Rapport: Little effect on children’s, adolescents’, and adults’ statement quantity, accuracy, and suggestibility

Abstract

Rapport building is widely recommended in eyewitness interview situations and is a critical component in some interview protocols. However, empirical evidence on the effectiveness of rapport building on memory performance is scant. The current experiment examined the effects of different levels of rapport (none, minimal, extensive) on children’s, adolescents’, and adults’ memory (N = 229). Participants viewed a video of a mock theft and received one of three possible rapport manipulations. They then provided a free narrative of what they had seen, followed by 18 cued (suggestive and nonsuggestive) questions. In general, we found limited evidence of positive effects of rapport building on statement quantity and accuracy across age groups. Adolescents did profit more from extensive rapport building compared to no rapport. In line with the idea of a linear development of memory measures with age, adolescents generally fell in-between the other two age groups across different memory measures. The current study encourages systematic experimental research on the effect of rapport building on eyewitness memory.

Keywords:

• Developmental trend
• eyewitness
• interview
• memory performance
• retrieval support

Putting a witness at ease in an interview situation and thereby creating rapport is an essential part of interview guidelines developed with the aim to elicit accurate and complete reports about potentially stressful events. Two commonly used and well-validated structured interview protocols start with an initial rapport building phase: The National Institute of Child Health and Human Development Investigative Interviewing (NICHD) protocol (Lamb, Hershkowitz, Orbach, & Esplin, 2008) used with children and the Cognitive Interview (Fisher & Geiselman, 1992) used primarily, but not exclusively with adult witnesses (McCauley & Fisher, 1995).

Concrete definitions of rapport are, however, rather vague and vary, as do operationalizations (Saywitz, Larson, Hobbs, & Wells, 2015; Vallano & Schreiber Compo, 2015). According to a recent survey, law enforcement interviewers agree that rapport refers to establishing a positive relationship between interviewer and interviewee by the use of verbal (e.g., discussing neutral events not related to the target event) and nonverbal techniques (e.g., showing empathy; Vallano, Evans, Schreiber Compo, & Kieckhaefer, 2015). Given the consensus between scientists and practitioners that rapport should be established in an interview situation, it is surprising that there is a lack of knowledge concerning its effectiveness as a whole or its individual components (e.g., verbal or nonverbal techniques) on memory performance measures (e.g., Larsson & Lamb, 2009; Saywitz et al., 2015).

Thus far and to the best of our knowledge, only four studies experimentally manipulated different levels of rapport and tested its effects on memory performance in adults and three in children. In adults, open rapport building questions (e.g., Tell me about your family) that conveyed interest in the interviewee (Vallano & Schreiber Compo, 2011), together with active listening and using the interviewee’s name (Kieckhaefer, Vallano, & Schreiber Compo, 2014) reduced the effects of misinformation on memory, compared to no-rapport conditions in which interviewees were only asked demographic questions as an opening. These positive effects, however, were only observed when rapport was established prior to the delivery of misinformation, not afterwards (Kieckhaefer et al., 2014). Free recall, elicited by friendly, relaxed interviewers who addressed the interviewee by name led to more correct information—but equal amounts of incorrect information—compared to abrupt and neutral rapport conditions (Collins, Lincoln, & Frank, 2002). In the abrupt condition, the interviewer used a harsher tone, did not address the interviewee by name, maintained a stiff body posture, and displayed lack of interest. In the neutral condition, the interviewer displayed a neutral voice tone, body posture, etc. Similarly, speaking in a friendly tone and encouraging participants to talk openly increased the number of accurate details (Nash, Nash, Morris, & Smith, 2016, Experiments 1 & 2) and decreased the number of incorrect details (Experiment 1), compared to a no rapport condition, in which the interviewer spoke with a neutral tone and simply noted participants’ responses without probing.

