Expanding Fröebel's garden: the effect of play on adolescents’ cognitive development

Abstract

This study explored the relationship of play and cognitive development in adolescents. The research design was a quantitative study that consisted of two groups of participants who engaged in either a computer program (work condition) to test cognitive skills or a computer game (play condition) that contains the same tests. The participants were 167 adolescent male students aged 14–19 from urban mid-Atlantic Catholic high schools. The results indicated that adolescent males performed the cognitive tasks of memory and reading better in the play condition than they did in the work condition and performed the cognitive tasks of logic and mathematics better in the work condition than in the play condition. Differences in performance related to age were not present. The findings suggest that play may have a significant effect on adolescents’ cognitive development.

Keywords::

• play theory
• adolescence
• cognitive development
• emotional development
• zone of proximal development

View correction statement:

Corrigendum

Introduction

Throughout the literature, different definitions of play activity have been developed. Gray (2009) synthesised the current discussion about the definition of play and created the following five characteristics of a play activity: self-chosen and self-directed; intrinsically motivated; structured by mental rules; imaginative; and produced within an active, alert, and non-stressed mental state. These characteristics generally hold true across play theory, with slight variations and interpretations of each of the characteristics depending on the perspectives of individual play theorists.

Play has been traditionally viewed as having a developmental purpose in childhood (Froebel, 1887; Groos, 1901; Piaget, 1976a, 1976b, 1976c; Smilansky, 1990; Vygotsky, 1976). Vygotsky (1976) argues that play must have a developmental purpose rather than a pleasure-seeking purpose because play can lead to suffering (through the loss of the game). Much of the play research relates to the developmental nature of play, particularly a phenomenon known as the zone of proximal development (Vygotsky, 1976, 1978). A serious gap therefore exists in the literature because it is evident that adolescents play, but no research has been conducted to determine the psychological nature of play in adolescents.

The purpose of this study was to explore the developmental nature of adolescent play. A theoretical framework for adolescent play was developed from the literature by combining seminal theories of developmental play in early-childhood, theories of adolescent development, and research into cognitive processing. A methodology was then designed using one of the earliest studies that validated the theories of developmental play (Istomina, 1975) and the findings of a group of replications of that study which offered guidance on how to correct for methodological mistakes in the original study. Data resulting from the new methodology were gathered, analysed, and interpreted using the same theoretical framework.

Theoretical framework

Vygotsky (1976) argues that play begins as a method of wish-fulfilment. A child is denied access to a desired activity or object (e.g. such as car) so the child uses play as a method of subverting the denial of access (e.g. using a plate as a steering wheel). What is so important about this development is that a child is able to separate the meaning of words away from the objects the words indicate. In other words, the rudimentary skills needed for logical abstractions and linguistic development are being set through a child's play. This process led Vygotsky to the belief that learning precedes development. Therefore, between the two developmental spaces is a zone of proximal development (Vygotsky, 1976, 1978). The zone of proximal development allows learners to accomplish tasks with the help of social interaction and other cultural tools. Play facilitates learners to reach the top of their current zones in a similar way as social interaction.

Adolescents do not develop in the same way as children. The development of adolescents takes place in three different domains: cognitive, socio-emotional, and physical (Rice & Dolgin, 2005). The cognitive development of adolescents appears to focus on logical abstraction and the ability to develop empathy (Piaget, 1991; Vygotsky, 1931). Also during adolescence, a host of cognitive changes are occurring inside the brain that are related to a development of the thinking capacities associated with adulthood (Sousa, 2005). These changes include the final development of linguistic structures (Pallier, 2007), memory capacity (Radavansky, 2011, Sousa, 2005), and the ability to initially respond to new situations with logic rather than emotion (Sousa, 2005).

Socio-emotional development is a particularly rich aspect of development in adolescence. During this time adolescents deal with the identity crisis described by Erikson (1991). It is at this juncture that adolescents must ‘try on’ new persona to ultimately find the one that suits who they are as individuals (Marcia, 1966; Muuss, 1991). This period of identity formation is particularly traumatic for adolescents as independence from parental figures is being formed as well. Also at this time, adolescents are determining the strengths and weaknesses of their person. Research has shown that emotional states during learning can dramatically shape adult self-perceptions, particularly in the realm of how well one learns (Chaffar & Frasson, 2004; Feick & Rhodewalt, 1997; Salzberger-Wittenberg, Henry, & Osborne, 1983).

