Lay perceptions of child witnesses: effects of presentation of rapport building in forensic interviews

Abstract

Rapport building is an important component of forensic interviews with children, especially about alleged sexual abuse. Because interviews are presented as evidence in trials, an important consideration is whether seeing interview rapport affects juror perceptions. We addressed this by examining how rapport, child age and participants’ gender affect perceptions of a child witness and defendant guilt. Participants (N=478, M=36.6 years, SD=12.3, 47% female), recruited via Prolific, read a forensic interview transcript of a child’s disclosing sexual abuse that varied interviewer rapport (high, moderate, minimal) and child age (5, 9, 13 years). Participants in the high rapport condition rated the child as more credible and felt more positively about the interviewer, and the 5-year-old was perceived most favourably. Overall, positive effects of rapport building extend beyond their influence on children’s accounts. Presenting rapport building to jurors may help bolster a child’s credibility, though other factors (e.g., child age) remain important.

Introduction

Rapport building is widely viewed as a key component of forensic interviews, especially with suspected child victims of abuse whose reports may be incomplete, contain errors or not be perceived as credible or accurate (Saywitz et al., 2015). Indeed, an impressive body of research has demonstrated beneficial effects of rapport building in terms of enhancing children’s accuracy and disclosures of maltreatment (Hershkowitz & Lamb, 2020; Hershkowitz et al., 2014; Roberts et al., 2004) and productivity of those disclosures (Karni-Visel et al., 2019; Lavoie et al., 2021; Price et al., 2013). What remains less clear, though, is how rapport building is perceived by others, particularly adults charged with evaluating the veracity of children’s claims. Interviews with a suspected child victim may be presented at trial, and adults, serving as jurors, must interpret children’s statements and ultimately render decisions about guilt or innocence of a defendant. In fact, in criminal trials involving allegations of child abuse, children’s statements comprise a primary (and at times the only only) piece of evidence (Walsh et al., 2010), meaning that jurors’ decisions are likely heavily influenced by their perceptions of the quality of the interview and the accuracy of children’s responses. How rapport building affects these perceptions is crucial first to ascertain whether laypersons recognise the benefits of rapport, and second to inform legal professionals of potential consequences of presentation of rapport building in cases with a child witness.

The overarching goal of the present study was to evaluate these very issues. Specifically, we systematically investigated the effects of rapport building with a suspected child sexual abuse victim on laypersons’ perceptions of the child’s credibility and defendant guilt. We also examined whether laypersons’ perceptions of rapport varied as a function of the child’s age and their (i.e. participants’) gender, given that age can affect how laypersons evaluate a child’s credibility and honesty (e.g. Davies & Rogers, 2009; Tabak & Klettke, 2014; Voogt et al., 2020), and women routinely view children’s claims of abuse more positively (i.e. as more likely to be true) than males (Davies & Rogers, 2009; Golding et al., 2003; Jones et al., 2021; McCauley & Parker, 2001; McCoy & Gray, 2007; Voogt & Klettke, 2017). Our specific interest concerned whether rapport building interacted with child age to influence their perceptions of the child and case decisions (e.g. defendant guilt). Before describing the study, we briefly review what effective rapport building looks like in forensic interviews with children. We then review the ways in which viewing rapport building might affect laypersons’ perceptions, followed by the potentially interactive role of child age and gender on these perceptions.

Rapport building with children

Rapport building is valued across many settings in which adults must interact with children: educational (Frymier & Houser, 2000; Lim et al., 2013); clinical (Brown et al., 2014; H. Fernandes, 2022; Ovenstad et al., 2020); and forensic (Saywitz et al., 2015). The purpose of building rapport in these settings can be somewhat similar, to increase comfort and trust. In forensic interviews, the specific aims of rapport building with suspected child victims also include ensuring that children understand the purpose of the interview and having them practise answering questions before they are asked substantive questions about the topic of interest (Saywitz et al., 2015). Typically, rapport building is divided into two general components: ‘getting to know you’ questions and ground rules. Best practice ‘getting to know you’ questions are phrased in an open-ended manner (e.g. ‘Tell me about’) to encourage children to provide narrative details in their responses, setting up an expectation that children should provide narrative answers throughout the interview (Yi & Lamb, 2018). Questions at times target both positive and negative topics (e.g. a child’s likes and dislikes) and a prior experience (e.g. a prior birthday; Henderson et al., 2022; Lamb et al., 2018; Lyon, 2021), giving children practice narrating details of episodic experiences and discussing positive and negative information. Ground rules (e.g. the importance of telling the truth, that the interviewer does not know what happened) give children guidance on the purpose of the interview and establish children as the expert (Brubacher et al., 2015; Fessinger et al., 2021).

As mentioned, consistent benefits have emerged when interviewers build rapport using these best practice strategies with children. In experimental studies, for example, rapport building is related to increases in the amount of detail children provide about a transgression with no adverse effects on accuracy (Lyon et al., 2014; Roberts et al., 2004). In forensic interviews with suspected child victims of sexual abuse, rapport building is associated with increased likelihood of disclosure (Hershkowitz & Lamb, 2020; Hershkowitz et al., 2014). Given these benefits, it is unsurprising that forensic interviewers routinely receive training in rapport practices, although their actual use still varies (Cronch et al., 2006; D. Fernandes et al., 2024), raising questions about whether that variability, which may well affect what children report, also affects what others think of the interview and of what children report. This is important, considering that legal professionals must evaluate children’s statements before trial (e.g. police when making investigative decisions, prosecutors deciding to move forward with prosecution), and laypersons may be presented with interview videos or transcripts as hearsay evidence, for example.

