This article was downloaded by: [University of Texas Libraries]On: 15 October 2014, At: 11:07Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

The Journal of Positive Psychology: Dedicated tofurthering research and promoting good practicePublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rpos20

The challenge–skill balance and antecedents of flow: Ameta-analytic investigationCarlton J. Fonga, Diana J. Zaleskib & Jennifer Kay Leachc

a Department of Educational Psychology, The University of Texas at Austin, One UniversityStation D5800, Austin, TX 78712, USAb Illinois State Board of Education, Springfield, IL, USAc The University of Texas at AustinPublished online: 15 Oct 2014.

To cite this article: Carlton J. Fong, Diana J. Zaleski & Jennifer Kay Leach (2014): The challenge–skill balance andantecedents of flow: A meta-analytic investigation, The Journal of Positive Psychology: Dedicated to furthering research andpromoting good practice, DOI: 10.1080/17439760.2014.967799

To link to this article: http://dx.doi.org/10.1080/17439760.2014.967799

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

The challenge–skill balance and antecedents of flow: A meta-analytic investigation

Carlton J. Fonga,1*, Diana J. Zaleskib and Jennifer Kay Leachc,2

aDepartment of Educational Psychology, The University of Texas at Austin, One University Station D5800, Austin, TX 78712, USA;bIllinois State Board of Education, Springfield, IL, USA; cThe University of Texas at Austin

(Received 17 January 2014; accepted 11 September 2014)

Flow is an intrinsically motivating state of consciousness characterized by simultaneous perception of high challengeand skill. The position that challenge–skill balance is the primary antecedent for achieving a flow state is unclear, andmore research is needed to examine its impact on flow within multiple domains. Therefore, a meta-analysis wasconducted on 28 studies examining the challenge–skill balance related to flow and intrinsic motivation in a variety ofcontexts. The results indicated that the relationship between challenge–skill balance and flow was moderate, and smallerwith intrinsic motivation. Moderator analyses revealed weaker correlations when individuals were from an individualisticculture, in work or education contexts, using experience sampling method, and self-reporting state flow vs. trait.Compared to other theorized antecedents, challenge–skill balance was a robust contributor to flow along with clear goalsand sense of control.

Keywords: flow; challenge–skill balance; antecedents; meta-analysis; intrinsic motivation

Csikszentmihalyi’s claim that

in flow, the demands of a situation match the individ-ual’s ability, and the individual is engaged fully in theact of doing the activity. In flow, the person loses self-consciousness and a sense of the passing of time andenters into a different level of experience. (2003, p. 38)

Most intuitively understand this phenomenon of being ‘inthe zone’ or ‘in flow’ – a state of total immersion andmerging of action and awareness (Beard & Hoy, 2010).This highly motivating state raises the question: Why aresome people highly committed to and engaged in activitieswithout obvious external rewards? Although others haveexplained this behavior (e.g. DeCharms, 1968; Deci &Ryan, 1980; White, 1959), Csikszentmihalyi describedthis ‘intrinsically’ motivated behavior as consisting of aflow state or optimal experience.

Flow is considered to be an optimal state associatedwith positive emotional, motivational, and cognitive expe-riences (Csikszentmihalyi, Abuhamdeh, & Nakamura,2005; Hektner, Schmidt, & Csikszentmihalyi, 2007;Waterman et al., 2003). Csikszentmihalyi (1975) definedoptimal experience or flow as a positive and intrinsicallymotivating state of consciousness associated withperception of high challenge and personal skills adequateto meet those challenges (see also Bakker, 2005;Csikszentmihalyi, Rathunde, & Whalen, 1993; Hodge,

Lonsdale, & Jackson, 2009). A large number of studieshave identified flow experiences in the lives of peoplefrom diverse cultural and economic backgrounds (seeCsikszentmihalyi & Csikszentmihalyi, 1988; Massimini &Delle Fave, 2000). Also, the importance of flow hasspread to fields such as education (e.g. Bassi & DellaFave, 2012) or work (e.g. Moneta, 2012) given that flowcan lead to greater concentration, determination, persis-tence, and motivation, which in turn contributes toincreased performance (see Aube, Brunelle, & Rousseau,2014).

Theoretically, flow should be related to enhanced per-formance for numerous reasons. First, flow is a highlyfunctional state, which should in itself foster higher per-formance. Second, individuals experiencing flow areintrinsically motivated to re-engage in future activities(Engeser & Rheinberg, 2008). In addition, in order toexperience flow again, there is greater desire to take onmore challenging tasks (Nakamura & Csikszentmihalyi,2005). Thus, flow could be understood as an internallymotivating force for achievement and enhanced perfor-mance. Not only is the idea of an optimal experientialstate an intriguing topic, but also deeper understanding offlow has the potential to raise productivity, to betterhuman life, and to foster life satisfaction and happinessacross the lifespan (Csikszentmihalyi, 1997). The conceptof flow has had a prominent status in the field of positive

*Corresponding author: Email: carlton.fong@utexas.edu1Department of Educational Administration, The University of Texas at Austin, One University Station D5400, Austin, TX 78712,USA.2Oregon State University, Academic Success Center, Corvallis, OR, USA.

© 2014 Taylor & Francis

The Journal of Positive Psychology, 2014http://dx.doi.org/10.1080/17439760.2014.967799

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

psychology, and research encourages maximization offlow experiences (Keller, Bless, Blomann, & Kleinbohl,2011); however, much debate exists regarding theexistence and strength of flow’s antecedents.

The concept of flow: the challenge–skill balance andother antecedents

The concept of flow has been difficult to define and op-erationalize (Lovoll & Vitterso, 2012). Csikzentmihalyi(1990) himself cautioned defining flow too precisely lestit break the spirit of this dynamic construct. Yet one ofthe most common and accepted conceptualizations offlow is ‘the balance between perceived challenges andperceived skills’ (Csikzentmihalyi, 2009, p. 398).Csikzentmihalyi argued that the challenge–skill balanceleads to the optimal experience and maintaining suchbalance in itself is intrinsically rewarding.

Flow’s dynamic structure of the perceived matchbetween high challenge and adequate personal skill hasbeen described by four channels of daily experience:flow (high challenge and high skill), boredom or relaxa-tion (low challenge and high skill), apathy (low chal-lenge and low skill), and anxiety (high challenge andlow skill) (Csikszentmihalyi, 1975; Csikszentmihalyi,et al., 1993; Deichter, 2011). Therefore, if an activity iseither very easy or very difficult in comparison to one’sskill level, the experience of flow will be weak. In thestate of flow, one feels optimally challenged and confi-dent. This has a strong functional aspect and explainswhy people in flow are committed to tasks despite thelack of foreseeable results. Csikzentmihalyi andNakamura (2010) further discussed how the ratio ofchallenges to skills should be around 50/50 for optimalexperience, and even a slight imbalance can induceanxiety and displeasure.

Previous research has indicated the centrality of thechallenge–skill balance to the induction of flow. In anexperimental study, Keller and Bless (2008) supportedthe challenge–skill balance by testing three conditions: abalanced condition vs. two controls of high challenge orlow challenge. Participants reported more positive sub-jective experiences and had higher performance in thebalanced condition compared to the control conditions.Moneta and Csikszentmihalyi (1996) measured the bal-ance between challenge and skill with an adolescentsample, using the experience sampling method (ESM).Across multiple contexts and domains, they found thatthe challenge–skill balance had a positive effect on ado-lescents’ perceptions of concentration, wishing to do theactivity, involvement, and happiness. However, thesefindings were not found within all contexts and domains,and across all dimensions of experience. For example,the challenge–skill balance may have a positive effect onone dimension of experience within one context and no

effect on others. A different context may yield differentresults. Similarly, in a later study, Moneta andCsikszentmihalyi (1999) showed that quite a significantamount (47%) of the variance in self-reported concentra-tion was explained by the balance of skills andchallenges.

On the other hand, there has also been a great dealof research that suggests that the challenge–skill balanceis not a salient predictor of flow experiences. Some stud-ies have shown that the challenge–skill balance explainsas little as 2–4% of the variance of emotional experience(Lovoll & Vitterso, 2012; Voelkl, 1990). Experimentalresearch also supports the greater importance of animbalance in challenge and skill compared to a balance(see Clarke & Haworth, 1994). For example, a study onchess players revealed that levels of enjoyment werehighest when playing better opponents compare to equal-ranked opponents (Abuhamdeh & Csikszentmihalyi,2009). Essentially, when perceived challenges werehigher than skills, the games were more enjoyable thanwhen the challenge matched one’s skills.

Other arguments have contested the original opera-tional definition of flow as a balance between skill andchallenge (Engeser & Rheinberg, 2008). One of the firstproblems is that people vary in the extent to which one’sskills and the perception of challenge are related.Furthermore, the construct of perceived challenge com-pounds both perceived difficulty and skill; for example,an easy task could be highly challenging because of alack of skill. Theoretically, this is a problematic issue;however, empirically, comparing the balance of chal-lenge-skill and difficulty-skill yielded no substantialdifferences (Pfister, 2002).

Another problematic issue to the challenge–skillbalance and flow relationship is that some people morefrequently experience flow when they are engaged inchallenging activities (Engeser & Rheinberg, 2008);therefore, an imbalance in skill and challenge is positedto have a greater association with flow. Empirically,Moneta and Csikszentmihalyi (1996) found that thechallenge–skill balance was not compatible with certainflow indicators or dimensions of experience such aswishing to do the activity and happiness. On the otherhand, other research has supported that relatively chal-lenging tasks were no more enjoyable than easy tasks(Haworth & Evans, 1995; Shernoff, Csikszentmihalyi,Schneider, & Shernoff, 2003).

Since flow’s original conception, Csikszentmihalyi(1990) has also identified eight other dimensions of theflow experience beyond the challenge–skill balance, withnine antecedents all together: (a) challenge–skill balanceor engaging in challenges that meet one’s current skilllevel; (b) action-awareness merging; (c) clear goals; (d)unambiguous feedback; (e) concentration on the task athand; (f) sense of control; (g) loss of self-consciousness

2 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

or self-awareness; (h) transformation of time or thedistorted sense of time; and (i) the autotelic experience(Kawabata & Mallet, 2011; Payne, Jackson, Noh, &Stine-Morrow, 2011). Kawabata and Mallet (2011)described the components as follows. The challenge–skillbalance refers to the perception that an activity’s chal-lenge is matched or balanced with one’s ability. Action-awareness merging is involvement in the flow activity toa point of spontaneity or automaticity. Clear goals referto one’s perception of the goals of the activity before orduring the activity. Unambiguous feedback refers to themonitoring of one’s behavior that provides immediateand clear feedback concerning the activity. Concentrationis the complete and intense sense of focus on the activityat hand. Sense of control refers to the perception thatone is able to respond to any challenge while engaged inthe activity. Loss of self-consciousness refers to the lackof concern about the perception of others. The transfor-mation of time involves a sense that time has passedeither faster or slower than normal. The autotelic experi-ence refers to the experience of the activity being intrin-sically rewarding and enjoyable, or that the task has apurpose in and of itself.

These nine dimensions do not necessarily occursimultaneously. Hypothesized by the Quinn Model ofFlow (Quinn, 2005), certain dimensions may be requiredin order to enter the flow state (i.e. challenge–skill bal-ance, clear goals, and unambiguous feedback), whileothers are necessary characteristics of being in a flowstate (i.e. concentration, merging of action and aware-ness, sense of control, loss of self-consciousness, andtransformation of time), or the result of the flow experi-ence (i.e. autotelic experience). These additional anteced-ents and components support how the challenge–skillbalance may not be the most salient contributor toachieving a flow state (Shin, 2006; Wang & Hsiao,2012), despite receiving the greatest attention amongconditions for entering flow according to the literature.