If Vygotsky is correct in his assertion that play is done for developmental purposes, then play must affect the development of adolescents as well, even though this play is different in presentation from that of early childhood. Adolescents typically engage in games with rules; however, Vygotsky argues that games with rules and fantasy play are similar activities because the mental processes involved in both activities are structured by rules that are negotiated by the participating children. Without fantasy, games with rules would be pointless activities.

Theoretically it is plausible that play has an effect on all three domains of adolescent development. Since adolescents have not completed cognitive development, then play may have the similar role as it does in early childhood, depending on the particular skill. It is also likely that play and identity formation are linked in adolescence. Identity formation is closely linked to the development of language. It is how words are used that create self-meaning. Since play initially supports linguistic development, it is possible that play would continue to have an effect with socio-emotional development related to language. Similarly, play, particularly games with rules, serves a purpose in the physical development of adolescents. Like the mastery play theorised by Piaget (1976a), adolescents may play physical games (e.g. football, marching band) as a way to practice and refine motor skills and muscular development.

Method

To test the effect of play on the development of adolescents, a narrow focus had to be employed. It would be impractical to simultaneously test the effects of play on all three developmental domains in adolescence. Rather, a specific domain was chosen, cognitive development, for several reasons. The first reason was related to the body of literature that already exists on the effects of play on childhood cognitive development. The literature in this domain is robust and offers much guidance on how to apply the theories of childhood developmental play into an adolescent population. It was determined that a cognitive developmental study would more clearly define the effect of play on development than if it was conducted on the other domains.

The research questions for this study were as follows:

• To what extent do adolescents perform basic cognitive tasks – specifically memory, reading, logical reasoning, and mathematical reasoning skills – differently during fantasy play situations than in instructional situations?

• Is there a difference in the age group of the adolescent (middle: 14–16; late: 17–19) and how basic cognitive tasks – specifically memory, reading, logical reasoning, and mathematical reasoning skills – are performed during fantasy play situations?

• Do adolescents’ age, socio-economic status, and overall academic performance predict their cognitive skills in memory, reading, logical reasoning, and mathematical reasoning during fantasy play situations versus instructional situations, respectively?

Participants

The participants of the study were purposively sampled from a Mid-Atlantic state's Catholic school system. The sampling process included three schools: two single gender (male and female, respectively) high schools and a co-educational middle school. Owing to issues with testing, only the data from the single gender male high school were viable (n = 167). Information was collected about socio-economic status (defined by ZIP codes), racial diversity, and general academic performance. This information was statistically analysed to determine that the population was homogeneous in relevant domains.

Design

The basis for the methodology of this study derives from the work of Zamora Istomina (1975). Istomina's study into voluntary memory laid the groundwork for the research of play being related to the zone of proximal development and the motivational aspects of play. In her study looking at voluntary memory (the earliest an individual can choose to remember versus general impressions), Istomina examined two groups of children. She informed one group that there was a list of words to be memorised and that they would be presented with a quiz at the end of the study period. The other group received the same list of words, but rather than being told to memorise the list of words they were informed that the words were a part of a grocery list and they were going to pretend to be in a store shopping for the items. At the end of the learning period for each group, Istomina compared how well the children were able to memorise the lists of words. What she found was that the group who were asked to engage in fantasy play had higher levels of retention than the group who were just asked to memorise.

Since Istomina's original study, many replications have been completed (Hasselhorn & Schneider, 1994; Meacham, 1977; Oyen & Bebko, 1996; Paris, 1978; Schneider & Brun, 1987; Smirnov & Zinchenko, 1969; Weissberg & Paris, 1986). The findings from this group of replications are mixed, with some being able to replicate the study and others failing to do so. The group of replications provided guidance to how a new replication can reliably replicate Istomina's (1975) study for a different population. The first issue had to do with relevance. The play condition must be culturally relevant to the participants and the tests must hold some sort of attractiveness to the participants (Reese, 1996). Other issues had to do with internal threats to validity relating to experimenters being too friendly in the play condition and treatment order effects as the participants in most of the studies participated in both conditions. Finally, a replication would incorporate recent findings into cognitive science. The methodology would expand the narrow focus of the studies on short-term memory into other cognitive domains of importance to the development of adolescents.

