Evidence for a relation between executive function and pretense representation in preschool children

Abstract

Several theoretical formulations suggest a relation between children’s pretense and executive function (EF) skills. However, there is little empirical evidence for a correlation between these constructs in early development. Preschool children (N = 104; M age = 4-0) were given batteries of EF and pretense representation measures, as well as verbal, memory, and appearance-reality control tasks. Confirmatory factor analysis revealed two separable but overlapping aspects of EF (Conflict and Delay). EF was significantly related to pretense after accounting for all controls. Understanding the pretend-reality distinction was strongly related to Conflict EF, whereas performing pretend actions was more strongly related to Delay EF. These results, although correlational, are consistent with the claim that EF skills are implicated in pretense, such as inhibiting reality and flexibly manipulating dual representations, and offer a potential mechanism by which pretend play interventions may enhance childhood EF.

1. Introduction

One of the most surprising and charming features of early childhood is the frequency with which children – who have much to learn about the real world – engage in pretend play, where reality is distorted on purpose. There may nevertheless be important cognitive benefits of pretense beyond suggested social-emotional benefits (Bretherton, 1989; Singer & Singer, 1990). One such benefit concerns executive functioning (EF), the cognitive processes that aid in self-control of thought and action (Carlson, Zelazo, & Faja, 2013). Important developments take place in children’s EF during the early preschool years, at the very time when fantasy and pretense are most prevalent. The development of EF skills, mediated by prefrontal and other cortical changes, may play a crucial role in children’s ability to entertain multiple, conflicting mental representations, which are displayed prominently in pretense. Moreover, pretense clearly involves an inhibitory component: Children must inhibit the tendency to refer to an actual, veridical state of affairs so that non-reality representations can arise and be maintained. Conversely, experience with pretend play may contribute to gains in EF by providing a way to reframe real-world events and problems and think more flexibly. We therefore aimed to examine the strength and nature of a relation between preschool children’s EF and pretense representation skills.

1.1 Pretense

Pretense follows a consistent developmental timeline (Carlson & Zelazo, 2008; Fein, 1981; Garvey, 1991). In the second year of life, toddlers attribute living characteristics to a nonliving object (e.g., a stuffed animal) and transform objects, such as pretending a banana is a phone. By age 2, children appear to understand the pretend overtures of others, such as wrapping a towel around a teddy bear when anadult “spills” imaginary tea over it (Harris & Kavanaugh, 1993; Lillard & Witherington, 2004). Gradually during the preschool years, pretending becomes less dependent on props and more abstract. For example, children will knock on an imaginary door of an imaginary house and serve imaginary food to imaginary people. The preschool period is considered the apex of make-believe play, when an estimated 40% of children in Western cultures have imaginary companions (Gleason, Sebanc, & Hartup, 2000; Singer & Singer, 1990; Taylor, 1999). As children enter middle childhood, organized games become more common than overt pretense, although children (and many adults) continue to have active imaginations (Singer & Singer, 1990; Taylor, 1999).

1.2 Executive Function

Concurrent with the development of pretense skills is the rapid acquisition of EF, which entails goal-directed behavior, including inhibition, working memory, and set-shifting (Garon, Bryson, & Smith, 2008; Hughes, 1998; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000; Zelazo, Carter, Reznick, & Frye, 1997). EF is increasingly recognized as a critical component of children’s cognitive and social functioning (for review see Carlson et al., 2013). One category of EF tasks targets ability to control impulses, often in the face of tempting rewards (Kochanska, Murray, Jacques, Koenig & Vandegeest, 1996; Reed, Pien, & Rothbart, 1984). For example, Kochanska et al. (1996) gave children a Gift Delay task, in which an examiner instructs them not to peek while she noisily wraps a present for them. Children’s waiting ability on such “delay” tasks improves across the preschool period. The second category, “conflict” EF, includes measures calling for children to respond a certain way in the face of a highly salient, conflicting response option, but there is no extrinsic reward and the task rules are relatively decontextualized from daily behavior (Diamond & Taylor, 1996; Frye, Zelazo, & Palfai, 1995; Gerstadt, Hong, & Diamond, 1994; Reed et al., 1984). For example, following Luria’s (1966) pioneering work, Gerstadt et al. (1994) presented 3- to 7-year-olds with cards depicting either the sun or moon and instructed them to say “night” in response to sun cards and “day” in response to moon cards. Children’s performance improved gradually from age 3 to 7 years. Significant improvements were found from age 3 to 5 years on an adaptation of this task in which children needed to point to a white card when the examiner said “grass” and to a green card when the examiner said “snow” (Carlson, 2005; Carlson & Moses, 2001).

