Numerous studies have found that perspective-taking reduces stereotyping and prejudice, but they have only involved negative stereotypes. Because target negativity has been empirically confounded with reduced stereotyping, the general effects of perspective-taking on stereotyping and prejudice are unclear. By including both positively and negatively stereotyped targets, this research offers the first empirical test of two competing hypotheses: The positivity hypothesis predicts that perspective-taking produces a positivity bias, with less stereotyping of negative targets but more stereotyping of positive targets. In contrast, the stereotype-reduction hypothesis predicts that perspective-taking reduces stereotyping, regardless of target valence. Three studies support the stereotype-reduction hypothesis. Perspective-taking also produced less positive attitudes toward positive targets, with reduced stereotyping mediating this effect. A final study demonstrated that perspective-taking reduced all stereotyping because it increased self-other overlap. These findings help answer fundamental questions about perspective-taking's effects and processes, and provide evidence that perspective-taking does not improve attitudes invariantly.