Developmental differences in children's learning and use of forensic ground rules during an interview about an experienced event.

Interviewers

Three female psychologists conducted the interviews. All had received training in use of the NICHD Investigative Interview Protocol. Deirdre A. Brown conducted frequent fidelity checks for adherence to the Protocol. CWIDs were recruited, in the main, from specialist schools, and so interviewers were typically aware of whether children had an intellectual disability (although not of the severity of impairment). Whether children were in the MA or CA matching group was evident from their age and class groups.

Interview

The interviews were conducted at school one week after the event by the same interviewer who administered the cognitive tests. Each interview began with presentation of the ground rules (see below) and then proceeded to a combined rapport building and narrative practice phase using open-ended questions exploring what the children had done prior to the interview.

Ground rules

Each of the ground rules was explained in turn using the scripted instructions from the NICHD Investigative Interview Protocol ( Orbach et al., 2000 ; see http://nichdprotocol.com/ ). The three rules were to (a) only tell the truth, (b) say “I don’t know” when appropriate, and (c) correct the interviewer if she made a mistake (see Table 2 ). The explanation of each rule was accompanied by an example, and then the child was asked to practice using each rule. Any children who did not respond correctly with the appropriate rule were given a second practice using the rule. If children still failed to correctly apply the rule after two practices, the rule was repeated verbally, and the interviewer progressed to the next rule (or the beginning of the interview if it was the final rule). Children were also instructed to tell the interviewer if they did not understand a question but this rule was not practiced.

Interview about the staged event

Focus was shifted to the staged event using a series of progressively informative prompts to help orient the children to the event the interviewers wanted them to talk about. The interview progressed using the prompts and structure outlined in the NICHD Protocol. Open invitations (e.g., “tell me about that time”) were used to encourage children to provide as much detail as possible. Children were encouraged to report further details using a variety of different prompts. Information reported by the children was used to form cued invitations (e.g., “you mentioned you got to choose a plaster; tell me more about choosing the plaster”), and children were also asked direct questions (open ended “wh-”, e.g., “which plaster did you choose?”), and option-posing questions, if needed, to clarify unclear or contradictory information (e.g., “did you or your partner wear the bandage first?). Interviewers were trained to follow responses to directive or option-posing prompts with open prompts (e.g., “tell me more about that”). Suggestive prompts reflected interviewer error (e.g., introducing information the child had not provided).

Scripted suggestive prompts

After the children indicated that they could not recall anything further they were asked 16 suggestive (leading and misleading) questions. Questions also varied depending on whether they were closed (e.g., “Were you in the blue group?”) or open , requiring the children to generate the information (e.g., “What color was the group you were in?”). Finally, questions varied depending on whether they assessed central or peripheral details about the event. Children were asked one of 12 sets of questions; across the sets we counterbalanced whether each topic (central or peripheral) was probed with a leading or misleading and closed or open questions.

Every interview was transcribed verbatim from the digital video recordings. All interviewer and child utterances were included.

Ground rules

Two coders (blind to the hypotheses) coded the children’s responses to the ground rules. Children were given a score of 2 if they correctly responded to the practice item for a rule (for a maximum score of 6 across the three rules). One point was given if children required the second practice example, but then correctly responded to that item. A score of zero was given to children who did not demonstrate correct use of the rule after two practice trials. Thus, the range of possible scores totaled across the three rules was 0–6. Each coder coded every transcript, and differences were resolved by discussion. Reliability was again high with a mean kappa of .91.

A different research assistant (also blind to the study hypotheses and to the accuracy of children’s responses) coded the use of the ground rules during the interviews and the suggestive questioning. Children’s use of the explicit language or behavior targeted by each rule was scored, as were alternative behaviors that could be considered as implicit, rather than explicit, demonstrations of the rule. The range of responses coded, with associated definitions and examples, are presented in Table 3 . A second assistant coded 40% of the transcripts for reliability. Reliability was high (Cohen’s κ = .88 during the interviews and .85 in response to suggestive questions).

Relationships Between Ground Rules Performance

We examined whether performance on each ground rule was related to the other two. There were no significant associations (in all cases, r < .08) between children’s success at responding to the practice trials of each rule. Concerning Research Question 2, grasping one rule appeared not to be related to understanding another, supporting the second of our proposed relationships. The bottom two rows of Table 5 report the children’s accuracy both when describing the experienced event when interviewed using the NICHD Protocol, and in response to the 16 scripted suggestive questions at the end of the interview. They show that, with some variations between groups, 85% of the children’s interview responses were correct but that accuracy dropped to 59% for responses to the suggestive questions.