Rapport building and perceptions of children

To date, very few studies have examined the effects of rapport building on laypersons’ perceptions of children’s abuse disclosures and credibility. Findings, moreover, are mixed. Kräahenbüuhl (2012), for example, presented participants with a transcript of a forensic interview with a 4- or 6-year-old child that contained either a full description of the rapport-building phase or a statement that rapport building had taken place. The rest of the transcript, which contained a child’s disclosure of physical assault, was identical between rapport conditions. Regardless of the child’s age, when the actual rapport building was presented, participants evaluated the child witness and the interviewer more favourably, were more likely to say that the case would proceed to court, and were more likely to report that the defendant would be found guilty. Insofar as laypersons understand the purpose of rapport building (e.g. to increase children’s comfort and teach them about what to expect in the interview) or know about its benefits on children’s reports, positive evaluations of the veracity of children’s statements when rapport is present would be expected. However, the opposite pattern of results was reported in another study. Waterhouse et al. (2020) presented participants with a video of a forensic interview with a 7-year-old child that began with rapport building or with a statement that rapport building had occurred (the video also included a single interview with a break in the middle or two separate interviews, a week apart; rapport was only shown once). When rapport was presented, participants rated the child as less believable and credible than when rapport was not presented, although no effects on defendant guilt emerged.

On the one hand, the studies’ opposing results make it unclear whether rapport has any effect on laypersons’ perceptions. On the other hand, however, several differences between the studies and limitations in their designs are likely contributing to the varying results and highlight the need for additional and more comprehensive investigations of how laypersons view rapport building with suspected child victims.

First, the studies’ methodologies varied. Krähenbühl’s (2012) transcript included ground rules and truth and lie discussion, which, as noted by the author, could have affected laypersons’ perceptions in ways separate from or additively with the other rapport-building strategies. Second, in both studies, participants either saw/read or were told about rapport, meaning all participants may have believed that rapport was built. Knowing that an interviewer built rapport, even if it is unclear how it was actually done, might be better than no rapport at all or than poor rapport building if presented. That is, in actual forensic interviews, virtually all interviewers engage in some form of rapport building (Collins et al., 2014). What varies is how interviewers build rapport, such as whether they rely on best practice strategies or not (Carson & La Rooy, 2015; Fessinger & McAuliff, 2020; Lee & Kim, 2020; Powell et al., 2005). A crucial question, therefore, concerns whether laypersons can discern among different rapport strategies, and how different strategies (more than just whether rapport is or is not present) affect laypersons’ perceptions.

Third, the format of the interview – transcript or video recording – may have led to conflicting findings. Visual cues from both the interviewer and the child could have swayed participants’ feelings about rapport building (e.g. as suggested by Waterhouse et al., 2020). Behaviours perceived of as overly supportive can appear coercive (Bottoms et al., 2007), reducing trust in what children later say. And fourth, children’s age may need to be considered more carefully in conjunction with rapport. Krähenbühl (2012) found no effects of child age on perceptions, but the children were fairly close in age, 4 or 6 years, and fall into young ages where memory capabilities are more limited and concerns about suggestibility are often greatest (Ceci & Friedman, 2000; Goodman & Melinder, 2007). The child in Waterhouse et al.’s (2020) investigation was only slightly older, 7 years. When laypersons’ general perceptions of wider age ranges of child victims are examined, age is often related to perceptions (Davies & Rogers, 2009; Tabak & Klettke, 2014; Voogt et al., 2020; though see McCauley & Parker, 2001). Young children are perceived as honest, especially in cases of sexual abuse, and as too young to knowingly fabricate or lie (Bottoms & Goodman, 1994; Call & Wingrove, 2022), although they are also seen as suggestible and potentially less believable, at least when compared to children in middle childhood (e.g. ages 8–10 years; Nunez et al., 2011; Wright et al., 2010). Once children move into the adolescent years (e.g. age 13 years and older), ratings of their credibility at times decrease, as concerns about their potential for deception or partial culpability increase (Back & Lips, 1998; Font, 2013; Voogt et al., 2020). Given these age trends, it would be of considerable interest to ascertain whether laypersons perceive younger or older children, in combination with rapport, as being more or less credible.

Insofar as laypersons recognise young children’s limitations (Quas et al., 2005), they may see rapport building as especially valuable with young children, such as those in pre-school years; although because laypersons also recognise these children’s susceptibility to errors (Antrobus et al., 2012), too much rapport (even if considered best practices) could be viewed negatively. With slightly older children and adolescents, who laypersons recognise as more competent but possibly still needing of support, rapport building could have consistent positive benefits. With adolescents, though, perhaps rapport is seen as excessive.

As a final note, although layperson gender has not been examined in prior studies as a predictor in conjunction with rapport of laypersons’ perceptions, women often view child abuse victims more favourably and are more likely to rate defendants guilty than men are (Davies & Rogers; 2009; Golding et al., 2003; Jones et al., 2021; McCauley & Parker, 2001; McCoy & Gray, 2007; Voogt & Klettke, 2017). When rapport is also considered, perhaps women have a greater appreciation for the value of rapport building, given that they tend to express greater empathy towards child victims than men (Bottoms et al., 2014), leading to women’s more positive perceptions when present. These tentative ideas were explored here.

