Self-regulation and self-efficacy as mediators of achievement goals and leisure time physical activity: a proposed model

Purpose: The purpose of this study was to examine the mediating role of self-efficacy and self-regulation between achievement goals and leisure time physical activity among Turkish high school students. Material: Four hundred and sixty-four high school students (216 male; 248 female) enrolled in physical education classes voluntarily participated in the study. To assess whether the data fit the proposed model structural equation modelling was used. Results: Results revealed that self-efficacy mediated the effects of mastery-approach, mastery-avoidance, and performance-approach goals on leisure time physical activity. Self-regulation was not a mediator of achievement goals and did not predict leisure time physical activity. Conclusions: This study revealed the importance of self-efficacy in physical education to promote students’ leisure time physical activity.

Introduction 1 Extensive data supports the importance of moderate to vigorous physical activity (MVPA) among children and youth. The (former) National Association for Sport and Physical Education (NASPE) concludes that regular PA is vital to preventing childhood obesity [1] and the World Health Organization (WHO) lists regular PA as an important goal for combating obesity and obesity-related diseases [2]. Despite compelling data, children and youth frequently fall below the recommended ≥50% level of MVPA [3,4].
School physical education (PE) programs represent an important conduit for promoting PA. Given that long time outcome of PE is encouraging children and youth to engage in PA outside the gymnasium setting (i.e., during their leisure time), physical educators should understand factors contributing to PA outside of PE settings. Students must first want to engage in activity outside the school setting. Motivation represents an important contributor to promoting leisure time physical activity (LTPA). Appreciating students' motivations, cognitions and affects such as effort and self-efficacy may provide important knowledge and strategies for teachers to encourage PA in student leisure time. This study draws from self-regulation (SR) and achievement goal (AG) theoretical frameworks to examine the impact of high school students' AG, selfefficacy, effort and engagement in LTPA.
Self-Regulation Perhaps one of the most inclusive theories about learning is SR. While McBride & Xiang [5] identified the lack of a consistent operational definition of SR in the literature, most agree it encompasses a process whereby individuals "activate and sustain cognitions, affects and behaviors that are systematically oriented toward the attainment of personal goals" [6; p.10]. SR individuals are active; they dichotomous model) were task and ego. A task-oriented individual is self-referenced and motivated to develop competence and ability through learning or task mastery. Ego oriented learners focus on demonstrating superiority and outperforming others or show avoidance to being judged incompetent. Mixed research findings about relationships between the two AG motivational patterns led to the 2 x 2 model with four categories: masteryapproach (MAp; "I want to learn as much as possible"), mastery-avoidance (MAv; "I worry that I may not learn all that I possibly could"), performance-approach (PAp; "It is important for me to do better than other students"), and performance-avoidance (PAv; "I just want to avoid doing poorly") [ Finally, a PAv goal orientation is driven by a fear of failure and represents the least positive goal orientation. One factor linked to PAv orientations is perception of overly externally regulated environments [24].
Self-efficacy Bandura [25] defined self-efficacy (SE) as "beliefs in one's capabilities to organize and execute the courses of action required to produce given attainments" [p. 3]. When determining self-efficacy individuals judge their skills and capacities, then convert those skills into actions. Selfefficacy addresses beliefs about perceived competence or beliefs when approaching and performing tasks. Learners with high perceived self-efficacy levels tend to be more willing to engage in challenging tasks, work harder and persevere longer than their less efficacious counterparts.
Wigfield et al.
[10] present a profile of high selfefficacy that include setting ambitious goals, choosing challenging tasks, and being mindful of requisite strategies needed to complete thought-provoking activities. Early correlational studies found positive relationships between self-efficacy and SR [see 26 for a more in-depth review). Classroom results also reveal self-efficacy mediated the effects of AG on academic outcomes. Bong [27] identified strong support for subject specificity of self-efficacy and AG among 424 Korean middle and secondary school students. PAp goals and PAv goals recorded strong crosssubject associations, while mastery goals recorded the weakest relationships. They concluded that students with PAp or PAv goal orientations in one academic subject would likely pursue the same goal orientation in other subject areas.
In the PA setting, self-efficacy has repeatedly predicted both adoption and maintenance of PA [28]. Among a population of young healthy adults, Kwan and Bryan [29] reported those who responded more favorably to exercise had greater self-efficacy, and Dishman et al.
[30] recommended self-efficacy as a mediator variable.
Gao, Lochbaum, and Podlog [31] tested the mediating effect of self-efficacy among middle school students on targeted mastery AG and students' in-class PA. Students' self-efficacy was significantly and positively related to the four AG and perceived motivation climates in addition to PA levels in PE class. Similarly, Gao and colleagues [32] provided additional support for self-efficacy as a mediator finding that middle school students' self-efficacy fully mediated the effect of MAp goals on a fitness performance and partially mediated the effects of MAv and PAp goals.
Leisure Time Physical Activity Since adolescence may be a pivotal time for decisions about continued PA engagement, promoting LTPA takes on added importance. Aarnio, Winter, Peltonen, Kujala, & Kaprio [33] reported adolescents who actively engaged in PA were more likely to continue being physically active into adulthood. While research exploring the association of participation in PE and LTPA are mixed, one congruent link identified is the type of motivation students encounter [34]. Mastery-oriented and autonomy-supporting climates tend to foster greater intrinsic interest and future intentions to be physically active [35,36].
In summary, SR, AG orientations and self-efficacy represent important contributors to our understanding of learning and motivational behaviors in PE and PA settings. While empirical evidence provide links to self-efficacy, little information exists on the role self-efficacy and SR might play as mediators of AG and increased participation in LTPA. Cecchini-Estrada and Mendez-Gimenez [37] agree and provide support that AG orientations might yield implications for SR learning. Since PE is required in Turkey's high schools, SR and self-efficacy as mediators of LTPA could contribute to future interests in PA and health benefits accrued from such participation.
This study utilizes both SR and self-efficacy as mediators of AG and LTPA. Following the recommendation of Dishman et al. [30] and the impact of Gao and colleagues [32], we examine the mediating role of self-efficacy and SR between the 2 x 2 AG framework and LTPA among Turkish high school students. Specifically, this study asks: 1) Does the hypothesized model fit the data? 2) Do self-efficacy and SR mediate the effects of MAp, MAv & PAp and LTPA and, 3) Do SR and selfefficacy predict LTPA? Figure 1 presents the hypothesized model. Based on the previous studies we first hypothesized that approachoriented goals (MAp and PAp) would positively predict self-efficacy, while avoidance-oriented achievement goals (MAv and PAv) would negatively predict self-efficacy [e.g. 38]. Second, we hypothesized students' self-efficacy and SR would positively predict LTPA. Third we hypothesized that self-efficacy and SR would mediate the relationship between AG and LTPA. In line with earlier research [39,31] self-efficacy was hypothesized to mediate AG and LTPA. When experiencing success students may be more inclined to continue in that achievement behavior [40].