In children (3–9 years), direct rapport (using closed questions such as How old are you?) vs. open-ended rapport (using open-ended prompts such as Tell me about yourself) did not lead to differences on a statistical level in statement quantity (Roberts, Lamb, & Sternberg, 2004). Open-ended rapport did have a beneficial effect on the accuracy, that is, the proportion of accurate information reported, of answers given to cued questions, but not free recall accuracy. Of note, the open-ended rapport building phase was substantially longer (16 min on average) than the direct rapport building phase (6 min on average). We will discuss possible detrimental effects of relatively long rapport building phases further below. In another study, rapport consisted of a narrative rapport building phase with open questions that was or was not supplemented by event recall practice or a non-narrative rapport building phase using directive questions (e.g., who and where questions) supplemented by event recall practice,¹ resulting in three conditions (Brown et al., 2013). Children (5–7 years) reported more information per prompt in the most elaborate rapport condition (narrative rapport building + event recall practice), compared to narrative rapport only, while the moderate rapport condition (non-narrative rapport building + event recall practice) did not statistically differ from either of the other two. Accuracy was not consistently affected as a function of rapport condition. The third study suggested that different types of rapport building might be beneficial for different age groups (Hardy & van Leeuwen, 2004). More specifically, younger children’s (3–5 years) reports, compared to older children's (5–8 years), were less detailed and less accurate when indirect rapport prompts (e.g., Can you tell me who was there?) were used. This age difference could be compensated by the use of direct prompts (e.g., Who was there?) that is, age differences were not observed in this condition.

The mechanisms underlying the assumed positive effects of rapport building are still debated. Proposed candidates include increasing interviewee’s motivation to perform well and facilitating memory retrieval by inducing a talkative mood early in the dyadic interaction (e.g., Vrij, Hope, & Fisher, 2014). These mechanisms, however, could also account for detrimental effects of rapport building found in other studies. For example, questions that targeted neutral information about children’s lives and longer duration spent on the rapport building phase led to less informative accounts in young children in one study (Teoh & Lamb, 2010) and may have undermined positive effects of open-ended rapport in another (Roberts et al., 2004). A relatively long and cognitively demanding rapport phase may have interfered with interviewees’ attention span, leading to a diminished subsequent account of the target event. In fact, in one study, a rapport building phase of more than eight minutes resulted in less informative accounts than a shorter rapport building phase (Davies, Westcott, & Horan, 2000). Additionally, an extensive rapport building phase might lead to acquiescing behavior to please the interviewer and therefore to less accurate reports, especially in children (Saywitz et al., 2015). A recent study provides support for this idea (Wright, Nash, & Wade, 2015): when adults were interviewed by a friendly, interested interviewer, which could be—to some extent—considered as extensive rapport building, they more readily falsely corroborated a co-participant to have cheated than when interviewed by a cold uncommunicative interviewer.

In summary, rapport building has the potential to be a useful tool to improve children’s and adults’ eyewitness recall. However, the current database is still thin and recent reviews have identified several gaps in the literature that need addressing (Saywitz et al., 2015; Vallano & Schreiber Compo, 2015). One of those concerns the systematic study of the effects of rapport on adolescents’ memory. Indeed, adolescents have not received much attention in the eyewitness literature (see also Brackmann, Otgaar, Sauerland, & Howe, 2016; Jack, Leov, & Zajac, 2014; McGuire, London, & Wright, 2015). In the current study, we therefore set out to investigate the effect of different levels of rapport on the quality and quantity of statements made by children, adolescents, and adults. Participants watched a stimulus film depicting a mock-crime. Following a short delay of less than five minutes, they were interviewed using either no, minimal or extensive rapport building. Based on the literature, the three conditions differed on critical issues concerning verbal and nonverbal rapport dimensions that have been used in previous work in the field. These included the way the participant and interviewer were seated (opposite or sideways) and the interviewer’s posture (closed, neutral, or open; Saywitz et al., 2015; Vallano et al., 2015). Furthermore, we manipulated the way the interviewer introduced themselves (mentioning their name: yes or no; Rush et al., 2014), addressed the participant (by their name: yes or no; Collins et al., 2002; Kieckhaefer et al., 2014; Vallano et al., 2015), and whether they offered sweets during the interview (yes vs. no). Prior to a free recall prompt, the use (yes or no) and format of rapport building questions was varied (none vs. closed vs. open-ended questions; Kieckhaefer et al., 2014; Lamb & Brown, 2006; Roberts et al., 2004; Saywitz et al., 2015; Sternberg et al., 1997; Vallano & Schreiber Compo, 2011). Finally, different explanations were used regarding the purpose of the audio recording (evaluation purposes vs. none vs. cannot write that quickly). An overview of the operationalization of the three conditions can be found in Table 1.

