The Effects of Rapport Building on Information Disclosure in Virtual Interviews.

Rapport building has been identified as an effective tool when interviewing victims and witnesses of events that may be sensitive in nature. The objective of this study was to examine the rapport-building process within a virtual interviewing context. Participants (N = 94) were shown a sexual education video and then questioned about the content of the video in a live virtual interview using either a rapport (e.g., empathy, personalization, smiling) or no-rapport (e.g., flat tone, no smiling, no personalization) approach. Results showed that perceived rapport was much higher in the rapport condition compared to the no-rapport condition (d = 1.47). Participants in the rapport condition also provided substantially more dialog (d = 0.85) and reported more accurate details (d = 0.42) in the substantive phase of the interview than those in the no-rapport condition. Implications of this study for investigative interviews conducted virtually will be discussed.

Investigative interviews are a key source of information for law enforcement when investigating a potential criminal event (Fisher et al. 2011; Lassiter 2004; Milne and Bull 1999). Gathering details from interviewees can be difficult, however, given that the events of interest may be sensitive in nature, embarrassing, and/or traumatic to recall (Risan et al. 2016, 2017). One strategy that has been shown to facilitate the disclosure process is the building of rapport with the interviewee—defined as a productive working relationship based on mutual attention, positivity, and coordination—using verbal and non-verbal behaviors (e.g., self-disclosure, smiling; Gabbert et al. 2020; Tickle-Degnen and Rosenthal 1990). While the rapport process has been studied extensively within in-person interviews, it is unclear how or if rapport-building behaviors lend themselves to a virtual context. The current study sought to examine how rapport building functions in virtual interviews with the hope of informing investigative interviewing practices within this context. We begin by reviewing the concept of rapport within investigative interviewing before presenting the findings of a direct comparison between a rapport and non-rapport interviewing approaches in a virtual interview.

Rapport Building

Definitions of rapport building vary widely according to the context in which rapport is established and the nature of the relationship sought to be established (Gabbert et al. 2020). Among the most influential definitions found in the literature is Tickle-Degnen and Rosenthal’s (1990), which states that rapport is a combination of mutual attention, positivity, and coordination. Mutual attention refers to how behaviors such as nodding and utterances like “um hmm” allow for a cohesive, involved, and mutual interaction. Positivity refers to a certain level of caring, warmth, and respect in an interaction. Lastly, coordination refers to how synchronized an interaction between two people is in a way that it allows for a pattern of predictability and fluidity. Gabbert et al. (2020), in discussing rapport within investigative interviews, have further conceptualized rapport in this context as having three components: (1) personalizing the interview to the context as well as to the interviewee, (2) presenting an approachable demeanor, and (3) paying attention to what is being said to respond appropriately.

Rapport building is a key component of virtually all investigative models for both for adults and children (e.g., cognitive interview, NICHD protocol, PEACE model, Reid technique; Geislman and Fisher 2014; Inbau et al. 2013; Lamb et al. 2007; Milne and Bull 1999; Snook et al. 2010). Some key procedures for building rapport within investigative interviewing contexts include immediacy behaviors (e.g., open body posture, eye contact), active listening (e.g., minimal encouragers such as um-hmm), meeting practical needs (e.g., bathroom, water), self-disclosure (e.g., sharing personal anecdotes), demonstrating empathy (e.g., expressing understanding about the nature of the target event), and establishing common ground (e.g., identifying and pointing out common goals for the interaction; Abbe and Brandon 2014, 2013; Gabbert et al. 2020).

