The power of words: the impact of police interviewer’s judgment error and apology on sexual violence victims in simulated interviews

This experiment examines how a police interviewer ’ s judgment error and apology affect a sexual violence victim ’ s secondary victimization, trust in the interviewer, rapport, and willingness to provide information. Ninety students from Peru were asked to imagine being sexual violence victims and were interviewed online by a mock police interviewer. Participants were randomized into three conditions: judgment error without apology, judgment error with apology, and a no-error control. The judgment error involved suggesting the victim was partly responsible for their victimhood, citing women ’ s clothing choices and questioning men ’ s masculinity. Our findings show judgment errors reduced trust, rapport, and willingness to provide information while increasing secondary victimization; apologizing improved trust and rapport but did not affect secondary victimization or willingness to provide information. These outcomes indicate that judgment errors in sexual violence victim interviews can negatively impact the interviewing process and outcomes and highlight the importance of an apology.


Introduction
The prevalence of sexual violence has not dramatically changed globally for over two decades.Between 1993 and2017, around 1,155-1,194 males and 2,750-2,816 females per 100,000 persons were victims of sexual violence annually (Borumandnia et al., 2020).That is, around 1% of men and almost 3% of women become victims of sexual violence across the world each year.Although these numbers are alarming, the actual prevalence may be significantly higher because survivors of sexual violence often do not disclose to law enforcement what happened to them.Reasons for non-reporting are diverse, and include feeling ashamed, a perceived lack of evidence, having concerns about the legal processes of investigation and trial, and adherence to gender norms and cultural/religious taboos (Borumandnia et al., 2020;Ceelen et al., 2019;Depraetere et al., 2020).It is, therefore, important that victims of sexual violence who report to the police are treated with care and understanding to prevent secondary victimization.
Secondary victimization refers to additional harm victims experience due to interactions with individuals or institutions because of their initial victimization (Campbell & Correspondence: Miriam S. D. Oostinga, Section of Psychology of Conflict, Risk, and Safety, University of Twente, PO Box 217, 7500AE, Enschede, The Netherlands.Email: m.s.d.oostinga@utwente.nlRaja, 1999;Orth, 2002).For example, Patterson (2011) note that some police officers threatened victims who reported being sexually abused that if they were lying, they themselves would be prosecuted instead.Other sexual assault victims state that their experience with the police was emotionally damaging because it was characterized by stigmatization and insensitivity, and they felt they were being blamed for what happened (McQueen et al., 2021).
Thus, the phenomenon of secondary victimization is about people who should ensure the safety of the victims creating, unintentionally or on purpose, harmful situations for the victim.Webster and Oxburgh (2022) conducted a qualitative appraisal of the effect of being interviewed as a victim of sexual violence with genuine victims of sexual violence.They showed that the process of being interviewed, even when conducted well, can result in secondary victimization.However, secondary victimization becomes much more likely when the interviewer fails to engage with the victim in a humane and empathetic manner.Hence, secondary victimization can occur for several reasons.For example, the interviewer may interpret the situation differently from the victim; have different values, or lack knowledge on the subject.Oostinga et al. (2018a) proposed that subjective errors, meaning a misrepresentation of the thoughts and feelings of the interviewee, made in the communication by police, are referred to as judgment errors.In the context of suspect interviews, Oostinga et al. (2018b) found that judgment errors negatively affected the interview effectiveness in terms of the trust in and rapport with the interviewer.However, Oostinga et al. (2018a) found that police officers found that such errors could be ameliorated by using an appropriate response strategy (e.g. an apology), yet, they did not assess the effect of errors and response strategies when interviewing victims.
In this article, we experimentally and qualitatively (through follow-up interviews) explore the effect of judgment errors and apologies on a victim's experience of secondary victimization, trust in the interviewer, rapport, and willingness to provide information.Our experimental approach allows us to seek cause-effect relationships between what is being said to (simulated) victims of sexual violence by police interviewers and a victim's stance in the interview.This study, therefore, complements existing field research in this area which has described the impact of interviewer behavior on victims qualitatively.
In the following paragraphs we focus first on communication error management, specifically judgment errors and apologies, in victim interviews, after which we connect these concepts with secondary victimization, trust in the interviewer, rapport, and willingness to provide information.
Communication error management: the judgment errors and apologies Law enforcement interactions with the public are often long, complex discussions about sensitive topics.Even skilled interviewers and crisis negotiators are unlikely to avoid errors in perpetuity (Oostinga et al., 2018a;Russano et al., 2014).Consequently, communication errors commonly occur in law enforcement interactions.Oostinga et al. (2018a) proposed three types of errors in law enforcement interactions: First, factual errors are mistakes made concerning an objective factfor example, confusing the name of the interviewee.Second, judgment errors occur when the officer does not interpret the feelings or thoughts of the other person correctly; these are subjective mistakes.Finally, contextual errors happen when there is an error related to police procedures: for instance, when the officer uses police or technical terminology that the interviewee does not understand or says something regarding the procedure, assuming the interviewee is familiar with this when they are not.Here we focus on judgment errors, since they impacted interviewees most negatively compared to other types of errors (Oostinga et al., 2018b).
After a communication error occurs in a police interaction, the interviewers' response strategy will depend on whether they recognize an error has been made, and on individual differences in how to address recognized errors (Weiner, 1985); in this sense, the handling of the error varies from person to person.In the context of crisis negotiations, Oostinga et al. (2018a) identified four types of response strategies: (a) contradiction, where the police do not accept the error and so do not take responsibility for it; (b) attribution, when the police hold someone else responsible for the error, for example, by saying 'This is what someone else told me'; (c) apology, which refers to taking responsibility for what was said and asking for forgiveness; and (d) acceptance, where the police officer does not ask for forgiveness for their error but gives assurance that it will not happen again.For this research, we focused on apology, as saying sorry to the 'wronged' person creates a sense of being worthy in error recipients (Oostinga et al., 2018b;Sharma et al., 2022).This is important in victim interviews, since victims of sexual violence likely already feel vulnerable and stressed when approaching the police (McQueen et al., 2021;Webster & Oxburgh, 2022).