Participants.
Four hundred and sixty-four high school students (216 male; 248 female; age range = 14-17 years, (M age = 15.04 ± 0.62 years) enrolled in PE classes voluntarily participated in the study. Participants attended five public high schools located in a central southwestern Turkish city and were of middle to upper middle-class SES. All programs delivered traditional sport-based lessons such as basketball, soccer, volleyball and games to meet state curriculum guidelines. Students typically participated for two hours one day per week and classes typically were teacher-centered. Distribution of gender and grade level are presented in Table 1.
Research Design Prior to data collection permission was obtained from all participants and the Turkish Ministry of Education and Ethics Committee. During regularly scheduled PE classes, students responded to a questionnaire pack. Teachers were not present. Participants were informed there were no right or wrong answers, to respond as honestly as possible and participation in the research project (or not) would not impact their grade. Responses were kept confidential and took approximately 40 minutes to complete.

Variables and Measures
Questionnaires included demographic information and items assessing AG, SR, self-efficacy, and LTPA. Demographic information included name, age, gender, grade classification and course currently enrolled. Achievement Goals. This construct was assessed using the 2x2 Achievement Goal Scale [41] and adapted to Turkish by Ağbuğa [21]. The questionnaire consists of 21 items beginning with the stem "in my physical education classes…." followed by a 7-point Likert scale (1 = not at all true of me; 7 = very true of me). Item examples included "It is important for me to do better than other students" (PAp), "I just want to avoid doing poorly" (PAv), "I want to learn as much as possible" (MAp), and, "I worry that I may not learn all that I possibly could" (MAv).
Self-Regulation and Self-Efficacy. Related subscales from the Motivated Strategies for Learning Questionnaire [MSLQ; 42] was adapted for Turkish populations by Üredi [43] and confirmed by Erturan-İlker, Arslan and Demirhan [44] assessed students' SR and self-efficacy. The SR subscale consists of 9 items and the self-efficacy subscale consists of nine items. All items are followed by a 7-point Likert type scale (1 = not at all true of me; 7 = very true of me). Example items included "When work is hard I either give up or study only the easy parts" (SR) and "Compared with other students in this class I expect to do well" (self-efficacy).
Leisure Time Physical Activity (LTPA). We used the adapted Turkish short form [45] of the International Physical Activity Questionnaire [IPAQ ; 46]. It consists of 7 items that measure frequency, duration, and intensity of PA across the preceding seven days.
Statistical Analysis

