The dynamics of two-session interviews with suspected victims of abuse who are reluctant to make allegations

Abstract Supportive forensic interviews conducted in accordance with the National Institute of Child Health and Human Development Revised Protocol (RP) help many alleged victims describe abusive experiences. When children remain reluctant to make allegations, the RP guides interviewers to (a) focus on rapport building and nonsuggestive support in a first interview, and (b) plan a second interview to allow continued rapport building before exploring for possible abuse. We explored the dynamics of such two-session RP interviews. Of 204 children who remained reluctant in an initial interview, we focused on 104 who made allegations when re-interviewed a few days later. A structural equation model revealed that interviewer support during the first session predicted children's cooperation during the rapport-building phase of the second session, which, in turn, predicted more spontaneous allegations, which were associated with the interviewers’ enhanced use of open-ended questions. Together, these factors mediated the effects of support on children's free recall of forensically important information. This highlighted the importance of emphasizing rapport with reluctant children, confirming that some children may need more time to build rapport even with supportive interviewers.

(Received 1 October 2020; accepted 1 October 2020) Although our program of research on developmentally appropriate interviewing began in the late 1980s, the second author's awareness of and scholarly interest in child maltreatment began in the early 1970s, when he was a graduate student at Yale. Ed Zigler had just returned from Washington, where he had, as head of the Children's Bureau (soon renamed the Office of Child Development and later the Administration for Children and Families), ensured creation of the National Center on Child Abuse and Neglect, which pioneered the development of policy and practice designed to reduce the incidence of this underappreciated problem. As in so many other areas, Zigler's vision was broad and his agenda ambitious; as a mentor, he pushed his students to address research questions that were practically important, however complex and difficult they might be to answer. When the author and Zigler last met, decades later, Zigler eagerly discussed the research on forensic interviewing, and was delighted that our research group was still focused on the "tough questions": specifically, how to identify, understand, and help maltreated children who were unable to disclose what had happened to them. We hope Zigler would have approved of the latest study in that program of research, reported for the first time in this special issue highlighting his indelible contributions to applied developmental science and the enhancement of children's well-being.
As Zigler and his contemporaries realized (e.g., Gerbner, Ross, & Zigler, 1980), many suspected victims of child abuse do not make allegations when formally interviewed, with the numbers doing so varying depending on their age, gender, and relationship with the suspected perpetrator (Hershkowitz, Horowitz, & Lamb, 2005;London, Bruck, Wright, & Ceci, 2008). Specifically, children are particularly reluctant to report abuse by parents, guardians, or adults on whom they depend (Malloy, Brubacher, & Lamb, 2011), especially when they are under 6 years of age or boys. Although some children do not report suspected abuse because it did not occur, as evidenced by higher disclosure rates in substantiated cases (e.g., Leach, Powell, Sharman, & Anglim, 2017), many children still fail to disclose abuse when there is clear evidence that it did occur (Hershkowitz, Lamb, & Katz, 2014;Rush, Lyon, Ahern, & Quas, 2014).
Researchers have highlighted how rapport and emotional support can facilitate children's reports of abuse (for a review, see Saywitz, Larson, Hobbs, & Wells, 2015), albeit noting concerns that false allegations may be elicited more frequently following suggestive support (Saywitz et al., 2015), or selective reinforcement of specific content (Lyon & Dorado, 2008). The challenge when forensically interviewing reluctant children is therefore to provide nonsuggestive interviewer support without referring to specific allegation-related details (Lamb, Brown, Hershkowitz, Orbach, & Esplin, 2018).
Interviewers are often ineffective and even counter-productive when they encounter suspected victims who are hesitant to make allegations (e.g., Hershkowitz, Orbach, Lamb, Sternberg, & Horowitz, 2006;Orbach, Shiloach, & Lamb, 2007). Studying interviews conducted in accordance with the Standard National Institute of Child Health and Human Development (NICHD) Protocol, Hershkowitz et al. (2006) reported that interviewers withdrew support, explored the possibility of abuse prematurely, and asked reluctant children intrusive questions, thereby ignoring their emotional needs. In turn, the children became less responsive to the interviewers' prompts, revealed fewer personally meaningful details about neutral experiences during a rapport-building phase, and provided fewer abuse-related details when substantive issues were explored (Hershkowitz et al., 2006;Lewy, Cyr, & Dion, 2015). The researchers thus stressed the importance of recognizing early signs of reluctance, which signal the need to develop further rapport in order to enhance children's engagement (Hershkowitz et al., 2006;Katz et al., 2012). They recommended that interviewers (a) make enhanced efforts to establish rapport when children are uncooperative, (b) use open rather than closed-ended questions, and (c) refrain from exploring the possibility of abuse before children appear trusting and engaged.
