Given the importance of combating the physical inactivity and obesity epidemics, researchers have targeted PE as a logical context to study whether teaching style may impact the eventual health of youth. To date, a quantitative review of this important literature does not exist. Though certainly PE teacher autonomy-supportive instruction framed within SDT is not hypothesized to directly impact student outcomes past the three basic needs, the relationship may be direct. In the health care context, Ng and colleagues (2012) reported in the SDT health care context small to medium effect sizes with the three basic needs and motivational processes and regulations with physical activity (Ng et al., 2012). The effect size between health care provider perceived autonomy support and physical activity was small (met- analyzed r = .23 from 30 samples); yet, greater in magnitude than a number of SDT constructs (i.e. basics need of autonomy, autonomous regulation) that theoretically should be related more to an outcome such as physical activity. Hence, it is important to know the meaningfulness of perceived PE teacher autonomy-supportive instruction on all student motivated physical activity based outcomes is of great value to help direct future research endeavors.

Our hypotheses were very straightforward. We first hypothesized that perceived PE teacher autonomy-supportive instruction would be positively related to desirable student outcomes such as satisfying the three basic needs, intrinsic motivation, positive emotions and thoughts of self, physical activity thoughts, intentions, and behavior and negative related to undesirable student outcomes such as external regulation and negative emotions. We also hypothesized that effect sizes between perceived PE teacher autonomy-supportive instruction would be more meaningful in interpretation overall than those from LT. Last, we hypothesized that within the PE and LT contexts, the effect sizes between perceived PE teachers autonomy- supportive instruction would be more meaningful for the three basic needs and intrinsic motivation than more distal outcomes such as physical activity behavior.

The literature search for published studies was systematic. It included electronic databases, reviewing reference lists of three past published review articles, and search of references from retrieved articles (see Figure 1 for PRISMA diagram). The electronic database search was conducted in EBSCO with the range of individual databases specific to sport (Sport-Discus), psychology (PsychINFO, PsychARTICLES), and education (ERIC). Key word combinations to locate published studies were based on the following terms: autonomy-supportive climate, autonomy-support, physical education, physical education teachers, physical education climate, student motivation, and Self-Determination Theory. The key word combinations were established and refined throughout the article retrieval process.

Figure 1
PRISMA flow chart

Articles retained for the current meta-analysis met the following inclusion criteria: (a) papers must be published in a language that the authors were fluent and if not fluent could obtain assistance from a native speaker; (b) papers must published up to the stop of the search, March 1, 2015; (c) papers must be original data published in peer-reviewed journals, and not theses, book chapters, and or conference proceedings; (d) papers must contain a measure that was clearly in regards to perceptions of the PE teacher’s autonomy-supportive climate behaviors; (e) papers must have at least one student outcome variable related to the PE teacher’s autonomy-supportive behaviors; (f) papers must report sufficient statistical information such as raw correlations and sample sizes or means, standard deviations, and sample sizes; and (g) if insufficient statistical data were present, contacted authors must provide sufficient statistical information via email correspondence. The two authors with the aid of a research assistant worked diligently discussing the final set of papers meeting all eligibility criteria.

The Comprehensive Meta-Analysis (CMA) version-2 software version 2.2.064 (July 27, 2011) was used for this meta-analysis. Based on Hedges and Olkin’s (1985) suggestion, the mean weight correlation (rw) was chosen as the measure of effect size as all extracted data were reported as correlations (see sample summary for reason for exclusion of experimental investigations). Cohen’s (1977) criteria were used for interpretation of each rw as follows: 0.10 to 0.30 as small, 0.30 to 0.50 as medium, and > 0.50 as large. Positive effect sizes should be interpreted as perceived autonomy-supportive behaviors facilitating the specific student outcome variable, whereas a negative effect size should be interpreted as the perceived autonomy-supportive instruction thwarting the specific student outcome variable.