01
Descriptive statistics for all variables were computed and Cronbach's alphas assessed the internal reliability of the multi-item subscales. Z scores for IPAQ scores were calculated. Pearson correlation analysis examined the correlations among all the variables used in the study. To assess whether the data fit the proposed model path analysis with the version 21 of the AMOS software [47]. In each analysis, we initially evaluated the multivariate normality of the data using Mardia's multivariate kurtosis coefficient.
Several indices were used to assess model fit. The overall fit of the model to the data employed the chisquare test. The standardized root means square residual (SRMR) and the root mean square error of approximation (RMSEA) were used as indicators of absolute fit. Values less than .08 and .06 respectively advocate a model with a good fit good fit between the proposed model and the data [48]. Comparative fit index (CFI) was to be less than .90 to indicate adequate model [48].

Preliminary Analyses
After initial data screening, three univariate outliers were removed from the dataset. Means, standard deviations, and alpha reliability coefficients for the variables are recorded in Table 2. The observed alpha coefficients indicated that the scales used in the study demonstrated acceptable internal reliability (i.e., a ≥ 0.70), except for PA, PAv goal, and SR.
Correlations among Variables Pearson's product moment correlation analysis results are shown in Table 3 below. Bivariate correlations suggest that all the significant relationships are positive.
Path Analysis The hypothesized path model was examined using the maximum likelihood method. The data in the measurement model did not display multivariate normality (Mardia's Multivariate kurtosis = 12.94). Multivariate normality implies that the sampling distributions of means of the   [49]. One approach to handling multivariate non-normal data set is the bootstrap technique [50] that was employed in all further analyses and the calculation of model statistics, parameters, and standard errors are derived from the bootstrap sample distribution. Because the AGs were interrelated their associated error terms were correlated, the indices of fit suggested the revised model adequately fit the data [X 2 (sd) = 2.55, p <0.05, CFI = 0.98, NFI = 0.96, SRMR = 0.00, RMSEA = 0.043]. See Figure 2.
According to the path model, MAv goals a revealed negative relationship with SR and self-efficacy, while the MAp goal revealed a positive relationship with SR and self-efficacy. The PAp goal showed only a positive association with self-efficacy. Lastly, the PAv goal produced no relationship with variables. Self-efficacy is an important variable in the model due to its direct link with LTPA. It positively predicted LTPA and was a mediator between AG and LTPA. Contrary to our hypotheses, SR did not significantly predict LTPA. The percentage of variance (squared multiple correlations) accounted for in each dependent variable were: Self-efficacy: 27%, LTPA: 9%, and SR: 5%. All R 2 were statistically significant (p < .01).