Informed by research underlining the importance of meaningful rapport and emotional support as well as the counterproductive tendencies of some forensic interviewers, the Standard NICHD Protocol (SP) has been updated over the last several years (Lamb et al., 2018). The Revised Protocol (RP) provides interviewers with more guidance about how to (a) build rapport effectively, especially when children are reluctant, and (b) behave supportively yet nonsuggestively.
Several studies have illustrated that, following intensive training  in use of the RP rather than the SP, interviewers are more supportive and that this is in turn associated with reduced reluctance during the presubstantive (Hershkowitz, Lamb, Katz, & Malloy, 2015) and substantive (Blasbalg, Hershkowitz, & Karni-Visel, 2018;Blasbalg, Hershkowitz, Lamb, Karni-Visel, & Ahern, 2019) phases of individual forensic interviews, especially when the support is directly responsive to signs of reluctance (Ahern, Hershkowitz, Lamb, Blasbalg, & Winstanley, 2014;Blasbalg et al., 2018). In addition, when addressed supportively, children are also more responsive and emotionally expressive . These reactions are associated with the provision of more information about the suspected abuse Karni-Visel et al., 2019) in the context of more coherent narratives . Such findings imply that decreased reluctance, increased responsiveness, and increased emotional expressiveness reflect increased cooperation, and therefore they were treated as indices of cooperation in the current study. Most importantly, changes in the interview dynamics were accompanied by corresponding increases in the likelihood that suspected victims of abuse would make allegations (Hershkowitz & Lamb, 2020). This was especially true for suspected victims of intrafamilial abuse , who, as noted above, are particularly reluctant to make allegations. Nevertheless, many reluctant children fail to make allegations even when there is good reason to believe that they have indeed been abused.

Repeated Interviewing as a RP Practice
Single interviews characterized by nonsuggestive support, both cognitive and emotional, represent best practice, but some children may benefit from additional interviewing sessions (Lamb et al., 2018), either for the purpose of overcoming socioemotional barriers to disclosure (Faller, Cordisco-Steele, & Nelson-Gardell, 2010;Waterhouse, Ridley, Bull, La Rooy, & Wilcock, 2016) or so that children can elaborate upon and clarify previous statements (Hershkowitz & Terner, 2007;Katz & Hershkowitz, 2013). The latter group of repeated interviews mostly involve cognitively based strategies to promote recall through reminiscence or hypermnesia (La Rooy, Katz, Malloy, & Lamb, 2010), issues that are not discussed here. Instead, we examined emotion-focused strategies for building trust and reassuring anxious and reluctant children in order to facilitate valid disclosures. When children display reluctance in early phases of the interview, the RP includes structured guidelines for conducting multisession interviews focused solely on extended rapport-building and socioemotional support (Lamb et al., 2018). This guidance is informed by research stressing the vital roles of rapport and trust in the abuse disclosure process (see Saywitz et al., 2015), as well as by the recognition that some children may need more time to work through suspicion or mistrust and establish rapport with interviewers (Carnes, Wilson, & Nelson-Gardell, 1999). Multiple interviews may allow trust to develop gradually as interviewer support and reassurance help dissipate negative emotions such as fear, embarrassment, or general distress associated with disclosure and/or the abuse itself . Thus, multiple supportive interviews may enable reluctant children to form relationships with interviewers at their own pace, making them feel more comfortable to talk about abusive experiences. We review the extant research in this section, focusing narrowly on nonsuggestive repeated interviews in light of abundant research showing that repeated suggestive interviews have adverse effects on accuracy (see La Rooy et al., 2010).