The methodology of this study attempted to answer these issues. A computer program was developed to test the cognitive processing skills of adolescents. The program consisted of four cognitive tests: short-term memory, working memory, spatial-logical reasoning, and mathematical reasoning. The short-term memory test consisted of the presentation of 11 food items followed by a 15-second interference period. The participants were then asked to recall by typing answers to as many responses as they could remember. The working memory test was a modified version of Daneman and Carpenter's (1980) Reading Span Test. The version used was a shortened version derived from van den Noort, Bosch, Haverkort, and Hugdahl's (2008) computerised version of the test. The test would present a series of sentences in groups (of two to six) and would then ask the participant to recall the final word of each sentence in the group. The instrument tests how much a person can ‘work with’ in the slave memory systems (the aspect of memory that stores different information types before being transferred to long-term memory storage). This instrument is also a fairly accurate predictor of reading comprehension (Barr, Kamil, & Mosenthal, 1996). The spatial–logical reasoning test consisted of a Tower of Hanoi logic test. The test presents three poles with a series of ‘blocks’ of diminishing size placed on top of each other placed on the first post. The goal of the test is for the participant to reproduce the tower on the final post by moving one block at a time and not being able to place a larger block on top of a smaller block. The mathematical reasoning test comprised of a number series test. The number series test is a series of six numbers that have a relationship. Three of the numbers are left blank and the participant must determine the rules of the series in order to fill-in the missing numbers. This type of test was chosen because it is one of two mathematical reasoning tests that do not rely on a participant's linguistic ability (Carroll, 1996).

Procedure

After the participants indicated Institutional Review Board assent and entered basic demographic information, they were randomly assigned to one of two conditions in the program. The first condition presented the four cognitive tests in a fairly standard manner. A set of directions were given and the test was presented. The second condition presented the same tests, but the tests were embedded into a fantasy turn-based game. In order to be culturally relevant and important to the adolescents participating, the game was designed for the sampled population. The game consisted of the teens creating a character (see Figure 1) who will throw a party to impress a potential ‘crush’ selected at the beginning of the game (see Figure 2). The participants then navigated through the game on a map to make the arrangements for the party (see Figure 3). At set points during the game, the cognitive tests were presented as mini-games. The directions for each of the tests were virtually identical (some changes were made to insure the game context did not contaminate the work condition), but with the exception of a storyline to explain why the game was being completed.

Figure 1 Character creation screen.

Figure 2 Character sample.

Figure 3 Interactive game map.

The program was piloted in three ways. The first included an expert panel including experts in cognitive testing, play, curriculum development, and an experienced classroom teacher. The expert panel made corrections relating to the context of presentation and wording of the cognitive tests. The second piloting consisted of having adolescents (n = 6) complete the program and offering recommendations of how to make the play condition more engaging. The final pilot consisted of actual participants (n = 23). Corrections were made to directions and how the program linked to the server as a result of this piloting procedure.

Data collection and analyses

The measure taken from the short-term memory test was the number of items correctly recalled. For the working memory test, the total number of correct responses was recorded. The measures taken for the spatial–logical test were the time needed to complete the test and the number of moves a participant needed to correctly solve the puzzle. The mathematical reasoning test collected data on the time needed to complete the test and the total number of correct responses. The data taken were sent in real-time to a server for later analysis.

Before the data were analysed, they were corrected and vetted. Since the original tests for short-term memory and working memory were oral, it was possible for a participant to write a correct answer, but misspell it. Corrections were made when the spelling intent of the participant was clear. Inclusion of the working memory test offered a particular issue. The test consisted of 10 groups of sentences. If a participant did not answer an entire group, it could indicate he was not participating in the test or he could not remember. Therefore, to ensure that only data which indicated participation was included, a participant had to complete 60% of the working memory test for the data to be included.

Any participant who did not complete the testing in its entirety was removed from the dataset. This decision was made because it relates to engagement. In order for an activity to be regarded as play, a participant must be fully engaged (Gray, 2009). While it could be argued that a participant was engaged in play but chose to opt out of a particular mini-game, total removal was preferred as it more thoroughly eliminates the threat to validity. Also, any participants who did not answer demographic questions were removed from the dataset as well. The total number of qualifying participants in the dataset was 91.

