This study examined the relationship between psychosocial resources (perceived physical competence and social support for PA), environmental resources for PA, and measures of PA. Consistent with hypotheses and previous research, higher levels of social support and perceived competence were associated with higher levels of PA (assessed via in-school and out-of-school sports participation, accelerometry, and self-reported PA recall). Interestingly, environmental resources did not display a direct relationship with any of the PA variables in this study. However, there was a significant interaction between environmental access and social support. Specifically, among adolescents with high levels of environmental resources, greater social support was associated with students participating in a greater number of sports in school, whereas no such relationship emerged among adolescents with low environmental resources.
Evidence that adolescents with high levels of both social support and environmental PA resources participated in the most school sports, whereas adolescents with low levels of one or both of these variables were significantly less likely to participate in school sports has important implications. This finding supports the calls by Leonard Epstein  and others [see [24, 49]] for interventions aimed at increasing PA that address both environmental influences on PA and mobilization of support for PA from members of participants' social support networks. The fact that adolescents with high levels of either environmental resources or social support were no more active than adolescents with low levels of both suggests that social support and environmental resources may each be necessary but not sufficient conditions for promoting adolescent PA. Thus, interventionists may see improved success in promoting PA by addressing the dual goals of increasing social support for PA and increasing access to PA opportunities.
One surprising result was that the relationship between perceived competence and PA was not moderated by access to environmental PA resources. Adolescents with higher levels of perceived physical competence showed higher levels of PA than adolescents with lower levels of perceived competence whether they had access to environmental resources or not. One possible explanation is that individuals with high perceived physical competence may not see environmental deficits as insurmountable roadblocks and may be more willing to go out of their way to be active. This result was unexpected given the recent reports of interactions between environmental variables and self-efficacy, a construct similar to perceived competence, in relation to PA [27, 28].
However, the present study differed from these earlier studies in important ways beyond examining competence as opposed to self-efficacy. The current study objectively measured the environment in the United States, whereas these other studies measured the environment subjectively, and examined residents of Australia and Belgium. It is possible that these differences and others may help explain the differing results. It is also worth noting that the sample size of the present study is substantially smaller than the samples studied by Cerin and colleagues  and Deforche and colleagues , and that the relatively smaller sample size may help explain why the present study did not produce statistically significant links between the environment and PA as reported in these other studies, despite comparable effect sizes across all 3 studies.
Another surprising finding was that although all four measures of PA had strong positive relationships with both psychosocial measures, only one of these relationships (i.e. social support and school sports participation) was moderated by environmental resources. It is possible that school sports participation is a unique and important predictor. Indeed, research suggests that involvement in organized sports as a child and adolescent increases the likelihood for a physically active lifestyle into adulthood . It is also possible that idiosyncrasies of each of the other PA assessments, discussed below, could make school sports participation the most representative measure of PA among these adolescents and could help explain why the hypothesized moderation was seen for school sports participation but not for any of the other PA measures used in the present study.
The self-reporting involved in the 3DPAR may have been subject to recall bias or inflated reporting of time spent engaging in PA. Adolescents self-reported about 3 hours more MVPA per day on the 3DPAR than was recorded via accelerometry (see Table 1). Additionally, the 3DPAR and accelerometer data reflect only a small slice of time in the lives of these adolescents, which may not be highly representative of their year-round PA. Further, the outside-of-school-sports questionnaire elicited a range of responses (e.g. paintball, surfing, horseback riding) that likely varied a great deal more in terms of intensity and frequency than did participation in a school sport.
It is also worth noting that the 3DPAR and accelerometer data reflect PA patterns during the summer, which may differ from year-round PA behavior. Summer in Southern California is marked by high beach traffic, and it is possible that these teens were traveling to beaches for some of their PA (e.g. surfing, swimming) rather than using environmental resources nearer to their homes. Further research could investigate this possibility by objectively recording (e.g. via Global Positioning System; GPS) physical activity location. Participation in water sports has additional ramifications for the data collected by the accelerometers, which were not waterproof. Many participants reported removing the Actigraph® during times that they were active, thus affecting the representativeness of the accelerometer data. However, analyses were also run using an accelerometer variable that adjusted for missing data based on participant report of when the accelerometers were removed to participate in water sports, and the results were not significantly different from those reported here.
Taken together, these factors may help explain why school sports participation (which reflects PA behavior during the majority of the year) was the PA measure in this study that behaved as hypothesized whereas the other PA measures did not; however, the measure of school sports used here was also not without flaws. The measures of sports participation did not gauge the frequency or intensity of PA and thus were not directly translatable to minutes of MVPA to compare with Actigraph® and 3DPAR data. Nor did these measures distinguish between training activities for a sport (e.g. weight-lifting) and playing the sport itself, so participants may have inconsistently categorized their-sport related training as part of the sport or as another activity entirely. However, it should be noted that there was objective biological evidence suggesting that those adolescents participating in more school sports were, in fact, more active than those who participated in fewer or no school sports; specifically, participation in school sports was correlated with both physical fitness (ergometer-determined VO2peak) and body composition (DEXA-determined percent body fat).
The present study provides insight into the ways in which environmental and psychosocial factors can impact adolescent PA, but there are limitations to consider beyond the possible lack of year-round representativeness of some of the PA measures. First, the median household income for the participants in this study was $87,133 (range: $41,591 to $182,300), which is higher than the national median household income ($51,425) as well as the median for California ($61,154); however it is closer to the median for Orange County, where the participants resided ($75,176) . In addition, the racial composition of this sample does not mirror the racial distribution across the U.S. (although the percentage of Caucasian participants in this study [68%] was not drastically dissimilar from the percentage of Caucasians living in California [61%], where the data were collected, or from the national average [74%; ]).
These demographic characteristics limit the generalizability of these results to other groups of adolescents; however, as reported elsewhere , these adolescents are quite representative of the American adolescent population in other important ways (e.g. body composition and cardiovascular fitness). It should be noted that SES was assessed at the aggregate level (i.e. each participant's SES was determined by the median household income in his or her Census block group), and likely differed from any individual participant's family income. However, there was a moderate correlation between self-reported income and median household income (r = 0.46, p < 0.01) among the final cohort of participants, from whom individual-level income information was collected.
Parent PA was reported by the child and may not have been an accurate representation of parental PA. In addition, although parental perceptions of PA benefits for their child were provided by a parent, these responses were only collected from one parent per participant, and thus may not reflect the perceptions of both parents. It is important to recognize that youth PA may relate to perceptions held by both parents as well as other unassessed family variables like living situation and family structure . Future research might benefit from assessing these family variables as well as from collecting data from two parents when possible to evaluate additional questions regarding the associations between PA and parent-child relationships (e.g. do PA-related behaviors/attitudes of the same-sex parent have a stronger association with youth PA than those of the opposite-sex parent?). These questions could not be effectively addressed in the present research as only one parent completed a questionnaire per adolescent participant, and only 17.7% (n = 34) of parental respondents were male.
In addition, PA and psychosocial data were collected from 4 cohorts of adolescents over 4 consecutive years (2005 - 2008), but environmental data were all gathered at one time (2008). It is possible that some gyms, trails, parks and fields were opened or closed during the four years that individual-level data were collected. These changes in environmental access to PA resources cannot be examined in the present data. Relatedly, only proximity was assessed in relation to environmental PA resources. Usability was not measured in any other way. It must be acknowledged that factors other than proximity (e.g. safety, attractiveness) can impact use of environmental PA amenities. Finally, in light of the present results, future research assessing social support provision related to specific types of PA (e.g. transportation to school sport activities vs. non-school sport activities) might be a fruitful line of inquiry.