To study how roles shape the construction of narrative accounts, participants read a story while identifying with its victim or perpetrator and then retold the story in their own words. In Study 2, participants told the story twice, the second time after a 3-to 5-day retention interval. In Study 3, some participants were given instructions to tell as accurate a story as possible. In all three studies, victims and perpetrators distorted to about the same extent, significantly more than control participants and especially by selective omission of uncongenial facts. Specific distortions suggested self-serving, role-based motivations. Perpetrators embellished mitigating circumstances and their own benevolent actions while omitting facts that emphasized severity and responsibility. Victims concentrated on details that described the severity of the offense and downplayed the perpetrator's positive actions.