Secondary victimization and communication error management
Secondary victimization refers to the retraumatizing effect of inadequate treatment of victims by justice operators (Campbell & Raja, 1999).This phenomenon often occurs without any malicious intent (Albert ın, 2006)for example, through the excessive use of legal technicalities that the victim does not understand, slowness in the judicial process, lack of information, or confusion about the legal process that victims need to follow if they want to report.Secondary victimization may also arise because the police interviewer focuses on whether the case will likely reach court (i.e. the organization's needs), rather than focus on the victim's needs (Campbell & Raja, 1999).
In addition to the potential harm from procedural aspects of the justice system, secondary victimization can also be a consequence of the attitudes and prejudices of those working within the justice systemfor example, because society stereotypically attributes specific characteristics to victims, and those who do not conform to such stereotypes are often not acknowledged as being genuine victims.In the case of sexual violence, this means that the victim is expected to show a certain level of passivity and helplessness to be perceived as a genuine victim of sexual exploitation or abuse (van Dijk, 2009).People who do not match this 'ideal' victim are often perceived as bearing responsibility for their own victimhood, which can impact on their access to justice.Genuine victimhood is also often contingent on the victim's behavior before and during the assault.For example, women under the influence of a substance or wearing clothing considered 'provocative' at the time of sexual assault may be held responsible for the sexual violence themselves (Grubb & Turner, 2012).In addition, men are often assumed to be incapable of holding victim status because it is believed men could have defended themselves if they had wanted to (Borumandnia et al., 2020;Jackson et al., 2017).Importantly, these beliefs have been shown to be present in police officers.Specialist training can reduce the impact of such attitudes within police officers, but even after accounting for specialist training, endorsement of rape myths and sexist sexism beliefs impact upon the extent to which police officers attribute blame to perpetrators or victims (Murphy & Hine, 2019;Sleath & Bull, 2012).These stereotypical beliefs may, in turn, have a direct effect on legal decision makingthat is, if officers do not believe the victim's story, they are less likely to follow up on the story, charge the perpetrator, or ultimately go to court (Parratt & Pina, 2017;Sleath & Bull, 2017).
The expression of the stereotypical beliefs of the police officer may also directly lead to secondary victimization in the victim of sexual The Power of Words violence during the (intake) interview.A victim is often already reluctant to share what happened, because they are afraid that they will not be believed nor taken seriously by the police interviewer (Patterson, 2011;Webster & Oxburgh, 2022).In cases where police interviewers (indirectly) express prejudice towards the victim, they reinforce these concerns.Moreover, the process of interviewing requires victims of sexual violence to relive the often traumatizing event to provide the details that the interviewer needs for investigation (Webster & Oxburgh, 2022).When victims are not comforted or believed during this process, this may result in direct and long-term psychological and physical health distress (i.e.secondary victimization; Campbell et al., 2001).An apology may, however, repair this situation by both giving the victim the sense of being worth saying sorry to (Oostinga et al., 2018b;Sharma et al., 2022), but also because an effective apology acknowledges the harm caused to the victim and clearly indicates the error maker as responsible for this harm (Lewicki et al., 2016).An effective apology might then be expected to restore trust lost through making an error.