A few field studies suggest that multisession interviews may make reluctant children more willing to make allegations of abuse in a subsequent interview than in a first one. De Voe and Faller (1999) reported that some 5-to 10-year-old children (11% of 77) made allegations in a second interview despite having been reluctant to do so in the first. Relatedly, Leander (2010) and Azad and Leander (2015) reported that children were less avoidant, made fewer denials, and provided fewer avoidant responses in a second or third interview than in the first. Although not directly tested, the latter authors attributed decreased reluctance in later sessions to the better rapport established across multiple sessions.
Most relevant to the current study, a structured model of multiple preplanned interviews has been implemented and tested in NCAC centers in the United States (Carnes et al., 1999;Carnes, Nelson-Gardell, Wilson, & Orgassa, 2001). A multisite study (Carnes et al., 2001) comparing interviews conducted over four and eight sessions showed that professionals were equally able to decide whether abuse had occurred in the two conditions. However, a more thorough examination of these cases, including credibility assessments , showed an advantage for the longer interview sequence.
Other laboratory analog studies have manipulated repeated interviews in combination with factors that affect disclosure by reserved or reluctant children: rapport, support, and question types. In a recent study, Stolzenberg, Williams, and Lyon (2018) examined the type of prompts that elicited children's disclosure of transgressions that they had been asked to keep secret. About half of the children (4-to 9-year-olds) disclosed in their first interview and another fifth disclosed in their second interview. Prompting about conversations was especially effective in eliciting disclosures, although many children still failed to disclose the transgression in either interview.
Goodman, Bottoms, Schwartz-Kenney, and Rudy (1991) manipulated both repeated interviewing (following 4 weeks) and interviewer-provided support. Children (4-to 7-year-olds) were less suggestible in their second than in their first interviews and provided more accurate information from free recall when given support in a second interview. Davis and Bottoms (2002) found that 6-to 7-year-old children who were interviewed immediately by a supportive interviewer were less suggestible. In a second interview performed one year later (Bottoms, Quas, & Davis, 2007), children interviewed by supportive interviewers were more accurate in their free recall and, again, less suggestible. By contrast, another study of very young children (3-to 4-yearolds) failed to show effects of support in either of two interviews conducted 4 weeks later (Imhoff, 2000).
Recently, Brubacher et al. (2019) tested the effects of support (and familiarity) on children's reports of an adult's transgressions during repeated interviews. They examined 160 interviews conducted with 5-to 9-year-old children who participated in a staged event involving six transgressions. Children who were interviewed in a supportive rather than a neutral style reported more transgressions in the second interview, especially in response to open-ended invitations, and were also more consistent across interviews. However, support in the first interview did not directly affect the overall number of details reported.
Relatedly, Waterhouse (2016) reported that the length of rapport building in a first interview, rather than the mere existence of a rapport building session, was positively associated with children's performance in multiple interviews. Nevertheless, the amount of rapport building in second or third interviews was also correlated with performance in those interviews. Interestingly, children in Waterhouse's study reported less anxiety in the second interview than in the first one, suggesting that rapport had positive effects on psychological well-being over multiple interviews. Thus, contrary to concerns that second interviews may have negative effects on children's well-being (Connell et al., 2009), it appears that children can benefit emotionally from prolonged rapport-building across multiple sessions. Relatedly, in a study of adults, Holmberg and Madsen (2014) showed the emotional benefits of rapport across sessions: In a first interview, a 'humanitarian' interviewing style was associated with lower anxiety among interviewees, whereas in a second interview, rapport was associated with an improved sense of coherence.
Direct evidence for the effectiveness of the Revised NICHD protocol repeated interview model was recently provided by Blasbalg, Hershkowitz, Lamb, and Karni-Visel (in press). They sought to determine whether emphasizing rapport and support during a first interview, while avoiding intrusive questions, would increase children's willingness to make allegations in a second interview. They studied all repeated forensic interviews conducted in Israel by RP-trained interviewers using this approach within a specified period of time. Of 202 3-to 14-year-olds who did not make allegations in the first interview and were re-interviewed, 104 made allegations during the second interview while 98 children still did not. When interviewers offered more support during the first interview and asked more open-ended and fewer closed-ended questions about the abuse in the second interview, children were more likely to allege abuse in their second interview. This association was mediated by children's enhanced cooperativeness in the second interview, validating the repeated interview RP guidance for interviewing suspected victims of abuse who refrain from making allegations. However, Blasbalg et al.'s (in press) study focused only on whether the children made allegations and did not fully explore the dynamics of interviews with those who did make allegations.