Both delay and conflict EF are conceptualized as drawing on the same top-down control processes, and they are correlated in preschoolers (Carlson & Wang, 2007), but dissociations have been reported in which relatively “hot” or affective EF is delayed relative to “cool” EF, perhaps because emotional arousal can undermine control processes (Zelazo & Carlson, 2012). Furthermore, although there is a literature showing one latent factor for EF in the preschool period using largely cool EF measures (Wiebe, Espy, & Charack, 2008; Wiebe, Sheffield, Nelson et al., 2011), several studies of preschoolers by Carlson and colleagues have found overlapping but distinct factors for conflict EF measures, which are relatively cool and require considerable working memory (Grass/Snow), and delay EF measures, which are relatively hot because they involve a delay of gratification for a tempting reward or response (Gift Delay; Carlson & Moses, 2001; Carlson, Moses, & Breton, 2002). The dissociation between conflict and delay EF appears as early as toddlerhood (Bernier, Carlson, Bordeleau, & Carrier, 2010; Morasch & Bell, 2011).

1.3 Linking Pretense and Executive Function

A long-standing question about pretend play is how young children are capable of generating and understanding pretense representations despite such underdeveloped cognitive and brain systems? Different answers have been proposed, but leading developmental theories of pretense appear to have a common focus, which is the ability to hold two (or more) different representations of the same thing in mind without becoming confused regarding which represents reality. Scholars including Olson (1991), Friedman and Leslie (2005), Harris (1994, 2000), and Lillard (1993, 2001) have offered explanations for how children maintain the boundary between pretense and reality. Each of these theories has in common the claim that pretense takes place at a level of representation that is somehow separated or distant from reality, and that meta-awareness of such representational acts may not come until relatively late in the preschool years (for review, see Carlson & White, 2013). Yet we are left with the question of how this occurs – what are the psychological or neural mechanisms involved? In our view, EF provides a plausible domain-general mechanism to help account for how children manage dual representations in pretense.

As noted earlier, pretend play becomes more decontextualized, that is, distant from reality. Pretend themes result from children’s increasing ability to imaginatively manipulate various event scripts in a broad representational system (Carlson & Zelazo, 2008). Older children are capable of imagining highly fantastical scenarios, such as a world in which miniature aliens speak and perform all activities backward. This decontextualization is akin to Werner and Kaplan’s (1963) theory of psychological distancing (see also Dewey, 1931/1985; Hegel, 1807). Sigel (1970) adopted this term to mean “behaviors or events that separate the child cognitively from the immediate behavioral environment” (p. 111), which lead to representational competence —“the individual’s awareness and understanding that an instance can be represented in various forms and still retain its essential meaning” (Sigel, 1993, p. 142).

Similarly, Vygotsky (1967) suggested that pretense is instrumental to internalization, that is, the development of internal systems of representation that assist children in freeing themselves from external stimulus control and permit thinking about objects and events not immediately present. These systems include language (private speech) and what might be referred to today as EF. Vygotsky also noted a paradox in children’s make-believe play: Rather than being spontaneous and “free”, it requires the suppression of impulses so that social rules for behavior can be followed (Nicolopoulou, 1991). Golomb and Kuersten’s (1996) research has shown that even 3-year-olds are well aware of the boundary between pretense and reality, and they object when reality intrudes on make-believe play, such as when a researcher really bites into a Playdoh cookie.

Flexible executive control over mental representations and prepotent responses may be the central feature that binds these theories regarding the cognitive underpinnings of pretense. We suggest that young children become increasingly adept at managing conflicting mental representations in pretense (and eventually with metacognitive awareness) in concert with EF development.

1.4 Prior Evidence

Empirical evidence of a connection between EF and pretense is sparse. Only a handful of correlational studies have uncovered relations between EF and symbolic play skills. In a short-term longitudinal study, Elias and Berk (2002) observed 53 3- and 4-year-olds during free play and clean-up periods in a preschool setting. They found that the complexity and duration of play with a partner in the beginning of the school year predicted increases in compliance on the cleanup task observed 8 months later, independent of age and vocabulary level. In another study, Albertson and Shore (2008) presented 32 preschoolers with an object (e.g., a block) and explained its pretend identity (e.g., a phone). Children’s ability to later recallboth the real and pretend identities of an object significantly correlated with scores on a set of three conflict EF tasks. Finally, after controlling for mental age, Kelly and Hammond (2011) found a relation between structured pretend play and a version of the Day/Night inhibitory control task among preschoolers. Although these studies reveal a link between pretense and self-regulation, they were limited by several factors, such as small sample sizes for correlational analyses and few behavioral measures of EF and control variables. Moreover, each of the EF tasks used in these studies focused on relatively “cool” regulatory abilities, leaving the relation between pretense and hot or delay EF unaddressed.

A broader spectrum of measures in a large sample is needed to apprehend the full nature of the relation. It is especially important to examine the links between pretense and both the cool and hot facets of EF. Although not yet explored, it is possible that pretense representation relates to hot/delay EF just as strongly as cool/conflict EF. Such correlational evidence can serve as a basis for deciding whether pretense might serve as a useful tool for helping to improve EF in young children.