Trust in the interviewer and communication error management
Trust in the interviewer is important in interviews with victims of sexual violence because victims are placing themselves in a position of vulnerability, and additional disclosure is more likely when victims believe that the interviewer listens to them and takes them seriously (B€ ogner et al., 2007;Campbell et al., 2009).Trust in this context is therefore defined as the victim accepting to be vulnerable to the actions of the police interviewer (Mayer & Davis, 1999).Oostinga et al. (2018b) found that judgment errors negatively affect the trust between suspect and interviewer.These effects may be expected to be even stronger with victims of sexual violence interviews because of the capacity for reporting a sexual crime to be harmful for victims (B€ ogner et al., 2007;Campbell & Raja, 1999).In that sense, it is expected that making a judgment error during a victim interview harms the trust established with the victim.Yet, an apology is expected to repair trust, because by taking responsibility for their actions, interviewers can be seen as willing to improve and caring about the victims (Oostinga et al., 2018b).For an apology to be effective, research has shown that taking responsibility is the most important component (Lewicki et al., 2016).

Rapport and communication error management
Tickle-Degnen and Rosenthal (1990) define rapport through mutual attention, positivity, and synchrony or coordination between interlocutors.Rapport has been shown to be effective in eliciting more complete accounts from witnesses in laboratory studies (Gabbert et al., 2021).In victim interviews, police interviewers may, however, focus on asking questions rather than on maintaining rapport with the victim.This may lead to victims feeling rushed and uncomfortable (Patterson, 2011).Oostinga et al. (2018b) found that judgment errors affect the rapport between suspects and interviewers.We expect similar effects in victim interviews, as misjudgments of the situation may make police interviewers be seen as unsympathetic, which inhibits rapport (Maddox et al., 2011).Consequently, we expect rapport to be negatively affected after a judgment error.On the other hand, an apology is expected to help repair the connection, because it acknowledges the harm caused by the error.This requires a demonstration of mutual attention and attempts to reinitiate positivity in the interaction.

Willingness to provide information and communication error management
One way to obtain information in a criminal investigation is to interview the actors involved in the criminal event, such as the suspects, witnesses, and victims.The impacts we have described on trust and rapport have a tangible consequence on investigators' ability to elicit information from an interview.A poor relationship between the interviewer and interviewee directly impacts the information provided by that interviewee (Collins et al., 2002).Next to that, it is important that the police interviewer is perceived to be open and listening to the story of the victim (Webster & Oxburgh, 2022).In other words, negative experiences with the police directly affect willingness to share information with them (Koster et al., 2018).
In the case of the present research, a judgment error is expected to negatively impact interviewees' evaluation of the interviewer, causing them to be less willing to provide information.On the other hand, in line with Oostinga et al. (2018b), it is expected that an apology effectively repairs willingness to provide information.We predict this restorative effect of an apology, since it addresses the need for attention and care of victims (Webster & Oxburgh, 2022), as well as our prior expectations that an apology would also restore trust and rapport.

Current study
In this study, we complement existing qualitative research on interviews with victims of sexual violence (Jackson et al., 2017;Patterson, 2011;Webster & Oxburgh, 2022) by taking the experimental work of Oostinga et al. (2018b) that assessed the effect of judgment errors and apologies on trust in the interviewer, rapport, and willingness to provide information in suspect interviews to the victim interviewing setting.In addition, we added one further potential impact of error management relevant to victim interviews, i.e. secondary victimization.In summary, we expect that willingness to provide information has a positive relationship with both trust and rapport.We further hypothesize that when a police interviewer misjudges the situation and suggests that the victim bore culpability for their own victimhood because of stereotyped gender rolesin other words commits a judgment errorthen this error leads to more secondary victimization and less trust, rapport and willingness to provide information in comparison to an interview where no error was made.Finally, previous research shows that an apology restores trust and the relationship by taking responsibility for the transgression and showing remorse (Kim et al., 2004;Oostinga et al., 2018b;Sharma et al., 2022).It is, therefore, hypothesized that an apology removes the negative effects of a judgment error on secondary victimization, trust, rapport, and willingness to provide information.

Method Design
Participants played the role of victims of sexual assault and were randomly allocated via a random number generator to one of the three experimental conditions (apology, error, control).The conditions varied regarding the presence or absence of a judgment error and an apology.The apology group experienced both a judgment error and an apology.The error group experienced a judgment error without an apology.The control group experienced neither a judgment error nor an apology.Our dependent variables were participants' experience of Secondary Victimization, Trust in the Officer, Rapport, and Willingness to Provide Information.