Table 1. Overview of three rapport building conditions.

	No rapport	Minimal rapport	Extensive rapport
Seating arrangement	Sitting opposite of each other	Sitting opposite of each other	Sitting sideways
Posture	Closed posture (arms and legs crossed)	Neutral posture (legs parallel, arms in lap, script on table)	Open posture (legs parallel, holding script in one hand, perhaps gesturing)
Offering sweets	None	None	Smarties offered
Introduction; addressing participant	“Hello. I am from the university, and I am interested in your memory.”	“Hello. I am from the university, and I am interested in your memory.”	“Hello [participant name]. I am [interviewer name] and I am from Maastricht University. I am interested in your memory. Please take a seat. You can take some smarties if you like to.”
Rapport building questions	No rapport building question (longer filler task: Tetris)	Direct rapport building questions	Open-ended rapport building questions
Justification audio recording	“I’ll audio record the testimony, because this makes it easier for me to evaluate it.”	“I’ll audio record the testimony.”	“I’ll audio record the testimony, because I can’t write it down that quick.”

There are some indications that adolescents may have different needs than other age groups when it comes to establishing rapport with the goal to facilitate memory retrieval. Legal practitioners have pointed out that interviewing adolescents can be challenging (Collins, Doherty-Sneddon, & Doherty, 2014). They can be reluctant to talk to police officers who try to build rapport with them and sometimes only disclose little information voluntarily. Their behavior seems to be guided by emotional arousal and skepticism and more so than for example adults. In addition to these social particularities, adolescents’ memory capacity, especially in stressful and emotional situations such as witnessing a crime, may diverge from that of children and adults (Quas et al., 2016).

Based on these considerations, we expected more intense rapport building to generally have a positive effect on report quantity and accuracy (hypothesis 1). For adolescents, however, extensive rapport may lead to poorer outcomes (less quantity and accuracy) than minimal rapport, compared to the other two age groups (hypothesis 2). Furthermore, in an attempt to please the interviewer, children may display a stronger tendency to give in to suggestions in the extensive rapport condition than adolescents and adults (Hypothesis 3; Ceci & Bruck, 1993; Lamb & Brown, 2006; Saywitz et al., 2015). Finally, we expected to replicate the effect of age on memory outcome measures (i.e., quantity and accuracy), with performance increasing as a function of age (Bjorklund & Muir, 1988; Fitzgerald & Price, 2015; Goodman & Reed, 1986; Howe & Brainerd, 1989; Jack et al., 2014; Sternberg et al., 1997; Hypothesis 4).

Method

Participants

Two-hundred-twenty-nine participants took part in the current experiment. The participants consisted of 75 children (range: 6–10 years, M_age = 7.82, SD = 1.20), 77 adolescents (range: 12–17 years, M_age = 14.33, SD = 1.26), and 77 adults (range: 18–27 years, M_age = 20.51 years, SD = 1.77). Children (41 females) were tested in a German primary school. Adolescents (29 females) were tested in a German secondary school. Adults (68 females) were students who were tested at Maastricht University. An a priori power analysis using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007) with a power of .85 and an estimated medium effect size of f = .25 resulted in an anticipated sample size of 220. We tested slightly more participants to reach equal group sizes and to account for possible dropouts. All adult participants or legal guardians (i.e., of children and adolescents) provided consent prior to participation in the experiment. Children and adolescents provided oral consent prior to being tested. The study was approved by the local ethical committee of the faculty. All data are available via https://hdl.handle.net/10411/P65LNI.

Materials

Stimulus film

Participants were shown a 3:20-minute stimulus video depicting the nonviolent theft of a wallet in a university building. The content of the video can be described as follows: A man (man 1) pins a world map to one of the walls. When he takes a step backward, a woman passes by, so they run into each other. In the next scene, another young man (man 2) gets coffee out of a vending machine. The subsequent scene depicts a workplace. The woman and man 2 are now sitting at two different tables. Both are working quietly. Man 3 is standing in front of a window and typing on his cell phone. He is looking around. When the woman receives a phone call, she leaves the workplace. Man 3 approaches her bag, takes her wallet, and leaves. In the next scene, he takes the money out of the wallet, and throws the wallet into a garbage bin. Back at the workplace, to which the woman returns, she looks for her wallet and cannot find it.

