Peers are an important determinant of physical activity in children and adolescents (McPherson, Smith-Lovin, & Cook, 2001). Previous studies have shown that young people are more similar in physical activity to their friends than those who are not their friends (Long, Barrett, & Lockhart, 2017; Stearns et al., 2018). More specifically, children and adolescents both select new friends on the basis of physical activity and influence the amount of physical activity of friends (de la Haye, Robins, Mohr, & Wilson, 2011; Gesell, Tesdahl, & Ruchman, 2012). The current study is particularly interested in these two processes that explain the similarity between adolescents’ friendships in physical activity. More specifically, this study examined changes in nominated friendships and objectively measured physical activity levels at three time points and test whether similarity in physical activity exists prior to the formation of friendships (selection), or whether friends become more similar over time (influence). To test for selection and influence effects of friendship networks, stochastic actor-based models are used (i.e., RSiena; Snijders, van de Bunt, & Steglich, 2010). Longitudinal social network analysis simultaneously investigates changes in friendships and physical activity, and are capable of disentangling whether adolescents select friends based on a similar level of physical activity and whether friends influence each other’s physical activity. In total, 726 adolescents (47% male, Mean age = 10.88, SD age = 1.16, range = 8-15 y/o) in 38 school classes participated for three separate weeks between February and June of 2018. Physical activity was measured by wrist-worn accelerometer (Fitbit Flex) for five days (M = 9,669 steps/day, SD =2,872), and reduced to four categories. The friends networks were determined based on the question “With whom in your class do you spend time during the breaks?” on the research smartphone. Social network modelling revealed that, after controlling for endogenous network effects, both selection (b=.44, SE= .12) and influence (b=.18, SE= .06) effects were statistically significant. However, after controlling for the effects of sex, age and zBMI, the results slightly changed. The influence effect was still observed (b=.15, SE= .06), but the selection effect was not significant anymore (b=.23, SE= .14). Instead, the analysis suggest that adolescents select each other based on sex and zBMI. Together, these results indicates that adolescents do not initiate friendships with others that have similar physical activity levels but friends do become more similar in physical activity to that of their friends. Mostly in line with previous studies (de la Haye et al., 2011; Gesell et al., 2012; Long et al., 2017; Shoham et al., 2012; Simpkins, Schaefer, Price, & Vest, 2013), these findings show that friends are similarity in physical activity. That is, adolescents do not select friends based on similarity in physical activity, but do influence the physical activity of friends.