Although memory for actual events tends to be forgotten over time, memory for misinformation tends to be retrieved at a stable rate over long delays or at a rate greater than that found immediately after encoding. To examine whether source monitoring errors contribute to this phenomenon, two experiments investigated subjects' memory for the source of misinformation at different retention intervals. Subjects viewed a slide presentation, read a narrative containing misinformation, and, either 10 minutes or 1 week later, completed a recognition test about details seen in the slides and about the source of these details. After the longer retention interval in both experiments, participants were more likely to agree that they had seen misleading information and were also more likely to incorrectly associate the misinformation with the slide event. Theoretical implications of these findings are considered.