The aim of this paper was to describe and test a social cognitive model of physical activity tailored for adolescent girls. Participants were 1518 girls (aged 13.6 ± 0.02 years) from 24 secondary schools in New South Wales, Australia. Useable accelerometer (≥10 hours day(-1) on at least 3 days) and questionnaire data were obtained from 68% of this sample (N = 1035). Participants completed questionnaires assessing psychological, behavioural, social and environmental correlates of activity. The theoretical model was tested using structural equation modelling in AMOS. The model explaining accelerometer counts per minute was an adequate-to-good fit to the data (Tucker-Lewis Index = 0.89, the comparative fit index = 0.97 and the root mean square of approximation = 0.098; 90% confidence interval = 0.075-0.122) but explained only 5% of the variance in activity. There were significant model pathways from self-efficacy (r = 0.11, P = 0.01), school environment (r = 0.07, P = 0.02) and physical self-worth (r = 0.07, P = 0.04) to accelerometer counts. Although the proposed model provided an adequate-to-good fit to the data, it explained a small portion of the variance. Shared method variance may explain the larger portions of variance explained in previous studies. Future studies are encouraged to evaluate theories of physical activity behaviour change using objective measures of physical activity.