Intrinsic motivation, flow, and the challenge–skillbalance

Intrinsic motivation, the propensity to engage in a taskout of interest or enjoyment, for its own sake, or withoutany external incentive or reward (e.g. Ryan & Deci2000), has been shown to be highly related to flow(Csikszentmihalyi & LeFevre, 1989; Keller, Ringelhan,& Blomann, 2011; Jackson, 1995). By definition, flow isunderstood as an intrinsically motivating state; in fact,some researchers have coined flow to be a model ofintrinsic motivation (Keller & Bless, 2008). Moreover,individuals who experience a challenge–skill balance aremore likely to freely choose to reengage in activities, abehavioral indicator of intrinsic motivation. In an experi-mental paradigm, Keller, Ringelhan, et al. (2011) found

that compared with individuals not in flow, individualsin flow were more intrinsically motivated to perform afree-choice activity. They found that the degree towhich they indicated interest (self-reported measure ofintrinsic motivation) mediated the extent to which theyengaged in the activity (behavioral measure of intrinsicmotivation).

Self-determination theory (SDT; Deci & Ryan,1985), a prominent view of intrinsic motivation, hasbeen linked with flow. SDT posits that feelings of com-petence, autonomy, and relatedness undergird intrinsicmotivation, and research has supported the link betweenthese three determinants and flow (Kowal & Fortier,1999). In a study of Canadian swimmers, Kowal andFortier found that intrinsic motivation and two determi-nants, competence and relatedness, were significantlypositively correlated with flow and with challenge–skillbalance as well. Bassi and Della Fave (2012) argued thatoptimal challenge supports the self-determinationperspective given the competence need as a basis forintrinsic motivation.

Given the theoretical and empirical relationshipbetween intrinsic motivation, flow, and the challenge–skill balance, we also wanted to assess its magnitude anddirection in the present study. Inconsistent resultsreported in the above literature, issues in operationalizingchallenge and skill, and the alternate antecedent modelsof flow call for further understanding of how the chal-lenge and skill balance really predict flow experiences.In addition, a meta-analysis has yet to be conductedexamining this seminal yet debatable topic. Moreover,systematic variants or moderators to this relationshiphave not been assessed across a larger body of research.

Moderators to the challenge–skill balance

Additional variables may also differentially impact howthe challenge–skill balance influences flow experiences.In the present study, we systematically explored theoreti-cal and methodological factors that may moderate thisrelationship. First, individual differences, such asachievement motivation, have been found to moderatethis dynamic (Engeser & Rhineberg, 2008). For example,individuals with low need for achievement perceivemoderately difficult tasks as daunting. For the highlyachievement-motivated individuals, they prefer tasks ofmedium challenge, or when there is an optimal balanceof difficulty and skill. Similarly, Moneta andCsikszentmihalyi (1999) argued that individuals of highability or talent are expected to ‘express the closestapproximation to the theoretical model,’ that is, the chal-lenge–skill balance predicting flow (p. 630).Csikszentmihalyi (1975) even acknowledged the possi-bility of an autotelic personality. Autotelic individualsoften have greater curiosity about life, engaging in

The Journal of Positive Psychology 3

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

activity for their own sake rather driven by external pres-sure. This characteristic has obvious consequences totheir response to flow states and its antecedents.

Age

Demographic characteristics such as age may influencethe relationship between flow and the challenge–skillbalance. In a study comparing subjective experiences ofyounger and older individuals, results indicated that theolder participants were more alert and able to concentratethan younger participants (Prescott, Csikszentmihalyi, &Graef, 1981). With regard to domain, younger partici-pants were more relaxed in leisure settings such as thehome compared to the older group; whereas older partic-ipants were more interested and relaxed at work contextscompared to the younger group. One explanation couldbe the development of career challenges for the youngerparticipants and less enjoyment of leisure or recreationfor the older participants. Alternatively, some researchhas indicated that age may not differentiate the dynamicsof flow. Bye, Pushkar, and Conway (2007) revealed thatyounger traditional age college students had the samelevels of intrinsic and extrinsic motivation as oldernon-traditional age college students, suggesting the samedifficulty in experiencing flow across age groups.Although these study outcomes do not directly tap intothe flow construct, they provide evidence that age mightplay a moderating role in the relationship betweenchallenge–skill balance, flow, and intrinsic motivation.

Culture

Culture may play a moderating role in the experience offlow (Delle Fave, Massimini, & Bassi, 2011). Early criti-cism of the flow concept came from its supposedly biastoward Western culture as flow focused on more activeand goal-directed processes, suggesting that flow mayoperate differently among various cultures. For example,in a study comparing Chinese college students withGrade 12 students from the USA, Moneta (2004) founda cultural variation in which the Chinese students weremore motivated when there was an imbalance of chal-lenge and skill, favoring lower challenges. He suggestedthat it was partially due to the Chinese students internal-izing collectivist values. However, Csikszentmihalyi andCsikszentmihalyi (1988) argued that what causes flowmay differ from culture to culture, but the dynamics ofthe flow experience are universal.

Domain

Also, an important moderator to examine is whether thechallenge–skill balance relationship with flow variesdepending on domain or context. Given how flow has

been studied in numerous contexts, assessing whetherwork/academic contexts vs. leisure contexts is a criticalissue for applied researchers when examining flow andpractitioners who want to increase flow experiences.Csikszentmihalyi and LeFevre (1989) found that thegreat majority of adults were experiencing flow whenworking and not in leisure despite being more motivatedin leisure. Boredom and lack of engagement is a chronicissue in the workplace and classroom, and the applicabil-ity of flow to working and learning environments is diffi-cult given the compulsory nature of job and learningactivities (Kiili & Lainema, 2008; Marzalek, 2006;Shernoff et al., 2003). Other contextual factors such asenvironments that support autonomy or that aid in focus-ing attention or removing distractions can foster moreflow-related activities (Nakamura & Csikszentmihalyi,2005; Schmidt, Shernoff, & Csikszentmihalyi, 2007).Studies also assess flow during personal activities thatindividuals indicate are meaningful or salient to them intheir everyday experience.

Methodology

Lastly, there are methodological characteristics that wewant to examine as potential moderators. How researchershave formalized the challenge–skill balance has variedfrom study to study (see Moneta & Csikszentmihalyi,1999), and results revealed a differential impact on flowexperiences depending on how the skill-challenge balancevariable is calculated. For example, in a study with tal-ented high school students, Moneta and Csikszentmihalyi(1999) compared three methods of calculating theskill-challenge balance: cross-product, absolute difference,and quadratic effects following a rotation of the predictoraxes. Their results indicated that the cross-product and theabsolute difference models were preferable (determined bymodel fit).

Another methodological concern is how flow is oper-ationalized and measured (see Martin & Jackson, 2008).A study may use experience sampling method (ESM;see Csikszentmihalyi & Larson, 1987), which recordsmultiple temporal measurements of flow over a period oftime. More frequently, a single measurement is used suchas the Flow State Scale (Jackson & Marsh, 1996) orDispositional Flow Scale (Marsh & Jackson, 1999),which includes the nine antecedents of flow. Other self-reported measures include just one or two items assess-ing concentration or related topics. Given the range ofmethods to assess flow states, a moderator analysis mayfurther distinguish the validity of such techniques.

In addition, as described earlier as an autotelic per-sonality, flow can be conceptualized as a trait or a state(see Marsh & Jackson, 1999). Flow as a state involvesfeeling certain subjective experiences after engaging inan activity; however, flow as a trait, involves a more

4 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

enduring sense of flow, often measured by how often anindividual experiences flow. Whether the challenge–skillbalance is more strongly related to flow as a state or traitis both a theoretical and methodological concern.

The present study

Over 30 years of research has accumulated on theconstruct of flow across a variety of domains (seeCsikszentmihalyi, 1990). Flow theory posits that intrin-sic motivation peaks in activities characterized by thesimultaneous perception of high challenge and skill. Inparticular, the challenge–skill balance hypothesis offlow theory has been a center of much debate withempirical evidence supporting both sides (see Engeser& Rhineberg, 2008). Pockets of research have con-cluded that the subjectively perceived fit between thechallenge of an activity and the skills of the individualis the most important prerequisite to experiencing flow(e.g. Schiefele & Raabe, 2011). Therefore, a meta-analysis on the relationship between the challenge–skillbalance and flow is not only timely, but also essentialin empirically assessing the overall theoretical basis ofthis important flow construct, its relation to intrinsicmotivation, and the moderators that influence theserelationships. Second, assessing how strongly thechallenge–skill balance relates to flow in comparison toother factors (i.e. nine-factor model) was measured.Lastly, we also examined the relationship between thechallenge–skill balance and intrinsic motivation.

Method

The following section describes the procedures used toconduct this meta-analysis, including subsectionsaddressing study inclusion criteria, literature search andinformation retrieval, coding procedures, effect sizecalculations, data integration, search outcomes, andmoderator analyses.

Literature search procedures

Studies were collected from a wide variety of sourcesand included search strategies meant to uncover bothpublished and unpublished research. In order to locatethe most exhaustive set of studies, we searched ERIC,PsycINFO, Proquest Dissertation and Theses Full Text,Social Science Citation Index, and Google Scholar elec-tronic databases using a broad array of subject termsincluding ‘flow’ and ‘optimal experience,’ while exclud-ing keywords ‘cash flow,’ ‘optic flow,’ and ‘blood flow’to reduce the number of irrelevant results. The referencesections of relevant documents were examined to deter-mine if any cited works might be relevant to our topic.In addition, Social Sciences Citation Index was searched

for documents that had cited several seminal works onflow: Csikszntmihalyi, 1975, 1990. These searches com-bined located a total of 355 unique, potentially relevantdocuments.

Each title and abstract was examined by the authors.If the abstract provided and indicated that the documentcontained data relevant to the relationship on flow andthe challenge–skill balance, the full document wasobtained for further examination.

Criteria for including studies

To be included in the meta-analysis, a study wasrequired to meet several criteria. First, studies need tohave reported data to derive the bivariate relationshipbetween the challenge and skill balance and a measureof flow or intrinsic motivation. Many studies haveincluded measures of both perceived skill and challenge,but did not calculate a match or balance between thetwo; these studies were not included (e.g. Abuhamdeh &Csikszentmihalyi, 2012). Research studies conducted inany context with participants of any age were included.

Information retrieved from studies

Numerous different characteristics of each study will beincluded in our data. These characteristics encompassedsix broad distinctions among studies: (a) publication sta-tus (published or unpublished); (b) the flow variable(how flow was measured and/or calculated); (c) thedomain (work/education-related activities, leisure activi-ties, or self-selected personally salient activities relatedto one’s identity; (e) the sample characteristics (age andcountry of origin); (f ) the measure of the challenge–skillbalance; and (g) the estimate of the relationship.

Methods of data integration

Before conducting any statistical integration of the effectsizes, the number of positive and negative effects wascounted. Next, the range of estimated relationships wascalculated. We examined the distribution of sample sizesand effect sizes to determine whether any studies con-tained statistical outliers. Grubbs’s (1950) test wasapplied and if outliers were identified, these values wereset at the value of their next nearest neighbor.

Both published and unpublished studies wereincluded in the synthesis. There is still the possibilitythat not all studies investigating the relationship betweenflow and challenge–skill balance were obtained. There-fore, Duval and Tweedie’s (2000) trim-and-fill procedurewas employed. The trim-and-fill procedure tests whetherthe distribution of effect sizes used in the analyses wasconsistent with that expected if the estimates werenormally distributed.

The Journal of Positive Psychology 5

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Effect size calculation

We collected correlation coefficients between challenge–skill balance and flow (often represented by r or thePearson product moment coefficient). When only meansand standard deviations were provided for a flow groupand a non-flow group, we estimated a correlation. Sincesome of the study’s sample sizes were small, and wewanted to improve normality, we conducted Fisher’sr-to-z transformations, a rather effective normalizingtransformation (see Meng, Rosenthal, & Rubin, 1992).Meta-analytic methods assume that the sampling distri-bution of the observed outcomes is (at least approxi-mately) normal. Weighted procedures were used tocalculate average effect sizes across all comparisons inwhich each independent effect size is first multiplied bythe inverse of its variance and then the sum of theseproducts is then divided by the sum of the inverses (seeCooper, Hedges, & Valentine, 2009). Also, 95% confi-dence intervals were calculated for average effects toassess significance.