The original design of the study included a series of questions to determine the nature of the interactions the participants had. However, the response rate was 16.6% (28/169), but still within acceptable limits for an informal qualitative analysis. The responses were terse and usually a single word or a few-word phrase. Those in the work condition often described it using words like ‘boring’ or ‘frustrating’. One positive remark was made about the work condition. The play condition had mixed results. Some found the game ‘gay’ or ‘uninteresting’; however, many of the responses from the game condition were positive: ‘I like spending and making money’, ‘I felt good’, or ‘dissapoint [sic] that the party blew’. These statements indicate a mixed preference for gaming types, but overall an engagement in the game as play. For example, respondents in the work condition consistently referred to the tests as ‘tests’; however, the respondents in the game condition referred to the tests solely as ‘mini-games’.

The data were analysed using a one-way factorial analysis of variance (ANOVA) to answer the first and second research questions. The data met the statistical assumptions of the test and the purpose of the first two research questions. Effect size tests were also completed to determine the practical significance of the findings. The second method of data analysis used was a series of simple regression to answer the third research question. This method was appropriate because the sample size was not sufficiently large for a step-wise regression and the number of factors was low enough to ensure a type-I error did not occur through the multiple regressions that were performed. Additionally, the use of regression is appropriate in determining predictive variables to performance.

Results

Research question one

Before the ANOVA test was conducted to determine differences in the means between the play and the work condition, descriptive statistics were taken to determine the nature of the data. Table 1 shows the results from the cognitive tests in the work condition.

Table 1 Descriptive statistics of cognitive tests in work and play condition.

Condition	Test	n	Mean
Work	Memory Test	55	6.75 (2.33)
Play	Memory Test	36	8.61 (2.18)
Work	Math Test	55	4.09 (1.86)
	Math Time	54	106.26 (60.36)
Play	Math Test	36	3.03 (2.21)
	Math Time	35	101.37 (56.73)
Work	Logic Moves	55	32.36 (13.89)
	Logic Time	55	99.44 (61.74)
Play	Logic Moves	33	43.18 (21.36)
	Logic Time	35	141.12 (84.13)
Work	Reading Span Test	54	29.70 (6.64)
Play	Reading Span Test	36	32.81 (5.31)

Note: n = 91. Time to complete the task is given in seconds. Numbers in parenthesis indicate standard deviation.

Because of significant variation in the logic test [LS(90) = 5.02, p = 0.03)], the data relating to the logic test were removed from the one-way factorial ANOVA of the other tests. A t-test was conducted for the logical reasoning test results instead. The ANOVA was conducted to determine differences in performance in the differing treatment conditions with the other tests (see Table 2).

Table 2 ANOVA comparing test condition performance results.

Source	Degrees of freedom	F	p
Memory	1	14.69	0.000*
Mathematics score	1	6.12	0.02*
Mathematics time	1	0.146	0.70
logic time	1	7.32	0.01*
Reading Span Test score	1	5.51	0.02*

Note: n = 91. *p <  .05 level.

The findings from the ANOVA test comparing work condition results with play condition results were significant in each of the testing domains. The memory test had significant differences between the two conditions (p <  0.000). The descriptive statistics indicate that the differences in means support the hypothesis that adolescent males perform short-term memory tasks better in a play condition than in a work condition.

The mathematics test had significant results when relating to score (p = 0.02), but not relating to the time to complete the task. The descriptive statistics suggest that adolescent males performed better in the work condition than in the play condition; however, the differences were not statistically significant. These findings reject the null hypothesis that there would be no differences between the conditions, but they do not support the alternative hypothesis that adolescent males would perform better in the play condition.

The logic test had significant results relating to time to complete the task (p <  0.01). The descriptive statistics fail to support the research hypothesis that adolescent males would perform logical tasks better in a play condition than in a work condition. The participants in the play condition took longer to complete the task than the participants in the work condition.

The Reading Span Test also had significant results (p <  0.05). The statistically significant differences between the conditions and the descriptive statistics confirm the research hypothesis that adolescent males will perform better on working memory tasks in a play condition than in a work condition.

To determine the differences in the logic testing, an independent-samples t-test was conducted assuming that variances were not equal between the groups (n = 88). A significant effect for logic was present [t(48) = –2.60, p <  0.01], with participants in the work condition performing better than those in the play condition.

The logic test had significant results relating to the score (p <  0.01). The descriptive statistics fail to support the research hypothesis that adolescent males will perform logical tasks better in a play condition than in a work condition. The participants in the play condition both used more moves to complete the task than the participants in the work condition and needed longer time for completion.

