The present research evaluated an intervention, derived from the "extended contact hypothesis," which aimed to change children's intergroup attitudes toward refugees. The study (n=253) tested 3 models of extended contact among 5- to 11-year-old children: dual identity, common ingroup identity, and decategorization. Children read friendship stories based upon these models featuring in- and outgroup members. Outgroup attitudes were significantly more positive in the extended contact conditions, compared with the control, and this was mediated by "inclusion of other in self." The dual identity intervention was the most effective extended contact model at improving outgroup attitudes. The effect of condition on outgroup intended behavior was moderated by subgroup identity. Implications for theoretically based prejudice-reduction interventions among children are discussed.