Present study

In the present study, we examined laypersons’ perceptions of rapport building in forensic interviews with a suspected child victim of sexual abuse. Participants read a case summary followed by a transcript of a forensic interview in which a child alleged sexual abuse by her uncle. The age of the child (5, 9 or 13 years old) varied, as did the extent to which the interviewer followed best practice rapport building recommendations at the outset of the interview (minimal, moderate or high). This resulted in a 3 (age) by 3 (rapport) between-subjects design. After rapport, the interviewer then elicited a statement from the child about the alleged abuse. A separate hanging control condition, in which only the interviewer’s questions and child’s statement about the alleged abuse were included, was included to evaluate statement credibility separate from the pre-interview rapport. After reading the transcript, participants rated their perceptions of the child and interviewer and rated how guilty they thought the defendant was.

This design allowed us to test several hypotheses, all preregistered on AsPredicted (https://aspredicted.org/5wyx-37cz.pdf). First, we expected main effects of child age. As child age increased, at least from 5 to 9 years, participants would perceive the child more positively (e.g. more believable) and with less scepticism, and be more likely to believe the allegation (i.e. rate the perpetrator as guilty). With younger children, there may be concerns about suggestibility and their ability to remember a past event accurately (McCauley & Parker, 2001; Nunez et al., 2011; Ross et al., 2003; Wright et al., 2010); thus, while a 5-year-old might be on the cusp (i.e. a bit older than pre-school age), we expected perceptions to be swayed in a more negative direction than that of the school-aged child due to these potential concerns. We did not have any specific predictions about differences between the 9- and 13-year-old; although the latter age falls in the teen years, 13 is still sufficiently young that scepticism about adolescents’ honesty might not be strong enough to lead to negative perceptions relative to perceptions of the 9-year-old.

Second, although we did not predict a main effect of rapport on perceptions of allegation believability, we expected rapport to affect perceptions of the interviewer, with more extensive rapport (i.e. high and moderate) leading to more positive views of the interviewer. Third, we expected child age to interact with rapport. Perceptions of the youngest suspected child victims were expected to be more strongly influenced by rapport than perceptions of the oldest child victims, potentially overcoming concerns about the former’s suggestibility and cognitive ability. Specifically, when the 5-year-old was questioned by an interviewer who began with extensive rather than minimal rapport, the child would be evaluated more favourably, and guilty ratings would be more likely. With the 13-year-old, we expected that extensive rapport may be seen as excessive and thus not lead to the same positive effects as with younger children, given that laypersons are likely to believe that adolescent-aged victims do not need additional support and may even be partially responsible for their victimisation (Bottoms et al., 2004). Finally, women often hold more positive views of child sexual abuse victims than men (Voogt & Klettke, 2017), which was expected to emerge here. Also, although we did not have a strong prediction regarding an interaction between participant gender and rapport (as gender was not a main variable of interest), we explored the possibility that women would be more strongly influenced by rapport in terms of their view of the child and beliefs about defendant guilt than would men.

Method

Participants

The final sample included 478 participants, recruited via Prolific, an online research platform that allows for the recruitment and management of large representative samples of participants (Douglas et al., 2023; Eyal et al., 2021). Inclusion criteria consisted of participants being aged 18 years or older, a U.S. resident, and able to read and write English. Twenty additional participants completed the survey but were not included in the final sample: eight participants completed the survey but were eliminated for failing attention checks, and 12 participants were missing demographic data (e.g. they wrote their age when asked gender). A priori power analysis, conducted via the ‘pwr’ package in RStudio (Champely, 2020), indicated that N = 400 was sufficient to detect medium-sized effects, power = .80, alpha = .05.

Participants’ ages ranged from 18 to 73 years (M = 36.6, SD = 12.3); 46.7% identified as female, 48.8% identified as male, and 2% identified as nonbinary, genderqueer or transgender. A majority identified as White (63.9%), and the remaining identified as Black or African American (15.9%), Asian (6.9%), mixed race (i.e. two or more races) (7.3%), Hispanic or Latino/a/x (3.7%), American Indian (0.8%) and Middle Eastern (0.2%). Reported education varied as well: 12.9% completed only high school, 21.6% completed some college, 0.8% completed a post-secondary non-degree programme, 10% had a 2-year degree, 37.1% had a 4-year degree, 15.5% had a master’s degree, and 1.6% had a doctoral degree. A total of 37.1% of participants reported being a parent, and 10.4% reported having a job where they worked with children.

Materials and procedures

All materials and procedures were approved by the Institutional Review Board at the authors’ university. Following consent, participants completed an anonymous survey configured in Qualtrics. The order was pre-set, and participants were not allowed to return to sections after they were completed. Participants first read a case summary based on a criminal case in California and were given the California Penal Code’s definition of the criminal charge discussed in the interview: ‘lewd and lascivious acts with a child.’

Case introduction

The summary explained that Isabella, the suspected victim, her 4-year-old cousin, Maya and the defendant’s wife were all interviewed about alleged sexual abuse. Further, it stated that Maya had difficulty answering questions in the interview because of her age, and the defendant’s wife did not have any reason to suspect harmful behaviour. It also mentioned that other children were at the house where the alleged incident took place on the day of its occurrence but was ambiguous as to whether these children were present when the incident would have occurred.