Discussion
This study tested a hypothesized model examining the mediating role of self-efficacy and SR on the effects of the 2 x 2 AG model and LTPA among Turkish high school students. After initial lack of model fit, follow-up analysis techniques yielded an acceptable fit for the final model. Results revealed self-efficacy partially mediated the effects of three of the four AG on LTPA. SR was not a mediator of AG and did not predict LTPA. However, a MAp orientation revealed a positive relationship with SR while a MAv goal orientation produced a negative relationship with SR.
The first hypothesis stated that MAp, MAv, PAp and PAv would significantly predict self-efficacy and SR. The MAp coefficient was significantly and moderately related to self-efficacy, followed by PAp and MAv goal orientations. The MAv relationship with self-efficacy was statistically significant but negative in direction and recorded the smallest absolute value. These findings correspond to AG Theory where MAp and PAp positively associate with perceived competence, while MAv has a negative association with perceived incompetence [51]. Gao et al. [31] also confirmed this goal and suggested individuals endorsing MAp and PAp goals over MAv goals may expedite their own self-efficacy. Whether competence is perceived as mastery or task-oriented, the approach valence may be assumed to relate positively to self-efficacy. Teachers may benefit from these results through the creation of an 'approach-centered' climate that encourages and supports their student's efforts.
Only the mastery goals predicted SR. The obtained result adds further support that MAp and PAp goal orientations are positively linked to perceived competence and can operate simultaneously in the same environment, while MAv is negatively linked with perceived incompetence [52]. Since both goal orientations' positive valence (i.e., approach) yielded significant and positive predictions of self-efficacy, there are implications for practice. Though a general mastery orientation to learning is positively related to learning and self-motivation, some students respond to competition and assessing performance relative to others. Teachers might provide activities offering a blend of both approach orientations while still emphasizing task mastery and personal improvement.
That self-efficacy would mediate the effects of the four AG goals on SR and LTPA was partially met. Except for PAv, all were significant predictors of self-efficacy. MAv was not a significant predictor of self-efficacy. Specifically, the MAp coefficient was significantly related to self- efficacy, followed by PAp. The MAv relationship with self-efficacy was statistically significant, but negative in direction. The results add further support to the theoretical underpinning that both MAp and PAp goal orientations are positively related to perceived competence and MAv is negatively correlated with perceived incompetence [32,52]. For PE practitioners, both mastery and performance approach goal orientations can operate simultaneously in a perceived mastery-involving climate where students believe they are capable of performing activity [31,17].
Self-efficacy fully mediating the effects of PAp and MAv goals and partially mediating MAp and LTPA is supported by related research. Gao et al. [32] reported that students' self-efficacy fully mediated the effect of MAp on a fitness test and partially mediated the effects of MAv and PAp on the same fitness test. Using the dichotomous AG model, Li, Shen, Rukavina, and Sun [53] found that both mastery and performance goals positively predicted perceived competence among a sample of middle school students. The current results offer further documentation that students exhibiting an ego-involved goal orientation with an avoidance valance may not automatically display maladaptive behaviour in achievement settings [54]. That both high-task/low-ego and high-task/high ego configurations have been found to be motivationally adaptive bodes well for PE. Motivationally adaptive students work harder, record higher levels and selfefficacy and attribute success to their effort [55]. Teachers should be prepared for multiple goal orientations in their classes and include a variety of mastery and task-oriented activities.
Though the performance avoidance goal orientation revealed no significant paths, students nevertheless recorded higher than average subscale scores (M=5.07). According to AGT, a negative valence indicates behaviour initiated via deleterious or undesirable events or possibilities and are problematic in achievement situations [23]. Fear of failure is often an underlying motive for performanceavoidance learners. A performance-avoidance associated with surface processing of material, avoiding help seeking, and a perception that the achievement context is overly evaluative.
These characteristics are of potential concern because Wolters et al. [55] found that students reporting a high extrinsic goal focus tended to have lower levels of selfefficacy. As seen in the results, a MAv goal orientation negatively predicted self-efficacy. Physical educators, along with a mastery approach, might establish an environment that fosters student self-efficacy while simultaneously reducing external regulators that might enhance a performance avoidance goal orientation.
Contrary to the proposed model, SR was not a mediator of AG and LTPA and did not predict LTPA. One reason for the lack of mediation/predictability of SR may be due to the instrument. Research findings on reliability and factor validity of the MSLQ are mixed. While early research on the MSLQ found acceptable internal consistency reliability, confirmatory factor analysis demonstrated modest fits only [56]. Cook et al. [56] also produced results where Confirmatory Factor Analysis did not demonstrate a good fit but a follow-up Exploratory Factor Analysis suggested a five-factor model. Though adapted for and validated in a Turkish setting, the MSLQ is designed primarily for traditional achievement-oriented environments and may not have captured the context afforded in this setting.
High school PE in Turkey follows a traditional sportsoriented curriculum model delivered through teachercentered instruction. Under these conditions, students may not be exposed to SR opportunities and thus may not have associated the SR items with their PE classes and intentions for future LTPA. Like any skill set, SR skills must be taught and students provided with opportunities for application. Both an examination of the contextspecificity of the items and opportunities for SR in the environment may provide more definitive clues for the lack of predictability SR played here. Finally, the model proposed that self-efficacy and SR would predict LTPA. That self-efficacy was positively related to LTPA supports previous research identifying efficacy as a predictor of PA. McAuley and Blissmer's [57] review of the predictive role of self-efficacy to exercise yielded powerful relationships. The authors reported that changes in self-efficacy over time directly related to changes in exercise behavior and predicting long-term adherence to PA. To foster belief about capabilities to engage in PA, teachers can provide students with a variety of co-operative and competitive activities as well as opportunities for success. Providing a selection of activities combined with successful engagement can enhance the likelihood of students continuing PA after their high school years.
There are, however, limitations to note. First is using self-report measures. Cross-sectional nature designs may not permit the testing of reciprocal links which may appear over time. Future research might employ a longitudinal design whereby serial measurements are made in the same individuals over time. Another limitation was the LTPA assessment. Since IPAQ requires respondents to recall PA over the past week, the reliability of the data depends on adolescents' memory. To overcome this limitation, future studies may use specific PA measures during leisure time to assess LTPA.

Conclusion
Despite these limitations the study contributes to a better understanding of the mediating role of self-efficacy in the relationships among the 2 x 2 AG orientations and LTPA by including SR as a mediator of LTPA. Though numerous studies include self-efficacy as a mediator among a variety of outcome variables, SR has largely been ignored. To the authors' knowledge, no study has included SR as a mediator AG to include and predict LTPA employing path analysis techniques.
Though SR did not mediate AG nor predict LTPA, there is a long-established relationship between goal orientations and SR. Since this relationship was not confirmed among these Turkish high school students,