This study evaluated a physical activity intervention for children that comprised peer-modelling, pedometer step goals and tangible rewards. A version of the intervention without the reward component was also tested. Participants (n = 386) were from three primary schools, which were randomly assigned to one of three conditions: (i) full intervention, where children received "Fit 'n' Fun Dude" peer-modelling materials and were given daily pedometer goals to receive rewards, (ii) no-rewards intervention, where children received peer-modelling materials and pedometer goals but rewards were not used and (iii) control, where children wore pedometers with no further intervention. Physical activity was measured at baseline, intervention and at the end of a 14-week 'taper' phase. During the intervention, the full intervention school showed the largest increase in physical activity relative to baseline (+2456 steps per day, p < 0.001). There was a smaller increase in the no-rewards school (+1033 steps per day, p < 0.03), and no significant change in the control. At the end of the taper phase, physical activity in the no-rewards school continued to increase (+2030 steps per day, p < 0.001) but had returned to baseline in the full intervention school. The intervention that used only peer-modelling and pedometer goals produced better effects over time.