One problem that arises in calculating average effectsizes involves deciding what constitutes an independentestimate of effect. Here, we used a shifting-unit-of-analysisapproach (Cooper, 1998). This approach involves codingas many effect sizes from each study that exist as a resultof variations in characteristics of the intervention, sample,setting, and outcomes within the study. However, whencalculating the overall effect size, the multiple effect sizeswere averaged to create a single effect size for each study.To calculate an overall effect size of the intervention, aweighted average of all effect sizes was computed andentered prior to analysis, so that the study will only con-tribute one effect to the assessment of the overall effects ofthe intervention on achievement. The shifting-unit-of-anal-ysis approach maximizes the amount of data from eachstudy without violating the assumption of independentdata points.

Moderator analyses

We conducted moderator analyses when tested usinghomogeneity analyses (Cooper et al., 2009). Effect sizesmay vary even if they estimate the same underlying pop-ulation value; therefore, homogeneity analyses wereneeded to determine whether sampling error aloneaccounted for this variance compared to the observedvariance caused by features of the studies. We testedhomogeneity of the observed set of effect sizes using awithin-class goodness-of-fit statistic (Qw), which followsa chi-square distribution with k − 1 degrees of freedom(k equals the number of effect sizes). A significant Qw

statistic suggests that sampling variation alone cannotadequately explain the variability in the effect size esti-mation; it follows that moderator variables should beexamined (Cooper, 1998). Similarly, the Qb statistic

indicates that average effect sizes vary between catego-ries of the moderator variables more than predicted bysampling error alone.

Analyses were conducted using both fixed- and ran-dom-error assumptions (Cooper et al., 2009). In a fixed-effects model of error, each effect size’s variance isassumed to reflect only sample error or differencesamong participants in the study. In a random-effectsmodel of error, a study-level variance component also isassumed to be an additional source of random variation.Due to the potential to over- or underestimated errorvariance in moderator analysis (Hedges & Vevea, 1998),we conducted all the analyses twice using both modelsof error in order for sensitivity analyses to examine theeffect of different assumptions (Greenhouse & Iyengar,1994). All statistical analyses were conducted using theComprehensive Meta-Analysis statistical software pack-age (Borenstein, Hedges, Higgins, & Rothstein, 2005).

Results

Overall findings

The literature search uncovered 28 studies that reporteda relationship between optimal challenge–skill balanceand flow and 18 studies that provided a relationshipbetween challenge–skill balance and intrinsic motivation.For flow, the 28 studies reported 37 effect sizes based on34 separate samples with a total N of 9620 participants.For the relationship between challenge–skill balance andintrinsic motivation, the pool of 18 studies reported 51effect sizes from 25 samples with a total N of 4270. Thecharacteristics of the included studies are reported inTable 1.

Regarding the pool of studies that assessed challenge–skill balance and flow, the studies were published betweenthe years 1996 and 2013. The sample sizes ranged from 51to 1231, with a median sample size of 270. The averagesample size was 277.9, with a standard deviation of 230.8,suggesting a normal distribution. Two included studies uti-lized ESM (Chen, 2000; Fullager, Knight, & Sovern,2013). Chen (2000) assessed three time points foreach participant, yielding 1215 momentary assessments.Fullager et al. (2013) measured 1031 momentary assess-ments. There were also no significant outliers among thecorrelations, so all were retained for analysis as reported.The effect sizes of the correlations (Fisher’s z) ranged from−0.25 to 1.42. They were all positive correlations, exceptfor one.

Under a fixed-error model, the overall relationshipbetween challenge–skill balance and flow (a normallydistributed and weighted correlation or Fisher’s z) was0.56 with a 95% CI from 0.55 to 0.58, indicating a mod-erate relationship (see Table 2a). Under a random-errormodel, the weighted average correlation was 0.52 with a

6 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Table

1.Characteristicsof

includ

edstud

ies.

Autho

r(year)

Typ

eof

document

Sam

ple

size

(ESM)

Age

Cou

ntry

Culture

Flow

measure

Flow

type

Balance

measure

Dom

ain

Correlatio

ncalculation

Fisher’sz

Bakker( 200

5)J

120

41the

Netherlands

ISurvey

Trait

Scale

Work/Edu

cSeparate

F:0.22

IM:0.23

605

19F:0.04

IM:0.07

BassiandDelle

Fave( 201

2)J

268

(343

2)17

Italy

IESM

Group

ing

.High/High

ratio

Work/Edu

cSeparate

IM:0.29

CejaandNavarro

( 2011)

J60

(698

)38

Spain

CESM

S×C

Work/Edu

cSeparate

IM:0.21

IM:0.31

S+C

IM:0.21

60IM

:0.37

S×C

IM:0.37

IM:0.43

S+C

IM:0.49

IM:0.55

ChanandAhern

( 199

9)J

80Over

18USA

ISurvey

State

Sub

scale

Work/Edu

cSub

score-

Global

F:1.29

Chen(200

0)D

405

31USA

IESM

State

Scale

Leisure

Separate

F:0.04

Collin

s( 200

6)D

5577

.64

USA

IESM

State

Sub

scale

Personal

Separate

F:1.04

Csikzentm

ihalyi

andFevre

( 198

9)J

78(343

2)36

.5USA

IESM

Group

ing

.High/High

ratio

Work/Edu

cSeparate

IM:0.20

Deichter(2011)

T18

639

Canada

ISurvey

Trait

Sub

scale

Work/Edu

cSub

score-

Global

F:0.78

Fullagaret

al.( 201

3)J

2721

.71

USA

IESM

State

|S–C|

Work/Edu

cSeparate

F:0.73

Hod

geet

al.( 200

9)J

5122

.9Canada

ISurvey

Trait

Sub

scale

Work/Edu

cSub

score-

Global

F:0.91

Jackson( 199

6)J

394

22USA

ISurvey

State

Sub

scale

Leisure

Sub

score-

Global

F:1.37

Kaw

abataandMallet(2011)

J63

520

.5Japan

CGroup

ing

State

Sub

scale

Leisure

Sub

score-

Global

F:0.95

413

20.4

F:0.91

KellerandBless

( 200

8)J

7220

Germany

ISurvey

.Sub

scale

Work/Edu

cSeparate

IM:0.46

KellerandBlomann( 200

8)J

7220

Germany

IGroup

ing

.High/High

ratio

Work/Edu

cMeans/SD

IM:0.42

IM:0.44

Keller,Bless

etal.( 2011)

J10

220

Germany

IGroup

ing

.High/High

ratio

Work/Edu

cMeans/SD

IM:0.23

IM:0.26

84IM

:0.81

IM:0.37

(Con

tinued)

The Journal of Positive Psychology 7

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Table

1.(Con

tinued).

Autho

r(year)

Type

ofdo

cument

Sam

ple

size

(ESM)

Age

Cou

ntry

Culture

Flow

measure

Flow

type

Balance

measure

Dom

ain

Correlatio

ncalculation

Fisher’sz

KiiliandLainema( 200

8)J

9220–30

Finland

ISurvey

State

Scale

Work/Edu

cSeparate

F:0.79

Kow

alandFortier( 199

9)J

203

36.4

Canada

ISurvey

.Sub

scale

Work/Edu

cSeparate

IM:0.60

Lee

( 200

5)J

262

20.02

Korea

CSurvey

.Sub

scale

Work/Edu

cSeparate

IM:0.31

Lee

andLaR

ose(200

7)J

388

19US

ISurvey

State

Median

split

Leisure

Separate

F:0.48

Lov

ollandVitterso

(201

2)J

64(698

)21

.2Norway

IGroup

ing

.High/High

ratio

Leisure

Means/SD

IM:−0.24

IM:−0.11

26IM

:0.26

(260

)23

.5IM

:0.12

IM:0.15

IM:0.19

Marsh

andJackson( 199

9)J

385

NA

Australia

ISurvey

State

Trait

Sub

scale

Leisure

Sub

score-

glob

alF:0.29

F:0.54

Marzalek( 200

6)D

1341

2813

USA

ISurvey

Trait

State

Sub

scale

Work/educ

Sub

score-

glob

alF:0.98

F:−0.25

Murica,

Gim

eno.

andGon

zales

( 200

6)J

413

13.7

Spain

CSurvey

Trait

Sub

scale

Leisure

Sub

score-

Global

F:1.02

IM:.48

Nah

etal.( 201

0)J

211

22USA

ISurvey

State

S+C

Leisure

Separate

F:0.16

Payne

etal.( 2011)

J19

772

.1USA

ISurvey

State

Sub

scale

Personal

Sub

score-

Global

F:0.76

Rezabek

( 199

4)D

108

20USA

IGroup

ing

.High/High

ratio

Work/Edu

cMeans/SD

IM:0.40

Rob

insonet

al.(201

2)J

30(349

)51

Ireland

IGroup

ing

.High/High

ratio

Personal

Means/SD

IM:0.25

IM:0.25

IM:0.26

Rod

rigu

ez-Sanchez

etal.( 2011)

J25

840

.2Spain

CSurvey

State

S×C

Work/Edu

cSeparate

F:0.63

Saville(200

6)D

3725

.5USA

ISurvey

Trait

Sub

scale

Work/Edu

cSeparate

IM:0.25

IM:0.56

IM:0.52

State

IM:0.26

IM:0.41

IM:0.54

Schiefele

andRaabe

( 2011)

J89

23.7

Germany

ISurvey

State

Twoitems

Work/Edu

cSeparate

F:0.51

Schuler

( 200

7)J

5725

Switzerland

ISurvey

State

Singleitem

Work/Edu

cSeparate

F:0.05

395

F:0.03

Schwartz

andWaterman

( 200

6)J

8718

.9USA

ISurvey

State

S+C

Personal

Separate

F:0.38

F:0.34

F:0.30

IM:0.16

IM:0.28

IM:0.19

(Con

tinued)

8 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Table

1.(Con

tinued).

Autho

r(year)

Type

ofdo

cument

Sam

ple

size

(ESM)

Age

Cou

ntry

Culture

Flow

measure

Flow

type

Balance

measure

Dom

ain

Correlatio

ncalculation

Fisher’sz

Shin( 200

6)J

525

18–22

Korea

CSurvey

State

S–C

Work/Edu

cSeparate

F:0.21

Sno

w( 201

0)D

176

Over

18USA

ISurvey

State

Sub

scale

Leisure

Separate

F:1.04

Stavrou

etal.( 200

7)J

220

19.95

Greece

CSurvey

State

Sub

scale

Leisure

Sub

score-

Global

F:1.16

vanSchaiket

al.(201

2)J

8325

Japan

CSurvey

State

Sub

scale

Work/Edu

cSub

score-

Global

F:0.55

Vlachop

oulous,Karageorghisand

Terry

( 200

0)J

1231

31.43

Eng

land

ISurvey

State

Sub

scale

Leisure

Sub

score-

Global

F:1.42

WangandHsiao

(201

2)J

122

Varied

USA

IGroup

ing

.High/High

ratio

Lesiure

Means/SD

IM:0.11

136

IM:0.17

102

IM:0.49

Waterman

etal.( 200

3)J

348

20USA

ISurvey

State

S+C

Personal

Separate

F:0.35

IM:0.41

270

F:0.44

IM:0.19

Waterman

etal.( 200

8)J

217

20USA

ISurvey

State

S+C

Personal

Separate

F:0.38

IM:0.54

202

F:0.28

IM:0.50

218

F:0.35

IM:0.39

Notes:J=Journalarticle,T:Master’sThesis,

D:DoctoralDissertation;

I:Individualistic

orindependentself-construal;C:Collectivistic

orinterdependent

self-construal;S=Skill,

C=Challeng

e;F:Flow,IM

:Intrinsicmotivation.