Of the two primary models to determine statistical assumptions of error, the fixed as opposed to random model was chosen. The fixed effects model assumes that all of the gathered studies share a common effect and differences are a result of within study error or sampling error. The random effects model assumes both within study error and between-study variation. The fixed effects model was selected because a priori student outcome variables were separated for physical education and leisure-time, categorized into groups (e.g. motivation processes) and then reported by identical subcategories (e.g. intrinsic motivation, external regulation) when applicable. Thus, the reported effect sizes should share a common effect given the a priori levels of separation. Even with the a priori categories and subcategories, heterogeneity was analyzed. Two indicators (Q and I2) were used to determine whether heterogeneity of variance existed. Moderators were not analyzed such as country of origin, type of autonomy- support questionnaire used as again a priori many distinct categories were formed. The Q test is a test of heterogeneity significance. This test is based on the critical values for a chi-square distribution. A significant Q value indicates that heterogeneity of variance exists across the individual effect sizes used to calculate the overall effect size. The Q value does not provide information on the magnitude of the individual effect size dispersion. The I2 statistic does provide this information as it is the ratio of excess dispersion to total dispersion. As explained by Higgins and colleagues (Higgins & Thompson, 2002; Higgins, Thompson, Deeks, & Altman, 2003), I2 may be interpreted as the overlap of confidence intervals explaining the total variance attributed to the covariates. Higgins and Thompson (2002) have provided a tentative classification of I2 values to help interpret magnitude of the heterogeneity of variance: 25 (low), 50 (medium), and 75 (high). Last, publication bias (i.e. only reporting hypothesis supportive results or only hypothesis supporting manuscript published) is always a concern in a quantitative review. CMA provides a fail safe N statistic that is interpreted as the number of samples required to change a significant effect size into a non-significant effect size. Hence, this statistic was calculated and reported.

Table 1
Summary of included studies.

Note: SD = Standard deviation; LCQ = Learning Climate Questionnaire; PASSES = Perceived Autonomy Support Scale for Exercise Settings; SCQ = Sport Climate Questionnaire; TASCQ = Teacher as Social Context Questionnaire; MCPEQ = Motivational Climate in Physical Education Questionnaire; ASCQ = Autonomy-supportive Coaching Questionnaire; IBS = Interpersonal Behavior Scale; HCCQ = Health Care Climate Questionnaire; UK = United Kingdom; USA = United States of America.

Table 1 (cont. 1)
Summary of included studies.

Note: SD = Standard deviation; LCQ = Learning Climate Questionnaire; PASSES = Perceived Autonomy Support Scale for Exercise Settings; SCQ = Sport Climate Questionnaire; TASCQ = Teacher as Social Context Questionnaire; MCPEQ = Motivational Climate in Physical Education Questionnaire; ASCQ = Autonomy-supportive Coaching Questionnaire; IBS = Interpersonal Behavior Scale; HCCQ = Health Care Climate Questionnaire; UK = United Kingdom; USA = United States of America.

Table 1 (cont. 2)
Summary of included studies.

Note: SD = Standard deviation; LCQ = Learning Climate Questionnaire; PASSES = Perceived Autonomy Support Scale for Exercise Settings; SCQ = Sport Climate Questionnaire; TASCQ = Teacher as Social Context Questionnaire; MCPEQ = Motivational Climate in Physical Education Questionnaire; ASCQ = Autonomy-supportive Coaching Questionnaire; IBS = Interpersonal Behavior Scale; HCCQ = Health Care Climate Questionnaire; UK = United Kingdom; USA = United States of America.

Table 1 provides a summary of the characteristics of the sample, the autonomy-support measure used, and the contributions to student outcomes coded and found in Table 2 and 3. Given the vast difference even within the located experimental manuscripts, we decided to only analyze the correlate studies for this review1. A total of 39 papers were located that met the inclusion criteria with a total population size of 23,554 of which 10,954 (46.50%) were identified as females and 9,970 (42.32%) were identified as males. The approximate, as not all studies reported sufficient data, mean age of the entire population was 14.30; SD = 1.01 years. Of the 36 papers, 15 countries totaling 51 country specific samples were represented with the United Kingdom (k = 9), USA (k = 9), and Spain (k = 6) accounting for nearly half of all samples contributed to analyses. There were eight autonomy-support measures across all of the included papers. Of those, versions of the Learning Climate Questionnaire (LCQ; Williams & Deci, 1996; k = 10), the Sport Climate Questionnaire (SCQ; Baard, Deci, & Ryan, 2000; k = 8), and Perceived Autonomy Support Scale for Exercise Settings (PASSES; Hagger et al., 2007; k = 7) were the most used.

