This study explored preschoolers′ eyewitness testimony under conditions designed to maximize (session 1) or degrade (session 2) the quality of their event reports. In session 1, thirty-nine 3- to 4-year-olds and twenty-nine 5- to 7-year-olds interacted with Mr. Science and were immediately interviewed using nonsuggestive techniques. The children did well in this immediate interview, and nonsuggestive prompts elicited substantial amounts of new accurate information. Three months later 21 of the children heard their parents read a story about Mr. Science that described experienced and nonexperienced events preparatory to an interview in which children were asked nonleading, leading, and source monitoring questions about their experiences with Mr. Science. The children made many erroneous reports in this second interview (e.g., 41% of the 3- to 4-year-olds spontaneously reported that Mr. Science had done things that were mentioned only in the story). Patterns of errors in response to free recall, leading, and source monitoring questions are described.