Was this an idiosyncratic error made by an expert who consciously misrepresented the content of her interview? The scientific literature suggests otherwise. Her error was not deliberately motivated, but reflected the limitations of our memory system. We do not remember events in detail, nor do we remember conversations on a word-for-word basis. When we try to reconstruct past experiences at a later time, no only do we fail to report some information, but our current beliefs and motivations guide our reconstructions. As a result, although much of what is recalled is accurate, significant errors can also inadvertently occur. Memories of conversations of interviews are a special instance of this phenomenon.

Recently, we and our colleagues have provided scientific support for these conclusions based on our research into adult-child interviews. We asked adults to interview children about a recently experienced event (about which the researchers had full knowledge but about which the interviewers were ignorant). These interviews were electronically recorded in order to obtain an accurate record of the exact statements made by the interviewer and by the child. Later, the interviewers were asked to recall the content of their interviews. The following consistent results have been reported in various laboratories.

Three days after interviewing their four-year-old children about a special event, mothers recalled only 35% of the details of the actual conversation.

Two weeks after interviewing four different children about special visitors to their school, mental health trainees made significant substantive errors: 40% claimed that the child had participated in a special event, when an examination of the transcripts revealed that the child had never made this report.

Ten minutes after interviewing children about a previously experienced event, highly trained and experienced interviewers did not recall a significant number of statements made by the children and they frequently reported statements that the child never made.

Thus, when asked to recall prior interviews with young children, interviewers of varying levels of expertise frequently omitted important details and also included details that were never stated by the child.

The situation regarding the accuracy with which an interviewer can recall the gist of what a child told them is worse than the above data indicate. This is because it is not sufficient for witnesses to recall only the essence or gist of what a young child told them (e.g., "According to my notes, he told me she made him drink urine.") Even if such a statement was made by the child, judges need to consider the full interrogative context in which such a statement was emerged. For example, was the statement a spontaneous disclosure to an initial, open-ended question, "Tell me everything that happened at school"? Or, was it prompted, the result of monosyllabic acquiescence to a series of suggestive questions? Was the statement a product of initial denials by the child that were eventually abandoned after repeated interviews or repeated leading questions? To make these assessments and to determine whether strategies recognized as capable of affecting the reliability and accuracy of children's reports were applied by interviewers, the trier of fact must examine a record that contains the exact wording and order of each question asked and each response supplied during each interview. This report should also contain the number of times questions are repeated and the tone of questioning.

When providing hearsay evidence, how easily can adults recall these important elements of interviews? The following example demonstrates the fallibility of reporting by one police detective whose written report provided details of his interview with a young boy who accused his parents and other adults of sexual abuse. "On other occasions, Britt said that a man would put his privates in his butt and that at the same time, a woman would make him put his mouth on her privates." The transcript of the audiotape of this same interview shows that although the detective accurately reported the gist of the interview, his hearsay testimony misrepresented the manner in which the statements were extracted from the child:

• Adult: Okay, when you were tied up Britt, on the floor, and a man was sticking his penis into your butt, was a lady doing something to you at the same time?

• Child: No.

• Adult: Would that ever happen?

• Child: Yes.

• Adult: What would the lady be doing?

• Child: I can't remember.

• Adult: Would she be doing anything with your mouth?

• Child: (pause) Yes.

In most of the studies conducted by our colleagues and ourselves, the interviewers were also asked to recall the exact words used and how the information was disclosed by the children. The results suggest that this detective's failure to report the manner in which the child's statements were obtained is common across interviewers, and that it reflects the rapid loss from memory of the exact words used, and the sequences of interactions between speakers. For example, mothers could not remember who said what (e.g., they could not remember whether they had suggested that an activity had occurred or if the child had spontaneously mentioned the activity). They could not remember the types of questions they had asked their children (e.g., they could not remember whether they had used an open-ended question or a series of leading questions to obtain a piece of information). The mental health trainees made similar types and numbers of errors when asked to recall their interviews with preschool children.

In addition, these trainees mixed up which children said what. That is, they often attributed the actual report of Child A to Child B. And the highly experienced interviewers who were questioned 10 minutes after an interview with a young child recalled that they had not asked leading questions or questions requiring any one-word answers. In fact, however, the transcripts of their audio taped interviews showed that they mainly used leading and specific questions.

In summary, serious errors occur in recall of conversations and interviews with children. These errors are made by interviewers with various levels of training and also with various levels of familiarity with the child. The errors include the omission of details (forgetting) and the commission of details (inserting facts that were not stated), as wells as misreporting the degree to which the child's answers were spontaneous or the result of suggestive techniques. In addition, interviewers often cannot recall the source of their hearsay statements; they cannot remember whether the child originally made the statement, whether the interviewer originally made the statement, and in some cases, whether another child made that statement. The last error is most likely to occur when investigators interview a number of children during the same investigation.

Our demonstrations of interviewer fallibility probably underestimate memory errors in courtrooms. First, unlike interviewers in these experimental studies who were asked to recollect a recent interview, experts and other witnesses often must reconstruct an interview that occurred months or even years previously. Memory fades with the passage of time, and therefore courtroom interviewers can be expected to do even more poorly than the interviewers in our experiments, who were tested only minutes to days after completing their interviews. Second, many experts have interviewed, evaluated, or treated hundreds of children. Our results suggest that their reports may at times reflect confusion among cases (a situation that is reported by many school principals and by pediatricians).

What are potential remedies to ensure the accuracy of hearsay testimony? One suggestion is to encourage each interviewer to keep notes or diaries that can be used to aid future testimony. However, notes and diaries are subject to a number of distortions that can include omission of important details, inclusion of inaccurate details, and, most importantly, the absence of verbatim record of each utterance produced in the interview. Usually notes only contain pieces of information that the investigator thinks are important at the moment. In fact, no interviewer can write down every word during an interview. They cannot and do not write down every question asked, especially ones that failed to produce a response, or produced an undesired response. If the investigator has a bias that the chid was a victim, participant, or observer of a crime, this could color his or her interpretation of what the child said or did; and it is this interpretation that appears in the notes/diary rather than a factual account of what transpired.

The results of the scientific studies reported above support these conclusions. First, explicitly warning interviewers to remember all details and words of an upcoming interview with a child does not influence the accuracy of their subsequent reports of the interview. Second, errors are made even when interviewers are encouraged to make careful notes. In the most naturalistic of the current studies, audio taped transcripts of investigatory interviews with sexually abused children were compared to handwritten notes that interviewers took during the interview. The notes were inaccurate; there were omissions of important details, and there were frequent misreports that the child's statements were spontaneous when, in fact, they were produced by repeated questions.

The most significant message to be drawn from this work is that interviewers should be mandated to electronically preserve all (and especially the very first) of their interviews with children. If courts are interested in historical accuracy, there is simply no substitute for a tape that can be played to verify the accuracy of the witness's recall and the details of the discussion that took place between the interviewer and child. Although there may be times when it is not feasible to electronically record interviews (specifically when parents question their children at home or in the car), it is nevertheless important for jurors and judges to know how to interpret hearsay testimony and to consider the potential for different types of errors, even though the testimony may be compelling and be offered in good faith. Finally, it might be argued that electronic records should be mandated for interviews of adults as well as children. At present, however, there is only sufficient scientific evidence and legal cause to support the recommendation in the case of children - - who often cannot and do not provide courtroom testimony to support or rebut the hearsay testimony provided by their adult interviewers.