical Background

P. K. Smith (2010) laid out three theoretically possible relation- ships between pretend play and positive developmental outcomes.

3 Exceptions were made for particularly important unpublished studies reported in other published work by the author or thesis advisor, and for one article under review.

2 LILLARD ET AL.

The first is that pretend play is crucial to optimal development. The second, which Smith supported, is equifinality: Pretending helps some developments, but it is only one possible route. Other activities can work as well or better. The third possibility is that pretending is an epiphenomenon or byproduct of some other selected-for capability, but in and of itself makes no contribution to development; rather, the other activity or condition to which it is sometimes attached is the actual contributor. Two major devel- opmental theorists, Vygotsky and Piaget, align with the first and third views, respectively.

For Vygotsky (1978), pretense has a crucial developmental role, because it is the activity by which children learn to separate referent from object. In play, children first understand that actions (and objects on which one might act) can be separated from reality and can be based on the meaning of a given situation rather than on the physical properties of objects (Vygotsky, 1967). In this way, for example, a banana could become a phone in a pretense situa- tion and the child could act on it as if it were a phone, inhibiting how he or she would act on it if it were a banana. The upshot of this is that children develop abstract thought through pretend play (Vygotsky, 1967). In addition, because reality must be inhibited, children also develop inhibitory control through pretending (Bodrova & Leong, 1996). Because of these features, “in play, it is as though [the child] were a head taller than himself” (Vygotsky, 1978, p. 102); play takes a child to the upper end of his or her “zone of proximal development” (p. 86).

In contrast, for Piaget (1962), pretend play is more an index than a promoter of development.4 Its appearance around 18 months indicates the development of the semiotic function, which also allows for deferred imitation and language. The semiotic function separates an idea from its referent, a memory from its context, and an object from its label, allowing one to entertain and elaborate on mental content that is separate from the physical, present reality.

Here we consider which of these views is best supported by the evidence. Each view is compatible with a particular pattern of evidence from correlational and experimental (short-term and training) studies, shown in Table 1 (cf. P. K. Smith, 2010, Table 9.2, p. 187). These patterns assume methodologically sound stud- ies including sufficient duration and sample sizes. First, if pretend play does crucially cause positive developments (Vygotsky’s po- sition), then strong positive correlations between pretend play and those developments should consistently be found; if a child pre- tends more, whether naturally (in a correlational study) or due to an experimental manipulation, the development should increase. If pretend play causes creativity, then children who pretend more will generally be more creative. Additional predictors, like intelligence, are also possible, but if pretending is truly the important causal factor, the unique and important relationship to pretend play should hold even when those other predictors are partialled out.

Conversely, if Smith’s equifinality position is correct, then one would generally expect positive relationships between play and the outcome but also correlations with other predictors that engender the outcome. For example, if social pretend play develops theory of mind and so does adult talk about mental states, then correla- tions should be found for both variables. Interventions increasing mental state talk and pretend play might have an additive effect when combined, which could lead to even larger effects (but not if there was a ceiling on development for that age). There could be cases when although equifinality is the best model, pretend play

fails to evince a significant effect. This might occur, for example, when there is substantial multicollinearity, or when an alternate predictor’s effect is much larger, masking pretend play’s effect. Thus equifinality does not insist on 100% consistent results, but it generally expects them.

The third, or epiphenomenon, position is supported if pretend play coincides with some other causal circumstance; in such cases pretend play might mistakenly be considered causal. For example, if social pretend play is related to theory of mind because adults who engage in a lot of mental state talk also happen to encourage pretend play, then perhaps what is actually leading to the increased theory of mind is not the pretend play, but the mental state talk; social pretend play is secondary or epiphenomenal to the mental state talk–theory of mind relationship. If pretend play is an epiphe- nomenon then one might find inconsistent correlations with outcomes (because pretend play does not always go along with the real predic- tors) but consistent correlations between real predictors and outcomes. Because different studies measure different possible predictors, the true predictors might not always be evident. Here we evaluate the patterns of evidence with an eye to each of these positions. Before beginning to do so, it is useful to note some recurring problems in this literature (see also Cheyne, 1982).