Rapport building

In the no rapport building condition, the interviewer greeted the participant in a formal, neutral way (not introducing her-/himself or addressing participants by their name), sat vis-à-vis to the participant, with a closed posture (arms and legs crossed), gave no verbal feedback, and focused solely on the target information. Interviewers asked no personal questions. Participants were informed that the statement was going to be audio-recorded to “evaluate it.” This was assumed to increase the perceived stressfulness of the situation.

In the minimal rapport building condition, interviewer and participant still sat opposite of each other, but the posture of the interviewer was neutral and open (legs and arms parallel and relaxed). The interviewer did not introduce her- or himself and did not address participants by their name, but did ask a number of predetermined personal questions about participants’ age, favorite food, movie, and game. These questions were selected from Roberts et al. (2004). Participants were informed that the statement would be audio-recorded without further explanation.

For the extensive rapport condition, chairs were arranged next to each other and the interviewer took an open and engaging body posture. This setup enabled participants to seek eye contact whenever they wanted to, but also allowed them to avoid it when they felt insecure or uncomfortable. Interviewers introduced themselves and addressed participants by their name during the course of the session. The interviewer provided nonverbal feedback (e.g., nodding), offered sweets, and asked four open-ended personal questions (e.g., prompt to tell something about themselves; question about preferred free-time activities, follow-up questions), based on the NICHD protocol (Lamb et al., 2008). Participants were informed that the statement was going to be audio-recorded because the interviewer “cannot write that quickly.” An overview of differences between the rapport conditions can be found in Table 1.

Interview questions

Participants were asked 18 cued questions regarding the incident. Eight of them were suggestive. Questions predominantly concerned the victim’s as well as the perpetrator’s appearance and actions. Three questions dealt with the environment and two with the witnesses. The questions can be found in Appendix A.

Questions on experimenters’ and participants’ subjective impression of interview

In one item, experimenters rated how they had experienced the atmosphere on a 5-point scale ranging from stressed to relaxed. Two items were used to assess participants’ subjective impression of the interview. The wording of the items was How friendly was the experimenter? (ranging from very unfriendly to very friendly, rated on a 5-point scale) and How comfortable did you feel during the experiment? (ranging from very uncomfortable to very comfortable, rated on a 5-point scale).

Design

A 3 (age group: children vs. adolescents vs. adults) x 3 (rapport condition: none vs. minimal vs. extensive) quasi experimental between-subjects design was employed. Participants in the three age groups were randomly assigned to one of three rapport conditions. Experimenters were trained by the principal investigator and adhered to a strict protocol.

Dependent measures

From participants’ free recall accounts, we examined the total number of details reported (free report quantity), number of accurate details reported, number of inaccurate details reported, and rapport accuracy (quantity correct/quantity total). For the cued questions, we looked at the number of questions answered accurately and the number of suggestive questions that were answered incorrectly. Finally, we looked at the effect of the independent variables on experimenters’ subjective impression of the atmosphere and participants’ subjective impression of the interview (experimenter friendliness and feeling comfortable during experiment).

Procedure

Throughout the testing session, the interviewer followed the procedure for establishing no, minimal or extensive rapport as previously described. Participants watched the stimulus film, followed by a filler task (playing Tetris). Then, participants were asked to imagine they were at a police station and that they had witnessed the incident depicted in the video. They were asked to imagine that the interviewer was a police officer, who would ask questions regarding the incident. A free recall account was prompted using the following instructions: “Report everything as detailed as possible. If you are not sure about something please answer with ‘I don’t know’. Take your time. Is everything clear to you? Time for free recall was unlimited. During participants’ report, interviewers did not interrupt the account. When the participant paused for some time, or indicated to be done, the interviewer asked if they wanted to add anything else. When participants had finished their free report, they were asked the 18 cued questions. Finally, the participant was provided with a sheet with two questions regarding participants’ subjective impression of the interview. The participant put the sheet into a box that was placed on the table. Experimenters rated how they had experienced the atmosphere. Participants were thanked for their participation and asked not to tell others about the content of the testing session. For children and adolescents, a verbal debriefing took place with the whole class at the end of every testing day and debriefing sheets were handed out. Any remaining questions were answered and participants were given an eraser as a token of appreciation. Adult participants received a written debriefing form after testing and were given the opportunity to ask questions for clarification (either immediately or by contacting the principal investigator). Adult participants were granted study credit or compensated with a 5€ gift voucher.