The Current Study
We thus focused on the quality of the forensic statements made by a group of reluctant children who did not make allegations in their first interviews but did so in second interviews. We anticipated that more interviewer support in the first session would predict the following child outcomes in the second session: (a) better rapport, operationalized by greater cooperation in the presubstantive phase in which rapport-building continues, (b) more spontaneous reports of abuse (i.e., allegations made in response to fewer prompts in the transitional phase, in which interviewers explore the possibility that abuse happened), and (c) increased informativeness in the substantive phase, when the abusive event/s are described. Following Blasbalg et al.'s findings, we predicted that the association between level of support in the first session and spontaneity of the allegations would be mediated by the level of rapport displayed in the second session while questioning strategies predict the level of informativeness once allegations had been made.

Sample
Between January 2014 and December 2016, 77 investigators from the Israeli Ministry of Welfare and Social Services conducted second investigative interviews of 202 suspected victims of abuse who did not make allegations when first interviewed formally. Those interviews were studied by Blasbalg et al. (in press). In 104 of those cases, the children made allegations of and described incidents of abuse when re-interviewed; those interviews were subject to examination in the current study. The children (49% girls), aged 3.50 to 13.63 years (M = 8.87, SD = 2.56) alleged physical (58%) or sexual abuse in the second session, allegedly perpetrated by either a family member (77%) or a nonfamily member. All interviews were performed by interviewers who had been extensively trained to use the RP , including for conducting multisession interviews.

Materials
The RP represents a revision of the SP and is characterized by an enhanced emphasis on continuous rapport building and supportive interviewing (Lamb et al., 2018). It is fully structured, covering all phases of the investigative interview. Both Protocols open with a presubstantive phase designed to establish rapport, prepare child-witnesses for their role as informants, and train narrative response styles by exploring nonabusive experienced events. In the Transitional phase, interviewers then switch focus to substantive issues (the possibility that abuse occurred), using a structured series of transitional prompts in which increasingly focused, nonsuggestive prompts are used when 'free recall' prompts fail to elicit an allegation. If the child makes an allegation, the interviewer seeks further information, primarily by using open-ended invitations, during the Substantive phase. The guide encourages interviewers to exhaustively probe children's memory using free recall prompts before asking directive questions, with few option-posing questions asked only when necessary to elicit critical information.
In order to promote children's emotional comfort, trust, and cooperation, rapport building in the RP (but not in the SP) precedes rather than follows explanation of the ground rules. Welcoming greetings and expressions of interest in the child's well-being are offered at the beginning of the interview alongside supportive utterances reflecting personal interest in the child ("I'm really interested in getting to know you"). Interviewers were guided to identify expressions of reluctance and emotion made by the child as they occurred, and to answer them using nonsuggestive supportive comments (see examples below). Additionally, the RP lists nonsuggestive yet supportive comments of several types: Expressions designed to promote rapport with the child ("Good to meet you," "I want to know you better"); emphasis on the interviewer as a trustworthy professional ("I am here to listen to you"); positive reinforcements of the children's effort ("You are being very clear," "Thank you for sharing with me"); expressions of emotional support (echoing/ acknowledging/ exploring children's feelings and anxieties: "You say you feel embarrassed to talk about that; please tell me about that") or encouragement ("It's important that you tell me everything you remember as best you can").
With especially reluctant children, the RP encourages interviewers to extend the process of rapport building and to conduct it over multiple sessions. Interviewers are guided to assess children's level of cooperation/or reluctance during the presubstantive phase. When encountering reluctance, interviewers are guided to avoid engaging in substantive topics pertaining to the suspected abuse and, instead, to continue building rapport with the child while responding supportively to verbal and nonverbal expressions of reluctance. When the child remains reticent, the Protocol recommends that the session be ended, with another planned within approximately one week. During the follow-up session, interviewers are instructed to continue developing rapport and acting supportively while assessing reluctance, moving on to explore the possibility that abuse might have occurred when reluctance declines and rapport has been established.