Participants
Participants were recruited using snowball sampling, and with the assistance from with the assistance from professors at the Pontifical Catholic University of Per u (PUCP) who advertised the study to their students.The inclusion criteria were being Peruvian (to prevent any impact of differences in interviewer interviewee culture biasing results, e.g.see Minhas et al., 2017), over 18 years-old, and The Power of Words being a Spanish speaker.The sample consisted of 90 individuals with exactly 30 participants per experimental condition.
The age of the participants was between 18 and 41 years old (M age ¼ 27.84, SD ¼ 4.05).One participant chose not to disclose their gender; 45 participants were male (50.0%) and 44 were female (48.9%).

Materials and measures
Gender identity, nationality, age, level of education, occupation, and participation in a feminist organization were measured via an online questionnaire.All materials are available via the Open Science Framework project website: https://osf.io/axthv/.

Judgment errors
To mimic real life as much as possible, interviews were conducted with four police officers who work in the Direcci on de investigaci on criminal de la Polic ıa Nacional del Per u [Direction of Criminal Investigation of the Peruvian National Police] (DIRINCRI), who had on average five years of experience.Specifically, they provided information about which judgment errors are common when interviewing victims of sexual assault at police stations in Peru.Based on these conversations, we framed our judgment errors around implying that the victims bore some responsibility for their assault.For female participants it was implied that their clothing had invited sexual assault: 'Oh well, you exposed yourself, you should have avoided dressing like that'.For males it was implied that a real man would have been able to prevent the assault from happening: 'But if you are a man, how could anyone do that to you?' In addition, the officers also gave their feedback on how to make the scenario and interviews more realistic from a Peruvian perspective.The interviews are described in the Procedure section.

Response strategy
Interviewers varied in how they responded after making the judgment error.In the apology scenario, the interviewer, after a moment's pause, corrected themselves: 'What I said was not the most appropriate thing to say.I should not have said that, I am sorry.'In the no apology condition the interviewer doubled down on their initial statement, saying: 'You have to be careful with how you dress' for female, and 'Look at you, you are man enough to be able to defend yourself, right?' for male participants.In the control condition, no judgment error had been made, and consequently there was no apology: 'Ok, I got everything, we can proceed'.

Dependent variables Secondary victimization
The 'Instrumento para evaluar victimizaci on judicial en v ıctimas durante la etapa de denuncia' (Instrument to Evaluate Judicial Victimization in Victims during the reporting stage) (Mantilla & Avendaño, 2020) was used to measure secondary victimization.The full scale has 18 items, divided into three factors corresponding to the three types of judicial processes that victims undergo when reporting a crime: attention (procedural justice), information (justifies information), and organization (intrapersonal justice).However, we used only the first factor consisting of 8 items, because this subscale measures how the respondent feels after interacting with the police at the time of filing the complaint.The second and third factors were disregarded, as they are related to the process around the victim interview that was not part of this study (e.g. the waiting time for assistance, or waiting time for information about a filed complaint).
The items in the secondary victimization scale were modified to be consistent with this research.For example, instead of referring to a 'paralegal', items referred to a 'police officer'.
In addition, the items 'The paralegal got angry when you told them that you did not understand what they were asking about,' and 'The paralegal implied that you wanted to take advantage of the situation' were deleted as being irrelevant.
Finally, the response scale was changed from a binary present-versus-absent measure to a 5-point Likert Scale that measured the extent of agreement with an item (1¼ strongly disagree to 5 ¼ strongly agree).We created a total secondary victimization scale using the mean score across all items, with high scores indicating more secondary victimization.

Trust in the interviewer
Mayer and Davis (1999) trust scale was adapted for our experimental setting by replacing references to 'management' to 'police officer'.This scale has 17 items across three subscales: 6 items for ability: 'The police officer was very capable of performing their job,' 5 items for benevolence: 'The police officer was very concerned about my welfare' and 6 items measuring integrity: 'The police officer had a strong sense of justice'.However, our version of the scale had only three ability items, leaving a total of 14 items.Items were removed from the ability subscale because they were irrelevant to our context.The excluded items were 'The police officer has specialized capabilities that can increase our performance,' 'The police officer is known to be successful at the things he/she tries to do,' and 'The police officer has much knowledge about the work that needs to be done'.Items were measured on a 5-point Likert scale (1¼ disagree strongly to 5¼ agree strongly).We created a total trust in the interviewer scale using the mean scores across all items, with higher scores indicating greater trust.