Coding of free recall statements

A coding scheme comprising 216 details was developed for coding the quantity (sum of correct and incorrect details) and accuracy (number of correct details divided by quantity) of participants’ free recalls. The statements were coded by four coders. To establish inter-rater reliability, for each age group, two of the four trained coders independently coded 19 statements (i.e., 57 statements in total, or 25% of all statements). Details were coded as correct or incorrect if they did or did not match the content of the stimulus film, respectively. Subjective details (e.g., ‘beautiful shirt’) were not coded. Interrater reliability was found to be almost perfect (Landis & Koch, 1977). Specifically, for children’s, adolescents’, and adults’ statements, Cohen’s κ were .84, .83, .82, respectively, ps < .001.

Results

Experimenters’ and participants’ subjective impression of the interview

To test whether experimenters’ experience of the atmosphere differed as a function of the experimental condition, we computed a two-way ANOVA with age group (adults vs. adolescents vs. children) and rapport condition (none vs. minimal vs. extensive) as factors. The main effect of rapport condition was statistically significant, F(2, 218) = 39.31, p < .001, η_p² = .27, whereas the main effect of age and the interaction were not, ps ≥ .291. Post-hoc comparisons using Tukey correction showed that all three rapport conditions differed statistically significantly from each other (ps < .001), with moderate to large effect sizes throughout. In the extensive rapport condition, the atmosphere was perceived as most relaxed (M = 3.09, SD = 0.81), in the no rapport condition as least relaxed (or most stressful; M = 1.96, SD = 0.80, d_{extensive-none} = 1.41). The minimal rapport condition fell in-between (M = 2.46, SD = 0.74, d_{extensive-minimal} = 0.82; d_minimal-none = 0.65).

Analogous analyses were computed for participants’ impression of experimenter friendliness and feeling comfortable. No significant main effects or interactions were found, all ps ≥ .089.

Of note, impression of the atmosphere or experimenter friendliness was not statistically significantly correlated with any of the outcome measures, ps ≥ .137. There was a significant positive correlation (with a small effect size) between feeling comfortable during the experiment and free report accuracy, r = .21, p = .002. All other effects were not statistically significant when correcting for multiple tests (alpha = .05/6 = .008; ps ≥ .049).

Effect of rapport and age on free report

To investigate free report performance as a function of age group and rapport conditions, we calculated two-way ANOVAs. Separate analyses were computed for free report quantity, number of reported accurate details, number of reported inaccurate details, and free report accuracy. In the following, we only report statistically significant main effects and interactions.

For free report quantity, F(2, 220) = 152.76, p < .001, η_p² = .58, and number of reported inaccurate details, F(2, 220) = 8.07, p < .001, η_p² = .07, a significant main effect of age was found. Post-hoc comparisons showed that all three age groups differed statistically significantly in report quantity (ps < .001), with large effect sizes throughout. Adults provided the most detailed (M = 37.12, SD = 7.73), children the least detailed reports (M = 13.95, SD = 7.78, d_{adults-children} = 2.99). Adolescents fell in-between (M = 22.91, SD = 8.43, d_{adults-adolescents} = 1.76; d_{children-adolescents} = 1.10).

Adults also reported significantly more incorrect details (M = 2.10, SD = 1.73) than children (M = 1.23, SD = 1.39, d = 0.55) and adolescents (M = 1.44, SD = 1.02, d = 0.46; ps ≤ .011). Effect sizes were of moderate magnitude. Children and adolescents did not differ significantly (p = .616, d = 0.17).