Common Methodological Problems

Several problems recur in the literature on whether pretend play helps development. Sometimes these problems occur because the

4 Others have claimed Piaget gives pretense a stronger role in develop- ment; for example, Singer and colleagues, citing Piaget (1962), claimed he “concluded that play was a vital component to children’s normal intellec- tual and social development” (D. G. Singer et al., 2009, p. 285). In our reading the closest Piaget (1962) comes to this is when he says it is undoubtedly “a preparation for imaginative aptitudes” (p. 155), where imagination (as in pretend play) is the assimilative pole of thought (in contrast to accommodation), and creative imagination arises only when one integrates the two. This is essentially the position taken by Harris (2000) and D. G. Singer and Singer (1992): Pretending assists imagination. But whereas for these modern authors this is a reason to centralize pretend play, our reading of Piaget’s text on play suggests that this role in imaginative development was a minor concern; pretend play was primarily an offshoot of the symbolic function. Perhaps confusion has arisen because elsewhere Piaget assigns manipulative activities (Piaget, 1929) and peer interaction (Piaget, 1932) as important to development, and pretend play often in- volves these other activities. But in considering manipulative activity, Piaget refers more to what is now referred to as embodied cognition (“manual work is essential to the child’s mental development”; Piaget, 1929, p. 383), and regarding peer involvement Piaget’s own focus on pretend play was particularly as a solitary activity. Piaget (1962) did think pretending served an egoistic function in that it allowed the child to fulfill wishes that he or she could not fulfill in reality. A child who wants to be a mother can simply pretend to be one. But Piaget was concerned with cognitive development, not personality development, and pretend play was pre-operational because it indicated what the child lacked. For Piaget, children outgrow pretending as they develop the ability to accommodate reality. Here he followed some major figures of his time in child psychol- ogy, such as Freud (1955, as discussed by Harris, 2000) and Montessori (1989). Aligning with our own reading, Sutton-Smith (1966) colorfully summarized Piaget’s view of pretending as “a buttress to an inadequate intelligence” (p. 108). For further discussion, see P. K. Smith (2010, pp. 31–37).

3PLAY DEVELOPMENT

research was conducted when experimental standards were not as high as they are today, pointing to the need to modernize the evidence base. In more recent studies, perhaps scholars did not apply more rigor because of a deep belief in the power of play (Elkind, 2007), what P. K. Smith (1988) dubbed the “play ethos” and Sutton-Smith (1995, p. 279) the rhetoric of “play as progress.” Here we strive to overcome the tendency to favor pretend play by holding all studies to a high standard.

One common problem in discussions of the impact of play on development is that correlational findings are often discussed as if they were causal. When children who play more do better on some other measure, of course it does not mean that the play definitely caused the outcome. Positive correlations between pretend play and a development are only a necessary precondition to pretend play being causal. Likewise prominent authors have described elaborate pretend worlds they constructed as children, and one might see the earlier behavior as causing their subsequent literary genius (Root-Bernstein, in press), but it is as plausible that their creativity led to conjuring up elaborate imaginary worlds at both time points.

A second recurring problem is failure to replicate. For example, one study shows increases in empathy associated with pretense training (Saltz & Johnson, 1974) and another does not (Iannotti, 1978), and typically only the positive finding is cited. If other key experimental factors are essentially equal, either the reported pos- itive result reflects a Type I error or the failure reflects a Type II error. Inconsistent findings in correlational studies contradict the causal view but would be expected with either the equifinality or epiphenomenalism. For equifinality, nonreplications would occur when the alternate route was stronger in one study, and including it masked the effect of pretend play; for the epiphenomenon position, nonreplications would occur because the underlying cause sometimes accompanied pretend play and sometimes did not. In the literature extolling play’s benefits, failures to replicate are often ignored.

A third problem concerns experimenter bias. Every under- graduate research methods course should impart the importance

of experimenters being “masked” insofar as possible: that is, unaware of (a) the hypotheses being tested and (b) participants’ conditions. Yet cognitive development research rarely uses masked experimenters. This might usually be fine: Child de- velopment researchers and the kinds of tests they give might not be vulnerable to experimenter bias under the usual circum- stances. For example, we know of no research suggesting that false belief or conservation errors occur at certain ages only when experimenters are unmasked. However, for research on the benefits of pretend play there are several cases where results obtained with knowledgeable experimenters went away when masked ones were employed (Christie, 1983; Guthrie & Hud- son, 1979; Pepler & Ross, 1981; Simon & Smith, 1983, 1985; P. K. Smith, Simon, & Emberton, 1985; P. K. Smith & Whit- ney, 1987). Nonreplications with masked experimenters make a strong case for being cautious about pretend play results ob- tained with knowledgeable experimenters.

