Objective: Examine relationships between changes in physical activity (PA) and personal, psychosocial and environmental factors among adolescent girls. Methods: Longitudinal data were analyzed from two cohorts of adolescent girls. Participants were placed into four categories based on PA levels. Active Maintainer-AM, Inactive Maintainer-IM, Adopter-A, or Relapser-R. Anthropomorphic, personal, psychosocial and environmental data were collected. Multivariate logistic regression was used to estimate parameters.

Results: In early adolescence, Free/reduced price lunch and PA Self-Efficacy were associated with AM; Race and Friend Support for PA with IM; and Friend Support for PA with R. In later adolescence, Distance to Nearest Park, PA Frequency with Friends, and Global Physical were associated with AM; BMI, Friend Support for PA, PA Frequency with Friends, and Age at Menses, with IM; BMI with A. Conclusion: PA predictors were different across the categories and the predictors were different for early adolescence and later adolescence.