Data coding
Transcribed recordings of the interviews were checked for their completeness and accuracy before coding using Atlas.ti software (Muhr, 1997). Coders determined whether the specific types of interviewer utterances and children's responses described below were present or absent in each conversational turn. Supportive interventions by the interviewers in the first session were defined as expressions designed to promote rapport with the child, to communicate the interviewer's trustworthiness, to positively reinforce the child's efforts, or to express emotional support and encouragement. Coders then identified substantive interviewer utterances in the second interview and classified their type as either open-ended (invitations) or narrow/suggestive (directives, option-posing, and suggestive) prompts using the definitions provided by Lamb et al. (2018).
Expressions of reluctance by the children in the second interview rapport-building phase included omissions (no answer, "nothing to say," "don't know"), expressions of resistance ("I don't want to tell you," "I'll answer only this last question"), or denials ("It didn't happen," "I didn't say that"), regardless of whether the turn was responsive or informative (see below). We sought to minimize the misidentification of reluctance by not coding omissions as reluctance when the child could not be expected to possess the knowledge requested, such as about others' thoughts or temporal information. Omissions were only coded in the rapport-building phase during which interviewers prompted for nonspecific personal information ("tell me about your school/about things you like to do" -"I don't know").
Coders also identified each child utterance as either responsive or not, depending on whether the child gave a reply relevant to the question asked, whether the information provided was new or repeated. In the substantive phase of the second interview, when the children were responsive, their responses were coded as either informative, when they provided new forensically relevant information (Lamb et al., 1996;Yuille & Cutshall, 1986), or noninformative.
The presence of internal content was coded whenever the children described their own emotions (e.g., "afraid," "happy"); thoughts ("I was planning to run away") feelings (e.g., "I felt like I wanted to bang my head against the wall"), or sensations ("my body was shaking").
Measures were computed as proportions of the total number of conversational turns in the relevant part of the interview, to account for variations in the length of the interview phases. Interviewers' behavior was represented by three measures: the proportion of utterances in the first session in which interviewers offered children support, the length of (number of utterances in) the transitional phase during the second session (a measure of spontaneity), and the proportion of open-ended questions posed during the substantive phase of the second interview. Children's level of cooperation during the second presubstantive phase in which rapport-building continued was represented by three indices, shown in previous research to reflect an enhanced willingness to provide substantive information: the proportion of prompts to which they replied responsively, expressions of reluctance, and references to internal content. The amount of forensically relevant information provided by the children was indexed by the proportion of informative replies, in which they provided substantive information in response to open-ended questions.

Inter-rater reliability
Four raters who first established inter-rater reliability on a separate set of transcripts coded the interviews. To ensure the maintenance of high levels of reliability throughout coding, all coders coded 20% of the transcripts. Coders were blind to the research hypotheses. K alpha inter-rater index (Hayes & Krippendorff, 2007) coefficients for support, reluctance, question types, responsiveness, informativeness, and internal content were 0.78, 0.77, 0.88, 0.79, 0.76, and 0.80, respectively.

Analytic strategy
Outlier detection analyses (using an IQR of 2.2; Hoaglin & Iglewicz, 1987) revealed no outliers on the aforementioned measures. Pearson correlations were conducted to assess simple associations between measures of the interviewers' and children's behavior. To test the hypothesized associations between interviewers' support during the first interview and children's cooperation and disclosure patterns in the second session, structural equation modeling (SEM) was performed in R (R Core Team, 2018) using the lavaan package (Rosseel, 2012) with the weighted least square mean and variance adjusted (WLSMV) estimator and robust standard errors to effectively account for nonnormality in the data (Brown, 2006). Missing values on the measures of responsiveness, reluctance, and internal content for the six cases in which narrative training was not conducted were replaced via multiple imputation (five iterations) using the mice package in R (van Buren, 2015).
We employed a two-step modeling process (Tabachnick & Fidell, 2013). First, we tested the initial theoretical model. Then, we examined modification indices to identify significant correlations absent from the model. Child's age, being a continuous index, was modeled as part of the SEM model. The number of cases in our sample was not sufficient to explore possible differences with respect to gender, identity of the offender (family member or not), or type of abuse (physical or sexual).
Model fit was tested using multiple indicators. In addition to the χ 2 test of model fit, the Tucker-Lewis index (TLI), the comparative fit index (CFI), and the root mean square error of approximation (RMSEA) score were examined. For the χ 2 test, a nonsignificant χ 2 test indicates a good model fit (McDonald & Ho, 2002). For both the TLI and CFI, a value ≥ .95 indicates a good model fit, and values ≥ .90 indicate an acceptable fit (Hu & Bentler, 1999). For RMSEA, values ≤ .06 indicate a good model fit, and values ≤ .08 indicate acceptable fit (McDonald & Ho, 2002).