Experimental evidence suggests that pretense representation serves to “cool” down and thus improve performance on a challenging “hot” task. Mischel and Baker (1975) first demonstrated this when preschoolers children were shown able to delay gratification for longer intervals when instructed to pretend that the tempting stimulus (e.g., a marshmallow) was something less tempting (e.g., a white fluffy cloud). Similarly, Carlson, Davis, and Leach (2005) introduced a symbolic context to facilitate children’s performance on Less is More, an EF task in which children need to point to a smaller number of treats in order to receive a larger number of treats. Degrees of symbolic distancing had incremental effects on 3-year-olds’ ability to override the prepotent response of pointing to the larger array. When symbols were used to represent the treats, the more abstract or decontextualized the symbol, the greater their control over their responses (see also Apperly & Carrol, 2009). This work suggested a relation between symbolic representation (as in pretense activities) and self-control over thought, action, and emotion. Further empirical evidence is nevertheless needed on the nature and strength of relation between EF and pretense in a large sample, independent of confounding variables.

We hypothesized that individual differences in EF would be significantly positively related to children’s representational skills in the domain of pretense, over and above differences that might be attributable to age, verbal ability, memory capacity or understanding of the appearance-reality distinction. We further predicted that the correlations would hold not only for cool EF, but also for relatively hot, affective measures of executive control.

2. Method

2.1 Participants

Participants included 104 typically developing preschool children (M age = 4-0, SD = 5.20 months, range 39–60 months, 48 girls). Ten additional children failed to complete the study. Children were recruited by telephoning parents from a university database and posting fliers in a major metropolitan area. Participants were primarily Caucasian (80%; remainder Asian, African American, or two or more races). Median education for mothers (M age = 37.68, SD = 5.40) and fathers (M age = 39.5, SD = 4.51) was a college degree. Parents received $8 at each session and children received small snacks and prizes.

2.2 Procedure

All children were tested individually in two 60-minute videotaped laboratory sessions approximately one week apart by the same female examiner. We administered measures of receptive vocabulary and short-term memory along with batteries of EF and pretense tasks. Session 1 included the Peabody Picture Vocabulary Test (PPVT-III), Standard Dimensional Change Card Sort, Appearance-Reality, Less is More, Tower Building, and Backward Digit Span. Session 2 inncluded Pretend-Reality, Forward Digit Span, Delay of Gratification, Pretend Actions, Grass/Snow, Bear/Dragon, and Gift Delay. All coders were blind to the hypotheses of the study.

2.3 Control Measures

Three measures were included to control for potential confounding variables that might account for any relations found between pretense representation and EF.

2.3.1 Verbal Ability

We planned to control for verbal ability because it is frequently correlated with both EF (Albertson & Shore, 2009; Garon et al., 2008) and pretense (Taylor & Carlson, 1997). The Peabody Picture Vocabulary Test-3^rd edition (PPVT-III; Dunn & Dunn, 1997), a normed measure of receptive vocabulary, was administered using the standard method. Raw scores (ceiling item – number of errors) were used in analyses where age in months was also included as a covariate.

2.3.2 Memory

Forward Digit Span was included as a control for memory capacity, as it might be implicated in both EF performance (Carlson, 2005) and the ability to hold in mind a pretense representation (Ervin-Tripp, 1991). To assess short-term memory capacity, children were asked to repeat a series of numbers in the Forward Digit Span task (Davis & Pratt, 1996). Children were given a two-digit practice trial, and then the examiner administered the test trials beginning with two digits and increasing by one digit (up to 7) upon each successful response. The task was terminated after three consecutive failures at a given level. Highest level achieved was recorded.

2.3.3 Appearance-reality

A large literature now exists demonstrating the association between individual differences in EF and theory of mind (for review see Moses & Tahiroglu, 2010), as well as evidence to suggest a relation between theory of mind and pretense (Astington & Jenkins, 1995; Meins, Fernyhough, Russell, & Clark-Carter, 1998; Youngblade & Dunn, 1995; Taylor & Carlson, 1997). We included the Appearance-Reality task (Flavell, Green, & Flavell, 1986) as a control measure because understanding the distinction between appearance and reality is often highly correlated with false-belief performance (Astington & Jenkins, 1995; Carlson & Moses, 2001) and because it bears an identical structure to the Pretense-reality task used here (Flavell, Flavell, & Green, 1987), thus serving as a stringent control in a test of the relation between EF and pretense representation. The examiner presented children with two different objects with ambiguous appearances. The first object was characterized by a discrepancy between the way it appeared and what it really was (a sponge crafted in such a way that it appeared to be a rock). The second object was characterized by a discrepancy between the way it looked and its authentic color (a red castle was covered with a green transparent cover to produce the appearance of a black castle). The examiner showed children the appearance and reality of the object and then asked, “When you look at this right now, does it look (like a sponge/red) or does it look (like a rock/black)?” Then she asked, “What (color) is this really and truly, a (sponge/red) or a (rock/black)?” Children received one point for answering both the appearance and reality question correctly for each object, yielding a maximum score of 2.

2.4 Executive Function Measures

Eight EF measures were administered, including four cool/conflict tasks and three hot/delay tasks. The final task, Less is More, can be considered from either facet, because it has the structure of a conflict task (activating a subdominant response) and was significantly correlated with a cool EF task (the DCCS) in previous research (Carlson et al., 2005), but the reward context of a hot task (playing to win food treats). Thus we included exploratory factor analysis to determine which construct this task would fit with most closely.