Case transcript

Following the summary, participants were presented a written transcript of the forensic interview with Isabella. It began with the pre-substantive phase and ended with recall and open-ended questions that resulted in the child’s disclosure. The pre-substantive portion varied the interviewer’s level of rapport building – high (best practice), moderate practice or minimal – and included a question in which the child explicitly mentioned her age, as 5, 9 or 13 years. A separate hanging control group was also included. This group was only shown the substantive portion of the interview that contained child’s disclosure, omitting the pre-substantive rapport phase and child’s age. Across conditions, the substantive portion and child’s disclosure were identical. Transcripts underwent successful pilot testing with 36 participants to ensure that our rapport manipulations were clear.

In the high rapport condition, the interviewer provided a thorough introduction and adhered to best practice rapport recommendations in forensic interviewing (Lyon, 2021). This included an explanation of ground rules (e.g. ‘it is okay to say “I don’t know”’) combined with practice applying the rules and open-ended prompts asking about the child’s likes (e.g. making bracelets), dislikes (e.g. reading) and previous birthday. One ‘tell me more’ prompt followed the child’s response to the birthday prompt. The child’s answers were identical across age.

In the moderate rapport condition, the interviewer began with a brief introduction and described the same ground rules as in the high rapport condition, but did not give the child the opportunity to applying the rules. Next, the interviewer asked closed-ended questions on the same topics as those targeted by the open-ended questions in the high rapport condition. The interviewer included the same open-ended birthday prompt, but did not follow with the ‘tell me more’ prompt after the child’s initial answer.

In the minimal rapport condition, the interviewer provided a brief introduction and instructed the child to tell the truth. There was no further discussion of ground rules. This was followed by closed-ended questions about the same topics in the other rapport conditions, with the closed-ended questions being identical to those in the moderate rapport condition, including the birthday prompt. It also included a few neutral (e.g. ‘you need to use your words’, ‘speak up’) statements from the interviewer to indicate how the child should be answering.

Post-transcript perception questionnaire

After reading the forensic interview, participants completed attention check items asking what the child’s age and gender were. Participants were then presented with the definition of the penal code and were asked about the case outcome: ‘Based on the interview, how likely is it that the defendant committed “lewd acts with a minor” against the witness?’ on a scale of 1 (definitely guilty) to 6 (definitely NOT guilty). This was followed by an open-ended question: ‘What led you to making your decision?’

Next were Likert-scale items concerning participants’ perceptions of the child and the interviewer, along with manipulation and additional attention check items (embedded within). Child victim perception questions included four sets of items. The first set asked about the child in general, specifically how believable, credible and suggestible (reverse coded) the child was, and the second set asked how evasive (reverse coded), forthcoming, and cooperative the child was, and the extent to which the child withheld important information (reverse coded). All were rated on a scale from 1 (not at all) to 6 (extremely). Third was a set concerning perceived child truthfulness, with four statements to which participants rated their agreement on a 1 (strongly disagree) to 6 (strongly agree) scale (e.g. ‘the witness was telling the truth about these events’, ‘the witness is intentionally lying about the abuse occurring [reverse coded]’). The fourth set concerned perceptions of the child’s statements: how complete, accurate and consistent the child’s statements were, all rated on a 6-point scale of 1 (not at all) to 6 (extremely).

The manipulation check questions followed. One asked participants to rate the extent to which the interviewer made an effort to get to know the witness from 1 (not at all) to 6 (completely), and the other asked how clear the interviewer’s instructions were from 1 (vague, no explanation and/or practice) to 6 (clear, explained and/or practised).

The final interview questions concerned participants’ perceptions of the interviewer. Six items asked how fair, friendly, competent, warm, supportive and suggestive (reverse coded) the interviewer was, all rated on a 6-point scale of 1 (not at all) to 6 (extremely). Responses were averaged to reflect how positively the interviewer was perceived.

After participants answered questions about the interview, they completed a separate questionnaire regarding their perceptions of suspected child sexual abuse reporting tendencies in general (Bottoms et al., 2014). The questionnaire included 12 statements (e.g. children do not lie about sexual abuse; many innocent people have probably been falsely convicted of child sexual assault). After reading each statement, participants rated their agreement on a scale from 1 (strongly disagree) to 6 (strongly agree). Ratings were averaged (half were reverse coded) to create a single score such that higher values indicate more positive general perceptions of suspected child victims’ credibility (α = .87). An analysis of these items is available to interested readers on Open Science Framework (OSF) (https://osf.io/wjgu5/?view_only=3683477599c54aa38022c747d0fe38a7).

Demographic questions were asked last. These asked about participants’ age, gender (open-ended), race/ethnicity, education, political affiliation, parental status and whether they currently work with children (all indicated via relevant multiple-choice responses). Upon completion of the survey, participants were thanked and compensated $6 through Prolific.

Data analysis plan

Three sets of preliminary analyses were conducted. First, although the four sets of questions regarding the child (witness) were designed to tap four domains, it was possible that the domains were not entirely separable. Thus, we conducted a factor analysis containing all items to identify separable domains of perceptions of the child. Second, we compared the control (statement only) condition to each of the nine other study conditions via a series of t tests. Dependent measures included participants’ guilt ratings and perceptions of the child (each domain considered separately). These analyses were conducted to ensure that participants’ perceptions significantly differed when any form of rapport was present versus absolutely none. Third we tested for demographic differences in the dependent measures.

Main analyses (see also our preregistration; https://aspredicted.org/5wyx-37cz.pdf) consisted of separate 3 (rapport) × 3 (child age) × 2 (participant gender) analyses of variance (ANOVAs). Outcomes of interest included: participants’ guilt ratings, domains of perceptions of the child, and perceptions of the interviewer. For significant effects, simple effects and planned comparison t tests were conducted to test hypotheses, and post hoc Tukey’s honestly significant difference (HSD) tests were conducted for other effects.