The Journal of Positive Psychology 9

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

95% CI from 0.38 to 0.62. Additionally, the tests of thedistribution of effect sizes revealed that the hypothesisthat the effects were estimating the same underlying pop-ulation could be rejected (Q(36) = 2282.4, p < 0.001), orthat the averaged correlation was greater than zero –potentially explained by the existence of moderators ofthis relationship. Next, trim-and-fill analyses were con-ducted. With both a fixed-effects model and a random-effects model, there was no evidence that effect sizesmight have been missing in the sample of studies.

Studies that assessed challenge–skill balance andintrinsic motivation were published between the years1989 and 2012. The sample sizes ranged from 26 to605, with a median sample size of 163. The averagesample size was 191.56, with a standard deviation of135.92, suggesting a normal distribution. Five of theincluded studies utilized experience sampling methodol-ogy (Bassi & Delle Fave, 2012; Ceja & Navarro, 2011;Csikszentmihalyi & Fevre, 1989; Lovoll & Vitterso,2012; Robinson, Kennedy, & Harmon, 2012). Theincluded studies widely varied in the number of momen-tary assessments: 5985 assessments (Bassi & Delle Fave,2012); 698 assessments (Ceja & Navarro, 2011); 3432assessments (Csikszentmihalyi & Fevre, 1989); 698assessments in Study 1 and 260 assessments in Study 2(Lovoll & Vitterso, 2012); 349 assessments (Robinsonet al., 2012). There were also no significant outliersamong the correlations, so all were retained for analysisas reported. The effect sizes of the correlations (Fisher’s z)ranged from −0.236 to 1.02. They were all positivecorrelations, except for two.

Under a fixed-error model, the overall relationshipbetween challenge–skill balance and intrinsic motivationwas 0.24 with a 95% CI from 0.22 to 0.25, indicating a

small relationship (see Table 2b). Under a random-errormodel, the weighted average correlation was 0.32 with a95% CI from 0.25 to 0.39. Additionally, the tests of thedistribution of effect sizes revealed that the hypothesisthat the effects were estimating the same underlying pop-ulation could be rejected (Q(24) = 526.94, p < 0.001).Next, trim-and-fill analyses revealed no evidence thateffect sizes might have been missing in the sample ofstudies.

Findings of the moderator analyses

Since the overall relationships between challenge–skillbalance and flow and intrinsic motivation were found tobe statistically heterogeneous, a series of moderator anal-yses were conducted to help explain variation amongeffect sizes. Table 3a and b presents the findings fromthe moderator analyses.

Publication status

First, we assessed the publication status (published vs.unpublished status) of the study report. For the flowmoderator analysis, 22 of the studies had been publishedas journal articles, and their results were compared to thefive studies that had appeared in dissertations, conferencepapers, and master theses. Under the fixed-error model,correlations from the unpublished reports, z = 0.43 (95%CI from 0.38 to 0.48), were just significantly differentfrom those from published sources, z = 0.58 (95% CIfrom 0.53 to 0.56), Q(1) = 37.50, p < 0.001. Under therandom-error model, there was no difference betweenpublished and unpublished reports, Q(1) = 0.04,p > 0.05.

Table 2a. Results of main analysis examining the relationship between flow and the challenge–skill balance.

95% confidence interval

k z Low estimate High estimate Q

Challenge–skill balance 37 2282.37***

Fixed model 0.56 0.55 0.58Random model 0.52 0.38 0.62

Note: All effect sizes were significantly different from 0 at a p < 0.001 value unless specified.***p < 0.001.

Table 2b. Results of main analysis examining the relationship between intrinsic motivation and the challenge–skill balance.

95% confidence interval

k z Low estimate High estimate Q

Challenge–skill balance 25 526.94***

Fixed model 0.24 0.22 0.25Random model 0.32 0.25 0.39

Note: All effect sizes were significantly different from 0 at a p < 0.001 value unless specified.***p < 0.001.

10 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

For the intrinsic motivation moderator analysis, pub-lished reports (k = 23, z = 0.24) had a significantly smal-ler relationship than unpublished reports (k = 2, z = 0.39)under fixed-error model (Q(1) = 9.17, p < 0.01). Therewere no differences under the random-error model, Q(1)= 1.37, p > 0.05.

Age

We next examined whether age would moderate thechallenge–skill balance and flow relationship. We codedage dichotomously and continuously. First, we formedtwo groups – age 30 and above, and below 30. Separat-ing the two groups at the 30-year mark followed previousliterature examining age groups and flow (e.g. Prescottet al., 1981). Second, we examined age as a continuousvariable to assess any linear trends, using the average ageof the sample if only a range was reported. Only twostudies did not report age characteristics of their samples.First, our findings indicated that for older participants(k = 7), the correlation between skill-challenge balanceand flow was z = 0.73 (95% CI = 0.71–0.74) comparedto z = 0.50 (95% CI = 0.48–0.52) for younger participants

(k = 27) under the fixed-error model. This comparisonwas significantly different (Q(1) = 260.14, p < 0.001).Under the random-error model, there was no significantdifference (Q(1) = 0.70, p > 0.05). Second, we conducteda meta-regression analysis to assess the impact of age asa continuous variable. Using maximum likelihood estima-tion, we found that age was only contributing a smallnon-significant linear effect of 0.007 (slope coefficient)on the relationship between skill-challenge balance andflow. So as age increases, the correlation was veryslightly increasing as well. The two age findings do notseem to reconcile together, suggesting a potential non-lin-ear relationship with age.

For the intrinsic motivation moderator analysis, therewere no significant differences between age groups infixed or random model of error. Similarly, the meta-regression indicated no significant age moderation onchallenge–skill balance and intrinsic motivation.

Cultural characteristics

We next assessed the moderation of country and culturalcharacteristics. We first compared samples from the USA

Table 3a. Results of moderator analyses for flow and challenge–skill balance.

95% confidence interval

k z Low estimate High estimate Qb

Publication status 37.50***

Published 29 0.58 (.51) 0.56 (0.36) 0.59 (0.63) (0.04)Unpublished 6 0.43 (0.54**) 0.38 (0.15) 0.48 (0.79)

Age 260.14***

Under 30 27 0.50 (0.48) 0.48 (0.35) 0.52 (0.60) (0.70)30 and over 7 0.73 (0.63**) 0.71 (0.27) 0.74 (0.83)

Country 88.76***

USA 18 0.47 (0.47) 0.45 (0.29) 0.50 (0.62) (0.47)Non-USA 17 0.61 (0.55) 0.60 (0.36) 0.63 (0.70)

Culture 59.14***

Individualistic 28 0.53 (0.46) 0.51 (0.30) 0.55 (0.61) (2.60)+

Collectivistic 7 0.64 (0.65) 0.62 (0.48) 0.66 (0.77)Domain 688.00***

Leisurea,c 11 0.73 (0.67) 0.71 (0.47) 0.74 (0.80) (5.10)+

Work/educa,b 16 0.32 (0.40) 0.29 (0.23) 0.36 (0.55)Personalb,c 8 0.40 (0.44) 0.36 (0.33) 0.44 (0.54)

Type of flow 39.28***

State 29 0.58 (0.48) 0.57 (0.34) 0.60 (0.62) (0.21)Trait 7 0.47 (0.56) 0.43 (0.29) 0.50 (0.75)

Measurement 182.96***

ESM 2 0.00^ (-0.31^) −0.10 (−0.79) 0.09 (0.41) (5.62)*

Single measure 34 0.58 (0.54) 0.57 (0.42) 0.60 (0.65)Balance measure 1118.60***

Global–subscore 15 0.75 (0.70) 0.74 (0.57) 0.76 (0.80) (16.29)***

Separate 20 0.30 (0.33) 0.28 (0.22) 0.33 (0.44)

Notes: All effect sizes were significantly different from 0 at a p < .001 value unless specified. Fixed-effects values are presented outside of parenthesesand random-effects values are within parentheses.*p > 0.05; **p < 0.01; ***p < 0.001; +p < 0.10; ^p > 0.05.Shared superscripts indicate significant pairwise comparisons:a and bpairwise comparison is significant under both fixed and random models of error.

The Journal of Positive Psychology 11

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

(k = 18) and outside the USA (k = 17). Results revealedthat samples from the USA had a correlation of z = 0.47(95% CI = 0.45–0.50) compared to samples outside theUSA, which had a correlation of z = 0.61 (95%CI = 0.60–0.63). The international difference was signifi-cant under the fixed-error model (Q(1) = 88.76,p < 0.001), but not under the random-error model (Q(1)= 0.47, p = 0.49). International samples appeared to havea stronger relationship between challenge-skill and flowcompared to USA samples.

To further understand this moderation, we codedeach sample as either a collectivistic or individualisticculture based on the country of origin. We determinedsuch categorizations by previous research on collectiv-ism–individualism (e.g. Hofstede, 2001). For example,countries such as the USA, Canada, the Netherlands,Finland, Switzerland, and Germany were consideredindividualistic (k = 28), whereas China, Japan, Korea,Greece, and Spain were coded as collectivistic (k = 7).Results indicated that collectivistic samples reported ahigher correlation of z = 0.64 (95% CI = 0.62–0.66) thanindividualist samples (z = 0.53; 95% CI = 0.51–0.55).However, this difference was only significant under thefixed-error model (Q(1) = 59.14, p < 0.001), but closeto marginally significant under the random-error model(Q(1) = 2.60, p = 0.098). Similarly, the correlationbetween challenge–skill balance and intrinsic motivationwas higher for collectivistic cultures (k = 3; z = 0.36)compared to individualistic cultures (k = 22; z = 0.20)

under fixed effects only (Q(1) = 80.71, p < 0.001).However, there were no significant differences betweenUS and non-US countries under both models of error forintrinsic motivation.

Domain

Next, the domain of each study was assessed as a mod-erator comparing studies of flow in a leisure context,work/education context, or a personal setting. For exam-ple, leisure contexts included online activity (e.g. surfingthe web), recreational sports, and video gaming. Work oreducation contexts involved job settings, school settings,professional sports, and taking exams. Flow in personalcontexts typically involved participants choosing a fewsalient or meaningful activities that take place throughouta given day. Overall for flow, there were significant dif-ferences among leisure (k = 11), work/education(k = 16), and personal settings (k = 8) under a fixed-errormodel (Q(2) = 688.00, p < 0.001) and marginally signifi-cant under a random-error model (Q(2) = 5.10,p = 0.083. Additional pairwise comparisons indicatedthat leisure contexts had a correlation z = 0.73 (95%CI = 0.71–0.74), and work/education contexts had acorrelation of z = 0.32 (95% CI = 0.29–0.35). This dif-ference was significant under both the fixed-error model(Q(1) = 650.33, p < 0.001) and the random-error model(Q(1) = 4.63, p = 0.031). Compared to leisure contexts,studies examining personal contexts had a weighted

Table 3b. Results of moderator analyses for intrinsic motivation and challenge–skill balance.

95% confidence interval

k z Low estimate High estimate Qb

Publication status 9.17**

Published 23 0.24 (0.32) 0.22 (0.24) 0.25 (0.39) (1.37)Unpublished 2 0.39 (0.39) 0.29 (0.29) 0.48 (0.48)

Age 1.06Under 30 19 0.23 (0.34) 0.21 (0.22) 0.25 (0.44) (0.66)30 and over 6 0.25 (0.28) 0.23 (0.22) 0.27 (0.35)

Country 0.38USA 10 0.24 (0.31) 0.22 (0.25) 0.27 (0.37) (0.13)Non-USA 15 0.23 (0.33) 0.22 (0.22) 0.35 (0.44)

Culture 80.71***

Individualistic 22 0.20 (0.30) 0.18 (0.23) 0.22 (0.36) (0.96)Collectivistic 3 0.36 (0.49) 0.33 (0.07) 0.38 (0.76)

Domain 44.22***

Leisurea 3 0.15 (0.36^) 0.11 (−0.15) 0.18 (0.72) (0.78)Work/educa 13 0.24 (0.33) 0.29 (0.25) 0.36 (0.40)Personala 9 0.29 (0.29) 0.26 (0.26) 0.31 (0.33)

Measurement 119.98***

ESM 6 0.18 (0.19) 0.16 (0.09) 0.20 (0.29) (6.20)*

Single measure 19 0.36 (0.37) 0.33 (0.27) 0.38 (0.46)

Notes: All effect sizes were significantly different from 0 at a p < 0.001 value unless specified. Fixed-effects values are presented outside of parenthe-ses and random-effects values are within parentheses.*p < 0.05; **p < 0.01; ***p < 0.001; ^p > 0.05.aShared superscripts indicate significant pairwise comparisons under fixed models of error.

