In three experiments, subjects read a story, followed by a summary of the story that included both true and false information about the story, then recalled either the story or the summary. Subjects occasionally incorporated details of misleading summary statements into their story recalls. However, these statements were recalled as parts of the summary more often than as parts of the story. When recall was delayed a week, subjects were more misled if they had read the summary a few minutes earlier, compared with one week before, and more misled than they were immediately after presentation of the story. However, if they read the summary a week after the story and then recalled after a second week, they were rarely misled. The effects of delay were not only due to decay of memories, but also to subjects’ decreased confidence in their memories for the original event and the increased likelihood of accessing the older original representation when the newer representation was no longer extremely recent. It is concluded that the misleading effects were not due to alteration of a single memory representation, but to the use of either of two separate records – of the original story and of the summary – in reconstruction of the story.