2.4.1 Standard Dimensional Change Card Sort(Zelazo, 2006)

Children were presented with two boxes with affixed target cards (red rabbit and blue boat). The examiner instructed children that rabbits go in the box with the red rabbit and boats go in the box with the blue boat in the “shape game” (pre-switch trials). After children successfully sorted five consecutive cards, the examiner introduced the “color game” (post-switch trials) in which all of the red things go in the box with the red rabbit and all of the blue things go in the box with the blue boat. Following Frye et al. (1995), two of the five post-switch trials were compatible with the pre-switch sorting rule and three trials were incompatible (i.e., if children reverted to the pre-switch rule they would respond incorrectly). The number of correct incompatible trials (of three) was used in analyses.

2.4.2 Backward Digit Span (Davis & Pratt, 1996)

The examiner asked children to repeat a series of numbers in reverse order. After demonstrating the task using a puppet, she administered a two-digit practice trial with feedback up to six times before moving to the test trials. The first trial consisted of two numbers and the series increased in length by one number with each successful trial (up to a maximum of 5 digits). After children failed on three consecutive trials, the task was stopped. We recorded the highest level achieved. (Children unable to complete two digits received a score of 1.)

2.4.3 Grass/Snow (Carlson & Moses, 2001)

In this Stroop-like task, the examiner first instructed children to place their hands on top of a pair of felt cut-out hands mounted on a black vinyl placemat, which also had a white card affixed to the upper left corner and a green card affixed to the upper right corner. She explained that they were going to play a “silly game” in which they were to point to the white card when she said “grass” and point to the green card when she said “snow.” Following two practice trials, 16 test trials (8 grass; 8 snow) were administered in a fixed interspersed order. Percentage of correct trials was used in analyses.

2.4.4 Bear/Dragon (Reed et al., 1984)

This task requires children to suppress certain actions while activating others in an alternating fashion. Children were introduced to a “nice” bear puppet and a “naughty” dragon puppet and instructed to do what the bear asks (e.g., touch your nose) but not what the dragon asks. After practicing with feedback (max 4 trials), the examiner administered 10 test trials, which alternated between bear and dragon commands. A rule reminder was given after the fifth trial. Performance on dragon trials was taken as an index of self-control and was coded as follows: 0 = full commanded movement; 1 = partial commanded movement; 2 = a wrong movement, or flinch; 3 = no movement. Trials were summed to create a total dragon score for each participant.

2.4.5 Less is More (Carlson et al., 2005)

In this reverse-reinforcement contingency measure of executive function, children were presented with a choice between a larger and a smaller quantity of treats (e.g., 2 v. 5 jelly beans). They were told that whichever tray they selected would be given away to a naughty puppet and they would keep the treats in the other tray. Therefore, children must inhibit the natural response to point to the tray they want for themselves. After a practice trial and rule check, children completed 16 trials with one rule reminder after the eighth trial. The left-right orientation of the treats was counterbalanced. The percentage of correct (smaller) treat choices (out of 16 test trials) was calculated.

2.4.6 Tower Building (Kochanska et al., 1996)

The examiner presented children with 20 blocks and instructed them to take turns with her to build a tall tower. After demonstrating how to take turns, the examiner began to build the tower by placing the first block and thereafter, the examiner did not take her turn unless the child explicitly directed her to do so by handing her a block or stating, “it’s your turn.” Scores were computed as the percentage of blocks placed by the examiner (ideally, 50%). Following Kochanska et al. (1996), five points were subtracted if children knocked the tower down at the end and five points were added if children removed the blocks with care. Final scores were averaged across two towers.

2.4.7 Delay of Gratification (Mischel, Shoda & Rodriguez, 1989)

The examiner presented children with a choice of treats (Frootloops, Goldfish crackers, or raisins). After children made their selection, she placed two treats on one plate and 10 treats on another and confirmed that children preferred the larger amount. She then informed children that she had to leave the room to do some work and, if they remained in their seat for a period of time without eating any of the treats, upon her return, they could have the large pile. If they did not want to wait, they were given the option of ringing a bell to make the examiner return, although they could only have the small pile if they did so. Following a rule check, the examiner left the room until the child ate the snacks, rang the bell, or 5-min passed (whichever came first). We recorded the reaction time to the first touch to the treat, bowl or bell, total number of touches (reverse-scored) and total waiting time. Given these variables were highly intercorrelated, they were converted to z-scores and averaged to create a total Delay of Gratification score for each participant.

2.4.8 Gift Delay (Kochanska et al., 1996)

In this delay of gratification measure, children were asked to sit in a chair with their back facing the examiner and asked not to peek while she noisily wrapped a present for them in a standardized manner for 60 sec. Following Kochanska et al., an aggregated z-score was computed by averaging a peeking score (0 = turning around to peek; 1 = peeking over the shoulder; 2 = no attempt to peek), total number of times peeked (reverse-scored), and latency to the first peek.

2.5 Pretense Battery

We administered two well-established tasks designed to tap pretense representation.