12 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

average correlation of z = 0.40 (95% CI = 0.36–0.44).This was also significantly lower than leisure contextsunder both the fixed-error model (Q(1) = 311.33,p < 0.001) and the random-error model (Q(1) = 4.07,p = 0.044). For only the fixed-error model, resultsshowed that relationship between challenge–skill balanceand flow significantly varied by whether the activity wasin a work/education context or personal context, Q(1) =467.004, p < 0.01. In sum, the relationship betweenchallenge–skill balance and flow is strongest for leisurecontexts, then personal contexts, followed by work oreducation contexts.

The domain was a significant moderator for the rela-tionship between challenge–skill balance and intrinsicmotivation, but only under fixed effects (Q(2) = 44.22,p < 0.001). Personal activities (k = 9) had the highestcorrelation (z = 0.29), then work or education-relatedactivities (k = 13, z = 0.24), followed by leisure activities(k = 3; z = 0.15). Interestingly, leisure activities had thesmallest correlation out of the three domains, in contrastwith the high correlation in the previous analysis.

Methodological characteristics

We next considered the influence of various methodologi-cal characteristics present in the included studies. Unfortu-nately, some of the moderators that we believe werepractically and theoretically relevant to the flow literaturedid not report enough data or used methods too heteroge-neous to meaningfully aggregate. For example, how stud-ies calculated the challenge–skill balance varied toowidely. Studies measured the challenge–skill balance invarious ways: a single-item assessing balance (e.g.Schuler, 2007), a scale (e.g. Bakker, 2005), the product ofa challenge measure and skill measure (e.g. Rodriguez-Sanchez, Salanova, Cifre, & Schaufeli, 2011), an absolutedifference between a challenge measure and skill measure(e.g. Fullager et al., 2013), or a sum of challenge and skill(e.g. Waterman, Schwartz, & Conti, 2008). Although themost common form of operationalizing challenge–skillbalance was to use a separate scale of subscale, other stud-ies computed the balance as either a difference, sum, orproduct of challenge and skill. Due to the large amount ofheterogeneity, a formal moderator analysis could not beconducted in the present study. See Table 1 for anoverview of how flow was measured.

ESM vs. single measurements. In the database of studies,flow was operationalized using a single survey (eitherthe sum of flow antecedents or a separate flow measurewhich included items assessing engrossment or involve-ment) or multiple momentary assessments using ESM.One methodological concern was comparing whetherflow assessed using a survey differs from flow assessedusing ESM, which would provide a more ‘real-time’ T

able

4a.

Relations

betweenmod

erator

variablesforflow

andchalleng

e–skill

balance.

Mod

erator

variable

Age

Cou

ntry

Culture

Dom

ain

Type

offlow

Cou

ntry

χ2(2,N=35

)=0.06

p=0.80

Culture

χ2(1,N=35

)=0.21

χ2(1,N=36

)=9.27

p=0.64

p=0.00

2Dom

ain

χ2(2,N

=35

)=1.28

χ2(2,N=36

)=10

.32

χ2(2,N=36

)=3.86

p=0.53

p=0.00

6p=0.15

Typ

eof

flow

χ2(2,N

=36

)=0.28

χ2(1,N=36

)=3.50

χ2(1,N=36

)=0.05

χ2(2,N=36

)=4.40

p=0.87

p=0.06

p=0.82

p=0.11

Balance

measure

χ2(1,N=35

)=.578

χ2(1,N=35

)=1.37

χ2(1,N=35

)=2.97

χ2(2,N=35

)=4.96

χ2(1,N=36

)=2.56

p=0.45

p=0.24

p=0.09

p=0.08

p=0.11

The Journal of Positive Psychology 13

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

measurement of optimal experience. Flow research hasmoved toward ESM to assess momentary variation insubjectively reported experiences in order to examineflow. Some researchers argued that this is more accurateassessment of the flow state (Csikszentmihalyi & Larson,1987). There were only two studies that employed ESMwhen assessing the challenge–skill balance and flow. Ina 10-week longitudinal study, Fullagar et al. (2013) mea-sured flow during every practice session for 27 musi-cians, assessing their momentary subjective experiences.They found that the relationship between challenge–skillbalance and flow was quite robust with an averageweighted correlation z = 0.73. Interestingly, the averageweighted correlation in the second ESM study was muchlower. In Chen’s (2000) study, participants engaged inonline activities and web browsing with a repeated pop-up questionnaire assessing flow. The average weightedcorrelation z between challenge–skill balance and flowwas only z = 0.04. The variability in measurements inthe ESM studies suggests further research examiningESM vs. survey measurements of flow. Despite thepaucity of studies using ESM, we attempted to testthis moderator. Single measurements of flow and thechallenge–skill balance (k = 34) had a significantly largercorrelation than ESM correlations (k = 2) under bothfixed effects (Q(1) = 182.96, p < 0.001) and randomeffects (Q(1) = 5.62, p < 0.05).

Similarly, the average correlation between challenge–skill balance and intrinsic motivation was significantlysmaller for ESM studies (k = 6) than single-measurementstudies (k = 19). The average correlation for ESM studieswas 0.14 under fixed effects and 0.27 under randomeffects, whereas for single measurement studies, the aver-age correlation was 0.28 and 0.35, respectively.

Subscore–global vs. separate measurements. A fairnumber of the correlations between skill-challenge andflow were calculated by comparing a challenge–skill bal-ance subscore to a global flow score (k = 15). This typeof correlation contains some shared variance because thechallenge–skill balance is a part of the flow measure.Other studies compared two separate measurements offlow and challenge–skill balance, respectively (k = 20).Under fixed effects, moderator analyses revealed thatstudies with a subscore–global correlation (z = 0.75; 95%

CI = 0.74–0.76) had significantly higher correlations thanstudies that calculated separate measurements of flowand the skill-challenge balance (z = 0.30; 95%CI = 0.28–0.33). This was significant under both fixed-(Q(1) = 1211.86, p < 0.001) and random-error models(Q(1) = 16.02, p < 0.001).

State vs. trait flow. Flow has been understood as either astate of subjective experience measured after engaging inan activity or the frequency of activity-specific flow,enduring over time (Jackson, 1996). The differencebetween trait (k = 7) and state (k = 30) types of flow wasassessed. The correlation between flow state and chal-lenge–skill balance was significantly lower in trait flow(z = 0.47; 95% CI = 0.43–0.50) than in state flow(z = 0.58; 95% CI = 0.57–0.60) under the fixed model oferror (Q(1) = 39.11, p < 0.001). This was not significantunder the random model of error (Q(1) = 0.30, p = 0.58).

Relations between moderator variables

The moderator analyses indicated that many variablessignificantly influenced the relationship between chal-lenge–skill balance and flow. However, when moderatorsare tested individually, they might be confounded withone another (see Cooper, 1998; Patall, Cooper, &Robinson, 2008). For example, both study location ofUSA or non-USA as well as cultural characteristics ofindividualistic or collectivistic self-construal were signifi-cant moderators, but it is possible that non-US countriesare more likely to be collectivistic whereas the USA isindividualistic. Therefore, we assessed the pairwise rela-tionships between the following significant moderators:age, country, culture, domain, type of flow, and correla-tion calculation. Using effect sizes as the unit of analy-sis, we conducted a series of chi-square tests forpairwise comparisons of each of the moderators sincethey were all categorical. The results of all tests for chal-lenge–skill balance and flow are reported in Table 4aand for intrinsic motivation in Table 4b.

Using a conservative p-value of 0.01, we found onecluster of confounded variables involving country, cul-ture, domain, and type of flow for the flow moderators.One way to describe this cluster of confounded studyvariables is as follows: Compared to studies in non-US

Table 4b. Relations between moderator variables for intrinsic motivation and challenge–skill balance.

Moderator variable Culture Domain Measurement

Domain χ2 (2, N = 25) = 1.53p = 0.47

Measurement χ2 (1, N = 25) = 0.16 χ2 (2, N = 25) = 3.54p = 0.64 p = 0.17

Balance measure χ2 (1, N = 25) = 1.92 χ2 (2, N = 25) = 2.00 χ2 (1, N = 25) = 0.672p = 0.17 p = 0.37 p = 0.412

14 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

countries, studies in the USA were more likely to repre-sent a culture with individualistic self-construal, weremore likely to measure flow in leisure settings, and weremore likely to assess flow as a state, rather than a trait.For intrinsic motivation, there were no confounds for thesignificant moderators.

Comparing the challenge–skill balance to other flowantecedents

Among the studies that met our initial inclusion criteria,13 studies also measured the correlations between flowand the other eight following factors of the nine-factormodel of flow: merging of action and awareness, cleargoals, unambiguous feedback, concentration, sense ofcontrol, loss of self-consciousness, transformation oftime, and autotelic experience. In order to assess therelationship between challenge–skill balance and flowrelative to the other antecedents, we also meta-analyzedthe correlations between flow and each of the eight otherfactors. Results are presented in Table 5.

Comparing the within-sample correlations to eachother (see Meng et al., 1992), we found that thechallenge–skill balance is a relatively robust flow anteced-ent compared to the other eight factors. Under the fixedmodel of error, its correlation to flow (z = 0.76) was signif-icantly larger at the p < 0.001 level than merging of actionand awareness (z = 0.56; t = 10.42), concentration(z = 0.65; t = 5.83), loss of self-consciousness (z = 0.46;

t = 14.11), transformation of time (z = 0.33; t = 19.18),and autotelic experience (z = 0.59; t = 9.59). Thechallenge–skill balance was also larger than unambiguousfeedback (z = 0.72; t = 2.48, p = 0.02), but to a lesserdegree. The relationship between flow and having cleargoals (z = 0.75) and a sense of control (z = 0.79) were notsignificantly different from the challenge–skill balance. Itis worth noting that sense of control was the most highlycorrelated antecedent with flow.

In addition, we conducted another moderator analysisto assess whether flow measured as a state or trait mod-erated the correlations of all the antecedents. Overall,measured as states, flow correlations with most of theantecedents were significantly larger compared to beingmeasured as a trait only under fixed model of error. Thefew exceptions were that concentration and a loss ofself-consciousness were not significantly different fromeach other under both fixed and random models of error,and trait transformation of time (z = 0.45) was signifi-cantly higher than as a state (z = 0.27; Q = 37.5,p < 0.001) under the fixed model of error.

Discussion

The results indicated that the relationship betweenchallenge–skill balance and flow was moderate, and thisrelationship was influenced by a number of moderatingvariables. This moderately large correlation reveals that

Table 5. Results of comparing correlations between flow and its antecedents and assessing trait vs. state moderator analysis.