The two authors made a list of all measured students outcomes. From this extensive and exhaustive list, outcome variables were categorized together for analyses. The grouping into categories was straightforward. Therefore, separate analyses were set up for the following physical education student outcomes: basic needs (autonomy k = 20; relatedness k = 19; competence k = 20), motivation processes (intrinsic motivation k = 14; identified regulation k = 9; introjected regulation k = 9; external regulation k = 9; amotivation k = 22; relative autonomy index k = 10; autonomous regulation k = 10), motivated behaviors (effort k = 7; physical activity k = 3), physical activity motivation (deficient effort k = 7; deficient ability k = 7; insufficient task values k = 14; intentions k = 4), emotions (positive k = 11; negative k = 4), and self-esteem/concept (physical k = 6; general k = 4).

For leisure-time student outcomes, separate analyses were set up for the following: basic needs (autonomy k = 2; relatedness k = 2; competence k = 2), motivation processes (intrinsic motivation k = 3; identified regulation k = 3; introjected regulation k = 3; external regulation k = 3; amotivation k = 1; relative autonomy index k = 8; autonomous regulation k = 5), and physical activity (intentions k = 18; self-reported k = 18; objective k = 1; attitude (positive > negative k = 13; perceived behavioral control k = 13). Additional outcomes variables that did not fit the above categories existed in the two domains, physical education and leisure-time. Every effort was made to provide similar data for comparison between PE and LT as well as not present constructs with just one sample and thereby simply making a long list of constructs. A complete list of all variables found in all of the 39 included manuscripts is available from this paper’s first author.

Table 2
Summary of the fixed effect sizes for correlations between perceived PE teacher autonomy-support and student PE outcome variables.

Note: All effect sizes are statistically significant from 0 (p < .001) except the insignificant negative emotion effect size. k = total number of correlations included in the analysis; N = total number of participants; rw = weighted correlation; CI = confidence interval; LL = lower limit; UL = upper limit; Z = test of null (2-tailed); I2 = I-squared test of heterogeneity.

Table 2 contains all of the variables meta-analyzed. Except for negative emotions, all of the effect sizes were statistically significant (p < .01) from 0. In addition, all of the effect sizes were calculated from total samples of at least 1,433 (sample range 1,433-12,180). For the three basic needs, the effect sizes for all needs and perceived teacher provided autonomy- support were medium (competence rw = .41; relatedness rw = .46) to large (autonomy rw = .57) in meaningfulness. The fail safe N values were extremely large with all greater than 10,000. For the motivation processes category, the effect sizes were small (introjected regulation rw = .20; external regulation rw = -.15; amotivation rw = -.19), medium (relative autonomy index rw = .42; autonomous regulation rw = .44), and large (intrinsic motivation rw = .54; identified regulation rw = .50) in meaningfulness. The fail safe N values ranged from 244 to 8,528. For the motivated behaviors category, the effect sizes were small (effort rw = .33) and small (physical activity rw = .10). The fail safe N for physical activity was only 15 while for effort it was 646. There were four constructs that were directly related to physical activity motivation. The effect sizes were small (deficient effort rw = -.18; deficient ability rw = -.15; intentions rw = .20) to medium (insufficient task values rw = -.26) in meaningfulness. The fail safe N values ranged from 46 to 1,146. The effect size for positive emotions was large (rw = .52) with the fail safe N being 4,415. As stated previously, the negative emotion effect size was not significant and also inconsequential in meaningfulness (rw = .03). Lastly, the effect size for physical self-esteem/concept was medium (rw = .32) while general self- esteem/concept was small (rw = .32). The fail safe N values were 135 and 652, for general and physical self-esteem/concept respectively.

Except for the very low I2 value for autonomous regulation (I2 = 8.20), and low to medium values for deficient ability (I2 = 31.6) and medium I2 value for deficient effort (I2 = 72.20), the rest of the I2 values were large (>75) suggesting that heterogeneity still exists in the data.

Table 3
Summary of the fixed effect sizes for correlations between perceived PE teacher autonomy-support and student leisure-time outcome variables.

Note: all effect sizes are statistically significantly (p < .001) in all cases except for introjected (p < .05) and external (p < .01) regulation different than 0. k = total number of correlations included in the analysis; N = total number of participants; rw = weighted correlation; CI = confidence interval; LL = lower limit; UL = upper limit; Z = test of null (2-tailed); I2 = I-squared test of heterogeneity.