2.5.1 Pretend-Reality (Flavell et al., 1987; Frye et al., 1995)

The examiner presented children with objects and asked them to differentiate between their actual and pretend identities. After a warm-up trial, she presented four objects and their respective pretend identities (string-snake, block-car, pencil-hammer, sunglasses-telephone). While pretending with the object (e.g., using a pencil as a hammer), she asked the following: “What is this really and truly – is it a (string/block/pencil/sunglasses) or a (snake/car/hammer/telephone)?” and “What are we pretending this is right now – are we pretending it’s a (string/block/pencil/sunglasses) or are we pretending it’s a (snake/car/hammer/telephone)?” Across trials, the reality and pretend questions were counterbalanced. The number of correct responses (of 8) was recorded.

2.5.2 Pretend Actions (Overton & Jackson, 1973)

Children were required to physically demonstrate a series of pretend actions designed to assess their level of representation of pretend gestures. After a brief warm-up, the examiner asked children to pretend to do the following 4 actions: brush your teeth with a toothbrush, comb your hair with a comb, drink out of a cup and put on sunglasses. For each trial, children’s responses were coded as using a body part (lower level, e.g., when asked to comb your hair with a comb, children used open fingers through the hair) or a symbolic object (higher level, e.g., they used a closed fist as though guiding a comb through the hair) to display the pretend action. Using a symbolic object represents a more sophisticated developmental level of pretense. The following coding system was used: body part = 0, symbolic object = 1, for a maximum score of 4.

3. Results

We present descriptive data, followed by results for the assessments of pretense and EF individually, and then analyses of the relations between them.

Less than 1% of EF task trials and 4% of pretense task trials were missing. A task was considered missing if >50% of trials were not completed. When creating EF composites, the small amount of missing data for EF tasks was handled by averaging scores on existing task data for that individual (i.e., if one EF task was missing, that individual’s average across the seven other EF tasks (z-scored) was computed for the Full EF Composite). Nine children refused to take part in the Pretend Actions task and one child’s data were dropped due to experimenter error (N_{PretendActions} = 94). No scores were missing for Pretend-Reality or EF composites (Ns = 104).

Descriptive statistics for control, pretense and EF measures appear in Table 1. Scores were within the normal range for children of this age. We found no significant sex differences on the EF or pretense measures; hence gender was not considered in subsequent analyses.

Table 1.

Descriptive Statistics for Verbal Ability, Memory Capacity, Executive Function and Pretense Measures

Task	Mean (SD)	N
Verbal Ability: PPVT-III (raw score)	70.07 (20.26)	104
Memory Capacity: Forward Digit Span (highest level)	3.97 (0.98)	101
Appearance-Reality (out of 2)	1.13 (0.86)	103
Executive Function
DCCS (out of 3)	1.33 (1.46)	104
Backward Digit Span (highest level)	1.71 (0.83)	89
Grass/Snow (percent correct)	69.91 (33.60)	97
Bear/Dragon (dragon score out of 15)	12.78 (4.60)	102
Less is More (percent correct)	67.04 (31.25)	100
Tower Building (max score of 55)	33.42 (16.03)	99
Delay of Gratification		103
RT to 1st touch (out of 300 sec)	132.69 (122.91)
Total touches (reversed prior to analyses)	3.58 (4.41)
Total wait time (out of 300 sec)	229.13 (113.47)
Gift Delay		104
Peeking score (out of 2)	1.13 (0.89)
Total peeks (reversed prior to analyses)	1.46 (1.99)
Latency to 1st peek (out of 60 sec)	36.72 (25.69)
Pretense
Pretend-Reality (out of 8)	6.88 (1.57)	104
Pretend Actions (out of 4)	1.84 (1.53)	94

3.1 Pretense Assessment

Pretend-Reality was modestly, although significantly, correlated with Pretend Actions (see Table 2).

Table 2.

Correlations among Individual Pretense and Executive Function Tasks Controlling for Age

	PA	DCCS	BDS	G/S	BD	LIM	Tow.	DoG	GD
Pretense
Pretend-Reality	.13	.32**	.29**	.32**	.51**	.38**	.01	.34**	.16
Pretend Actions		−.14	.17	.08	.16	.09	.18†	.19†	.18†
Executive Function
DCCS			.30**	.01	.18†	.24*	.03	.34**	.12
Backward Digit				.29**	.17	.26*	.11	.31**	.22*
Grass/Snow					.02	.19†	−.04	.20†	−.08
Bear/Dragon						.31**	.18†	.23*	.35**
Less is More							.08	.12	.13
Tower Building								.24*	.28**
Delay of Gratification									.20†

Note. N= 94–104. AR=Appearance-Reality; PA = Pretend Actions; PR = Pretend-Reality; DCCS = Dimensional Change Card Sort; BDS = Backward Digit Span; G/S = Grass/Snow; BD = Bear/Dragon; LIM = Less is More; Tow. = Tower Building; DoG = Delay of Gratification; GD = Gift Delay.

^†p < .10;

^*p < .05;

^**p < .01.