A statistically significant Age Group x Rapport Condition interaction was found for the number of reported correct details, F(4, 220) = 2.49, p = .044, η_p ² = .04. Separate analyses for each age group revealed a significant effect of rapport condition for adolescents, F(2, 74) = 3.39, p = .039, η_p²= .08, but not the other two age groups, Fs(2, 72/74) ≤ 1.62, ps ≥ .205, η_ps² ≤ .04. Post-hoc comparisons for adolescents revealed a tendency for more accurate details to be reported in the extensive rapport condition (M = 24.84, SD = 7.24), compared to the no rapport condition (M = 19.69, SD = 7.30), p = .058, d = 0.71. Comparisons with the minimal rapport condition (M = 20.00, SD = 9.00) were statistically nonsignificant, ps ≥ .080, ds ≤ 0.59. No statistically significant effects were found for free report accuracy, Fs(2/4, 218) ≤ 2.24, ps ≥ .070, η_ps² ≤ .04. Accuracy rates varied between 92.3% (SD = 8.4) in children and 94.4% in adults (SD = 4.4; adolescents: M = 93.5%; SD = 4.9) with a total mean of 93.4% (SD = 6.1).

Apart from a statistically significant Age Group x Rapport Condition interaction for the number of reported correct details, all main effects of rapport condition or interactions with age group were not statistically significant. To examine whether our data were more in line with the null than the alternative hypothesis, we conducted separate Bayesian ANOVAs for each dependent measure. Using JASP (JASP Team, 2018; version 0.8.4) and a default prior, Bayes Factors₀₁ varied between 4.62 and 20.74. This means that the data are between 4 and 20 times more likely to favor the null than alternative hypothesis.

Effect of rapport and age on cued questions

The final set of analyses concerned recall performance in response to the cued questions as a function of age group and rapport condition. Separate ANOVAs were computed for the total number of questions answered correctly and the number of suggestive questions answered incorrectly.

Statistically significant main effects of age group were found for both dependent measures, Fs(2, 220) ≥ 97.80, ps < .001, η_ps² ≥ .47. Post-hoc comparisons showed that all three age groups differed statistically significantly from each other in both dependent measures, with moderate to large effect sizes throughout (ps < .001). More specifically, adults answered about 13 cued questions accurately (M = 13.34, SD = 2.12), adolescents about 12 (M = 11.56, SD = 2.66, d_{adults-adolescents} = 0.74), and children only seven to eight (M = 7.48, SD = 2.87, d_{children-adolescents} = 1.47; d_{children-adults} = 2.32). Adults also gave in to only about one of the eight suggestive questions, (M = 1.30, SD = 1.08), adolescents to approximately two (M = 2.37, SD = 1.44, d_{adults-adolescents} = 0.84), and children to four to five (M = 4.45, SD = 1.34, d_{children-adolescents} = 1.34; d_{children-adults} = 2.26). To test whether the overall accuracy effect could be attributed to increased suggestibility in children, we reran the analysis, discounting of the eight suggestive questions. The effect remained the same (ps < .001), with adults answering most of the ten nonsuggestive cued questions accurately (M = 7.74, SD = 1.47, d_{adults-children} = 1.67; d_{adults-adolescents} = 0.70), whereas children answered least accurately (M = 4.85, SD = 1.95; adolescents: M = 6.60, SD = 1.77, d_{children-adolescents} = 0.94).

In an explorative analysis, we also tested the effect of age group and rapport condition on the frequency of don’t know answers provided in response to cued questions. This age effect was statistically significant, F(2, 220) = 5.23, p = .006, η_p² = .06. Post-hoc comparisons showed that adolescents (M = 2.26, SD = 1.35) provided significantly fewer don’t know responses than children (M = 3.05, SD = 1.75, p = .004, d = 0.51). Adults fell in-between and did not differ significantly from either children or adolescents (M = 2.60, SD = 1.46, ps ≥ .157, ds ≤ 0.28).

Bayes Factors₀₁ for the effect of rapport varied between 6.92 and 17.34. This means that the data are between 6 to 17 times more likely to favor the null than alternative hypothesis.

