The purpose of this study was to examine the longitudinal relationships between naturally occurring changes in leisure-time physical activity, depressive symptoms and self-efficacy in adolescent girls. We also aimed to test whether depressive symptoms would moderate the self-efficacy-physical activity relationship. Participants were 181 urban adolescent girls. Physical activity was measured using the 3-Day Physical Activity Recall. Self-efficacy and depressive symptoms were assessed using questionnaires. Body height and body mass were measured and body mass index (BMI) was calculated. Data were collected on three occasions over a 2-year period. There was a decrease in physical activity and self-efficacy and increase in depressive symptoms across three measurement occasions. There were statistically significant and negative relationships between initial level and change for physical activity and depressive symptoms. Initially higher levels of physical activity were related with initially lower levels of depressive symptoms, and change in physical activity across time was inversely associated with change in levels of depressive symptoms across measurements. There were statistically significant and positive relationships between initial level and change for physical activity and self-efficacy after controlling effect of BMI. Latent growth modeling (LGM) also indicated a moderating effect of depressive symptoms on the self-efficacy-physical activity relationship. Girls who had high initial levels of self-efficacy and smaller increases in depressive symptoms had the lowest decline in physical activity participation. Our results encourage the design of interventions that reduce depressive symptoms and increase self-efficacy as a possible of means of increasing adolescent girls' physical activity.