Results

Preliminary analyses

The factor analysis included participants’ agreement ratings with each of the 14 statements regarding the child (e.g. accuracy, suggestibility). Some ratings were reverse coded such that higher scores on all reflected more positive perceptions of the child (see Table 1 for all items). All items were significantly correlated (rs > .22, ps < .001) except for suggestibility, so it was removed from the factor analyses.

Table 1.

Factor analysis results.

	Factor loading		Communality	M	SD
	1	2
How believable was the witness?	.91	.03	.85	4.85	1.14
How credible was the witness?	.92	−.04	.82	4.61	1.19
This witness was telling the truth about these events.	.87	.07	.81	4.95	1.07
I would trust a typical witness of this age to be telling the truth.	.79	−.02	.61	4.68	1.08
This witness believes the abuse occurred.	.62	.24	.55	5.29	0.96
How complete was the witness’s statement?	.87	−.05	.72	4.27	1.20
How accurate was the witness’s statement?	.94	−.06	.84	4.42	1.13
How consistent was the witness’s statement?	.88	−.04	.75	4.24	1.35
How evasive was this witness? (R)	.06	.85	.76	4.76	1.23
To what extent did it seem like this witness withheld important information during the interview? (R)	−.04	.89	.76	4.57	1.25

Note: R = reverse coded. Bold indicates items included in each factor.

We used principal components extraction and Direct Oblimin rotation, as we expected extracted factors to be correlated. All items had sufficient communalities (.54–.83; Child, 2006). A two-factor solution resulted. The first factor, labelled ‘general credibility’ (α = .95) accounted for 62.15% of variance and reflected eight items: the child witness’s believability, credibility, truthfulness (3 items) and perceptions of the statement (e.g. completeness, accuracy, and consistency). The second factor, labelled ‘responsiveness’ (α = .70) accounted for 12.37% of variance and reflected two items: (a) the witness was evasive, and (b) the witness was withholding information. Three items were removed for cross-loadings: the witness was intentionally lying about the abuse, cooperativeness, and forthcomingness. See OSF for analyses conducted with the four removed items.

Next, independent samples t tests compared the control (statement only) condition to the nine remaining conditions on three outcomes: perceptions of guilt and perceptions of the child witness’s general credibility and responsiveness. Table 2 contains the means across conditions and outcomes. When guilt ratings were entered as the dependent variable, three of the nine comparisons emerged as significant. Participants who read an interview with a 5-year-old in the minimal [t(90) = −1.73, p=.04, d=1.44], moderate [t(87) = −2.94, p=.002, d=1.37] and high [t(88) = −2.44, p=.01, d=1.37] rapport conditions rated the defendant as more likely to be guilty than those who read the statement only.

Table 2.

Means and standard deviations of outcomes across conditions.

	Minimal rapport			Moderate rapport			High rapport			Control
	5 years N = 50	9 years N = 47	13 years N = 48	5 years N = 47	9 years N = 42	13 years N = 52	5 years N = 48	9 years N = 45	13 years N = 47	No age N = 42
Guilt	2.12* (1.37)	2.53 (1.52)	2.48 (1.40)	1.79* (1.22)	2.17 (1.15)	2.23 (1.35)	1.94* (1.21)	2.24 (1.25)	2.43 (1.41)	2.64 (1.53)
General credibility	4.77* (0.83)	4.38 (1.00)	4.40 (1.15)	4.90* (0.86)	4.73* (0.96)	4.67* (0.93)	4.93* (0.89)	4.82* (0.86)	4.68* (0.96)	4.14 (1.16)
Responsiveness	4.88* (0.98)	4.35 (1.04)	4.48 (0.90)	4.81 (1.14)	4.65 (1.25)	4.43 (1.14)	5.04 (1.01)	4.67 (1.16)	4.62 (1.08)	4.57 (1.07)

Note: Ns reported here do not include the 2% (n = 10) of participants that identified as nonbinary, genderqueer or transgender. Means are shown with standard deviations in parentheses.

^*Statistically significantly different from the control condition mean.

When the mean scores for general credibility and responsiveness were entered, eight of the 18 comparisons emerged as significant. Participants perceived the 5-year-old in the minimal [t(90) = 3.03, p=.002, d=0.99], moderate [t(87) = 3.53, p = .001, d=1.01] and high [t(88) = 3.65, p = .001, d=1.02] rapport conditions as more credible than in the statement only condition. The same pattern emerged for the 9-year-old in moderate [t(82) = 2.54, p=.01, d=1.06] and high rapport [t(85) = 3.13, p=.001, d=1.01] and the 13-year-old in moderate [t(92) = 2.44, p=.01, d=1.04] and high rapport [t(87 = 2.38, p=.01, d=1.06] conditions relative to the statement only condition. Though not all comparisons were statistically significant, patterns indicate that some rapport is better than none whatsoever in terms of increasing favourable perceptions of the child’s credibility. In terms of responsiveness, the 5-year-old in the high rapport condition was perceived as more responsive than the child in the statement only condition [t(88) = 2.13, p=.02, d=1.04].