The ultimate purpose of creating rapport within investigative interviews is to increase the amount of information that can be generated from interviewees, and research findings to date support the idea that the process is effective in accomplishing this goal (Collins et al. 2002; Holmberg and Madsen 2014; Kieckhaefer et al. 2014; Vallano and Schreiber Compo 2011; for conflicting results see Kieckhaefer 2014) and Villalba (2014). For example, in a laboratory study by Collins et al. (2002) in which they compared participants’ information disclosure about a stimulus video across three conditions (i.e., rapport, neutral, and abrupt), participants in the rapport condition recalled a greater number of accurate details compared to participants in other conditions. Similarly, in their review of 418 videotaped intelligence interviews, Alison et al. (2013) found that a rapport-based approach led to increased information disclosure from interviewees in this context. A survey of 221 police officers in the UK also found that they view rapport as among the most effective interviewing techniques when working with witnesses (Dando et al. 2008), and Gabbert et al. in their systematic review found that more than 90% of research on rapport building reported that it had a positive effect on information provided during investigative interviews (Gabbert et al. 2020).

Virtual Interviews

In the context of the ongoing COVID-19 global pandemic, many investigative interviews have had to make their way online into a virtual platform (e.g., interviewing via Zoom or similar videoconferencing software; Vieth et al. 2020). While the impetus for the change may have been the pandemic, virtual interviews may continue to be popular moving forward. Interviewers have pointed to the fact that interviewing witnesses remotely, as opposed to bringing them to the police station, has allowed them to save time and resources as well as to question witnesses within shorter delays of the incident in question (Hager 2020). Additional advantages include removing the cost and difficulty associated with travel for interviewees and allowing interviewees to be questioned in the comfort of their own environment (Hoogesteyn et al. 2020).

While virtual investigative interviews may have benefits, they also raise potential challenges related to creating rapport. For example, common non-verbal elements of rapport building such as posture mirroring, eye contact, and physical proximity or touch (e.g., handshake) are difficult or impossible to replicate in virtual settings. Control over the interviewee’s physical space and the ability to facilitate practical needs (e.g., offer food or water) are also limited. In addition, the virtual nature of the interview can lead to technology and connection-related problems, which could result in interruptions and issues with visual cues (e.g., lagging, freezing). It is therefore unclear whether (1) rapport can be built successfully in a virtual context and (2), if so, whether it will lead to increased disclosure for difficult-to-discuss events as has been found in in-person contexts.

The only study that appears to have addressed the topic of rapport-building within virtual interviewing was conducted by Meijer et al. (2021). In this study, the authors compared in-person and online rapport building using the chat function on Skype and virtual reality (VR) stimulus video. They found that participants interviewed using online chat reported lower levels of certain aspects of rapport (i.e., attentiveness, trust/respect, and expertise; measured on Duke et al. (2018) Rapport Scales for Investigative Interviews and Interrogations-Source – RS3i-S) compared to those interviewed in-person. They also found, however, that participants interviewed using online chat and those interviewed in-person did not differ in terms of crime-relevant details provided about the VR mock-crime video. While an interesting first step, the virtual condition consisted of chat/written communication only, making it difficult to compare against live virtual video interviews.

The Current Study

The purpose of the current study was to examine the rapport process within mock virtual investigative interviews. Specifically, we sought to answer two questions related to rapport in this context. First, given the constraints of the virtual setting, is it possible to build rapport effectively within a virtual interview? Second, does increased perceived rapport within this context lead to increased disclosure of a difficult-to-discuss event? While research within the virtual context is limited, based on the research from the in-person literature, we hypothesized that:

Participants within the rapport condition would report higher levels of perceived rapport with the interviewer compared to those in the no-rapport condition.

Participants within the rapport condition would provide more information disclosure than participants in the no-rapport condition.

Methodology

Participants

Participants consisted of 96 undergraduate students from a Canadian university. Two interviews were dropped due to technology issues (one interview was not recorded, and one participant was not able to view the stimulus video), meaning that 94 interviews (Mage = 21 years old, range = 17–39 years old) were kept for the final sample. Of the 94 participants, 51 identified as female and the remaining 43 identified as male. The self-report demographic breakdown was as follows: Indigenous (1%), Arab (3.1%), Black (14.6), White (35.4%), Chinese (2.1), Filipino (5.2%), West Asian (4.2%), South Asian/East Indian (30.2%), Southeast Asian (1%), person of mixed origin (2.1%), and other (6.3%). Participants were assigned randomly to one of two conditions: rapport (46 participants) or no-rapport (48 participants).