Besides correlational data, nonreplication, and unmasked ex- perimenters, other recurrent problems are very small sample sizes, nonrandom assignment, confounding implementer with intervention (particularly concerning when there is only one implementer per condition and interventions last for several months), control conditions that differ beyond pretend play, confounding content with pretend play, and unsound statistical practices like using subsets of data and one-tailed tests without prior rationale.

Methodological problems are so prevalent in this literature that meta-analysis is precluded. E. P. Fisher (1992) did a meta-analysis of the impact of play (generally) on development, despite aware- ness of these limitations (see “Shortcomings of the Studies,” pp. 164–168), but he also did not have a consistent even-handed approach to which statistics he included, and further, he used some wrong statistics that inflated his result. As a particularly egregious illustration of this, from Christie (1983) he used the statistic pertaining to a variable named variable (F � 257.67), reflecting the overall sample scores on five variables, when the far smaller Variable � Time � Condition statistic (F � 0.49) is what should

Table 1 Three Possible Relationships Between Pretend Play and Development

Expected pattern of results Crucial (Vygotsky) Equifinality (P. K. Smith) Epiphenomenon (Piaget)

Correlational studies Strong, unique, and consistent. Generally consistent but not unique. Including other causal variables could mask pretend play’s effects, so correlations could be inconsistent.

Inconsistent, but consistent with other variables that are causal. For example, if presence of certain toys increases pretending in children who are more creative, but other objects have no impact, then correlations between pretending and creativity will be seen only in environments with those toys.

Experimental (short-term and training) studies

Strong, unique, and consistent. Strong and consistent but not unique, so other conditions could also affect development. For example, skills training and pretend play training could both increase the development.

Effects found only if the crucial underlying factor(s) is (are) influenced by the intervention. For example, suppose pretend play only assists development when intensive adult interaction is part of that training; when children pretend but there is no intensive adult interaction, the pretend play does not increase the development; in addition, another condition might show that intensive adult interaction alone increases the development, even in the absence of play.

4 LILLARD ET AL.

have been used. Careful reading reveals many more problems, yet this article is often cited (126 times, Google Scholar, as of May 28, 2012) as evidence that play helps development (e.g., Bergen, 2009; Ginsburg et al., 2007; Wyver & Spence, 1999).

Because so many studies in this area are methodologically unsound, the current literature base is best suited to a descriptive review, on which we now embark. In each section, we begin with theoretical and construct issues, then review studies. A series of 10 tables compiles the studies pertinent to each domain or subdomain of development.

After reviewing the studies, each section concludes by dis- cussing the evidence with respect to the three views (summa- rized in Table 12). In these discussions, we sometimes rely on the absence of evidence to support a position. We do this with caution, since one can never prove that a relationship does not exist (Altman & Bland, 1995). However, inconsistent correla- tion patterns across studies with similar samples and methods and reliable coders are against a causal view. Likewise, when sound experimental methods yield null effects or even effects showing play is less positive than the alternative, this is also relevant. Finally, doubt is also cast on a causal view when masking experimenters or equalizing other aspects of interven- tions nullifies previous findings.

A final note before treading into the evidence concerns the “straw person” element of the crucial causal view. When put on the stand, perhaps few would endorse the position that pretend play is crucial (in the sense of essential or vital) for various aspects of development. Yet the quotes with which we opened and additional references throughout this review show that this stance is taken in the literature, so we consider it here.

Nonsocial Cognitive Aptitudes

As was seen in our opening paragraph, many scholars have asserted that pretend play produces cognitive benefits. One way pretend play could help cognition is by predisposing children to a generally playful attitude (Dansky & Silverman, 1973) that could lead to production of unusual ideas, creative problem solving (Vandenberg, 1980), and then to other cognitive aptitudes. This view is compatible with Fredrickson’s (2001) broaden-and-build theory of positive emotions, with play eliciting joy, which in turn leads to a broadening of individuals’ thought–action repertoires. Vygotsky’s ideas on symbolic and abstract thought, just reviewed, also suggest how pretend play could assist cognitive abilities. Here we discuss evidence that pretend play assists development in five subdomains: creativity, problem solving, intelligence, conserva- tion, and reasoning.

