Abstract
Intergroup contact is arguably the prejudice reduction intervention with the most existing empirical support. However, recent meta-analyses of experimental contact interventions find signs of publication and reporting biases. To mitigate such bias, I carry out a meta-analysis of 41 preregistered contact experiments, considering only treatment effects on preregistered primary outcomes. I find that ( a ) the average effects of intergroup contact are smaller than indicated by previous findings, at roughly one-tenth of a standard deviation; ( b ) the subset of in-person interventions that satisfy Allport's four desirable scope conditions (e.g., common goals) are no more effective; and ( c ) generalization is limited: Contact is more effective at changing behavior and attitudes toward people met than toward the outgroup as a whole. I offer suggestions for how researchers might make progress on the problem of generalization through careful measurement of its extent and the consideration of moderating factors beyond those emphasized by Allport.