In addition to the ANOVA test, a Cohen's d test for effect size was conducted to determine the impact of the differences in addition to statistical significance. Cohen's d value is presented because it has been demonstrated to give a more realistic representation of actual effect. Pearson's correlation coefficient is also presented because it is a highly conservative representation of actual effect (see Table 3).

Table 3 Results of Cohen's d test for effect size for cognitive tests comparing test conditions.

Test	Work mean^a	Play mean ^b	d	r
Memory	6.75	8.61	0.82	0.38
Mathematics test	4.09	3.03	0.52	0.21
Logic moves	32.36	43.18	0.60	0.29
Logic time	99.44	141.12	0.56	0.27
Reading Span Test	29.70	32.81	0.52	0.25

Note: n = 91. ^an = 55. ^bn = 36. Time is given in seconds.

Research question two

To answer research question two, a 2 × 2 ANOVA was employed. Table 4 reports the differences between age and test condition with performance on the different cognitive tests (see Table 4).

Table 4 Descriptive statistics of cognitive tests in work and play conditions as a function of age.

Test	Condition	Age group	Mean
Memory	Work	14–16^a	6.39 (2.55)
		17–19^b	7.32 (1.99)
	Play	14–16^c	7.93 (2.59)
		17–19^d	9.06 (1.89)
Mathematics	Work	14–16^a	4.10 (1.87)
		17–19^b	4.18 (1.92)
	Play	14–16^c	2.93 (2.57)
		17–19^d	3.11 (2.06)
Mathematics time	Work	14–16^a	108.44 (58.14)
		17–19^b	106.18 (64.43)
	Play	14–16^c	91.32 (60.10)
		17–19^d	105.12 (55.05)
Logic moves	Work	14–16^a	31.81 (13.11)
		17–19^b	34.14 (15.33)
	Play	14–16^c	47.00 (19.01)
		17–19^d	38.83 (22.69)
Logic time	Work	14–16^a	98.12 (54.00)
		17–19^b	104.82 (73.83)
	Play	14–16^c	152.76 (86.97)
		17–19^d	115.78 (71.20)
Reading Span Test	Work	14–16^a	29.03 (7.03)
		17-19^b	30.55 (6.25)
	Play	14–16^c	31.21 (7.18)
		17–19^d	34.61 (3.31)

Note: n = 85. ^an = 14. ^bn = 22. ^cn = 14. ^dn = 18. Time is given in seconds. Numbers in parenthesis indicate standard deviation.

The final dataset after the removal of outliers decreased the size of the participating population for the 2 × 2 ANOVA. The treatment group that contained the participants of the 14–16 age group and engaged in the play condition went from meeting the statistical assumption of a large enough population size (n = 15) to being insufficiently large (n = 14). A series of estimations comparing the expected and observed values was created for each of the cognitive tests and was compared against an expected normalised value to determine normality, and found the data from the participants could be included. Levine's test was also conducted and determined no statistically significant variance existed between the groups. Owing to the results of the expected and observed variables comparison and Levine's test, the dataset from the 14–16 age group could be included because the findings determined an appropriate amount of statistical normalcy.

The 2 × 2 ANOVA confirmed the statistically significant findings of the 1 × 1 ANOVA conducted for the first research question. The results vary slightly due to differences in calculation (see Table 5).

Table 5 Results of 2x2 ANOVA comparing the effects of test condition.

Source	Degrees of freedom	F	p
Memory score	1	9.990	0.000*
Mathematics score	1	5.86	0.02*
Mathematics time	1	0.45	0.50
Logic moves	1	6.65	0.01*
Logic time	1	4.41	0.04*
Reading Span Test score	1	4.89	0.03*

Note: n = 91. *p <  0.05.

When test conditions and age were compared, several interactions were present. Interactions occurred between test condition and age in the performance of the participants on the following cognitive tests: mathematics score, mathematics time, logic score, logic time, and Reading Span Test. Even though the majority of the tests demonstrated interactions, none of them were significant. The main effects were significant as they related to the treatment condition, but not age. These findings are consistent with the results of the first research question.