Creativity

Although creativity has been operationalized in a number of ways, in the studies on play it has typically been defined as the ability to produce original content relevant to a particular task (Wallach & Kogan, 1965). The most commonly used measure of creativity in this literature is the alternate uses task (R. C. Wilson, Guilford, & Christensen, 1953), in which participants give possible uses for common objects, like a paper towel or a paperclip (Dan- sky, 1980a). Task responses are typically coded for fluency (num- ber of uses named) and originality (number of uses not given by

any other participant). Another common task is the Torrance Thinking Creatively in Action and Movement Test or TCAM (Torrance, 1981), which includes several subtests, including hav- ing children move like trees in the wind, and also alternate uses.

Correlational studies. Several studies have addressed the claim that pretending makes children more creative (Ginsburg et al., 2007; J. L. Singer, 1973) by looking for correlations between naturalistic play and creativity, since they should exist if more frequent pretenders have become more creative via their pretend activity. Of course correlations are not evidence of causation, but if causation exists, correlations should be consistently found.

Naturalistic classroom play has been categorized differently in different studies, but a combination of Smilansky (1968) and Parten’s (1932) schemes has been used most often (Rubin, 2001). Smilansky’s scheme (derived from Piaget) divides play into cog- nitive categories (functional play when a child repeats motor actions on objects, construction play when a child builds things, dramatic or symbolic play when the child substitutes an imagined world for reality, and games-with-rules like Checkers). Parten’s scheme is focused on the social aspect of children’s play: coop- erative when children are actively interacting in a common group endeavor; associative when they interact but not toward a single common endeavor; parallel when they play similarly but side by side, with little interaction; solitary independent when they play alone at their own games; onlooker when they watch others play; and unoccupied.

We found eight studies correlating pretend play and creativity (see Table 2). Seven concerned preschoolers. Typically children’s play in preschool was coded for 1–5 min per day for a period of 20 days or more, and then alternate uses with two to four objects (or in some cases the TCAM or another test) was administered. Results were inconsistent.

Johnson (1976) found that, controlling for IQ, amount of social but not solitary fantasy play was related to fluency. This would suggest that something about the social element, rather than pre- tending in and of itself, was related to creativity. However, John- son (1978) later failed to replicate this finding in a very similar study, showing no relationship between pretend play (social or solo) and alternate uses. A different study with the same age range also found no relationship between creativity and social pretend play (L. Dunn & Herwig, 1992) but found a negative relationship between originality of responses and solo pretend play that disap- peared when IQ was partialled out. Pellegrini and Gustafson (2005) also did not find children’s frequency of pretend play to be related to creativity (r � –.18) yet did find frequency of construc- tion play was related. On the other hand, a different study found all pretend play was significantly related to creativity in a sample of 15 high-IQ preschoolers (Moran, Sawyers, Fu, & Milgram, 1984). Wyver and Spence (1999) found that particularly fantastical pre- tend play was related to semantic creativity (naming all the objects one could think of) but not figural creativity (making objects from shapes); two other categories of pretend play that they coded were not related to either. The remaining preschool study, Lloyd and Howe (2003), was not useful regarding pretend play because they combined it with functional play, which was twice as common as pretend play, but it is worth noting that in contrast to Pellegrini and Gustafson (2005) they found construction play (which was coded separately) was not related to creativity. In the eighth and final correlational study, looking at somewhat older children, Russ and

5PLAY DEVELOPMENT

Grossman-McKee (1990) examined play with Russ’s Affect in Play Scale (APS), which uses 10 min of play with puppets and blocks coding generativity. Alternate uses also taps generativity, so it is not surprising that positive correlations were found, both concurrently and over time (Russ, Robins, & Christiano, 1999).

The eight correlational studies show an inconsistent pattern of relationships that does not support the causal model. The other models seek alternate routes or third variables that could underlie relationships when they are found. With this particular set of studies, no obvious other variable emerges. IQ is sometimes re- lated to creativity (Johnson, 1976; Wyver & Spence, 1999) but sometimes not (L. Dunn & Herwig, 1992; Lloyd & Howe, 2003; Moran et al., 1984; Wyver & Spence, 1999); Johnson (1978) did not test IQ. Johnson worked with low-socioeconomic status chil- dren, whereas others involved middle-class ones, but this cannot explain the discrepant results across Johnson’s own studies. An- other third variable that fits either view is environment (i.e., types of toys supplied), which can drive the types of play children engage in (McLoyd, 1983). Because environment was not mea- sured, we cannot evaluate if this could underlie the inconsistencies. Regardless, distinguishing the second and third views requires experimental studies.