Rapport
To measure rapport, the Rapport Scales for Investigative Interviews, and Interrogations (RS3i) Interviewee Version (Duke et al., 2018) was administered to the participants. 1It has 21 items, measured with a 5-point Likert Scale (1 ¼ strongly disagree to 5 ¼ strongly agree).The questionnaire consisted of 5 scales of rapport (attentiveness, trust/respect, expertise, cultural similarity, connected to flow) and the Commitment to Communication scale.We created a total rapport scale using the mean scores across all items, with high scores indicating higher rapport.

Willingness to provide information
Following Beune et al. (2009), participants were asked the following 3 items to indicate to what extent they will give further information to the interviewer, to what extent the information they will provide is truthful, and to what extent they will tell everything to the interviewer.Responses were measured on a 5-point Likert scale (1 ¼ strongly disagree to 5 ¼ strongly agree).We created a total willingness to provide information scale using the mean score across all items, with high scores indicating that the participant was more willing to provide information.

Procedure
An emotional support protocol was created in case any of the participants experienced discomfort during the interview.This protocol details the indications for the interviewer if any interviewees experience distress due to the questions or our scenario.It was not necessary 1 Note that this scale includes several items that overlap with our measure of trust because they assess, in our opinion, the perceived competence and integrity of the interviewer.We also ran all our analyses using the rapport measure with these items excluded.We include this measure of rapport that omits the items that overlap with trust in our open data.Omitted items were: 'I think the interviewer is generally honest with me.' 'The interviewer did his/her job with skill during the interview.,''The interviewer performed expertly during the interview.,''I think that the interviewer can generally be trusted to keep his/her word,' 'I feel I can trust the interviewer to keep his/her word to me,' 'The interviewer made an effort to do a good job' and 'The interviewer acted like a professional.'No analyses were meaningfully affected by removing these items.
to employ this protocol, and no participants indicated that they were unduly distressed by the experimental procedure.Nonetheless we have also made this protocol available via the project Open Science Framework project [OSF] page, as are all our other materials.
Participants were asked to read the consent form, which explained that the research concerns how people interact in an investigative interview, and that participation is voluntary, and confidentiality will be maintained as far as possible.This information also notified participants that the content of the study addressed violent crime.Participants' right to refuse to answer any question and to withdraw was also stated.Finally, we asked for permission to record the interview for analysis.
After giving consent, participants completed the online demographic questionnaire.They then read the scenario where they were asked to imagine as vividly as possible that they were victims of sexual assault.Specifically, after going to a party with their friends and getting drunk, they started kissing someone, and then that person wanted to have sex.The participants were told that they did not give consent for this, but when they communicated this to their abuser, the latter ignored this and started sexually rubbing against them.After days of considering the situation, they decided to make a police complaint against the offender.
Participants were then randomly divided into one of the three conditions (apology, error, and control).The researcher followed an interview script, and the interview itself took, on average, 10 min.The interview scripts in all three experimental conditions had the same questions and the questions were asked in the same order each time, though we did allow the interviewer to adapt the wording to match the responses given by participants or to respond to unanticipated questions in order not to artificially inhibit the development of rapport through unnaturally structured conversation.The scripts began with a welcome from the police officer, then proceed with routine questions about who the victim was with and what they were doing before and during the attack, and then concluded with the interviewer saying they will notify the participant about any progress in the case.Errors (and the apology) were made during the routine questioning when asking the participant what they had been wearing.In the control condition the question about the clothing was introduced as a routine question for collecting evidence to make sure that the question did not come across as a judgment error (for the exact wording of these messages see the 'Independent Variables section above).Finally, participants were asked to complete the post-hoc online questionnaires to measure secondary victimization, trust in the interviewer, rapport, and willingness to provide information.The experiment was performed online, which meant that the participant as well as the researcher logged in from home.The interviewer was always the same researcher, wearing professional clothing, with a neutral background.The total experiment took approximately 20 min2 (M ¼ 20.2, SD ¼ 22.9), after which participants were thanked for their cooperation and debriefed.The BMS ethics committee: domain Humanities and Social Sciences of the University of Twente approved the present study (Reference Nr. 220068).