Results
The first session averaged 90.75 interviewer-child exchanges (SD = 40.60; See Table 1). Interviewers made supportive comments in 0.45 (SD = 0.13) of their utterances in the first session. The second session averaged 128.05 interviewer-child exchanges (SD = 78.37). A full rapport-building phase was carried out during the presubstantive phase in 98 of the second interviews, consisting of 15.93 (SD = 14.98) exchanges on average. Children were responsive to 0.60 of those rapport-building utterances (SD = 0.28) and expressed reluctance and internal content in 0.28 (SD = 0.23) and 0.13 (SD = 0.15) of the utterances, respectively.
The transitional phases of the second interviews averaged 18.20 (SD = 29.83) conversational turns, and the substantive phase averaged 80.59 (SD = 68.46) interviewer-child exchanges. In the substantive phase, the proportion of utterances involving open-ended questions asked by interviewers was 0.35 (SD = 0.18); children provided forensically relevant information in response to 0.40 (SD = 0.23).
The proportion of utterances in which interviewers offered children support during the first session was positively and significantly correlated with the proportion of utterances to which children were responsive (r = 0.34, p < .001) in the presubstantive phase of the second session, during which rapport-building took place. Interviewers' support was also correlated with the proportion of open-ended questions they asked during the substantive phase (r = .24, p = .01). The number of responsive replies during the second session's presubstantive phase was negatively correlated with the number of conversational turns during the transitional phase (r = −.30, p < .01), but positively correlated with the interviewers' use of open-ended questions during the substantive phase (r = .36, p < .001), and with the production of substantive information (r = .29, p < .01). The use of open-ended questions was also correlated with the production of substantive information (r = .36, p < .001). There was no significant association between support in the first interview and either the number of turns during the transitional phase or the children's informativeness, indicating the need to test for possible mediation.

Testing a mediation model
First, SEM was performed to examine the relationships among measures of the interviewers' behavior and children's responses. The model included the proportion of supportive expressions offered during the first session as the predictor, the proportion of children's statements that were forensically informative as the outcome, and three mediators: children's cooperation during the rapport-building phase in the second session, the number of utterances during the transitional phase, and the proportion of questions asked by interviewers that were open-ended. The number of utterances during the transitional phase served also as an intermediate outcome, representing the spontaneity of the allegation. Children's cooperation was modeled as a latent variable defined by the proportions of their utterances that were responsive, contained internal content, and expressed reluctance.
Modification indices indicated that four associations should be added to the original model: (a) between two measures of children's cooperation (responsiveness and internal content), (b and c) between each of these indices of children's cooperation and their substantive informativeness, and (d) between the length of the transitional phase and the proportional use of open-ended questions. The resultant model, revised using the modification indices, fitted well, χ 2 (16, N = 104) = 4.129, p = 1.000, TLI = 1.330, CFI = 1.000, RMSEA = 0.000, 95% confidence interval [0.000, 0.001]. Figure 1 displays standardized parameter estimates for the structural model. Tests of the SEM model revealed that the provision of interviewer support during the first session was positively associated with children's cooperation during the rapportbuilding phase of the second session. In addition, cooperation during the rapport-building phase of the second session was negatively associated with the length of the transitional phase that followed. The length of the transitional phase was negatively associated with the proportion of questions during the substantive phase that were open-ended, which in turn positively predicted children's informativeness in response to open-ended questions. Together, these findings supported the prediction that interviewer support in the first session was positively associated with children's cooperation at the beginning of the second session. The latter predicted more spontaneous disclosure, enhanced use of open-ended questions by interviewers, and the production of more informative statements by the children.
Further examination of the model sought to test the hypothesized patterns of mediation. We computed the indirect effect as the product of four coefficients reflective of the four pathways between the predictor (support) and the outcome (substantive informativeness). The results showed significant indirect (β = .12, p = .04, 95% CI [0.01, 0.25]) and total (β = .39, p = .04, 95% CI [0.02, 0.76]) effects between the provision of support during the first session and the children's substantive informativeness. Because the direct effect between interviewer support and children's substantive informativeness was not significant initially, these findings suggest a full mediation of the relationship between the two variables.