3.2 Executive Function Assessment

Inter-correlations among the EF tasks are shown in Table 2. Several tasks were significantly correlated over and above age. On the basis of an exploratory factor analysis showing a 1-factor solution, EF Composite scores were formed. These consisted of the average of the z-scores on DCCS (factor loading = .65), Backward Digit Span (.74), Grass/Snow (.46), Bear/Dragon (.63), Less is More (.63), Tower Building (.47), Delay of Gratification (.63), and Gift Delay (.52). We next used confirmatory factor analysis to examine whether the tasks we selected did indeed reflect separable conflict and delay facets of EF. As shown in Table 3, both a 1-factor model and a 2-factor model fit the data well (following modification for covariance). Figure 1 illustrates the results of the 2-factor model in which there are 5 measures with a conflict structure and three measures with a delay structure. Although parsimony would point toward the 1-factor model, it was not a significantly better fit than the 2-factor model – an issue of special interest in the present study given the goal to extend previous findings on the relation between EF and pretense. Hence, in addition to the overall composite EF scores, we computed separate Conflict EF and Delay EF scores for each participant.

Table 3.

Summary of Confirmatory Factor Analysis Results on 8 Executive Function Tasks

	Model	Chi-square	df	p	TLI (> .95)	CFI (> .95)	RMSEA (< .06)	SRMR (< .08)	Significance of factor loadings
1	1-factor	24.25	18	.147	.934	.957	.058	.056	All
2	2-factor	24.56	18	.138	.930	.955	.059	.057	All

Note. The cutoff criteria for each index are in parentheses.

Figure 1. Confirmatory Factor Analysis of the Executive Function Measures.

Note. Dragon = Bear/Dragon; BDS = Backward Digit Span; LIM = Less is More; GS = Grass/Snow; DCCS = Dimensional Change Card Sort; Gift = Gift Delay; DoG = Delay of Gratification; Tower = Tower Building.

3.3 Relation between Executive Function and Pretense

3.3.1 Correlations

Correlations between individual pretense and EF measures appear in Table 2. As indicated in Table 4, EF was significantly correlated with both Pretend Actions and Pretend-Reality after controlling for age. More precisely, Delay EF was significantly related to both pretense measures, whereas Conflict EF was significantly related only to Pretend Actions.

Table 4.

Correlations among Pretense and Composite EF Measures Controlling for Age

	Full EF Composite	Conflict EF	Delay EF
Pretend-Reality	.53**	.60**	.24*
Pretend Actions	.21*	.11	.26*
Full EF Composite		.87**	.78**
Conflict EF			.37**

Note. N = 104.

^†p < .10;

^*p < .05;

^**p < .01.

3.3.2 Regression Analyses

To more specifically assess the unique contribution of pretense to individual differences in EF, and vice versa, we conducted a series of hierarchical multiple regression analyses controlling for age, PPVT-III, Forward Digit Span (Block 1) and Appearance-Reality (Block 2).

First, we regressed EF on pretense (see Table 5). Pretend-Reality and Pretend Actions were entered together in Block 3 of each model. Pretend-Reality scores significantly predicted EF Composite scores, and Pretend Actions predicted EF Composite scores at a marginal level. Separate regressions predicting Conflict and Delay EF revealed divergent relations with the two pretense variables: Pretend-Reality scores predicted Conflict EF scores, whereas Pretend Actions predicted Delay EF.

Table 5.

Standardized Regression Coefficients from Hierarchical Multiple Regression Analyses of Variables Predicting Executive Function Scores

	Full EF Composite	Conflict EF	Delay EF
Block 1
Age	0.28**	0.37**	0.06
Verbal Ability	0.37**	0.27*	0.38**
Memory	0.21*	0.19*	0.16
R2	.48**	.45**	.27**
Block 2
Appearance-Reality	0.32**	0.36**	0.12
ΔR2	.08**	.11**	0.01
Block 3
Pretend-Reality	0.22**	0.33**	0.01
Pretend Actions	0.14†	0.04†	0.24*
ΔR2	.05**	.08**	.05†

Note. N= 94–104 with pairwise deletion of missing cases.

^†p < .10;

^*p < .05;

^**p < .01.

Given the likelihood of reciprocal relations between these constructs, we also investigated the relation by regressing pretense on EF (see Table 6). The full EF composite predicted both Pretend Actions and Pretend-Reality over and above control variables. Also mirroring the above findings, when the EF subscales were entered together in Block 3 of an additional regression, Conflict EF predicted Pretend-Reality and Delay EF predicted Pretend Actions.

Table 6.

Standardized Regression Coefficients from Hierarchical Multiple Regression Analyses of Variables Predicting Pretense Scores

	Pretend- Reality	Pretend Actions
Block 1
Age	0.08	0.35**
Verbal Ability	0.19	0.03
Memory	0.27*	0.10
R2	.19**	.17**
Block 2
Appearance-Reality	0.44**	−0.06
ΔR2	.17**	0.00
Block 3A
Full EF Composite	0.36**	0.31*
ΔR2	.06**	.04*
Block 3B
Conflict EF	0.50**	0.09
Delay EF	−0.03	0.26*
ΔR2	.11**	.06*

Note. N= 94–104 with pairwise deletion of missing cases. Block 3A and Block 3B variables were entered into two separate regression models, each with identical Blocks 1 and 2.