Table 3 contains all of the variables analyzed. As previously stated, the priority was to match the LT categories and or constructs to those reported on in PE. Given the differences in the some of the research questions addressed, not all categories and constructs were matched. Except for introjected and external regulation and the one objective report of physical activity, all of the effect sizes were statistically significant (p < .01) from 0. Given fewer samples when compared to the physical education effect sizes, the effect sizes were calculated from smaller total samples (range 491-6,692. For the basic needs, the effect sizes for all three needs and perceived teacher provided autonomy-support were small (competence rw = .22) and medium (autonomy rw = .44; relatedness rw = .45) in meaningfulness. Given each construct had only two samples, fail safe N values were not calculated. For the motivation processes category, the effect sizes ranged from insignificant and inconsequential (introjected regulation rw = .06; external regulation rw = -.08), small (amotivation rw = -.25; relative autonomy index rw = .22; autonomous regulation rw = .20), and medium (intrinsic motivation rw= .30; identified regulation rw = .32) in meaningfulness. The fail safe N values ranged from 1 to 193. Lastly, for the physical activity category the effect sizes were inconsequential (objective rw = .00) and small (intentions rw = .25; self-reported rw = .21; attitude rw = .24; perceived behavioral control rw = .15). The calculated fail safe N values were fairly robust (range 214-1,587) relative to the samples per each construct.

Heterogeneity varied greatly. Heterogeneity was small (I2 < 25) for autonomous regulation and external regulation. Heterogeneity was medium (I2 < 50) for intrinsic motivation, identified regulation, self-reported physical activity, attitude, and perceived behavior control. For the rest of the constructs, heterogeneity was high (I2 > 75). Thus, unlike the physical education data, the leisure-time data had less heterogeneity.

The main limitation of the present meta-analysis was actually the large variety of student outcomes, again a full report of all outcome variables are available from the first author, that were measured across the 39 included studies. Many of student outcome variables were measured only once. In addition though initially this body of research was focused on the trans-contextual model and grounded in SDT, a variety of research approaches have been taken; thus, increasing the number of measured student outcome variables. Thus, unlike the Ng and colleagues (2012) meta-analytic review of SDT variables and health outcomes, the present meta-analytic review had a number of approaches found within the included articles. In addition, though a priori it was determined that student outcomes would be categorized and like constructs grouped, a great deal of heterogeneity still existed in the report effect sizes. Though, certainly, meta-analytic research across a number of disciplines has used a variety of possible moderator variables such as country of origin, the present meta-analysis did not. Inherently, there is not a logical reason that any one country or region of the world would moderate the reported effect sizes. Nearly all of the data were published with both sexes included. Hence, sex of sample could not be examined as a moderator that may or not moderate the reported effect sizes. Last, a major limitation was the over reliance on student self-reported physical activity. In today’s era with the great number of options for physical activity tracking technology, only Standage and his colleagues (2012) utilized such technology.

This meta-analytic summary provided important findings with regarding the state of perceived PE teacher autonomy-supportive instruction and student physical activity based PE and LT outcomes. Certainly, one may conclude that teacher autonomy-supportive instruction directly and very meaningfully benefits student basic needs and intrinsic motivations for physical activity. Unfortunately besides the benefits of such perceived instruction on physical activity based positive emotions, the direct impact on physical activity itself was small. To overcome this small benefit, future research must investigate how to integrate autonomy- supportive instruction with both verbalized and objectively measured physical activity goals to increase students’ physical activity; more specifically up to 60 minutes of MVPA on a daily basis. Certainly, an autonomy-supportive teaching style has many benefits. But perhaps without telling the students the necessity of actually going beyond feeling motived and or feeling positive about physical activity, actual increases in physical activity may never occur. Again, Ng and colleagues (2012) meta-analysis did not provide a great deal of hope that the basic needs and or regulations have more than a small to medium positive impact on physical activity behaviors.

In conclusion, the current meta-analysis uniquely adds to the autonomy-supportive instruction literature by reporting relationships with a number of important students’ motivation related physical activity constructs as well as with physical activity. Unequivocally, an autonomy-supportive instruction style is of great value, but this instructional style must be linked more meaningfully to actual engagement in physical activity or our world will continue to struggle greatly with the dire consequences of insufficiently physically active children and adolescents.