Qualitative procedures
To allow us to gain greater insight into our data, right after the experiment we performed semi-structured interviews that lasted 4 min with 12 randomly selected participants (M ¼ 3.4, SD ¼ 0.3).None of the 12 selected interviewees refused to take part in the qualitative part of the study, and the interviewees were equally distributed in gender (i.e.50% male, 50% female).We used these interviews to offer a contextual analysis of the relationships we establish in our quantitative analysis.
We explored the subjective experiences of the interviewees, the interpretations they made of their context, their expectations, feelings, and perceptions to better illuminate the participant experiences underpinning our quantitative results (Patton, 2015).Therefore, we employed an essentialist theoretical thematic analysis of our data (Braun & Clarke, 2006).
Our topic guide is available in our supplementary materials on the project OSF page.First, we asked questions evaluating the roleplay.Next, we posed questions that sought clarification regarding how participants were impacted by the judgment error and apology (where the participant experienced these).For example, in the errorno apology scenario, participants were asked to discuss how this error affected the relationship with the interviewer, whereas in the apology condition the participants were asked whether the apology helped repair the relationship with the interviewer.
The interviews were recorded and transcribed verbatim immediately after being conducted.Transcriptions are in the original language of the participants (Spanish).Illustrative quotes have been translated into English by a bilingual researcher.Data was organized using Atlas Ti 7.0 and Microsoft Excel.Our analyses organized data according to our independent and dependent variables, because our aim was to deductively expand on our understanding of the relationship between these variables rather than inductively generate new theory (Maguire & Delahunt, 2017).To achieve this, we emphasize the explicit meanings participants provided about their experiences within a semantic thematic analysis, rather than concern ourselves with the frequency with which ideas are expressed (Braun et al., 2016;Braun & Clarke, 2006).In practice, our analytical approach meant that we began our analysis with a thematic framework that incorporated our variables of interest as top-level themes: secondary victimization, trust in the interviewer, rapport, willingness to provide information, apology.When participants referred to these variables, we first coded participants' statements of how their experience of errors affected each outcome in their own words, alongside simple descriptive labels.After this step, internally homogenous subcodes per variable of interest were created by collecting individual codes that shared a common narrative regarding how errors impacted upon a specific outcome of interest.
Finally, themes and subthemes were labeled and described to narratively explicate the impact of errors upon each of our outcomes from the perspective of our participants.Data that did not address our research questions were excluded and were not included in the coding process.

Descriptive statistics
The mean scores, standard deviations, and Cronbach's alphas for all dependent variables are presented in Table 1.Table 1 also shows the correlations between all dependent variables.All variables were positively and statistically significantly (p < .01)correlated with each other except for secondary victimization, which, in line with predictions, was statistically significantly (p < .01)negatively correlated to all other variables.This indicates that secondary victimization is associated with reduced trust, rapport, and willingness to provide information.

Hypothesis testing
We used a one-way ANOVA to test for effects of our experimental conditions on all our dependent variables.Because we were interested in comparisons between all experimental groups, we followed up statistically significant main effects with post hoc tests using Tukey's correction for familywise error.The means and standard deviations for all experimental groups across our dependent variables are presented in  F(2, 87) ¼ 6.70, p ¼ .002,g p 2 ¼ .13.Post hoc tests showed that perceived secondary victimization was lower in the control group when compared to both the error group and apology group (ps < .001),but the difference between the error and apology groups was not statistically significant (p ¼ .059).This was similar for willingness to provide information where willingness was highest in the control group, and this difference was statistically significantly different to both the error (p ¼ .002)and apology (p ¼ .032)groups, but the difference between the error and apology groups was not statistically significant (p ¼ .592).
The control group had statistically significantly higher Trust and Rapport than either the error (p trust < .001,p rapport < .001) or the apology group (p trust < .001,p rapport < .001).For these variables, the difference between the error and apology groups were also statistically significant (p trust ¼ .002,p rapport ¼ .022).
Judgment errors therefore increase perceptions of secondary victimization, and lower trust, rapport, and willingness to provide information, regardless of whether or not there is an apology.Apologies do restore some trust and rapport but do not statistically significantly positively affect secondary victimization or willingness to provide information.

Exploratory quantitative analysis
As a robustness check, we also wanted to determine whether our experimental manipulation affected males and females differently.Therefore, we conducted a two-way ANOVA incorporating both experimental condition and Note: N ¼ 90.All scales are scored from 1 to 5, and all correlations were significant at the p < .01level (2-tailed).
a The correlation between trust and rapport remained extremely high even after we removed the items that overlapped with trust, r ¼ .92. gender3 on all dependent variables.The means and standard deviations across experimental condition split by gender are shown in Table 3.The statistically significant effect of experimental condition remained for all dependent variables, and there was no main or interaction effects of gender, with one single exception.For secondary victimization there was an additional main effect of gender whereby females (M ¼ 3.20, SD ¼ 0.97) did perceive a greater secondary victimization than did males (M ¼ 2.97, SD ¼ 0.89), F(2, 87) ¼ 5.83, p ¼ .018.We also tested whether membership of a feminist organization impacted on any of our results, but there were no main or interaction effects of membership across any of our four dependent variables (all Fs < 1.38, all ps > .243).