Materials

Online Questionnaire

An online questionnaire was created using the Qualtrics survey software. The survey consisted of 13 pages. The first page contained a consent form that outlined the purpose and procedures of the study. The second page consisted of three demographic questions (i.e., age, gender, and ethnicity), and the third and fourth page included instructions about how to watch the video as well as a request for the participant to pay attention to the video. The fifth page contained the embedded stimulus video, followed by two attention check questions for the stimulus video on page six. The seventh page consisted of instructions on how to join the meeting and the link for the Zoom virtual meeting room where they would undergo their interview.

The eighth page consisted of text thanking the participants for their participation in the interview and prompting them to continue working through the remainder of the survey. The ninth page asked participants to indicate their participant and group numbers as provided by the interviewer. The tenth page of the questionnaire consisted of a three-question manipulation check which asked participants whether they felt that the video content was awkward to talk about, were comfortable talking to the interviewer about the video, and reported everything that they could remember about the video. On the eleventh page, the questionnaire presented a definition of rapport and asked participants to indicate their degree of perceived rapport with the interviewer using a slider (0–100) as well as an open textbox asking them to justify their answer. The last two pages of the questionnaire consisted of questions about the device and setting with which they joined the meeting and a debriefing statement. See the Open Science Framework (OSF) portal for the questions and prompts included in the Qualtrics questionnaire (https://osf.io/hnq4g/).

Stimulus Video

The stimulus video consisted of an Irish Catholic sexual education video from the 1980s and was 2 min and 24 s in length. In the video, a woman briefly describes male and female reproductive anatomy as well as explains the concept of sexual intercourse to an intended audience of children and adolescents. The nature of the stimulus video was chosen to attempt and create a target event that was awkward and inherently uncomfortable to discuss, as rapport is likely to be most effective and necessary when discussing target events that involve a high level of unease (Risan et al. 2016). Furthermore, the use of the sexual education video allowed for the introduction of a topic and vocabulary (e.g., language pertaining to genital anatomy) that approaches that used in interviews with victims of sexual assault. The awkwardness of this stimulus video lies not only in the topic at hand but also in the delivery of the material, with the subject of the video making comments and hand gestures that participants expressed made them uncomfortable. See the OSF portal for the stimulus video links as well as a storyboard (https://osf.io/hnq4g/).

Interview Training

Prior to conducting the live virtual interviews, the first and second authors were trained by the third author. The third author is a content expert in the interviewing field who also has experience delivering interview training courses to law enforcement and investigative agencies. The training consisted of an individual practice interview between each interviewer and the third author, followed by a feedback session. Each interviewer then conducted two practice interviews with lab volunteers and received feedback on their recorded interviews in a feedback session with the third author. This training process also served to refine the interview guide that was developed for this study.

Interview Guides

Rapport Condition

The interview guide for this condition consisted of a rapport-building sequence followed by the substantive phase of the interview where the participants were asked to tell the interviewer everything that they would remember about the stimulus video. In this study, rapport was operationalized according to Tickle-Degnen and Rosenthal’s (1990) and Gabbert et al. (2020) definition of the concept. Rather than a strict script, a structured guide was used that allowed for flexibility depending on the individual interviewee’s behavior and responses. The guide was designed to ensure that (1) the interaction between the interviewer and interviewee was genuine and went beyond a pre-determined script (Collins et al. 2002) and (2) the rapport interactions occurred throughout the interview process and not just at the beginning (Holmberg and Madsen 2014; Walsh and Bull 2011). To build rapport, interviewers were instructed to start the interview with attempts to show empathy as well as with self-disclosure by asking participants about their experiences with online learning during the pandemic and speaking on their own experiences. Throughout the interview, the interviewer addressed the participant by their first name, used a gentle tone, smiled, and attempted to keep their gaze toward the screen and the webcam when possible (Quas and Lench 2007). The interviewer was also instructed to sit upright in a visibly relaxed and open posture. These behaviors were to be maintained throughout the entire interview. See the OSF portal for the specifics of the interview guide for both conditions (https://osf.io/hnq4g/).

