The current research adopted a multipronged mediational approach to test an associative self-anchoring account of automatic intergroup evaluation change following perspective taking. We contend that actively contemplating outgroup members' perspectives strengthens associative links between that outgroup and the self, enabling a transfer of positive automatic self-evaluations to the group. A first set of experiments, using both measurement-of-mediation and experimental-causal-chain designs, supported a model in which strengthened self–outgroup associations underlie perspective taking's positive effects on automatic intergroup evaluations. Additional experiments, using a moderation-of-process design, found that the benefits of perspective taking were attenuated when measured or manipulated automatic self-evaluations were relatively negative, preventing positive associative transfer. A final experiment uncovered a practical downstream implication of our causal model, as perspective-taking-induced changes in automatic intergroup evaluations were still evident 1 day later. Overall, these findings supported our associative self-anchoring account; additional analyses found no support for an alternative, empathy-based account.