Qualitative contextualisation
We further explored the impact the judgment error and apology had on secondary victimization, trust in the interviewer, rapport, and willingness to provide information through semistructured interviews.Because of the deductive nature of or analyses, our themes directly map onto our dependent variables and independent variables, and we explore each in turn.
Pointing the finger at the victim Female participants repeatedly referred to three emotions when describing how they felt after an interview in which errors were made: guilt, humiliation, and shame: 'The truth is that this comment about my clothes is something I saw coming, this is how the police blame the victim for what happened … the worst thing is that they really make you doubt, I feel that the mistake is mine' (Gabriela,4 apology condition).This participant explained how the error was, for them, predictable, because it tapped into her own beliefs about the police (that they tend to blame victims) but also that it was so hurtful because it would make her doubt her own version of events.That is, such errors could lead victims to question their own memories of events.
Almost all male participants reported feeling humiliated after the judgment error: 'Sometimes as men we feel that we can't go to the police station to report these types of incidents because they will make fun of us … after a comment like that, you feel even more humiliated, you feel like you don't want to go back' (Juan, error condition).The effect of the error on men was very similar to the effect on women, but while guilt was the predominant description given by women, men emphasised humiliation.How guilt differs from shame is debated, but one reasonable distinction is that guilt implies a self-evaluated moral failing while shame implies a self-evaluated failure to be one's ideal self (Miceli & Castelfranchi, 2018).That is, unlike guilt, shame does not require a moral judgment.Male participants were more likely after a victim blaming judgment error to feel as though they had fallen short of some ideal of masculinity, rather than that they had committed some moral transgression that provoked their victimization.Female participants also often invoked humiliation and shame, implying that they felt they would be judged to have fallen short of some expected standard of competence to avoid their own victimization.However, it was a far more typical female than male response to a victimblaming judgment error to invoke a failure to adhere to societal moral standards of chastity.

I (Don't) trust you
Participants indicated that they did not trust the interviewer after a judgment error.They did not think the interviewer cared about what happened to them, and so would not be motivated to help them: 'I do not trust her at all; I think she focused on getting answers; it was as if she did not care about what I said, maybe she is not even going to file a complaint to her boss to do something' (Pedro, error condition).This response, though unlikely to be exclusive to Peru, could be exacerbated by cultural beliefs about the police.For example, Santos (2020) found that when police officers made machista (expectations of how a man is supposed to behave in Hispanic cultures, Nuñez et al., 2016) comments toward people reporting a crime of sexual violence, the latter lost confidence in the police authorities and so did not want to continue with the complaint process.However, the belief that police will not act on information or bad prior experiences with the police is also a common reason for non-reporting in other countries, including the Netherlands (Ceelen et al., 2019).

Failing to bond
All participants indicated that the judgment error negatively impacted rapport.Participant explanations again fell into two types: after the error, they (a) perceived the interviewer as a person with little empathy, who was cold and cared little about what had happened to them, or (b) because of the seriousness of the crime, the error made participants feel more stressed about the interaction and closed off.
Regardless of the type of explanation offered by participants, the effect was similar.
Participants were less inclined to talk to interviewers who, they felt, did not understand their point of view or seemed to only complete a checklist of formalities rather than care about their story.Similarly, participants felt that the crime was already a difficult topic to talk about, and the judgment error minimized the severity of the crime and made the interview feel an unsafe space to discuss such a sensitive topic.The two forms of response also highlight that rapport is a two-way streetinterviewer behaviors must be appropriate in order to allow optimal conditions for rapport to develop (Alison et al., 2013), but it is the response of the interviewee to the interviewer's behavior that is most crucial in determining whether a cooperative relationship does or does not develop, since it is the interviewee that decides whether or not to disclose.
To talk more or to stop talking at all Interviewees' reactions to the judgment error on their willingness to provide information to the interviewer fell into two distinct types.
One response, mainly used by female participants, was to highlight the importance of continuing to provide the information requested by the police despite the error.The reasoning was that the police can catch the suspect and that no other women will be at risk from the suspect  dition).This behavior could be related to the idea of sisterhood (sororidad in Spanish), which is understood as an implicit obligation among women to confront machista behaviors, and which helps create safe spaces and support networks in situations of violence (Lagarde, 1992).A second dominant response type was to report a much-reduced motivation to continue to cooperate with the interviewer: 'Completely unnecessary, that question about my clothes, after that, I did not want to continue talking.If it had been in real life, I probably would have gotten up and walked out of the police station' (Gloria, error condition).In other words, after the error was made, they did not want to keep talking.This finding is in line with Koster et al.'s (2018) research, who pointed out that a negative experience with law enforcement officers, such as the feeling of being questioned, directly impacts the willingness to share information with the police.
How much an apology makes amends?Oostinga et al. (2018b) showed that apologies can be a successful response strategy after making a judgment error in crisis negotiations and suspect interviews.However, an apology was not sufficient to repair the situation for some participants: 'that comment completely jeopardized the interview.I would recommend that the interviewer be more assertive from the beginning, rather than apologizing at the end' (Gabriela, apology condition).Participants also argued that the effect of the apology depended on how it was said: 'A person does not change their belief system overnight, but I believe that if their apology is from the heart, then it does help repair the damage that was done' (Juan, apology condition).Bonensteffen et al. (2020) similarly showed that an apology must reflect genuine regret to be considered sincere.It is also noteworthy that most men (5 out the 6 male interviewees) indicated that apologies are sufficient to repair the bond between them and the interviewer, as opposed to women, who did not refer to apologies as a reasonable response strategy.This contrasts with the lack of any interaction effects with gender in our quantitative analysis may primarily reflect the main effect we observe of gender with secondary victimization, whereby female participants reported greater secondary victimization than did men.