Research question three

The first step in analysing the data to answer research question three was to determine how the data correlated with the selected independent variables of age, socio-economic status, and academic performance (GPA, Grade Point Average) within the work condition. Four correlations were found significant: income/mathematics score, GPA/memory score, GPA/mathematics score, and GPA/reading span score. Each of these correlations suggests that each of the independent variables of income and GPA may be predicting variables for performance on their respective tests. Participants’ GPA and performance were negatively correlated. A simple regression was completed for each of these significant correlations to determine how much of the overall performance was due to the predicting variable. The regression tests confirm the research hypothesis that socio-economic status and academic performance are predictors of performance on the cognitive tests related to mathematics scores, short-term memory, and working memory within the work condition. The predictive value of performance in relation to academic performance is negative, while in relation to income it is positive. The findings reject the research hypothesis that age is a predictor of performance in the work condition (Table 6).

Table 6 Coefficient results of regressions performed on significant correlations in the work condition.

Model	β	Standard error	t	p
Income/mathematics	5.61	< 0.01	2.47	0.02*
GPA/memory	− 0.04	0.02	− 2.01	0.04*
GPA/mathematics	− 0.08	0.30	− 2.92	0.01*
GPA/Reading Span Test score	− 0.04	0.02	− 2.78	0.01*

Note: *p <  0.05.

While the results confirmed that socio-economic status and academic performance were negative predictors, the power of the predictors is weak. Socio-economic status only predicts performance on a single test (mathematics). Socio-economic status explains 11% of the scores. While general academic performance is a stronger negative predictor as it contributes to four cognitive tests, it does not relate to the time taken on the mathematics test nor the time or scores of the logic tests. Even where academic performance is a predictor of performance on the cognitive tests, they do not predict a large number of the scores. Academic performance explains 8% of memory scores, 14% of mathematics scores, and 13% of working memory.

Correlations between participants’ academic performance (GPA) and age and socio-economic status were conducted within the play condition. Results show that only GPA and performance on the mathematics test are significantly, but negatively, correlated within the play condition. To determine the nature of the correlation and whether academic performance can be used as a predictor of performance on the mathematics test in the play condition, a simple regression was performed on the single significant correlation.

The GPA significantly predicted mathematics scores in the play condition. GPA also explained a significant proportion of variance in mathematics scores (R² = 0.27, F(31) = 11.16, p <  0.01). The negative relationship indicates that a higher GPA will result in a lower performance on mathematics scores in the play condition.

The results from the correlation and regression disprove the hypothesis that age and socio-economic status are predictors of performance in the play condition. Academic performance is a predictor of performance on the mathematics test in the play condition, but negatively.

Discussion

Findings for research question one support the hypothesis that adolescent males perform better while playing when completing memory and reading tasks, but they reject the hypothesis on the other tests because adolescents do better in the work condition when completing logical and mathematical reasoning tests with no difference in the time taken in either condition for completing the mathematics test.

The Cohen's d test indicates a substantial practical significance between work and play conditions with acceptable effect sizes. Between the two groups, the effect size on all of the tests demonstrates that the treatment conditions had a substantial effect. The large effect sizes point to the importance of play in the role of cognition of adolescents.

The differences between the performances on the test were unexpected. It is probable that play would either have a total effect on cognition or no effect on cognition in adolescents. Explaining the differences between the tests, however, may give insight into how play is interacting with different cognitive domains.

In his seminal article on play, Vygotsky argues that the primary purpose of imaginative play in children is the acquisition of language. A child is denied access to something he wants (e.g. a weapon) and therefore he or she uses an object as a representation of his or her want (e.g. a stick). It is by subsuming an object's use through renaming (e.g. ‘this stick is now a gun’) that a child begins to be able to abstract meaning of the names of words from the words themselves. This transition is the beginning of abstract language use and cultural concepts (e.g. justice).

Play and language formation have a relationship in early childhood (Vygotsky, 1976). Because of this early relationship, it is plausible that activating the centres of the brain that control play would also activate the centres of the brain related to language formation. Vygotsky (1931) argued that fantasy is play that is not acted out. It appears that this statement may be true because of the residual effect of play on language tasks.

This position also helps support Erikson (1991) and Marcia (1966) theories about identity formation. The development of linguistic and memory ability could be a skill within an adolescent's zone of proximal development during identity formation because identity formation involves verbal ability as the process involves compromise, sense-making of other people's intents, and cultural reasoning and are related to verbal competence (Piaget, 1976b; Rogoff, 2003; Vygotsky, 1976, 1978).