The last set of preliminary analyses evaluated associations between participant demographics and the study outcomes, designed to identify potential covariates to include in the main models. Of particular interest was participant gender, as prior research has shown that men and women evaluate child witnesses differently. Independent samples t tests revealed significant gender differences on general credibility, confirming our expectations about women holding more positive perceptions of child witnesses (see Table 3). This warranted our inclusion of gender as an additional factor in the main analyses and allowed us to test the possibility of Gender × Rapport interactions as well. Regarding other characteristics of participants (e.g. age, whether they were a parent or not), several significant associations emerged but were not considered further as we had no theoretical reasoning to include them in main analyses. Results are available in OSF for interested readers.

Table 3.

Gender Differences by outcome by participant gender.

	Male M (SD)	Female M (SD)	t	df	p	d
Guilt	2.35 (1.43)	2.15 (1.27)	−1.55	466	.06	1.35
General credibility	4.56 (1.04)	4.73 (0.91)	1.85	466	.03	0.98
Responsiveness	4.66 (1.12)	4.64 (0.70)	−0.13	466	.45	1.09

Main analyses

Main analyses consisted of four separate 3 (rapport condition: high, moderate or minimal) by 3 (age: 5, 9 or 13) ANOVAs with gender (women, men) entered as an additional factor (n = 426). The control condition was not considered in these analyses. Gender and number of participants were nearly equally split among the nine conditions. Table 4 contains results for all main effects, and estimated marginal means and d-based effect sizes are reported in the text.

Table 4.

Main effects from ANOVAs.

	df	F	p	η p 2
Guilt
Rapport	2, 423	1.62	.20	.01
Age	2, 423*	4.75*	.01*	.02*
Gender	1, 424	0.86	.35	.002
Rapport × Age	4, 422	0.07	.99	.001
Rapport × Gender	2, 423	2.22	.11	.01
Age × Gender	2, 423	1.00	.37	.01
Rapport × Age × Gender	4, 422	0.73	.57	.01
General credibility
Rapport	2, 423*	3.95*	.02*	.02*
Age	2, 423*	4.00*	.02*	.02*
Gender	1, 424	2.09	.15	.01
Rapport × Age	4, 422	0.45	.77	.004
Rapport × Gender	2, 423	0.02	.98	.000
Age × Gender	2, 423	0.30	.74	.001
Rapport × Age × Gender	4, 422	1.53	.19	.02
Responsiveness
Rapport	2, 423	1.46	.23	.01
Age	2, 423*	6.77*	.001*	.03*
Gender	1, 424	0.29	.59	.001
Rapport × Age	4, 422	0.57	.69	.01
Rapport × Gender	2, 423	0.27	.77	.001
Age × Gender	2, 423	0.83	.44	.004
Rapport × Age × Gender	4, 422	2.10	.08	.02
Interviewer Perceptions
Rapport	2, 423*	37.24*	<.001*	.15*
Age	2, 423	0.35	.78	.001
Gender	1, 424	1.45	.30	.004
Rapport × Age	4, 422	1.98	.10	.02
Rapport × Gender	2, 423	1.46	.23	.01
Age × Gender	2, 423	0.35	.70	.002
Rapport × Age × Gender	4, 422	0.88	.48	.01

Note. ANOVA = analysis of variance.

^*Statistically significant results at the p = .05 level.

Guilt ratings

Overall, a majority of participants (81%) categorised the defendant as guilty, reflected as a rating of between 1 and 3 on the guilt scale (1 = definitely guilty to 6 = definitely not guilty). However, the overall sample mean was 2.23 (SD = 1.35), suggesting variability in participants’ general feelings about the defendant’s likelihood of guilt.

Our first ANOVA tested the effects of child age, rapport, and participant gender on perceptions of guilt using the 6-point scale, as done similarly in other studies (Goodman et al., 2006). Only the main effect of age was significant. Post hoc Tukey tests revealed that participants perceived the defendant as more guilty when the child was a 5-year-old (M = 1.92, SE = 0.11) than a 13-year-old (M = 2.37, SE= 0.11; p= .02, d= 1.33). The difference between the 5- and 9-year-old (M = 2.29, SE = 0.12) child approached significance (p = .053, d= 1.29). We also conducted a binary logistic regression, with dichotomised guilt as the outcome, as preregistered. In jury trials, jurors are not asked to provide guilt ratings on a scale; thus, the dichotomised guilt variable represents a potentially more ecologically valid measure of guilt. However, no significant effects emerged, though there was little variability in the outcome (i.e. the model was moderately imbalanced). Using the 6-point guilt ratings in an ANOVA allowed for more variability in the outcome, and thus a better fitting model.

Perceptions of the child witness

Means were used for the two items derived from the factor analysis (general credibility and responsiveness). These were entered into separate 3 (age) × 3 (rapport) × 2 (participant gender) ANOVAs. When general witness credibility (i.e. means from the eight items identified in the factor analysis) was considered, significant main effect of both rapport and age emerged. Post hoc Tukey tests revealed that participants evaluated the witness more favourably in the high rapport condition (M = 4.83, SE = 0.08) than in the minimal rapport condition (M = 4.53, SE = 0.08; p = .03, d = 0.96). The moderate rapport condition fell in the middle (M = 4.77, SE = 0.08) and did not significantly differ from the other two conditions. Participants also evaluated the 5-year-old’s general credibility (M = 4.89, SE = 0.08) more favourably than the 13-year-old’s credibility (M = 4.59, SE = 0.08; p =.03, d = 0.94) with their perceptions of the 9-year-old falling in the middle (M = 4.64, SE = 0.08), but not significantly different from the other two ages at the p < .05 level. The main effect of participant gender was not significant, nor were the interactions.