Discussion
Although prior studies have shown that it may be beneficial to interview reluctant suspected victims of abuse more than once, the role of interviewer support in enhancing the quality of forensic statements across repeated interviews has not previously been studied. In this study, we assessed the effects of such support by examining interviews with children who did not make allegations of abuse in a first interview but did so in a second one. The data showed that higher levels of support during the first interview predicted more responsive replies during the second rapportbuilding phase (i.e., the next interaction between the interviewer and child). Although children were given support in the first interview but remained reluctant, in the second interview they seemed to soften their reluctance, making them more cooperative. The enhanced level of responsiveness and willingness to disclose personal information before substantive issues are raised can be viewed as indices of the amount of rapport achieved (Hershkowitz, 2011). Thus, the present results suggest that higher support in a first interview was associated with improved rapport in the second. Although researchers who have previously studied reluctant children across multiple interviews have speculated that cooperation in a second interview might have been the result of improved rapport (Azad & Leander, 2015;Leander, 2010), this was the first study to demonstrate this empirically. The RP guides interviewers to recognize early signs of reluctance, respond with general and reactive support, and continue monitoring reluctance to assess the need for further rapport building. When reluctance remains high in a first interview, the RP instructs interviewers to avoid exploring possible abuse, end the interview, and plan a second interview to allow further rapport building. Our study underlined the value of this strategy, showing that rapport continued to develop in the second interviews with children who were initially reluctant.
However, contrary to our hypotheses, support in the first interview did not directly predict either of the outcomes associated with the quality of disclosure-spontaneity and richness of the information provided-implying that indirect associations might be involved. In previous studies of repeated interviews, support was correlated with informativeness in Goodman et al.'s study (1991) but not in Brubacher et al.'s (2019), and the overall association was weak in Saywitz et al.'s meta-analysis (2015). The SEM model tested in the present study suggested that the association between support and informativeness might involve indirect mechanisms, which may partly explain the contradictory findings that have been reported.
The first indirect outcome of support revealed in the current study was the spontaneity of the allegation: Higher support in a first interview was associated with increased cooperation during the second rapport-building session, which in turn predicted greater spontaneity (i.e., the number of prompts in the transitional phase). This mechanism emphasizes the crucial role of establishing rapport with reluctant children, confirming that some children may need more time to build rapport with interviewers (Saywitz et al., 2015), even under supportive conditions. It is possible that more supportive interviewers allowed reluctant children to develop trustful relationships which made them better able to regulate their negative emotions, cope with internal conflicts about disclosure, and thus allowed them to disclose more readily (Azad & Leander, 2015). The fact that allegations were made following fewer prompts is noteworthy because more spontaneous disclosures are more likely to be accurate and also more likely to be deemed credible, thereby increasing the value of the testimony (Goodman et al., 1991;Orbach et al., 2007).
In a second step, using the relative spontaneity of disclosure as an independent variable, we sought to indirectly predict the richness of the allegation, through the use of open-ended questioning. The SEM model confirmed this prediction, showing that children who disclosed more readily were given more open-ended invitations to describe the abusive incidents, to which they provided more free recall information about those incidents. The amount of information from free recall reflects the quality of a forensic statement (Lamb et al., 2018), so this finding also illustrates the important role played by support in the course of repeated interviews with reluctant children.
Taken together, the data shed light on the dynamics of what may be deemed 'successful' supportive repeated interviews with reluctant children guided by the Revised NICHD Protocolthose interviews in which suspected victims of abuse did not make allegations in the first session but did so in the second session. The mediation model suggests a complex path through which initial interviewer support improves rapport and cooperation in a second interview, allowing children to make allegations more spontaneously, in turn enhancing interviewers' use of openended questions when investigating the abusive event/s, allowing them to obtain more free recall information. Moreover, the identification of both direct and indirect effects may help explain why previous findings were inconsistent. Finally, the findings support the rationale for the RP practice of repeated interviewing, adding to a growing body of research documenting the value of this Protocol in forensic contexts.