The findings of research question two support the hypothesis that there would be no differences in performance due to the participant's age. In each of the cognitive tests, there were significant differences as stated with the findings and discussion of research question one. When examining differences in performance based solely on age and not treatment group, the findings supported the hypothesis and the hypothesis is further supported because there were no significant interactions between age group and treatment group. This result was not entirely unexpected. The current research into cognition postulates that cognition in adolescents begins to set around the age of 14 (Sousa, 2005). It is not surprising that the two selected age groups (14–16 and 17–19) had no differences in performance.

Because cognition in these two groups is similar, the effect of play on both of these groups should also be similar. An examination of the interaction graphs show that in each of the cases the directional change between work and play was the same for each group on each test, as found in research question one. It is important that the changes went in the same direction because they indicate a similar effect across participant groups. The differences were due to treatment group, but not age.

Play has been shown to be a facilitator in this process of children accessing the highest level of their zones of proximal development. Since play facilitates learning in the zone of proximal development, differences in cognitive scores during play activities would be a clear indication that the skill is within zone of proximal development in the adolescent. A lack of differences in age and performance demonstrates in this study that, due to age, the participants did not access the zone of proximal development differently. This position about the findings is confirmed by the other research into adolescent development. Since the adolescents participating had already achieved typical development by the time they participated in the study, there would be no differences due to play, as related to age. Instead, these findings confirm that play, at least in the domains of memory and reading, have an overall effect on performance for the age groups tested.

The findings of research question three both support and reject the hypothesis. In the work condition, income predicted mathematics scores; and GPA negatively predicted memory scores and Reading Span Test scores. Income also significantly explains variance in mathematics scores. Likewise, GPA significantly explains variance in memory scores and Reading Span Test scores. The rest of the correlations between the selected factors and tests were not significant in the work condition. In the play condition, only GPA negatively predicted performance on the mathematics test and significantly explains variance in mathematics scores in the play condition. No other factors significantly predicted performance in the play condition.

The results of the regressions in the work condition that demonstrate the relationship between GPA and performance on the memory, the mathematical reasoning, and the Reading Span tests are not congruent with an expected result. The significant inverse relationship between participants’ GPA and performance on the cognitive tests could be explained using the theories of fluid and crystallised intelligences (Cattell, 1963, 1971, 1987). The theory of fluid versus crystallised knowledge posits that intelligence contains two different elements. Fluid intelligence is the ability to process and cope with new problems and information. Crystallised knowledge is information about previous solutions to problems that is applied to new situations through analogy.

What may have occurred in this study is that when a student has a higher GPA, he ‘knows’ how to solve certain types of problems; however, knowing is at the cost of being fluid. As the students learn, new patterns become more difficult. The cognitive tests are a type of problem that students do not often encounter. For example, most mathematical education focuses on how to use equations to solve types of problems, but not on the logic of mathematics (Ben-Zeev, 1996). Therefore, a test of mathematical logic would be an unfamiliar problem. The fact the student is trained in a certain type of mathematical skill (e.g. using equations) suffers in other ways (i.e. being unable to determine mathematical patterns). The development of Raven's Progressive Matrices and similar tests were designed as a method of testing fluid intelligence. The research into mathematics education supports the given interpretation (Carroll, 1993, 1996).

Furthermore, GPA is not necessarily a measure of ability. Rather, the type of work that goes into the creation of a GPA includes many dynamics including work ethic, social ability, locus of control, poverty, access to sources of help, and standards set by teachers and curriculum. The number series test focuses solely on mathematical reasoning ability separated from these other factors that impact grades. It could stand to reason that as a student develops these other skills related to good academic performance, it comes at the cost of raw ability.

The inability of GPA to predict most cognitive scores in the play condition may derive from play acting as an equaliser. The relationship between play and the zone of proximal development may have had an effect. If play allows learners to access a higher level of their zone of proximal development, it would mean that participants engaged in play would have a greater impact than those with a higher GPA. Therefore, play would create a more level performance between the two groups, even though both did better in the play condition on the tests of memory and reading. GPA may still have an effect on mathematical reasoning because of the intensity in which these skills are actively taught in the school environment whereas reading skills are usually more passively taught, particularly in the schools tested.

Sara Smilansky (1990) found in her research with children in poverty that play was able-related to academic performance. The better children were at play, the better they performed in school. She attributed this difference to play teaching the social skill necessary for success in a school environment. However, play may have allowed the participants to access the highest portion of their zones of proximal development, which would level out performance. At this time, the issue needs to be explored further to determine how play, cognitive function, and academic performance are related.

