Fifty-six 5- to 6-year-old children (29 girls, 27 boys) witnessed a confederate, acting as a janitor, either cleaning or playing with toys. An hour later, they were interviewed in succession by the janitor's "boss," by an experimenter, and by their own parent. Parents interviewed their child again 1 week later. The boss and experimenter interviewed each child in one of three ways: neutral (nonleading), incriminating (suggesting the janitor was bad and playing on the job), or exculpating (suggesting the janitor was good and doing his job of cleaning). When interviews were neutral, the children consistently gave accurate accounts of the janitor's behavior. When the interviews were suggestive, children's accounts shifted strongly in the direction of suggestion as the interview progressed. By the end of the suggestive interviews, children's accounts uniformly corresponded to the interviewers' suggestions, even when the suggestions were inconsistent with what actually happened. The effects of suggestion persisted during the two nonleading parent interviews. Notes, figures, references