The current study also illustrated the complex interplay between cognitive and socioemotional factors when interviewing alleged victims of abuse. Emotion-focused interventions (i.e., rapport and support) appeared essential for motivating reluctant children to make allegations more readily. However, once the child made an initial allegation, cognitively based strategies, with an emphasis on open-ended questioning, determined the richness and quality of the information provided.

Limitations
Because this was a field study of undocumented real-world events, it was not possible to assess the reliability of the statements or the accuracy of the details obtained from the children. However, the extant research suggests that supportive interviewing, and specifically supportive interviewing across repeated interviews, is likely to yield more rather than less accurate information and to improve children's resistance to suggestion (Bottoms et al., 2007;Goodman et al., 1991); risks are further reduced when support is correlated with the amount of freely recalled information, as in the current study. Nevertheless, the risks of contamination cannot be totally eliminated (Brubacher et al., 2019) and it is critically important for interviewers to avoid suggestive contamination when conducting repeated interviews with reluctant children.
Although the study provides insight into how support may have affected the behavior of reluctant children in repeated interviews, we recognize that the design does not permit us to establish causality. However, what happens in earlier phases is more likely to affect what happens in latter ones, rather than vice versa. Nevertheless, caution is warranted when interpreting the findings.

Implications for practice and policy
The reluctance of suspected victims of child abuse to report is a serious obstacle to the identification of victims, their protection, and their access to welfare and justice services (Cross & Hershkowitz, 2017;Gerbner et al., 1980;Kaufman & Zigler, 1996), especially because independent evidence of abuse is often lacking. However, despite the potential benefits of using multiple sessions with reluctant children, or conducting supportive interviews, those practices are not common (Cross et al., 2007;Hershkowitz et al., 2006;Lewy et al., 2015;Teoh & Lamb, 2013).
In the current study, the combined implementation of the two practices was made possible by interviewers following the NICHD RP guidance. Known as an effective tool for motivating reluctant victims to make valid allegations (Blasbalg et al., in press;Hershkowitz et al., 2014), the RP here has been shown to be valuable in multiple interviews with especially reluctant children. The findings show how gaining suspected victims' trust and cooperation occurred across multiple interviews was followed by allegations that were more spontaneous and richer. Thus, the RP may represent a useful technique for interviewers who often struggle with children's unwillingness to report their victimization.
The data showed how supportive interviewing was associated with the quality of the allegations made by reluctant children who were interviewed twice. However, many abused children remain resistant and decline to make allegations even after two interviews. Attempting to understand the dynamics that make some children, but not others, willing to make allegations, Blasbalg et al. (in press) reported similar dynamics in these two groups, suggesting a chain of effects in which initial support during the first interview, as well as rapport and open-ended questioning in the second interview, determined whether children would make allegations. Thus, the RP guidance to address reluctant children via effective emotionally based and cognitively based strategies seems as a promising practice, which might be useful in conducting additional interviews with children who do not initially make allegations.
However, just providing the Protocol to professionals does not improve their performance; in the absence of effective training, changes in interviewing practices are likely to be limited and temporary (for a review, see Lamb, 2016), perhaps even more so where the RP is concerned, because it requires the integration of various skills for addressing children's distress and facilitating detailed recall (Cross & Hershkowitz, 2017). Research shows open-ended questions as well as between references to internal content and substantive informative utterances were included in the model but omitted from the diagram in the interests of clarity. *p < .05, **p < .01, ***p < .001 that intensive and spaced training is an essential component of effective interviewer training and that continued supervision and feedback are needed to ensure the consolidation and maintenance of interviewer skills over time (Lamb, 2016). This costly and demanding method of training is rarely used in the field, where shorter but ineffective training methods predominate (Powell, Wright, & Clark, 2010). A responsible and evidencebased policy should recognize the indispensable role of training and ongoing supervision for delivering best practice.
Investigative and protective agencies must also structure interviewers' workload to allow time for multisession interviews when children are especially reluctant. The routine use of multiple interviews for reluctant children may raise concerns about costs and feasibility, as this practice is more time consuming. However, a cost-effectiveness analysis has suggested that the routine use of repeated forensic interviewing with children should improve decision making and result in a significant increase in conviction rates (Block, Foster, Pierce, Berkoff, & Runyan, 2013). Thus, it seems that multiple interviews, when needed, not only benefit the children involved, but also address societal goals.