Fig. 3

The moderated mediation model with group as predictor X (coded 1 0, for intervention and control conditions, respectively), similar body → positive contingency learning as mediator M, BISS post-intervention scores as criterion Y (with baseline scores as covariate), and BSQ body image scores as moderator W of the effect of contingency learning on BISS changes. The boot-strapped model confirmed that a positive effect of the intervention on contingency learning improved body dissatisfaction, but only among individuals with higher body image concerns