No-Rapport Condition

The interview guide for the no-rapport condition began with five close-ended filler questions that did not emphasize personal interest (e.g., “Is the webcam integrated into this computer?”) to lessen the time difference and replicate the presence of questions asked between watching the stimulus video and the substantive phase in rapport condition. This was followed by the substantive phase of the interview, which was identical to the rapport condition. In terms of mannerisms and approach, the no-rapport condition was designed to be neutral and professional as opposed to negative or confrontational. Throughout the interview, the interviewer did not address the participant by their first name. They used a flat tone of voice and made no effort to smile or attempt to keep their gaze toward the screen and the webcam (Quas and Lench 2007). The interviewer was sitting upright and was facing slightly offscreen. The interviewer was instructed only to make comments that helped progress the interview and ask clarification questions only when needed (Quas and Lench 2007).

Procedure

Participants were first sent the URL to access the Qualtrics survey. After reading the consent form, participants worked through the first six pages of the survey. Participants then clicked on the link within the survey to begin the live interview in Zoom (i.e., an online video conference platform). Interviews were conducted by either the first or second author, and interviewers were assigned randomly to both condition and interviewee. The interviewer then gathered the participants’ recall of the stimulus video using either the rapport or no-rapport interview guide. After the interview concluded, participants returned to their web browser to complete the remaining seven pages of the Qualtrics survey. Participants were granted one credit in an undergraduate psychology course for their participation in the study.

Measures and Coding

The interviews were screen and audio recorded using the Zoom videoconferencing software and transcribed verbatim by the first author. A coding guide was created using an iterative process in which the first author first read through the substantive phase of each interview (i.e., the participants’ free recall of the target event) and compiled an exhaustive list of every unique individual detail mentioned by participants. The finalized coding guide, consisting of 369 variables, was then used by the first author to code each participant’s recall of the stimulus video. Each detail mentioned was categorized as correct, incorrect, or unverifiable.

Inter-Rater

To assess the reliability of the coding, an undergraduate laboratory volunteer was trained by the first author to code the substantive portions of the interview. The volunteer first coded ten random transcripts, after which they received feedback from the first author. The volunteer then proceeded to independently code all 94 participant interviews. Cohen’s Kappa was used to measure inter-rater reliability, and the final value across all 34,686 comparisons was (K = 0.731, % agreement = 96%), suggesting substantial agreement between the two coders (Landis and Koch 1977).

Results

We found minimal differences between the two interviewers across as our dependent variables,1 as only the total length of interview (d = 0.45) had more than a small effect size and the 95% Confidence Intervals (CIs) for all comparisons overlapped.

There were also no meaningful differences in the dependent variables for participants’ demographic characteristics other than males rating the content of the video as more awkward to talk about than females (d = 0.42). Across all participants, the average rating for the awkwardness of the video, rated on a 5-point scale, was 3.34 (SD = 1.11, CI [3.11–3.57]).

Hypothesis 1

Participants within the rapport condition would report higher levels of perceived rapport with the interviewer compared to those in the no-rapport condition.

The overall mean rating on the question of perceived level of rapport with the interviewer was M = 72.37 (SD = 26.79, CI [66.88–77.86]). Participants in the rapport condition reported perceiving a greater level of rapport with the interviewer (M = 88.63, SD = 14.01, CI [84.47–92.79]), than did those in the no-rapport condition (M = 56.89, SD = 26.94, CI [49.10–64.65], Cohen’s d = 1.47, CI [1.01–1.93]; see Fig. 1). Participants in the rapport condition also reported a greater level of comfort talking about the video with the interviewer (M = 3.83, SD = 1.00, CI [3.53–4.12]), than did those in the no-rapport condition (M = 3.31, SD = 0.99, CI [3.02–3.60], Cohen’s d = 0.52, CI [0.10–0.93]; see Fig. 2).