Limitations and recommendations

The program used for the study creates the first major limitation to the study. How adolescents engage with video games and programs is still being investigated (Gee, 2003). This study proposed to look at how adolescents processed cognitive information in a play setting. Because the testing was done through a computer programmer and no data were taken by the proctors as to the affective physical responses of the participants, it is impossible to determine whether the participants were truly engaged in play.

Time creates another limitation in the study. The lengths of time of the two testing conditions were different. The length of the play condition program may have created a fatigue that did not exist in the work condition. To compensate for these effects, the two programs should have had a tighter time variance. The time difference creates issues in drawing conclusions comparing the two conditions, beyond fatigue. A non-uniform time for completion raises the question of engagement time (being given time respites in the play condition) affecting the results. Also, some participants may have felt pressure to rush when seeing other participants finish a treatment condition earlier than they did, regardless of the information given to them in the research protocol about different testing times.

A cultural limitation of the study relates to the presentation of the research materials. Modern America is much more technologically savvy than when the last major studies of this type were created (mid-1990s). Adolescents are immersed and engaged in technology in a way that is unparalleled in human history. The advantage of this cultural phenomenon is that the use of a computer program for testing would not cause undue stress or change in performance. The disadvantage is that adolescents now have an expectation of technology that did not exist before. Owing to practical and financial concerns, only a certain type of video game could be created for the play condition. These types of games are still fairly popular online, but it is highly possible that the participants did not fully engage with the play condition. Games that involve fluid actions and higher resolution graphics certainly would have more appeal to the participating group, but the creation of such a program would have been outside the scope of this study.

The first step would be to replicate this study to correct for methodological limitations that impact the clarity of the findings. An initial replication would be with a similar population, but reversing the order of the presentation of the cognitive tests. This replication would clear the issue surrounding the potential for fatigue causing differences in the logical and mathematical reasoning tests. This replication would also decrease the amount of ‘game’ to make the testing conditions more equal in requisite completion time.

Age is an important factor to consider with adolescent play. The research literature suggests that the impact on a younger adolescent group would be greater. A second research step would be to replicate the study a second time with a similar population, but expand the range of ages tested. Gender was not addressed in this research study. As children develop into adolescents, gender creates more significant differences between persons. Another replication of this study including a mirrored female population would provide information as to whether gender is actually a dividing issue with play.

Once the gaps have been closed with a similar population, replications of this study with different populations (e.g. minority, public school student, disadvantaged, or disabled) need to be conducted. These replications would have to be theoretical replications. The current study was a theoretical rather than a methodological replication of Istomina's (1975) study and thus prone to similar methodological issues. The game would have to be adjusted to be relevant to the different populations tested. The cognitive tests may have to be adjusted for appropriateness as well. Once a large enough pool of replications has been conducted, the nature of cognitive performance and play can be more solidly affirmed.

The current study demonstrated the nature of play in a new population – adolescents. New research paradigms could explore play as it relates to development across the entire life-span, not just early childhood or adolescence. Research is necessary to determine whether the effects found in this study carry over into an adult population or whether the nature of play in adults is radically different. Research could also be conducted with geriatric population to determine whether play serves a developmental purpose in slowing down the effects of cognitive degeneration. Play is not just the ‘the germinal leaves of all later life’ (Froebel, 1887, p. 55), but a lifelong endeavour.

Additional information

Notes on contributors

Ian Scheu

Ian Scheu is a recent doctoral graduate from Virginia Commonwealth University. His focus of study is adolescent play and development and how it relates to curriculum development. Currently, he is working with a private company to develop electronic delivery methods for curriculum focusing on teaching humanitarian perspectives.

Yaoying Xu

Yaoying Xu's research has been centred on social aspects of children with culturally and linguistically diverse backgrounds. Her specific research interests involve culturally appropriate learning contexts for young English language learners, the impact of social interactions of young children on their school performance, empowering culturally diverse families of young children with disabilities, and linking assessment and intervention for infants and young children who are at risk for disabilities/delays. Dr Xu's current research projects include using technology to increase social interactions and school performance of young English language learners, a longitudinal study of early intervention for young children who were born premature and low birth weight, and a meta-analysis on early identification of learning disabilities among young English language learners.