Overall Z 95% CI Flow as state 95% CI Flow as trait 95% CI Qb

(k = 13)1 (k = 9) (k = 5)

Challenge–skill balance 0.76 0.75, 0.78 0.79 0.78, 0.81 0.67 0.64, 0.7 56.64***

(0.70) (0.53, 0.82) (0.68) (0.39, 0.84) (0.69) (0.55, 0.79) (0.004)Merging of action & awarenessa 0.56 0.54, 0.58 0.59 0.57, 0.62 0.49 0.44, 0.53 17.31***

(0.54) (0.41, 0.65) (0.55) (0.34, 0.71) (0.50) (0.33, 0.64) (0.166)Clear goals 0.75 0.74, 0.77 0.79 0.78, 0.81 0.63 0.59, 0.66 94.07***

(0.69) (0.50, 0.82) (0.68) (0.38, 0.85) (.65) (0.49, 0.77) (0.059)Unambiguous feedbackb 0.72 0.70, 0.73 0.75 0.73, 0.76 0.64 0.60, 0.67 35.31***

(0.69) (0.53, 0.80) (0.68) (0.45, 0.83) (0.64) (0.42, 0.79) (0.106)Concentrationa 0.65 0.63, 0.67 0.65 0.63, 0.67 0.64 0.61, 0.67 0.225

(0.64) (0.53, 0.74) (0.64) (0.48, 0.76) (0.62) (0.41, 0.76) (0.05)Sense of control 0.79 0.78, 0.80 0.83 0.82, 0.85 0.66 0.62, 0.69 136.45***

(0.73) (0.55, 0.85) (0.74) (0.49, 0.88) (0.65) (0.43, 0.80) (0.453)Loss of self-consciousnessa 0.46 0.44, 0.49 0.46 0.43, 0.49 0.47 0.42, 0.51 0.22

(0.52) (0.39, 0.63) (0.50) (0.33, 0.64) (0.48) (0.25, 0.66) (0.03)Transformation of time a 0.33 0.30, 0.35 0.27 0.23, 0.30 0.45 0.41, 0.50 37.5***

(0.40) (0.24, 0.54) (0.30) (0.10, 0.47) (0.50) (0.26, 0.68) (1.79)Autotelic experiencea 0.59 0.57, 0.61 0.61 0.59, 0.64 0.55 0.51, 0.59 6.82**

(0.62) (0.59, 0.64) (0.63) (0.45, 0.76) (0.55) (0.45, 0.76) (0.39)

1Overall k does not add up to 13 because one study measured both trait and state flow with the same sample, so the independence assumption was notviolated.Note: Included studies were those indicated in Table in the column Correlation Calculation as “Subscore-Global.” All effect sizes were significantlydifferent from 0 at a p < 0.001 value unless specified. Fixed-effects values are presented outside of parentheses and random-effects values are withinparentheses**p < 0.01; ***p < 0.001.Superscripts denote statistically significant differences between challenge–skill balance and other antecedents.ap < 0.001; bp < 0.05.

The Journal of Positive Psychology 15

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

there is evidence for the fundamental notion that match-ing skill and challenge is an important flow indicator.However, the lack of a strong, robust relationship sup-ports the possibility of other theoretical antecedents offlow. The other eight theorized antecedents to flow var-ied in its relationship with flow relative to the strengthof challenge–skill balance. Additionally, the lack of rela-tionship between skill and challenge and the difficulty ofoperationalizing challenge (Engeser & Rhineberg, 2008)may explain this weaker than expected relationshipbetween challenge–skill balance and flow. Overall, thereis adequate support that matching skill and challenge isrobustly related with feelings of flow or optimal experi-ence. There is a similar finding with the intrinsic motiva-tion studies as well.

It is important to note that some of the meta-analyticfindings were based on a small number of effect sizesand studies. Caution should be taken when interpretingthe specific magnitude of the effects. Surprisingly, themajority of studies that reported correlational relationshipbetween challenge-skill and flow were single surveymeasurements of the related constructs with hardly anyexperimental designs where various levels of thechallenge–skill balance were manipulated. The studiesthat measured challenge–skill balance and intrinsic moti-vation represented a much more diverse set of designs,including ESM studies and experimental studies (seeTable 1). Other studies reported regression coefficientscontrolling for a diverse amount of variables but couldnot be statistically integrated together. Although thesestudies also investigated important questions related tothe present study, their data could not be practicallyaggregated in the meta-analysis.

Moderators

Driven by theoretical and methodological concerns in theliterature, moderator analyses revealed that the relation-ship between challenge–skill balance and flow varied byindividual characteristics, setting, and methodologicalcharacteristics. For example, published studies had a sig-nificantly higher averaged correlation compared tounpublished studies. There may be a bias in how thisconstruct is represented in the field.

Age

Assessing age dichotomously, we found that studies withsubjects aged 30 and over had a much stronger relation-ship between flow and the challenge–skill balance, butthis effect was only significant under fixed effects.Although flow in the majority of the literature seems totranscend any age group, our exploratory analysis sug-gested that as individuals get older, having their skillsmatch the level of challenge is more related with flow.

Perhaps as adults begin work, the initial excitement of anew job and career may dissipate as routine sets in.Wolfe and Kolb (1980) theorized that as individualsbecome more specialized in their fields, they experiencethe onset of routine and tasks becoming less challenging,and ultimately less satisfying. This is especially evidentwhen individuals reach the mid-life transition or ‘crisis’(see Brim, 1976). It follows that for older individuals toexperience flow, the importance of the challenge–skillbalance seems more salient. On the other hand, when weassessed age continuously in a meta-regression analysis,there was essentially no linear trend between age andcorrelation effect size. Either age really does have noeffect on how related the challenge–skill balance is withflow, or there is possibly a nonlinear relationship.Because of the uneven distribution of ages in our sampleof studies, creating equal groups to assess quadratic orhigher order function curves was unfeasible. Moreover,age did not moderate the relationship between chal-lenge–skill balance and intrinsic motivation, suggesting aconsistent influence of challenge–skill balance on intrin-sic motivation across the developmental lifespan.

Cultural characteristics

Because of the international scope of research on flow,we conducted two moderator analyses assessing whethercountry and self-construal affected the magnitude of thechallenge–skill balance and flow relationship. First, stud-ies with USA-based samples had a weaker relationshipbetween challenge–skill balance and flow compared tonon-USA countries. This followed previous researchsuch as a cross-cultural study comparing USA andItalian adolescents (Carli, Delle Fave, & Massimini,1987). They found that flow was much more congruentto challenge–skill balance in the Italian sample, but morediffused and less polarized in the USA sample. As dis-cussed earlier, this contrast was confounded by domain,type of flow, and self-construal. Self-construal contrastsindicated that the effect sizes were larger in samplesfrom collectivistic cultures compared to individualisticcultures. Our findings extended other cross-culturalresearch by adding other nations such as Spain andGreece, instead of limiting collectivistic nations to sim-ply China or other Asian countries. Previous researchactually showed that collectivistic nations might have amore prudent approach to challenges, and thereby biaspersonal skill in their optimal challenge/skill ratio com-pared to cultures with an independent or individualisticself-construal (Moneta, 2004). Our findings were con-trary to this: Collectivistic samples had a higher correla-tion with optimal balance and both outcomes of flowand intrinsic motivation compared to the individualisticsamples, which should theoretically be less challenge-avoidant. There is still little research in this area, and

16 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

future directions regarding this personal and culturalmoderation are suggested.

Domain

Given that flow–balance relationship is stronger in leisurecontexts, the weaker relationship in work/education envi-ronments and personal situations might be explained by avariety of reasons. Abuhamdeh and Csikszentmihalyi(2012) discussed that many everyday activities such asschool-related activities are not typically engaged in vol-untarily. Instead, students, out of obligation or necessity,might participate in academic or work-related activities(Graef, Csikszentmihalyi, & McManama Gianinno,1983). Therefore, optimal challenges do not seem to beas flow inducing in academic contexts than leisure con-texts (Bassi & Delle Fave, 2012). Csikszentmihalyi andLeFevre (1989) described that optimal experiences duringwork or school work involve low levels of happiness,freedom, and intrinsic motivation. In contrast, otherresearch pointed to how important one perceives theactivity to moderate how the challenge–skill balanceinfluences flow (Engeser & Rhineberg, 2008); however,assuming that work/education contexts are valued asmore important despite being less intrinsically motivating,the influence of task importance or value seems reversedaccording to the results. In addition, for work/educationactivities, the lower correlation between flow and chal-lenge–skill balance may be explained by the presence ofgreater challenge in this context. The high levels of chal-lenge reduce the likelihood for skill and challenge tomatch; instead of flow being induced, anxiety may bepresent. This demands greater attention on how to fittogether appropriately challenging assignments to theskill level of students and employees. However, even inthe typically highly extrinsically motivated classroom orworking contexts, it is possible for individuals to feelmore motivated while engaged in very challenging activi-ties (Csikszentmihalyi et al., 1993). How to create situa-tions in academic contexts that provide the types ofexperiences found in intrinsically motivated, goal-directedactivities is the challenge confronting intrinsic motivationresearch. Interestingly, leisure activities, such as web-browsing (see Shin, 2006), may be intrinsically motivat-ing, but the perceptions of challenge for such an activityare dubious. There is no built-in pursuit of goals;therefore, a sense of challenge is less relevant for taskabsorption or flow (Abuhamdeh & Csikszentmihalyi,2012).

Personal settings that consist of the activities salientto their identity or chosen by participants to be meaning-ful or important had significantly higher optimal balanceto flow and intrinsic motivation correlations than workand education activities. Practically, individuals mostlikely will choose personal activities that include both

leisure activities and work/education activities, whichmay explain this order of magnitude across domains.This includes both discretionary and obligatory activitiesthroughout a given day.

Methodological characteristics

The correlation between challenge–skill balance and flowwas higher when flow was measured as a state vs. trait.One potential explanation for this difference is with flowstate, measured by subjective experiences after a particu-lar task, individuals can more immediately, and arguablymore reliably, respond to flow antecedents. With flowtrait, individuals are responding to how often they expe-rience flow antecedents, essentially describing a moreenduring autotelic personality. Even in comparison withthe other flow antecedents, the trait correlations overallwere smaller or equal to state correlations because traitsare not exact, but based on situational factors and depen-dent on context (see Fridhandler, 1986). For example,there may be some contexts where an autotelic individ-ual may enjoy doing a task for its own sake, but alsoengage in other behavior out of duty or necessity. Suchtransient factors associated with trait measures mayexplain why trait correlations were smaller overall. Inter-estingly, concentration and loss of consciousness hadsimilar correlations when assessed as state vs. trait. Onenotable commonality between these two antecedents isthey are both aspects of being in flow, rather than a pre-cursor or outcome of flow based on the Quinn Model. Itfollows that these antecedents are equally likely to besalient as a trait, or the autotelic personality, compared toas a state. Csikzentmihalyi (1990) described autotelicindividuals as ‘more involved with everything aroundthem because they are fully immersed in the current oflife’ (p. 84). Concentration and loss of self-conscious-ness are also highly related as one is focused in the task,forgetting irrelevant concerns – in a way, to be truly con-centrated is to lose one’s sense of self. In addition, someresearchers argued that concentration would be correlatedwith both flow as a state or trait because flow’s definitionis so inextricably tied to intense-focused concentration.

Another interesting finding was that transformationof time was significantly higher when measured as a traitinstead of a state. In a similar way to loss of conscious-ness, the sense that time ‘flies by’ is a natural result ofbeing fully immersed in an activity. Thus, transformationof time seems more related to the autotelic personality(trait) vs. the feelings right after an activity.

One of the most robust findings was the calculationof the correlation between flow and optimal balance,using a global flow scale and one of its subscales ofchallenge-skill fit has a much larger correlation than ifthey are two separate, unrelated measures. Although weexpected shared variance between the global score and a

The Journal of Positive Psychology 17

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

subscale, the average weighted correlation was unexpect-edly high, and artificially inflated the average weightedeffect size. Examining the range of correlations betweenglobal scales and subscores, we found some studiesreporting low correlations (z = 0.55; van Schaik, Martin,& Vallance, 2012) and even negative correlations(z = −0.25; Marzalek, 2006). Not every inflated correla-tion appeared extremely high, suggesting adequate vari-ance in using this type of calculation. Also, many of theintercorrelations consist of one-ninth of the scale beingcorrelated to the global flow measure. However, in orderto more fairly compare apples to oranges, we examinedall the global to subscale studies that used all nine ante-cedents. In sum, the most conservative estimate of therelationship between challenge–skill balance and flow isusing the studies that used separate measurements ofboth; this results in a much smaller correlation ofz = 0.30, similar to the correlation of the challenge–skillbalance and intrinsic motivation, which suggests thatflow and intrinsic motivations are comparably relatedwith an optimal balance of challenge and skill.