Experimental studies. Four experimental studies found higher associative fluency when children first played with objects

for which they later named uses (Dansky, 1980b; Dansky & Silverman, 1973, 1975; Li, 1978). This fit with Sutton-Smith’s (1968) quasi-experimental study in which boys came up with more alternate uses for traditional boy toys than did girls (although for girl toys, they were equal). To check whether experience with the objects, rather than playing with them per se, was important, Dansky and Silverman (1973) included an imitation group, in which children gained experience by imitating the experimenter handling the objects. The imitators gave no more uses than the controls, suggesting that experience with the objects did not ex- plain the first results (see also Hughes, 1981, as cited in Hutt, Tyler, Hutt, & Christopherson, 1989).

Pretend play’s effect on creativity could be limited to the objects at hand. To test this, Dansky and Silverman (1975) used a different set of objects in the test phase, and here again the play group produced more uses, suggesting play’s hypothesized effect on creativity generalizes.

In the Dansky studies just described, although the theoretical rationale concerned pretend play, it is unclear whether children pretended with the objects or just manipulated them. Li (1978) tested whether pretend play might improve creativity above and beyond play generally. In the pretend play condition, the experi- menter told a short fantasy story, then showed children the stim- ulus objects and said, “Let’s make-believe or imagine that these

Table 2 Studies Examining the Effect of Play on Creativity

Type Citation � � � Masked Int Masked Exp Notes

C Johnson (1976) SPP Solo C Johnson (1978) SPP/Solo C L. Dunn & Herwig (1992) SPP Solo Solo disappears when partial IQ C Pellegrini & Gustafson (2005) Const. PP C Moran et al. (1984) PP Holds when partial IQ C/T Wyver & Spence (1999) PP PP/SPP No No C: One type of pretense related to one

task; T: effects work both ways; fantastical themes; finding for one of two measures for associative PP

C Lloyd & Howe (2003) Const. C Russ & Grossman-McKee (1990) Solo Creativity/play measures are redundant E Dansky & Silverman (1973) Play No No E Dansky & Silverman (1975) Play No No Extended to different objects E Li (1978) PP No No Two of four objects only E Dansky (1980b) Play No No Pretenders only E P. K. Smith & Whitney (1987) Play No Yes E Pepler & Ross (1981) Play No Yes See footnote 5 in text E Pellegrini & Greene (1980) Play No No Structured questioning better E Pellegrini (1981) Play No No Structured questioning better E Russ & Kaugars (2001) PP No No Scorer but not administrator masked E Howard-Jones et al. (2002) Play Irrel. Yes Control task very dull T Dansky (1980a) SPP Yes Yes Adult contact not controlled; “natural”

pretenders only T Feitelson & Ross (1973) PP PP No No Only on one measure; adult contact

not controlled T Christie (1983) PP � Skills No Yes Adult contact controlled T P. K. Smith & Syddall (1978) PP � Skills Yes No Adult contact controlled T P. K. Smith et al. (1981) PP � Skills No Yes Adult contact controlled T Moore & Russ (2008) Solo No Yes Adult contact controlled

Note. Type of study: C � correlational; E � experimental; T � training. Type of play: Play � pretense status unspecified; Const. � construction play (e.g., blocks); Solo � pretend play alone; SPP � social pretend play; PP � pretend play (social unspecified). �: positive relationship to play; �: no correlation or play � nonplay condition; � � negative relationship to play. Masking: Intervention (Int) or posttest experimenters (Exp). If masking status was not specified, we assume experimenters were not masked, since that is the unmarked case. Masking for correlational studies is omitted because it is rarely mentioned, even when it is likely (because play observations occurred several years earlier than testing, for example). Irrel. � irrelevant.

6 LILLARD ET AL.

objects could become anything you want them to be. Play with all of these things” (p. 33). Free play, imitation, and control condi- tions were similar to those used in Dansky’s studies. After 10 min, the same experimenter administered the alternative uses test with the three objects used in the intervention and a new fourth object. Significant differences were found for one of three old objects (a paperclip), for the make-believe group and the free play group, and for the new object (a screwdriver) only for the make-believe group versus the control. Overall, Li’s children came up with far fewer uses than children had in the previous studies, perhaps reflecting population variance.

Taking a different tack, Dansky (1980b) examined whether children who naturally engage in more pretend play would benefit more from a play intervention. Children were classified as pre- tenders if they engaged in pretense more than 25% of the time, or nonpretenders if they engaged in it less than 5%. They were then assigned to free play or control conditions. Suggesting the class- room classifications had validity, in the free play condition 88% of the pretenders but only 6% of the nonpretenders pretended with the objects. The alternate uses test was given with a different set of objects, and only the pretenders in the free play condition had higher fluency. Dansky concluded that play induces creativity only for those who are predisposed to pretend.