^†p < .10;

^*p < .05;

^**p < .01.

4. Discussion

The principal aim of this research was to investigate the relation between individual differences in EF and pretense representational skills in preschool-aged children. Both historical and contemporary scholars have speculated that such a relation exists on the grounds that inhibition of the real state of affairs is necessary for pretense representations to arise and be enacted, or alternatively, that pretense is a means by which young children develop self-regulatory skills. Surprisingly, however, empirical studies of this relation are sparse. We therefore sought to establish the nature and strength of the relation within the preschool period using a battery of EF and pretense tasks in a large sample. The results revealed there is indeed a robust, positive correlation between EF and the ability to manage dual representations in pretense. These findings held for relatively cool conflict measures of EF, as suggested by previous smaller studies (Albertson & Shore, 2009; Kelly & Hammond, 2011), but there was also new evidence for a relation between pretend actions and performance on relatively hot, delay-of-gratification measures. The results were reciprocal when we examined both EF—Pretense and Pretense EF—regressions.

These results, although correlational, are consistent with the idea that prefrontal cortex and concomitant EF development play a role in helping children manage conflicting mental representations in pretense. Inhibitory control is suspected to be a crucial aspect of representation in pretense. To entertain fantasy scenarios in one’s mind, prepotent reality-based cues must be suppressed (Nicolopoulou, 1991; Vygotsky, 1967). Moreover, self-control is believed to be an important aspect of successful pretend play. Preschoolers’ play is highly rule-based (Harris, 2000) and a child who is unable to inhibit inappropriate responses in the play scenario (e.g., actually biting into a mud pie) is at risk of exclusion from further play and peer rejection (Golomb & Kuersten, 1996).

The findings are also consistent with a bidirectional account, in which practice and experience with pretense might in turn contribute to hallmark EF skills of working memory, inhibition, set-shifting, and planning. Engaging in symbolic thought, as required during pretense, may allow children to psychologically distance themselves from salient or irrelevant stimuli, thus decreasing the pull of prepotent responses and in turn strengthening their ability to self-regulate (Mead, 1934; Vygotsky, 1967). Experimentally, increasing psychological distance from a stimulus through symbolic substitution (Carlson et al., 2005) or encouraging children to pretend that a rewarding stimulus is less appealing (Mischel & Baker, 1975) have been shown to increase children’s self regulatory abilities. Pretend play in early childhood thus may provide a means by which children learn to think in a more flexible and reflective manner, a necessary skill for successful cognitive and emotional regulation. The relations we found between pretense and EF support this claim. Indeed, this ability to think about alternatives is the overarching link within a bidirectional account of pretense and self-regulation. We propose that “playing” with representations in pretense and the growing ability to consider alternatives in thinking and acting through EF development may be rooted in the same core abilities, supported by neurological development.

For reasons that remain unclear, we found that Pretense-reality and Conflict EF were strongly correlated whereas Pretend Action performance was more strongly related to Delay EF. The former might be attributable to shared method variance (e.g., suppression of a dominant response and activation of a nondominant response in the Pretense-reality task). The link between the Pretend Action task and Delay EF could be explained by the fact that performing pretend actions, while seemingly rooted in the ability to think about objects symbolically, is strengthened by practice with pretending and imagining in daily life. Such experience might serve socioemotional self-regulation (as in delay-of-gratification) especially well, as] found by Singer (1961), who reported that children ages 6–9 who engaged in rich fantasy play waited twice as long as low-fantasy children when asked to wait quietly for 15-min.

The relation between EF and pretense representation is consistent with previous research showing a robust relation between preschool EF and representation of mental states in theory of mind tasks (Moses & Tahiroglu, 2010; Rakoczy, 2010). Some modicum of self-control may be necessary before children can appreciate someone else’s (or their own prior or future) mental state when it conflicts with present reality (Carlson & Moses, 2001). Pretense might be considered yet another domain of mental life that children are better able to access and reflect upon if they are well equipped with EF skills. This interpretation is consistent with a separate line of research showing a positive relation between individual differences in pretense and theory of mind, although the cognitive requirements for thinking about pretense may be less taxing than for desires or beliefs because the allure of reality is less potent (for review see Kavanaugh, 2006). For example, Lillard and Flavell (1992) presented preschool children with scenarios in which they were told that a character thought, wanted, or pretended X when in fact Y was the case. Children then were simply asked what the character thought, wanted, or pretended. Performance was significantly better for pretense and desire than for belief. Similarly, Gopnik and Slaughter (1991) set up conditions in which children thought, wanted, or pretended X at Time 1 but thought, wanted, or pretended Y at Time 2. They were then asked to recall their Time 1 mental state. Recall was significantly better for pretense than for belief, with desire intermediate in difficulty. Our research suggests an important triangle of EF, pretense, and theory of mind in preschool cognitive development (Friedman & Leslie, 2005). It is important to note, however, that the relations between EF and pretense representation held up in our study even after controlling for Appearance-Reality task performance, which is often viewed as a foundation for theory of mind (Flavell & Miller, 1998), suggesting that their common bond with theory of mind development does not entirely account for the relation.