Understanding Rule Use, Using Rules, and Accuracy in the Interview

The rest of the analyses unpacked Research Question 3. Performance on each of the three ground rule questions was correlated with the accuracy of the children’s accounts, r (98, 96, 1 and 98, respectively) = .23, .39, and .28, all p ≤ .02, for the three respective rules, and with the accuracy of the children’s responses to the suggestive questions at the end of the interview (respectively, r [98, 96, and 98, respectively] = .23, .25 and .25; all p ≤ .02). To reduce the variables, we collapsed the various response behaviors that were consistent with nonsubstantive responses to the interviewers’ questions (“don’t know,” “not sure,” “can’t remember,” and “don’t understand”), and corrective responses (corrects interviewer, and refutes interviewer), in part because many of the individual response categories were infrequently observed. Table 5 reports the findings for use by children in each group of these two overarching response categories in the main interview and in response to the suggestive questions. We next examined whether children’s nonsubstantive and corrective responses differed by Group and performance on each of the learning trials at the beginning of the interview. We conducted a series of repeated measures ANOVAs with response type (nonsubstantive vs. corrective) as the within-subjects factor and Group as the between-subjects factor. We then added performance on each rule as a continuous covariate. Given the positive skew for three of the four response categories included in Table 5 , preliminary analyses confirmed that parametric analyses could be performed (see footnote for Table 5 ; raw means are presented).

In the first analysis on the effects of utterance type and group, there was a significant difference in response type, F (1, 94) = 319.7, p < .001, η p 2 = .77: Children gave more nonsubstantive responses ( M = 10.41, SD = 8.00) than corrective responses ( M = 1.22, SD = 1.63). There was also a significant effect of Group, F (4, 94) = 3.45, p = .01, η p 2 = .13: Tukey-Kramer tests showed that children in the CA group gave fewer responses (irrespective of type) than children in the MA-Moderate group ( p < .05). When we added each of the measures of rule understanding, none made significant contributions to the model (Truth or lie: F [1, 93] = 0.04, p = .85, η p 2 = .00; Do not know: F [1, 93] = 0.48, p = .49, η p 2 = .01; Tell the truth: F [1, 93] = 0.00, p = .97, η p 2 = .00), nor were any interactions significant. Thus, our hypothesis relating to understanding of the ground rules and their subsequent use was not supported.

A similar set of analyses exploring children’s use of nonsubstantive and corrective responses during the scripted suggestive questions phase used raw data because the distributions allowed this. When Response type and Group were entered into the model, there was a main effect of Response Type, F (1, 94) = 19.7, p < .001, η p 2 = .17: Children made more corrective ( M = 5.14, SD = 2.4) than nonsubstantive responses ( M = 3.13, SD = 2.9). The effect of group was not significant, F (4, 94) = 2.21, p = .07, η p 2 = .09, but the Group × Response Type interaction was F (4, 94) = 2.6, p = .04, η p 2 = .1. To unpack this interaction, we conducted univariate analyses on each response type. For corrective responses, the main effect of Group was significant, F (4, 94) = 5.61, p < .001, η p 2 = .19, and Tukey’s tests showed that children in the Moderate ID group gave fewer responses than those children in each of the other groups (Tukey p s < .05). For the nonsubstantive responses, the effect of Group was not significant, F (4, 94) = 0.75, p = .56, η p 2 = .06. When we added each ground rule score separately, the main effect of Response Type and the Group × Response Type interaction remained significant. The Ground Rules “truth or lie” and “don’t know” did not add to the models (in both cases the main effects and interactions, F < 1), but the use of “correct me” × Response Type interaction did, F (1, 94) = 4.09, p = .03, η p 2 = .04. Follow up analyses showed that the main effect of corrective utterances showed an effect of a grasp of the “Correct me” rule, F (1, 94) = 5.04, p = .02, η p 2 = .11, but the children’s use of the nonsubstantive rule did not, F (1, 94) = 2.07, p = .15, η p 2 = .02. Thus we found partial support for our hypothesis that ground rule performance would be associated with higher rates of rule related responding.