Jeopardy in the Courtroom: A Scientific Analysis of Children's Testimony

The book begins with descriptions of seven cases involving children's testimony, from the Salem witch trials to the Little Rascals day care case in Edenton, North Carolina. The authors return to these cases throughout the book as illustrations of how not to interview children. They then review our inadequate knowledge of the frequency of childhood sexual abuse. There follows a brief minicourse on human memory, the basic point of which is that memory is a constructive process that can be influenced by suggestion at the time memories are retrieved. The authors then review the history of research on children's testimony and describe some typical laboratory experiments that investigate its accuracy. The basic procedure is to expose some children (the experimental group) to an unusual experience, such as a medical examination or observing a janitor deface a book, while others (the control group) are not exposed to this event. Later, children from both groups are interviewed using standardized techniques by experimenters who don't know whether the child was in the experimental or control group. The children's testimony is compared to the objective criterion of what actually happened.

The core of the book is seven chapters that describe what we know (and don't know) about interviewing children. The basic organizing principle is confirmatory bias (referred to by Ceci and Bruck as "interviewer bias"). Humans have a tendency to search for information that confirms their beliefs, rather than seeking information that might disprove their hypotheses (Snyder 1984). When interviewers start with a hunch that a child was sexually abused, they are likely to engage in a variety of behaviors that ultimately elicit answers that confirm their hunch. Consider this segment in the book from a grand jury hearing in the Kelly Michaels case (p. 121):

Prosecutor: Did she touch you with a spoon?

Child: No.

Prosecutor: No? O.K. Did you like it when she touched you with the spoon?

Child: No.

Prosecutor: No? Why not?

Child: I don't know.

Prosecutor: You don't know?

Child: No.

Prosecutor: What did you say to Kelly when she touched you?

Child: I don't like that.

By disregarding the child's initial no, the interviewer cues the child as to the desired response and creates confusion in his mind as to what his testimony had been. Once the child agrees that he was touched with the spoon, this misleading statement is incorporated into his memory of the event. The book contains many similar illustrations of suggestive questioning from interview and court transcripts. The confirmatory bias is not deliberate. It feels like a logical interviewing strategy and is characteristic of the behavior of experienced as well as naive interviewers. Researchers do not claim that children are "lying" when they respond to suggestive questioning by making false accusations.

Ceci and Bruck present laboratory and field studies that show, in detail, how confirmatory bias works. Here are some of the main points:

* Interviewers often ask leading questions that suggest a particular answer, such as "Did he touch your privates?" Such questions are more likely to elicit an incriminating response than nonsuggestive questions, such as "What did he do next?"

* When children deny having been molested, interviewers often repeat the question. When this happens, children are likely to conclude that their original denial was the "wrong answer" and switch to an affirmative response.

* Interviewers sometimes negatively stereotype the suspect by referring to the suspect as a bad person or someone who does bad things. This increases the likelihood that the child will make an accusation.

* When interviewers selectively reinforce accusations of abuse with attention and praise while ignoring denials, the child will repeat and expand upon those accusations.

* When children are told that their peers have already reported incidents of abuse by the suspect, they are more likely to accuse the suspect as well.

* Anatomically correct dolls focus the child's attention directly on what is different about these dolls. In one study (Bruck et al. 1995), when three-year-old children were given a routine physical exam that did not involve genital touching and questioned five minutes later using an anatomically correct doll, fifty percent of the children falsely reported that the doctor had touched their genitals.

* Having children imagine they might have been abused and tell a story about it ("guided imagery") may cause them to believe the story they made up.

These interviewing techniques could confuse adults, but they are particularly confusing to children. Research shows that young children are more suggestible than older children.

One irony is that the same manipulations increase the likelihood of accusations regardless of whether the event actually occurred. Suppose, Ceci and Bruck ask, you had a drug that, when given to people with cancer, would cure them but would cause cancer in healthy people. You don't know whether your patients have cancer or not. Do you give them the drug? With children's testimony, almost any procedure that increases the number of true positives increases the number of false positives as well. Should children be questioned aggressively? A pragmatist might say it depends on how strongly you suspect that the child was abused (analogous to the prior probability that the patient has cancer). But given the potential for false accusations, any use of these techniques violates the defendant's presumption of innocence.