Next, means from the second factor, responsiveness, were examined. Again, higher ratings reflect more positive perceptions. Only the main effect of age emerged as significant. Post hoc Tukey tests revealed that participants rated the 5-year-old as more responsive (M = 4.93, SE = 0.09) than both the 9-(M = 4.55, SE = 0.09; p= .02, d= 1.10) and the 13-year-old (M = 4.50, SE = 0.09; p = .004, d= 1.04). The main effect of rapport condition or gender was not significant, nor were the interactions.

Perceptions of the interviewer

Final analyses considered participants’ perceptions of the interviewer (M = 4.79, SD = 0.82, α = .81) – that is, how positively they viewed the interviewer. A mean score was used here, as we did not expect the items regarding the interviewer to tap into different domains of perceptions. When the 3 (age) × 3 (rapport) × 2 (participant gender) ANOVA was conducted, only the main effect of rapport was significant. As might be expected, participants perceived the interviewer more positively (e.g. warm, supportive, etc.) in the high (M = 5.06, SE = 0.06) and moderate rapport conditions (M = 4.97, SE = 0.07) than in the minimal rapport condition (M = 4.34, SE = 0.06; ps < .001, ds = 0.78). No other significant effects emerged.

Discussion

In the current study, we systematically varied the nature of rapport building present in a forensic interview with a suspected child victim of sexual abuse to understand how rapport affects laypersons’ perceptions of the child’s disclosure. Among the very small number of studies that have examined on the effects of rapport building on evaluations of child witnesses, findings have been mixed, with methodological differences potentially accounting for divergent patterns. Our preregistered study included a wide age range of suspected child victims; several rapport manipulations, which ranged from those that incorporated best practice approaches to those that were minimally supportive; a no-rapport presence condition; and multiple outcomes to ascertain how the variations in rapport affected laypersons’ perceptions of the child and interviewer and ratings of guilt. Findings support the notion that any rapport building bolsters the perceived credibility of the child victim, most noteworthy when rapport more closely adheres to best practices. Some subtle differences emerged based on the age of the child victim, but these did not alter the general benefit that best practice rapport seemed to have. Finally, although somewhat consistent layperson gender effects emerged with regard to their perceptions of the child, these were not altered by variations in rapport. Next, we discuss these trends in more detail, highlighting their potential value in legal settings.

First, turning to the quality and extensiveness of rapport building, while we did not expect main effects of rapport on guilt ratings, participants evaluated the child as more credible in the high rapport than in the minimal rapport condition, regardless of the child’s age. The high rapport condition adhered to best practice recommendations in child forensic interviewing (Lyon, 2021) and included an explicit explanation of ground rules, practice applying the ground rules, and open-ended prompts about the child’s likes, dislikes and a previous birthday. The minimal rapport condition only contained instructions for the child to tell the truth and closed-ended questions about the same topics in the other conditions, a condition that may well mimic what basic rapport building looks like among untrained forensic interviewers who, although they recognise the importance of and attempt to build rapport, lack adequate training in developmentally tailored effective rapport strategies (Cederborg & Lamb, 2008; Rivard & Schreiber Compo, 2017).

In line with our predictions, participants also felt more positively about the interviewer in the high and moderate rapport conditions. These findings suggest that laypersons are somewhat sensitive to variations in interviewers’ use of rapport-building strategies, with this sensitivity reflected in part in how they perceive the child’s credibility. Given that rapport building varies in interviews, but also in trials, with attorneys diverging in their use of specific rapport strategies as well (Bruer et al., 2022; Carr et al., 2024), legal professionals would be wise to consider not only the effects of rapport on children’s reporting but also, on laypersons’ perceptions of the child’s report.

In contrast to the effects of rapport building on laypersons’ perceptions of the child’s credibility, no effects of rapport emerged on perceptions of the child’s responsiveness, or whether the child was evasive or withholding information from the interviewer. Nor did rapport and age interact with rapport to influence such perceptions, as we predicted. Instead, results suggested that rapport alone, regardless of the child’s age, facilitates greater perceptions of credibility. Yet, rapport’s effects did not extend to considerations about responsiveness from a child. Instead, only child age predicted the latter. While this pattern parallels some prior work, it is not consistent with all studies and is inconsistent with not our prediction that the 9-year-old would be viewed more positively. Instead, participants rated the 5-year-old as more responsive than the 9- and 13-year-old. Children’s lie-telling ability increases with age (Talwar & Crossman, 2011, 2012). In addition, teens are savvier about concealing information, including abuse-related information that may be embarrassing to talk about with adults (Nogalska et al., 2023). Laypersons appear sensitive to these developmental shifts (Danby et al., 2022; Quas et al., 2005; Vrij et al., 2006), which can lead to age-related decreases in perceptions of child credibility, although primarily with adolescent victims. We thus were surprised that the 9-year-old was viewed as less responsive than the 5-year-old as well.

These age patterns may have emerged for several reasons. First, a key item typically linked to lower perceptions of young children’s capabilities, namely suggestibility, did not load on the responsiveness or credibility factor. We are unsure as to why. One possibility is that participants interpreted the suggestibility item as more reflective of the interviewer’s questions than of the child’s answers. Its exclusion, though, could have led to more positive views of 5-year-olds on the responsiveness factor. Second, had the child been a young preschooler, an age at which the largest memory errors and proneness to suggestibility often emerge (Ceci & Friedman, 2000; Peterson, 2012), laypersons might have felt less positive. And third, the interviewer’s questions were not especially suggestive. Such may also have explained why participants’ perceptions of suggestibility did not load on other factors, but had the questions been overtly leading, perceptions of the youngest child again might have varied. All of these possibilities are worth exploring in future work. With regard to defendant guilt, we predicted main effects of age, such that as child age increased, so would guilt likelihood ratings. Most participants, regardless of rapport condition, swayed towards guilty (81% rated the defendant 1–3 on the 6-point guilt likelihood scale). Child age was related to participants’ ratings, though in the opposite direction to our predictions. In line with the findings concerning perceptions of the child, participants’ ratings of defendant guilt were highest when the child was 5 years-old.

