Fourteen interviewers conducted 51 investigations of child sexual abuse with children ranging from 4.5 to 12.9 years of age. In 25 of the investigations, interviewers used a script including many open-ended utterances to establish rapport, whereas in 26 of the investigations the same interviewers used a rapport-building script involving many direct questions. Both scripts took about 7 minutes to complete. All children were asked the same open-ended question to initiate the substantive phase of the interview. Children who had been trained in the open-ended condition provided 2 1/2 times as many details and words in response to the first substantive utterance as did children in the direct introduction condition. Children in the open-ended condition continued to respond more informatively to open-ended utterances in the later (unscripted) portions of the interview. Two-thirds of the children mentioned the core details of the incident in their responses to the first substantive utterance and a further 20 percent mentioned core details more vaguely. These results demonstrate that children respond more informatively to an open-ended invitation when they have previously been trained to answer such questions rather than more focused questions, and demonstrate the sensitivity of children to the goals and expectations of forensic interviewers. Structured interview protocols also increase the amount of information provided by young interviewees. Tables, references, appendix