Several aspects of the present study limit our ability to draw firm conclusions regarding the relation between EF and pretense in early childhood. First, our results suggest that EF may be strongly related to the developing aptitude for pretense (i.e., the requisite skills of holding dual representations in mind, inhibiting reality), whereas questions remain regarding its relation to observed pretend play. It could well be that there is an increasingly strong reliance on EF skills as pretense becomes more decontextualized (e.g., pretending without props) and when one is called upon to reflect on its abstract properties, such as knowing that one has multiple ways of representing reality and consciously selecting among them (Carlson & Beck, 2009). Research in this area is hampered by a lack of reliable and valid measures of social pretense for preschool children that can be observed in a laboratory setting. Additional observational research along the lines conducted by Elias and Berk (2002) is needed to understand the link between social pretend play and EF.

Second, our sample was fairly homogenous demographically. Although we would expect the relation between individual differences in EF and pretense representation to pertain across a wide range of children, the relation between EF and observed pretense might vary as a function of experience and the content of pretense, such as violent themes (Berk, Mann, & Ogan, 2006; Hughes, White, Sharpen, & Dunn, 2000). More diverse samples should be included in future research to make more accurate generalizations about the relation between EF and pretense.

Third, as with any correlation between constructs in development, it is important to consider other factors that might mediate the relation. In addition to memory span and theory of mind, we undertook to control for verbal ability for this reason. The commonly found relation between EF and receptive vocabulary was substantiated.. As well, language and narrative skills are more advanced in high-pretense children (Taylor & Carlson, 1997; Trionfi & Reese, 2009). Although we found the relation between pretense representation and EF remained significant when controlling for PPVT-III scores, this measure is only a proxy for language development and verbal IQ, and thus it is possible that controlling for a more comprehensive set of language measures would have further attenuated the relation between EF and pretense. Nonetheless, it is worth keeping in mind the more general role of the language of thought’ (Vygotsky, 1934/1962), which serves to further both imagination and conscious control over thought, action, and emotion via self-speech (Fernyhough, 2009). Indeed, private speech could well be the central variable integrating EF and pretense given evidence for a domain-general shift to verbal mediation in early childhood (Al-Namlah, Fernyhough, & Meins, 2006). Hence, any effort to fully rule out language as a piece of this puzzle is likely to lose the very essence of the phenomena in question.

Fourth, the correlational design of the study cannot address the developmental ordering of the relation between EF and pretense. As noted earlier, we favor a bidirectional, iterative account in which these skills are mutually bootstrapping. In recent years, some educators have asserted that pretense promotes learning (Singer, Golinkoff, & Hirsh-Pasek, 2006), flexible thinking (Hirsh-Pasek, Golinkoff, Berk, & Singer, 2009), and self-regulation of emotion and behavior (Bodrova & Leong, 2006; Diamond, Barnett, & Munro, 2007). Yet, no study to date has provided evidence as to the direction of influence between pretense and EF. Longitudinal studies, even less than one year in length, have proven useful in showing that early pretense predicts later theory of mind (Jenkins & Astington, 1996), and such methods are now needed to trace intra-individual developmental trajectories of pretense and EF over time. Given that the first signs of EF emerge by 12 months (Diamond & Goldman-Rakic, 1989) and that pretense emerges around 18 months, it would be fruitful to begin such an investigation at an even earlier age than that of the children in the present study.

Fifth, experimental study is also necessary to tease apart the causal nature of the relation between EF and pretense. As noted earlier, studies of the beneficial effects of symbol substitution in the Less is More task of inhibitory control (Carlson et al., 2005), and of fantasy settings in logical reasoning tasks (Dias & Harris, 1988, 1990), have offered some clues. Although neither study directly addressed the effects of pretense on EF, their results indicate that fantasy-related manipulations can lead to improvements in EF tasks. Further experimental work of this kind will inform developmental theories of the nature and underlying mechanisms involved in pretense, imagination, and EF, as well as have immediate relevance to early childhood education and policy. For example, age-appropriate EF skills are crucial for the transition into school and beyond (Blair & Diamond, 2008). We have much to gain by better understanding the extent to which pretense is playing a key role in successful preschool EF intervention programs such as Tools of the Mind’ (Bodrova & Leong, 2006).

Finally, it will also be fruitful to examine the ways in which pretense interacts with other aspects of development that have also been shown to be related to EF, such as parenting and culture (Sokol, Mueller, Carpendale, Young, & Iarochi, 2010). Given the wide range of cognitive and socioemotional outcomes associated positively with both pretense and EF, it is valuable to further investigate the nature of the relation between them and the social contexts most likely to favor their co-development.

Highlights.

• A developmental relation between executive function (EF) and pretense is investigated in the preschool period

• A large battery of EF measures fit both a 1-factor and a 2-factor (Conflict and Delay) model in a confirmatory factor analysis

• Individual differences in pretense representation and EF were significantly related, over and above several controls

• Understanding the pretend-reality distinction was related to Conflict EF, whereas performing pretend actions was related to Delay EF

• A bidirectional between EF and pretense is proposed, with implications for pretend play interventions designed to foster EF development