When considering defendant’s guilt, jurors should weigh several factors, only one of which might be how an interviewer in a video or transcript built rapport. Such factors, like other corroborative evidence or witness accounts, the amount of detail in or quality of the child’s statements (though see Buck et al., 2011) or potential alibis all are likely important. Thus, it is not entirely unsurprising that we found no direct effects of rapport on guilt ratings. Instead, rapport affected perceptions more narrowly, of the specific outcome most directly linked to rapport, namely what the child said. Had the child’s report been less consistent or not as straightforward as the one presented here, perhaps the interviewer’s approach to eliciting that report including not only rapport but perhaps other questioning techniques (e.g. closed-ended prompts) in conjunction with one another would have become more central to laypersons. Their perceptions might also be shaped by how legal professionals build rapport (e.g. attorneys), given the potentially more salient effects of such strategies when presented live in person, all possibilities in need of direct exploration, along with other designs testing potential moderating effects of rapport building on a range of outcomes.

Finally, we considered participant gender, given past research consistently revealing women hold more positive views of children’s claims of sexual abuse than those of men (Davies & Rogers; 2009; Golding et al., 2003; Jones et al., 2021; McCauley & Parker, 2001; McCoy & Gray, 2007; Voogt & Klettke, 2017). However, layperson gender was not a particularly robust or consistent predictor of variations in perceptions: layperson gender was unrelated to perceptions of the child, the interviewer or defendant guilt. It would be of interest to replicate studies from past decades to see whether perceptions are shifting, particularly in men.

Limitations and future directions

Although the current study offered a number of novel and important manipulations relative to prior research on perceptions of rapport building with suspected child victims (Kräahenbüuhl, 2012; Waterhouse et al., 2020), limitations also need mentioning, as these highlight the need for additional investigations. First, best practice recommendations include both the provision of ground rules and rapport building – that is, asking open-ended questions that encourage narrative responding and establish expectations. Here, we took a holistic perspective and included both, as jurors would likely see both in the pre-substantive portion of the interview transcript. Yet, whether laypersons view these two components identically is not known. Laypersons may interpret the discussion of ground rules as establishing the child’s understanding and thus credibility, separate from whether they view the child as being comfortable talking about abuse; the latter may be reflected in the interviewer’s use of open-ended prompts. It would be of interest to ascertain whether laypersons view the two components similarly in terms of their effects on children’s reporting. Here, though, we assessed their overall impact first.

Further, the quality of rapport building indeed impacts the quality of the child’s statement in the substantive portion of the interview. Here, we kept the child’s statement constant across conditions for experimental control, but future work should assess how laypersons perceive statements of low quality (e.g. inconsistencies, lacking detail) combined with rapport building that is conducted according to best practices, for example.

Finally, participants read a transcript of a forensic interview that included variations in rapport rather than viewed a video of those same variations (e.g. Waterhouse et al., 2020). Videos allow laypersons to see the interviewer’s and child’s demeanour, which both have been shown to affect perceptions. For example, a sad (e.g. upset, crying) child is often perceived as more credible (Golding et al., 2003; Melinder et al., 2016; Rowsell & Colloff, 2022) than a child who appears unemotional. Interviewers’ nonverbal behaviour may also vary in ways that affect laypersons’ perceptions. Positive nonverbal behaviour (e.g. smiling, nodding, making eye contact), like the positive verbal behaviour shown here (e.g. asking about the child’s likes), has been shown to increase children’s informativeness, though not necessarily during rapport (Davis & Bottoms, 2002; Hershkowitz, 2009; Hershkowitz et al., 2015). Whether laypersons recognise variations in such nonverbal behaviour and believe that such variations affect children’s reports is worth exploring. Related would be examining how more complex interactions between rapport building and a child’s emotional expression (e.g. crying, sadness) at the outset affect laypersons’ perceptions, especially when those interactions seem to lead to different patterns of child disclosure (e.g. complete, piecemeal). Such investigations would offer novel insight into the complex ways that rapport building shapes, directly and indirectly, laypersons’ views on children’s credibility, deception and possibly even overall experiences of victimisation.

Conclusions

Until the present study, only two published studies examined how rapport building affects laypersons’ perceptions of children in forensic interviews, a surprising lack of work given the critical role that rapport building during the pre-substantive phase of an interview plays in shaping children’s disclosure, including completeness and accuracy (Hershkowitz & Lamb, 2020; Hershkowitz et al., 2014; Karni-Visel et al., 2019; Lavoie et al., 2021; Price et al., 2013; Roberts et al., 2004). Our findings indeed suggest that laypersons, to some extent, recognise and value quality rapport building, as seeing its inclusion led to more positive views of the child’s report and interviewer effectiveness. When considering broader influences on the occurrence of victimisation, though, other factors, like the child’s age, are important to consider. Overall, findings suggest that the presentation of rapport is valuable in helping laypersons understand the precise domains where it should exert an effect, namely on the child’s credibility.