Discussion

Using three different operationalizations of rapport building, we investigated the effect of rapport intensity on the quality and quantity of witness reports and suggestibility to suggestive questions from a developmental angle testing children, adolescents, and adults. We expected more intense rapport building to have a positive effect on report quantity and accuracy (Hypothesis 1). This effect was expected to be qualified in two respects: we hypothesized extensive rapport to lead to poorer outcomes than minimal rapport in adolescents (Hypothesis 2) and to increase suggestibility in children (Hypothesis 3). We also hypothesized performance to increase as a function of age (Hypothesis 4). In short, only hypothesis 4, the well-established effect of age on statement characteristics, was supported (e.g., Goodman & Reed, 1986; Jack et al., 2014). Generally, adolescents fell in-between the other two age groups across different memory measures, in line with the idea of a linear development of memory measures with age (Brainerd, Reyna, & Ceci, 2008; Jack et al., 2014). Adolescents’ free report contained more details than children’s, but fewer than adults’. They reported fewer incorrect details than adults, while the difference compared to children was statistically nonsignificant. This age-related increase in reporting inaccurate details explains why there was no age effect on free report accuracy. This finding is in line with previous work on developmental reversal effects. This work has revealed that because adults have a richer knowledge base than adolescents and children, they are most likely to make spontaneous (memory) errors (Brainerd et al., 2008; Otgaar, Howe, Merckelbach, & Muris, in press). In response to cued questions, adolescents answered more questions accurately than children, but fewer than adults. Likewise, adolescents gave in to more suggestive questions than adults, but to fewer than children; an effect which aligns well with the idea that suggestibility follows an age-related decrease across development with children being most and adults being least susceptible to suggestion and adolescents falling in between (Ceci & Bruck, 1993; but see Otgaar et al., in press).

The only outcome measure that did not show a linear age trend (apart from report accuracy, where no age effect was found at all) concerned the number of don’t know answers given in response to cued questions. Descriptively, adolescents gave don’t know answers least often, and this difference was statistically significant compared to children, but not compared to adults. This is in line with a study in which adolescents had difficulties in rejecting a lineup that contained a bystander, compared to adults and children (Brackmann, Sauerland, & Otgaar, 2018). This outcome may be due to reduced executive control processes (Luciana et al., 2005) and more impulsive decision making in adolescents than children (Gardner & Steinberg, 2005; Nigg, 2000; Steinberg et al., 2008).

Focusing on the effects of rapport building, neither the expected positive effects on memory outcome measures (Hypothesis 1), nor the expected moderations (Hypotheses 2 and 3) were evident. The only observed rapport effect concerned the number of accurate details reported by adolescents during free report. Adolescents profited more from extensive rapport building, compared to no rapport. One possible explanation for the lack of an effect of our rapport manipulation on memory measures might be related to our control group implementation. Possibly, building a positive or working relationship (i.e., building rapport) is beneficial only in comparison to a negative relationship, but not a neutral condition, similar to the one we adopted in the current study (Vallano & Schreiber Compo, 2015). In other words, the difference between a positive rapport vs. a neutral rapport condition may not be substantial enough to exert an effect on memory performance. Indeed, our no rapport control group still involved a generally friendly and accommodating interviewer. This might be regarded as some mild form of rapport building and is more likely to be seen in actual interviewing settings than control groups used in previous studies.

Indeed, the adult study reporting the largest effect sizes² (Collins et al., 2002) used an abrupt interviewer attitude condition in which the interviewer would display a harsh tone and lack of interest. In contrast, studies with adult participants with less extreme control groups reported small to moderate effects of rapport (Nash et al., 2016; Vallano & Schreiber Compo, 2011) that were qualified by interactions with other factors, such as timing of rapport (Kieckhaefer et al., 2014). Furthermore, research with children generally included milder differences across rapport building conditions (e.g., direct vs. indirect probes; open-ended vs. directive questions; with vs. without recall practice) and no study to date used harsh behavior towards child interviewees. Obviously, such an implementation would raise ethical issues. Similar to results with adult participants, performance differences across different rapport conditions with child interviewees were small to moderate (Brown et al., 2013) or limited to specific situations. For example, one study found an effect on accuracy for cued questions, but not free recall accuracy or free recall/question quantity measures (Roberts et al., 2004). In another study, rapport effects were qualified by other factors, such as age (Hardy & van Leeuwen, 2004).