Fig. 1

Mean participant reported level of perceived rapport with the interviewer on a 5-point Likert-scale. Error bars are 95% confidence intervals. Cohen’s d effect sizes for between-group comparisons are indicated

Fig. 2

Mean participant reported level of comfort talking about the video with the interviewer on a 5-point Likert-scale. Error bars are 95% confidence intervals. Cohen’s d effect sizes for between-group comparisons are indicated

Hypothesis 2

Participants within the rapport condition would provide more information disclosure than participants in the no-rapport condition.

The mean length of the substantive portion of the interview (in seconds) across all participants was M = 109.17, SD = 50.82, CI (98.82–119.52). The length of substantive phase of the interview (in seconds) was longer for participants in the rapport condition (M = 133.22, SD = 55.12, CI [116.85–149.59]) compared to participants in the no-rapport condition (M = 86.59, SD = 33.69, CI [76.91–96.27], d = 1.02, CI [0.60–1.45]; see Fig. 3). Participants in the rapport condition also spoke more words in the substantive portion of the interview (M = 237.96, SD = 122.48, CI [201.58–274.33]), than did those in the no-rapport condition (M = 153.63, SD = 71.17, CI [133.20–174.10], d = 0.85, CI [0.43–1.27]; see Fig. 4). There was a 43% difference in number of words spoken between participants in the rapport and no-rapport condition.

Fig. 3

Mean length of the substantive phase of interviews (in seconds). Error bars are 95% confidence intervals. Cohen’s d effect sizes for between-group comparisons are indicated

Fig. 4

Mean number of words spoken by the participant in the substantive phase of the interview. Error bars are 95% confidence intervals. Cohen’s d effect sizes for between-group comparisons are indicated

The mean number of details reported across all participants was (M = 26.63, SD = 11.34, CI [24.32–28.94]). Participants in the rapport condition reported a greater total number of details about the stimulus video (M = 29.57, SD = 12.18), CI [25.95—33.18]), than did those in the no-rapport condition (M = 23.88, SD = 9.84, CI [21.05–26.70], Cohen’s d = 0.52, CI [0.11–0.92]; see Fig. 5). Participants in the rapport condition also reported a greater number of accurate details about the stimulus video (M = 27.04, SD = 11.59, CI [23.60–30.50]), than did those in the no-rapport condition (M = 22.55, SD = 9.68, CI [19.77–25.33], Cohen’s d = 0.42, CI [0.01–0.83]).

Fig. 5

Mean number of accurate, inaccurate, and total details. Error bars are 95% confidence intervals. Cohen’s d effect sizes for between-group comparisons are indicated

Discussion

The objectives of this study were to examine the rapport-building process in a virtual interviewing context. Overall, our results suggest that (1) rapport can be built successfully in a virtual interview using many of the tactics suggested for in-person interviews and (2) building rapport led to increased information disclosure from interviewees regarding a difficult-to-discuss target event. These findings can help inform both the existing literature on rapport and professional practice for those conducting virtual interviews.

We found support for our first hypothesis as participants in our rapport condition reported substantially more perceived rapport with the interviewer compared to those in the no-rapport condition. There was more than a 30% difference on our perceived rapport scale (56.8% vs. 88.6%), the CIs did not overlap, and the standardized effect size was large (d = 1.47). While only a single study, these results provide strong evidence that rapport can be built within this context. We believe these findings are particularly impressive given the constraints present within virtual interviews (e.g., technological issues, absence of physical interaction).

Although speculative, we believe that the strong ratings of perceived rapport resulted from the flexible interview guide that we utilized in the study. Our guide allowed interviewers to personalize the approach for each interviewee and included key aspects of effective rapport building, including demonstrations of empathy, self-disclosure, and genuine expressions of interest in the interviewee throughout the interview process (Matsumoto and Hwang 2021). In our sample, this led to a discussion of topics ranging from personal experience with online learning to hobbies and career aspirations. This speculation is also supported by comments in the open-ended responses, as many participants in the rapport condition mentioned that they felt that the interviewer was personable, friendly, made them feel comfortable, and showed interest in getting to know them before asking about the video: “The interviewer opened the discussion by asking for basic information and random social conversation which made me comfortable talking about the video” and “The interviewer was very nice and made several connections with me. She connected what was going on in my life to her life. This made me feel very comfortable with her because I felt like we were alike because of our situations with school and so on.” Overall, these findings suggest that interviewers can build rapport effectively in a virtual environment using approaches like those found effective in in-person interviews.