Lastly, the use of ESM was compared with singlemeasurements of the challenge–skill balance. Whenassessing both intrinsic motivation and flow, ESM stud-ies on average reported lower correlations. This findingimplies that single measurements may inflate the influ-ence of the challenge–skill balance compared to momen-tary assessments, which may more accurately assess thelevels of flow and their antecedents.

Other antecedents

Compared to the other eight flow antecedents, challenge–skill balance remains a powerful precursor to flow. Thisfinding supported original conceptions of flow wherechallenge–skill balance must be met in order for flow tooccur. However, the challenge–skill balance as the solecatalyst of flow is also brought into question, and consid-eration for other antecedents is recommended. Results ofthe meta-analyses indicated that clear goals and sense ofcontrol were as powerful as challenge–skill balance assources of flow. Whereas the other antecedents (e.g. con-centration, merging of action and awareness, and feed-back) appear to be more cognitive of nature eitherinfluencing their thought processes or learning, sense ofcontrol and clear goals are more directly related tomotivation. Sense of control, or a sense of autonomy isone of the central components to SDT (Ryan & Deci,2000), and the importance of goals has been underscoredin a variety of motivation theories (e.g. Carver & Scheier,1982; Nicholls, 1975). Bassi and Delle Fave (2012)conducted a study of high school students using ESM toexamine optimal experience and self-determination. Theyfound that flow was associated with low levels ofself-determination, but that the quality of the experience

was better with moderate to high self-determination.Hodge et al. (2009) found that intrinsic motivation thatneeds satisfaction (competence, autonomy, and related-ness) were significant predictors of dispositional flow.Regarding goals, Novak, Hoffman, and Duhachek (2003)revealed that online users experienced greater flow whenthey were engaged in goal-directed activities vs. experi-ential activities, suggesting the importance of goals whenexperiencing flow. In sum, among the nine theorized flowantecedents, the challenge–skill balance is highly corre-lated with flow among other motivational antecedentssuch as control and clear goals.

Limitations

This study is not without limitations. Mainly, limitationsto generalizability are present. It is also important to notethat synthesis-generated evidence should not be inter-preted as supporting statements about causality (seeCooper, 1998). Thus, when exploratory moderators arefound to be associated with the effect sizes, these find-ings should be used to direct future researchers to exam-ine these factors. Finally, there were a number ofpotentially interesting and theoretically relevant variablesthat could not be examined as moderators. Gender aswell as other individual and personality variables wouldbe interesting to examine. Although we assessed agemoderation to some degree, a lack of data as well as abias toward older populations prevented further modera-tor analyses to measure curvilinear effect. Also, as notedearlier, there was a cluster of confounded moderator vari-ables. This prevents interpreting the effects of countryorigin from cultural characteristics, domain, and whetherflow was measured as a trait or state. In addition, someof the correlations in the study sample represented inter-correlations between the challenge–skill balance and flowmeasure, which caused some inflation in the effect sizes.

Assessing the differences between fixed- and ran-dom-error models, we found that most moderator analy-ses were significant under the fixed-error model, but notso in the random-error model. We caveat such findingsas limited in their generalizability of these particularmoderator variables (see Cooper, 1998).

Implications of flow antecedents in the real world

Every day, whether in work or school, in leisure activi-ties, or engaging in daily tasks, people prefer an optimalexperience of positive affect and attempt to avoid feel-ings of boredom, anxiety, and apathy. Across the life-span, the pursuit of happiness has become nearlyaxiomatic, and the importance of flow induction is inex-tricably a part of this ubiquitous endeavor (see Seligman& Csikszentmihalyi, 2000). How to create flow statesand to instill intrinsic motivation has been an important

18 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

question for researchers and practitioners as well as allindividuals who desire optimal experience, and ourmeta-analytic investigation shows that promoting achallenge–skill balance remains to be a robust contribu-tor. Assessing one’s set of personal skills or perceivedcompetence is critical as well as appropriately findingchallenges or scaffolding activities and tasks to match aprecise balance between the two. Other antecedents areimportant as well to engender flow: Clear goals and asense of control are also significant factors to consider.Goal-directed activities with clear instruction as well assupport and environments where the individual feelsautonomous and self-determined (e.g. providing choices)are motivating as well as flow-inducing (e.g. Patall et al.,2008; Su & Reeve, 2011).

Conclusion

When trying to create motivating and optimal experi-ences, what are some important factors to consider? Theresults of this meta-analysis suggested that finding abalance of challenge and skill is an important factor toconsider. Moreover, our findings indicated that this rela-tionship may be diminished with younger individuals,those with more of an individualistic self-construal, inwork or educational contexts. Overall, there are impor-tant and profound implications for the promotion ofhuman motivation, happiness, and thriving. Decision-makers might consider how to appropriately design chal-lenges, provide goals, and support autonomy to enhanceflow experiences for all individuals. We encourage morescholarship in this field to add greater validation andassessment of this important construct and how to bestinform practice in creating intrinsically motivating activi-ties in a diverse array of contexts.

AcknowledgmentsWe want to specially thank Erika A. Patall and Dale H. Schunkfor their helpful comments on previous versions of thismanuscript.

ReferencesAbuhamdeh, S., & Csikszentmihalyi, M. (2009). Intrinsic and

extrinsic motivational orientations in the competitivecontext: An examination of person-situation interactions.Journal of Personality, 77, 1615–1635.

Abuhamdeh, S., & Csikszentmihalyi, M. (2012). The impor-tance of challenge for the enjoyment of intrinsically moti-vated, goal-directed activities. Personality and SocialPsychology Bulletin, 38, 317–330.

Aubé, C., Brunelle, E., & Rousseau, V. (2014). Flow experi-ence and team performance: The role of team goal commit-ment and information exchange. Motivation and Emotion,38, 120–130.

Bakker, A. B. (2005). Flow among music teachers and theirstudents: The crossover of peak experiences. Journal ofVocational Behavior, 66, 26–44.

Bassi, M., & Delle Fave, A. (2012). Optimal experience andself-determination at school: Joining perspectives. Motiva-tion and Emotion, 36, 425–438.

Beard, K. S., & Hoy, W. K. (2010). The nature, meaning, andmeasure of teaching flow in elementary schools: A test ofrival hypotheses. Educational Administration Quarterly, 46,426–458.

Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H.(2005). Comprehensive meta-analysis (Version 2.1) [Com-puter software]. Englewood, NJ: BioStat.

Brim, O. G. (1976). Theories of the male mid-life crisis. TheCounseling Psychologist, 6, 2–9.

Bye, D., Pushkar, D., & Conway, M. (2007). Motivation, inter-est, and positive affect in traditional and nontraditionalundergraduate students. Adult Education Quarterly, 57,141–158.

Carli, M., Delle Fave, A., & Massimini, F. (1987). The qualityof experience in the flow channels: Comparison of Italianand U.S. students. In M. Csikszentmihalyi, & I. S.Csikszentmihalyi (Eds.), Optimal experience: Psychologicalstudies of flow in consciousness (pp. 288–306). New York,NY: Cambridge University Press.

Carver, C. S., & Scheier, M. F. (1982). Attention and self-regu-lation: A control-theory approach to human behavior. NewYork, NY: Springer-Verlag.

Ceja, L., & Navarro, J. (2011). Dynamic patterns of flow in theworkplace: Characterizing within-individual variabilityusing a complexity science approach. Journal of Organiza-tional Behavior, 32, 627–651.

Chan, T. S., & Ahern, T. C. (1999). Targeting motivation –Adapting flow theory to instructional design. Journal ofEducational Computing Research, 21, 151–163.

Chen, H. (2000). Exploring web users’ on-line optimal flowexperiences (ProQuest Dissertations and Theses). Syracuse,NY: Syracuse University. Retrieved from http://ezproxy.lib.utexas.edu/login?url=http://search.proquest.com/docview/304654237?accountid=7118

Clarke, S. G., & Haworth, J. T. (1994). “Flow” experience inthe daily lives of sixth-form college students. BritishJournal of Psychology, 85, 511–523.

Collins, A. L. (2006). Subjective well-bing in old age: Aninvestigation into the fole of flow and creativity. PhD Dis-sertation (ProQuest Dissertations and Theses). ChestnutHill, MA: Boston College.

Cooper, H. (1998). Synthesizing research: A guide for literaturereviews (3rd ed.). Thousand Oaks, CA: Sage.

Cooper, H., Hedges, L. V., & Valentine, J. C. (Eds.). (2009).The handbook of research synthesis and meta-analysis (2nded.). New York, NY: Russell Sage Foundation.

Csikszentmiahlyi, M., & LeFevre, J. (1989). Optimal experi-ence in work and leisure. Journal of Personality and SocialPsychology, 56, 815–822.

Csikszentmihalyi, M., & Larson, R. (1987). Validity and reli-ability of the experience-sampling method. The Journal ofNervous and Mental Disease, 175, 526–536.

Csikszentmihalyi, M. (1975/2000). Beyond boredom and anxiety.San Francisco: Jossey-Bass.

Csikszentmihalyi, M. (1990). Flow: The psychology of optimalexperience. New York, NY: Harper and Row.

Csikszentmihalyi, M. (1997). Finding flow: The psychology ofengagement with everyday life. New York, NY: Basic Books.

The Journal of Positive Psychology 19

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Csikszentmihalyi, M. (2003). Good business. New York, NY:Viking–Penguin Putnam.

Csikszentmihalyi, M. (2009). Flow. In S. J. Lopez (Ed.), Theencyclopedia of positive psychology, (Vol. 1, pp. 394–400).Malden, MA: Blackwell.

Csikszentmihalyi, M., Abuhamedeh, S., & Nakamura, J.(2005). Flow. In A. J. Elliot, & C. S. Dweck (Eds.), Hand-book of competence and motivation (pp. 598–608). NewYork, NY: The Guilford Press.

Csikszentmihalyi, M., & Csikszentmihalyi, I. S. (Eds.). (1988).Optimal experience: Psychological studies of flow in con-sciousness. New York, NY: Cambridge University Press.

Csikszentmihalyi, M., & Nakamura, J. (2010). Effortlessattention in everyday life: A systematic phenomenology. InB. Bruya (Ed.), Effortless attention: A new perspective inthe cognitive science of attention and action (pp. 179–190).Cambridge, MA: The MIT Press.

Csikszentmihalyi, M., Rathunde, K., & Whalen, S. (1993). Tal-ented teenagers: The roots of success and failure. NewYork, NY: Cambridge University Press.

DeCharms, R. (1968). Personal causation. New York, NY:Academic Press.

Deci, E. L., & Ryan, R. M. (1980). The empirical explorationof intrinsic motivational processes. In L. Berkowitz (Ed.),Advances in Experimental Social Psychology (pp. 39–80).New York, NY: Academic Press.

Deci, E. L., & Ryan, R. (1985). Intrinsic motivation and self-determination in human behavior. New York, NY: Plenum.

Deitcher, J. (2011). Facilitating flow at work: Analysis of thedispositional flow scale-2 in the workplace (ProQuestDissertations and Theses). Halifax, NS: Saint Mary’sUniversity.

Delle Fave, A., Massimini, F., & Bassi, M. (2011). Psychologicalselection and optimal experience across cultures. Dordrecht:Springer Science.

Duval, S., & Tweedie, R. (2000). A nonparametric “trim andfill” method of accounting for publication bias in meta-analysis. Journal of the American Statistical Association,95, 89–98.

Engeser, S., & Rheinberg, F. (2008). Flow, performance andmoderators of challenge-skill balance. Motivation andEmotion, 32, 158–172.

Fridhandler, B. M. (1986). Conceptual note on state, trait, andthe state-trait distinction. Journal of Personality and SocialPsychology, 50, 169–174.

Fullagar, C. J., Knight, P. A., & Sovern, H. S. (2013). Chal-lenge/skill balance, flow, and performance anxiety. AppliedPsychology: An International Review, 62, 236–259.

Graef, R., Csikszentmihalyi, M., & McManama Gianinno, S.(1983). Measuring intrinsic motivation in everyday life.Leisure Studies, 2, 155–168.