These experimental studies suggest that play might have a causal effect on creativity, at least for children who frequently pretend (see also Sutton-Smith, 1967). However, in these studies the experimenters administering the creativity test knew which condition each child was in, and perhaps their expectations influ- enced children’s responses. With alternate uses, the experimenter elicits answers until they think a child has run out of possibilities. More coaxing might inadvertently occur when children are ex- pected to produce more uses. In an attempted replication of Dan- sky and Silverman (1973), P. K. Smith and Whitney (1987) used different experimenters for the intervention phase and the posttest, with the latter masked to condition. Results showed no significant differences—in fact the control condition obtained the highest mean score. This finding is consistent with another study using a masked experimenter: The number of uses given was not signifi- cantly different for children in a play condition (Pepler & Ross, 1981, Study 2).5 A less direct test of possible experimenter effects occurs when a different hypothesis is being tested. Two studies tested the hypothesis that focused questioning would elicit more uses than would playing with objects (Pellegrini, 1981; Pellegrini & Greene, 1980). The hypothesis was upheld: Focused questioning led to more uses, and free play was not different from the control. Another way to check for experimenter biasing is to videotape task delivery, which could encourage standardization; Russ and Kaugars (2001) did so, and children who engaged in pretend play generated no more uses than children who did puzzles. All this suggests that the alternate uses task might be particularly vulner- able to experimenter bias.

A solution is to use a different creativity task, like making a collage, the creativity of which is assessed by masked judges (Howard-Jones, Taylor, & Sutton, 2002). Unfortunately the study that did this did not use a good control condition. Six- and 7-year-olds played with salt dough for 25 min or copied words from the board (and if they finished early, were told to start over). Afterward, they made collages, and the play group’s were more creative. However, perhaps forced copying had a deleterious effect

on creativity, rather than play having a positive one. Further research with a more neutral control condition is needed.

Training studies. Experimental studies (as defined here) examine short-term change; perhaps play does influence creativity but requires longer incubation periods. We located seven longitu- dinal play training studies. The first three produced effects that were in the expected direction, but four others controlled for adult contact and found play training itself had no effect.

Dansky (1980a) compared 36 low-income preschoolers in so- ciodramatic play, free play, and object exploration interventions over 3 weeks, with three 30-min sessions per week. The sociodra- matic play training involved enacting pretend play themes like going on a picnic. Children in the free play group could play as they wished, and they rarely engaged in pretend play. The explo- ration training group explored and discussed objects. All experi- menters were masked. The sociodramatic play group outperformed both other groups on alternate uses.

In a similar study, Feitelson and Ross (1973) compared 24 kindergarteners in play tutoring, music tutoring, toy play without tutoring, and control groups, with ten 30-min interventions over 5 weeks. Creativity was measured with the picture completion sub- test of Thinking Creatively With Pictures (Torrance, 1966), in which children complete up to 10 pictures in a way that “no one else will think of” and give each picture a title, and Dog and Bone (Banta, 1970), in which children make up different routes from a dog to a bone. The play tutoring group’s scores increased the most on the number of unusual titles given (but not number of pictures or picture elements) and trended toward better performance on Dog and Bone.

Wyver and Spence (1999) trained 38 children in three types of play based on Parten’s social categories and play type (associative fantastical, cooperative fantastical, or cooperative constructive/ everyday sociodramatic) for 4 hr over 4 weeks, with pre- and posttests of figural and semantic creativity; they included a no- intervention control. The former two groups increased their fan- tastical pretend play, and the third did not increase in any type of pretend play. However, only the third group showed increases in both semantic and figural creativity. Associative fantastical play training improved semantic but not figural creativity, and cooper- ative fantastical play did not improve in either. These results are quite mild then for the hypothesis that pretend play increases creativity, since it did not reliably do so for the fantasy groups and the sociodramatic group also engaged in construction play. Other experiments in this study tested the reverse direction by training children in creativity and then observing play; results suggested complex and bidirectional relationships.

The results of these studies suggest that the play training can increase children’s creativity. However, it is unclear whether ex-

5 Their first experiment, using a knowledgeable experimenter, yielded some significant differences

"Is this question part of your assignment? We can help"

ORDER NOW