We also found support for our second hypothesis, as the increased perceived rapport in the rapport condition led to participants providing increased information disclosure compared to those in the no-rapport condition. Both the length and the number of words spoken during the substantive phase of the interview were much higher in the rapport condition (d = 1.02 and d = 0.85, respectively), and participants in the rapport condition also provided a greater number of total and accurate unique details about the video to the interviewer (d = 0.52 and d = 0.42, respectively). We purposefully chose a stimulus (i.e., sexual education video) that would be uncomfortable to discuss, and the video was rated by participants as awkward to discuss. Anecdotally, interviewees were also often reluctant to use the words contained in the video (e.g., penis, vagina) and would instead paraphrase the content as opposed to reporting it verbatim. Although all participants found the video awkward to discuss, those in the rapport condition clearly were more willing to discuss the content of the video, as evidenced by the marked increase in interview length and words spoken.

Given the experimental nature of our design, we believe that the increased disclosure is a direct result of the rapport built with interviewees in this condition. This is supported by the open-ended responses as well, as participants in the rapport condition explicitly mentioned that having the interviewer take the time to try and connect with them made them more comfortable discussing the content of the video: “The interviewer was inviting and gave off very comforting energy, therefore I felt comfortable and trusting enough to talk about the video and situation”; “This level of rapport was [because] the interviewer made sure to talk with me beforehand and to get to know me better before asking me about the questions, therefore, I was more comfortable when talking about the video.” Overall, these findings support the idea that building rapport within virtual interviews will lead interviewees to be more comfortable discussing the details of uncomfortable or distressing events.

Limitations

This study has several limitations that may impact the generalizability of the findings. First, the fact that this study is a laboratory study means that the topic discussed is not a real event that the interviewee experienced and therefore lacks realism. Second, the interviewers were peers to the participants as opposed to a person of authority, as they would be in a real-life scenario, which may have made it easier for participants to discuss the event compared to being interviewed by a police officer. Third, while we purposely chose our stimulus to represent a difficult-to-discuss event, it was not a video of a crime; therefore, it is difficult to extrapolate directly to a real-world interview about a criminal event. We should note, however, that as the video did not contain a typical criminal event, many participants did not report details that they presumably would have if asked to recall a criminal event (e.g., a description of the main character(s) in the video). Although speculative, given the large disparity in length and words spoken in the substantive phase we found for our stimulus, we believe that the increase in the number of details reported would be even greater for a difficult-to-discuss criminal event.

Conclusion and Future Directions

A portion of investigative interviews have moved to virtual settings, and there is a potential for this practice to be maintained moving forward. This has raised questions regarding interviewers’ ability to build rapport in this novel context. Our results provide empirical evidence that (1) rapport can be established effectively within virtual interviews, and (2) increased rapport can lead to increased disclosure from interviewees about difficult-to-discuss events. Although a single study, we believe that this supports the idea that virtual interviews can be conducted effectively despite their limitations and that interviewers should continue to strive to build rapport within this context. This is particularly true where the topic discussed may be difficult to speak about (e.g., whether it be due to the event’s personal, taboo, or traumatic nature or topic). Not only do interviewees perceive the efforts to build rapport and feel more comfortable discussing difficult topics, but they also tend to provide more information about a target event, which, in turn, has the potential to aid investigations. While the goal of the current research was to examine the rapport-building process within virtual interviews, we believe that future research should also focus on comparing rapport within face-to-face interviews and virtual interviews directly to further explore the relative effectiveness of virtual interviews.