Greenhouse, J. B., & Iyengar, S. (1994). Sensitivity analysisand diagnostics. In H. Cooper, & L. V. Hedges (Eds.), Thehandbook of research synthesis (pp. 383–398). New York,NY: Russell Sage Foundation.

Grubbs, F. E. (1950). Sample criteria for testing outlying obser-vations. Journal of the American Statistical Association,21, 27–58.

Haworth, J., & Evans, S. (1995). Challenge, skill and positivesubjective states in the daily life of a sample of YTSstudents. Journal of Occupational and OrganizationalPsychology, 68, 109–121.

Hedges, L. V., & Vevea, J. L. (1998). Fixed and random effectsmodels in meta-analysis. Psychological Methods, 3, 486–504.

Hektner, J. M., Schmidt, J. A., & Csikszentmihalyi, M. (Eds.).(2007). Experience sampling method: Measuring the qual-ity of everyday life. Thousand Oaks, CA: Sage.

Hodge, K., Lonsdale, C., & Jackson, S. A. (2009). Athleteengagement in elite sport: An exploratory investigation ofantecedents and consequences. The Sport Psychologist, 23,186–202.

Hofstede, G. (2001). Culture’s consequences: Comparing val-ues, behaviors, institutions and organizations acrossnations. Thousand Oaks, CA: Sage.

Jackson, S. (1995). Factors influencing the occurrence offlow state in elite athletes. Journal of Applied SportsPsychology, 7, 138–166.

Jackson, S. (1996). Toward a conceptual understanding of theflow experience in elite athletes. Research Quarterly forExercise and Sport, 67, 76–90.

Jackson, S. A., & Marsh, H. W. (1996). Development and vali-dation of a scale to measure optimal experience: The flowstate scale. Journal of Sport and Exercise Psychology, 18,17–35.

Kawabata, M., & Mallett, C. J. (2011). Flow experience inphysical activity: Examination of the internal structure offlow from a process-related perspective. Motivation andEmotion, 35, 393–402.

Keller, J., & Bless, H. (2008). Flow and regulatory compatibility:An experimental approach to the flow model of intrinsicmotivation. Personality and Social Psychology Bulletin, 34,196–209.

Keller, J., Bless, H., Blomann, F., & Kleinböhl, D. (2011).Physiological aspects of flow experiences: Skills-demand-compatibility effects on heart rate variability and salivarycortisol. Journal of Experimental Social Psychology, 47,849–852.

Keller, J., Ringelhan, S., & Blomann, F. (2011). Does skills–demands compatibility result in intrinsic motivation? Exper-imental test of a basic notion proposed in the theory offlow-experiences. The Journal of Positive Psychology, 6,408–417.

Keller, J., & Blomann, F. (2008). Locus of control and the flowexperience: An experimental analysis. European Journal ofPersonality, 22, 589–607.

Kiili, K., & Lainema, T. (2008). Foundation for measuringengagement in educational games. Journal of InteractiveLearning Research, 19, 469–488.

Koval, J., & Fortier, M. S. (1999). Motivational determinantsof flow: Contributions from self-determination theory. TheJournal of Social Psychology, 139, 355–368.

Lee, E. (2005). The relationship of motivation and flow experi-ence to academic procrastination in university students. TheJournal of Genetic Psychology, 166, 5–15.

Lee, D., & LaRose, R. (2007). A socio-cognitive model ofvideo game usage. Journal of Broadcasting and ElectronicMedia, 51, 632–650.

Løvoll, H. S., & Vittersø, J. (2012). Can balance be boring? Acritique of the “Challenges Should Match Skills” hypothesisin flow theory. Social Indicators Research, 115, 117–136.

Marsh, H. W., & Jackson, S. A. (1999). Flow experiences insport: Construct validation of multidimensional, hierarchicalstate and trait response. Structural Equation Modeling, 6,343–371.

Marszalek, J. (2006). Computerized adaptive testing and theexperience of flow in examinees (Ph.D. dissertation). Uni-versity of Illinois at Urbana-Champaign, United States –Illinois. Dissertations & Theses: Full Text database. (Publi-cation No. AAT 3223665).

20 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Martin, A. J., & Jackson, S. A. (2008). Brief approaches toassessing task absorption and enhanced subjective experi-ence: Examining ‘short’ and ‘core’ flow in diverse perfor-mance domains. Motivation and Emotion, 32, 141–157.

Massimini, F., & Delle Fave, A. (2000). Individual develop-ment in a bio-cultural perspective. American Psychologist,55, 24–33.

Meng, X.-L., Rosenthal, R., & Rubin, D. B. (1992). Comparingcorrelated correlation coefficients. Psychological Bulletin,111, 172–175.

Moneta, G. B. (2004). The flow model of intrinsic motivationin chinese: Cultural and personal moderators. Journal ofHappiness Studies, 5, 181–217.

Moneta, G. B. (2012). Opportunity for creativity in the job as amoderator of the relation between trait intrinsic motivationand flow in work. Motivation and Emotion, 36, 491–503.

Moneta, G. B., & Csikszentmihalyi, M. (1996). The effect ofperceived challenges and skills on the quality of subjectiveexperience. Journal of Personality, 64, 275–310.doi:10.1111/1467-6494.ep9606164110.

Moneta, G. B., & Csikszentmihalyi, M. (1999). Models of con-centration in natural environments: A comparative approachbased on streams of experiential data. Social Behaviourand Personality, 27, 603–637.

Murica, J. A., Gimeno, E., & Gonzales, D. (2006). Motivacionautodeterminada y flujo disposicional en el deporte. Analesde Psicologia, 22, 310–317.

Nah, F. F., Eschenbrenner, B., DeWester, D., & Park, S. R.(2010). Impact of flow and brand equity in 3D virtualworlds. Journal of Database Management, 21, 69–89.

Nakamura, J., & Csikszentmihalyi, M. (2005). The concept offlow. In C. R. Snyder, & S. J. Lopez (Eds.), Handbook ofpositive psychology (pp. 89–105). Oxford: UniversityPress.

Nicholls, J. G. (1975). Causal attributions and other achieve-ment-related cognitions: Effects of task outcome, attainmentvalue, and sex. Journal of Personality and SocialPsychology, 31, 379–389.

Novak, T. P., Hoffman, D. L., & Duhachek, A. (2003). Theinfluence of goal-directed and experiential activities ononline flow experiences. Journal of Consumer Psychology,13, 3–16.

Patall, E. A., Cooper, H., & Robinson, J. C. (2008). The effectsof choice on intrinsic motivation and related outcomes: Ameta-analysis of research findings. Psychological Bulletin,134, 270–300.

Payne, B. R., Jackson, J. J., Noh, S. R., & Stine-Morrow, E. A.L. (2011). In the zone: Flow state in cognition in olderadults. Psychology and Aging, 26, 738–743.

Pfister, R. (2002). Flow in everyday life: Studies on the quad-rant model of flow experiencing and on the concept of theautotelic personality with the experience sampling method(ESM). Bern: Peter Lang.

Prescott, S., Csikszentmihalyi, M., & Graef, R. (1981). Envi-ronmental effects on cognitive and affective states: Theexperiential time sampling approach. Social Behavior andPersonality, 9, 23–32.

Quinn, R. W. (2005). Flow in knowledge work: High perfor-mance experience in the design of national security tech-nology. Administrative Science Quarterly, 50, 610–641.

Rezabek R. H. (1994). The relationships among measures ofintrinsic motivation, instructional design, and learning incomputer-based instruction (PhD dissertation). Norman,OK: The University of Oklahoma.

Robinson, K., Kennedy, N., & Harmon, D. (2012). The flowexperiences of people with chronic pain. Occupation, Par-ticipation, and Health, 32, 104–112.

Rodriguez-Sanchez, A., Salanova, M., Cifre, E., & Schaufeli,W. B. (2011). When good is good: A virtuous circle ofself-efficacy and flow at work among teachers. Revue dePsicologia Social, 26, 427–441.

Ryan, R. M., & Deci, E. L. (2000). Self-determination theoryand the facilitation of intrinsic motivation, social develop-ment, and well-being. American Psychologist, 55, 68–78.

Saville, J. F. (2006). Correlates of flow in post-collegiatewomen’s club rugby players (PhD dissertation). Philadelphia,PA: Temple University.

van Schaik, P., Martin, S., & Vallance, M. (2012). Measuringflow experience in an immersive virtual environment forcollaborative learning. Journal of Computer AssistedLearning, 28, 350–365.

Schiefele, U., & Raabe, A. (2011). Skills-demands compatibil-ity as a determinant of flow experience in an inductive rea-soning task. Psychological Reports, 109, 428–444.

Schmidt, J., Shernoff, D., & Csikszentmihalyi, M. (2007). Indi-vidual and situational factors related to the experience offlow in adolescence: A multilevel approach. In A. D. Ong,& M. van Dulman (Eds.), The handbook of methods inpositive psychology (pp. 542–558). Oxford: Oxford Univer-sity Press.

Schuler, J. (2007). Arousal of flow experience in a learningsetting and its effects on exam performance and affect.Zeitschrift fur Padogogische Pscyhologie, 21, 217–227.

Schwartz, S. J., & Waterman, A. S. (2006). Changing interests:A longitudinal study of intrinsic motivation for personallysalient activities. Journal of Research in Personality, 40,1119–1136.

Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positivepsychology: An introduction. American Psychologist, 55,5–14.

Shernoff, D. J., Csikszentmihalyi, M., Shneider, B., & Shernoff,E. S. (2003). Student engagement in high school classroomsfrom the perspective of flow theory. School PsychologyQuarterly, 18, 158–176.

Shin, N. (2006). Online learner’s ‘flow’ experience: An empiri-cal study. British Journal of Educational Technology, 37,705–720.

Snow, K. Y. (2010). Work relationships that flow: Examiningthe interpersonal flow experience, knowledge sharing, andorganizational commitment (ProQuest Dissertations andTheses). Claremont Graduate University. Retrieved fromhttp://ezproxy.lib.utexas.edu/login?url=http://search.proquest.com/docview/963995155?accountid=7118

Stavrou, N. A., Zervas, Y., Karteroliotis, K., & Jackson, S. A.(2007). Flow experience with athletes’ performance withreference to the orthogonal model of flow. The SportPsychologist, 21, 438–457.

Su, Y., & Reeve, J. (2011). A meta-analysis of the effectivenessof intervention programs designed to support autonomy.Educational Psychology Review, 23, 159–188.

Vlachopoulous, S. P., Karageorghis, C. I., & Terry, P. C.(2000). Hierarchical confirmatory factor analysis of theflow state scale in an exercise setting. Journal of SportsSciences, 18, 815–823.

Voelkl, J. E. (1990). The challenge skill ratio of daily experi-ences among adults residing in nursing homes. TherapeuticRecreational Journal, 24, 7–17.

The Journal of Positive Psychology 21

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4

Wang, L. C., & Hsiao, D. F. (2012). Antecedents of flow inretail store shopping. Journal of Retailing and ConsumerServices, 19, 381–389.

Waterman, A. S., Schwartz, S. J., & Conti, R. (2008). Theimplications of two conceptions of happiness (hedonicenjoyment and eudaimonia) for the understanding of intrin-sic motivation. Journal of Happiness Studies, 9, 41–79.

Waterman, A. S., Schwartz, S. J., Goldbacher, E., Green, H.,Miller, C., & Philip, S. (2003). Predicting the subjectiveexperience of intrinsic motivation: The roles of self-

determination, the balance of challenges and skills, andself-realization values. Personality and Social PsychologyBulletin, 29, 1447–1458.

White, R. W. (1959). Motivation reconsidered: The concept ofcompetence. Psychological Review, 66, 297–333.doi:10.1037/h0040934.

Wolfe, D. M., & Kolb, D. A. (1980). Beyond specialization:The quest for integration at midlife in work, family andcareer. In B. Derr (Ed.), New Frontiers in theory andresearch (pp. 239–280). New York, NY: Praeger.

22 C.J. Fong et al.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

exas

Lib

rari

es]

at 1

1:07

15

Oct

ober

201

4