As a whole then, the current database on the effects of rapport building in the context of eyewitness interviewing is mixed and inconclusive. This is not to say that rapport is not valuable. Clearly, the concept has been successfully employed across different disciplines including medical science, psychotherapy, and learning environments (i.e., schools and university). In these fields, evidence for the positive effects of good patient-physician, patient-therapist, and student-teacher working alliance on clinical, psychological treatment, and learning outcomes are accumulating (e.g., Derksen, Bensing, & Lagro-Janssen, 2013; Fuertes, Toporovsky, Reyes, & Osborne, 2017; Georgopoulou, Prothero, & D’Cruz, 2018; Horvath, Del Re, Fluckiger, & Symonds, 2011; Rogers, 2015; van Osch, van Dulmen, van Vliet, & Bensing, 2017). Thus, the status quo should inspire researchers to systematically approach the issue of rapport building in witness interviews in different age groups, testing the isolated and combined effects of different rapport building components (cf. Vallano & Schreiber Compo, 2015). Research into the usefulness of the different components of the Cognitive Interview (e.g., Davis, McMahon, & Greenwood, 2005; Memon, Wark, Bull, & Köhnken, 1997) could serve as an example. The variety of methodology across studies thus far is striking, making it very difficult to compare results and to draw conclusions. The field is still in its infancy, but researchers have begun to emphasize the importance of conducting research on the effectiveness of rapport (Saywitz et al., 2015; Vallano & Schreiber Compo, 2015) and there has been an increase in studies in the past few years.

We now turn to the limitations of the current study. Participants’ ratings of experimenter friendliness and feeling comfortable during the experiment did not differ as a function of rapport condition. Both items were scored relatively high on average (friendliness: M = 3.42, SD = 0.93; feeling comfortable: M = 2.97, SD = 0.88, on a scale ranging from 0 to 4). This could mean that the differences across rapport conditions were not strong enough in the current study. Alternatively, our questions did not constitute a valid way to measure rapport. The lack of an association between the ratings and the different memory performance measures points into this direction. Indeed, other studies used much more elaborate tools for assessing the interviewer-interviewee interaction and the presence of rapport-related characteristics (e.g., Kieckhaefer et al., 2014; Vallano & Schreiber Compo, 2011), more direct questions regarding the quality of rapport (Nash et al., 2016), or the nature of the interaction (Collins et al., 2002). Studies with children, however, did not include questions for assessing the rapport manipulation (Brown et al., 2013; Hardy & van Leeuwen, 2004; Roberts et al., 2004), probably because young children may find it difficult to answer such questions.

To recap, the current experiment examined the effects of rapport building on memory performance from a developmental perspective. Although widely used in eyewitness interviewing settings, we found scarce evidence that rapport building exerted any positive effects on memory. This does not mean that rapport building is useless in eyewitness interviews. On the contrary, contrasting our work with studies that do show positive effects of rapport buildings emphasizes the importance of investing in the operationalization of rapport building and conducting follow-up research in this limited field of psychology.

Acknowledgments

We thank Stella Ohnesorge, Elisabeth Reinking, and Henrike Steinbeck for their help with data collection.

Additional information

Funding

This research has been supported by a fellowship awarded from the Erasmus Mundus Joint Doctorate Programme The House of Legal Psychology (EMJD-LP) with Framework Partnership Agreement 2013-0036 and Specific Grant Agreement 2013-1438 and a student grant in aid by the American Psychology-Law Society to Nathalie Brackmann.

Notes

1 Event practice consisted of a probe to tell everything that happened today from the time the participant woke up until the arrival in the lab. Follow-up invitations and cued invitations followed.

2 d = 1.59 for free recall quantity and d = 1.09 for overall accuracy according to our own calculation, based on the provided F value and means using https://www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-SMD28.php.

Appendix A. List of cued questions

1. Where did the incident happen?

2. How many people were involved?

3. What was stolen?

4. Was the victim female or male?

5. Which hair color did the victim have?

6. The victim’s trousers were blue, right?*

7. Did the victim wear a necklace? If yes, was it golden or silver?

8. Was the victim’s sweater bluish or reddish?*

9. Did you see the victim drinking a Coke?*

10. Why was the victim’s bag left alone?

11. Was the offender female or male?

12. Which hair color did the offender have?

13. Was the offender’s sweater greenish or brownish?*

14. It was a bag that was stolen, wasn’t it?*

15. The perpetrator took the victim’s identity card out of the wallet, right?*

16. What did the perpetrator do right before the theft?

17. The man who was shown in the beginning attached an advertisement to the wall, didn’t he?*

18. He tripped the young woman, right?*

*marks suggestive questions