How physical activity shapes, and is shaped by, adolescent friendships

Abstract

The current study explored the role of school-based friendship networks in adolescents’ engagement in physical activity (PA). It was hypothesized that similar participation in PA would be a basis for friendship formation, and that friends would also influence behavior. Whether these processes were mediated through cognitive mechanisms was also explored. Self-reported participation in PA, cognitions about PA, and friendship ties to grade-mates were measured in two cohorts of Australian grade eight students (N = 378; M age = 13.7) three times over the 2008 school year. Interdependence between the friendship networks and PA was tested using stochastic actor-based models for social networks and behavior. The results showed that participants tended to befriend peers who did similar amounts of PA, and subsequently emulated their friends’ behaviors. Friends’ influence on PA was not found to be mediated through adolescents’ cognitions about PA. These findings show that there is a mutually dependent relationship between adolescent friendship networks and PA; they highlight how novel network-based strategies may be effective in supporting young people to be physically active.

Introduction

Physical activity (PA) tends to decline in adolescence with few youth meeting recommended guidelines (Gordon-Larsen, Nelson, & Popkin, 2004). From a public health and child development perspective this is problematic because engagement in PA has been found to have a positive impact on physical and mental health (Hallal, Victora, Azevedo, & Wells, 2006). Lack of vigorous PA has also been identified as a key risk factor for being overweight (Patrick et al., 2004). Evidence of a dose–response relationship between adolescent PA and health outcomes is lacking and so recommendations for participation in moderate to vigorous physical activity (MVPA) range widely, from 20 min three times a week (Sallis & Patrick, 1994) to 60 min every day (Biddle, Cavill, & Sallis, 1998). Nonetheless, health authorities worldwide recognize the importance of initiating and maintaining participation in MVPA in this age cohort. To date, multicomponent school-based interventions, that also involved family or community settings, have been the most successful in achieving these aims (van Sluijs, McMinn, & Griffin, 2007).

The success of school-based PA interventions may, in part, be due to the role of school peers in supporting these behaviors. Activities such as sports and active play are typically very social at this age, and young peoples’ friends seem to be important referents. Adolescents’ activity levels have been positively associated with perceived social support from friends (Duncan, Duncan, & Strycker, 2005) and participation in PA with friends (Voorhees et al., 2005). Friends’ modeling and social reinforcement of these activities may be an important determinant of PA (Bandura, 1977), and emulating friends’ behaviors could also be a useful strategy in establishing or maintaining relationships (Brown, Bakken, Ameringer, & Mahon, 2008).

Furthermore, friends may have a direct impact on young peoples’ attitudes and beliefs about PA. For example, adolescents’ perceptions of peer norms have been found to predict their attitudes toward PA, and intentions to engage in PA (Baker, Little, & Brownell, 2003). The theory of planned behavior (TPB) (Ajzen, 1991) proposes that the social environment influences individual behavior via perceptions of norms, and that these norms (as well as attitudes and perceptions of behavioral control) subsequently predict intentions and behavior. However, subjective norms are often a relatively weak predictor of behavioral intentions (Armitage & Conner, 2001), and support for this proposed cognitive pathway of social influence is weak.

These findings would suggest that adolescent friends are likely to be similarly engaged in PA, however studies looking at self-reported behaviors among friends have found interesting, albeit conflicting, associations. One cross-sectional study found that friends participated in similar amounts of sports and training (de la Haye, Robins, Mohr, & Wilson, 2010), yet a longitudinal study found that young people preferred friends with different sporting behaviors (Pearson, Steglich, & Snijders, 2006). Similarities in PA may not always be a characteristic of friendships, and whether similarities (or in fact differences) arise because youth form relationships with peers who have similar behaviors (i.e. selection processes) or because they change their behaviors to emulate their friends (i.e. influence processes), is not clear.

The conclusions to be drawn from the current literature are limited by a number of factors. First, many studies are cross-sectional and so cannot tease out the directionality of the effects. Second, few studies account for the role of PA in friendship formation, which may plausibly account, to some extent, for behavioral similarities (or differences) among friends. Third, although the literature has established the importance of adolescents’ perceptions of their friends’ behavior, additional research measuring self-reported behavior is needed to verify if friends’ actual behaviors are influential. And finally, many studies make assumptions about the independence of individuals or friendship dyads, when they are likely to be embedded in larger, interdependent, social structures (Snijders, Pattison, Robins, & Handcock, 2006). Research that accounts for the patterns of interdependent friendship connections, or friendship network, as well as individual-level variables, is likely to provide a more precise and rich account of behavioral outcomes (Robins & Kashima, 2008). Consequently, we have a limited understanding of the processes that lead to associations in the PA among adolescent friends, the magnitude of these effects, and the underpinning mechanisms.

To address these limitations, this study employed new methods to longitudinally model individual behavior in the context of complex friendship networks. These stochastic actor-based models for social networks and behavior (Snijders, Steglich, & Schweinberger, 2007) allow us to simultaneously test effects of friendship selection and influence, and to explore potential mechanisms driving these processes. Moreover, unlike conventional statistical techniques that assume independence among actors, this method explicitly models dependencies that arise through network structure. Studies using these methods have often found that health risk behaviors, like smoking and substance use, are relevant to the formation of adolescent friendships and are also influenced by friends (e.g. Mercken, Snijders, Steglich, & de Vries, 2009). Whether or not friends influence adolescent PA, controlling for similarities when friendships are formed, will be the primary focus of this study.

Delineating the processes that lead to associations in PA between friends may highlight avenues for behavioral interventions. Additionally, research looking at the role of social networks in obesity has suggested that obesity tends to “spread” as if it were contagious (Christakis and Fowler, 2007, Halliday and Kwak, 2009). Results of the current study will provide some insights as to whether this phenomenon is likely to be driven by the clustering and/or socialization of PA among adolescent friends.

It is hypothesized that participation in PA will be relevant to the formation of adolescent friendships, and that adolescents will be influenced by their friends’ behavior, resulting in friends’ PA becoming increasingly similar. Potential psychosocial mechanisms underpinning these selection and influence processes will also be explored. Based on social cognition models of behavior (Ajzen, 1991, Bandura, 1977) we anticipate that adolescents will emulate the behaviors of their friends, and that this process will be partially mediated via perceptions of peer norms and intentions.