Discussion
This study complements existing qualitative research by experimentally testing the effect of judgmental comments by police interviewers on (simulated) victims, their relationship with the interviewer, and their intention to cooperate during the interview.Our study suggests that judgment errors have a direct negative effect on trust, rapport, and willingness to provide information and increase the experience of secondary victimization.We also show that apologies have a restorative effect on trust and rapport, but not on secondary victimization or willingness to provide information.
Our study supports qualitative research in this area (Jackson et al., 2017;Patterson, 2011;Webster & Oxburgh, 2022) by showing a direct causal relationship between what is being said to the victim during the interview and their experience of secondary victimization.This shows that police officers must be careful not to convey stereotypical beliefs about sexual assault and sexual assault victims when interviewing these victims.Research shows that it is hard to alter these stereotypical beliefs, even with specialist training (Murphy The Power of Words & Hine, 2019;Sleath & Bull, 2012), which implies that the simplest route may be to carefully select interviewers to ensure they do not hold prejudiced beliefs.
The additional analysis furthermore shows that the overall feeling of secondary victimization is felt more by female than male victims.Our qualitative contextualization provides a possible explanation for this.That is, female victims experience a wide array of emotions, such as guilt, shame, and humiliation, while male victims predominantly refer to humiliation.This hints to the idea that females sense that they bear more responsibility for what happened, as they potentially see their victimhood as a moral transgression (Miceli & Castelfranchi, 2018).Whether these emotions indeed drive this gender-effect in judgment errors on secondary victimization needs further research.
The negative effect of judgment errors on both trust and rapport underlines the findings of previous research in the suspect interviewing domain (Oostinga et al., 2018b).The negative effect of judgment errors on willingness to provide information is, however, different, because Oostinga et al. (2018b) did not find any effect of an error on this variable.This difference can possibly be explained by how offensive the error was perceived to be by the victim.That is, the judgment error in this study may be considered more offensive than the error made in the interviewing context (bear responsibility for being sexually assaulted vs. committing exam fraud).Moreover, the context of this interaction may explain such differences.In a suspect interview, the interviewer is expected to have a somewhat adversarial tone, while in a victim interview, the interviewer is predominantly expected to help and support.A judgment error may therefore be considered more impactful in a victim interview.Relatedly, we found in the qualitative contextualization that the interviewees indicate that they start to question their own memories of events of who had been responsible for the act.This is worrisome, as police need to make sure that the victim shares their memory of events without confabulation and, with that, gather as much reliable information as possible, and research shows high confidence that suggestive questioning can introduce false memories into witness testimony (Scoboria et al., 2017).Yet, we need more research to determine whether it is indeed the offensiveness of the misjudgment and/or conversational context that drives the effect on willingness to provide information and potentially the quality of the information provided.
The reparative power of an apology in terms of trust and rapport confirms earlier research that shows the importance of addressing the other's sense of being worth saying sorry to (Oostinga et al., 2018b;Sharma et al., 2022).However, apologies were not sufficient to restore victims' willingness to provide information or to reduce the level of experienced secondary victimization.Combined, these results show that apologies can repair a relationship, but they are insufficient to remove the harm caused to victims, and this harm potentially reduces victims' cooperative intent.Sexual offences are especially difficult to prosecute, because often the victim is the only witness, and this makes securing victim cooperation during interview especially important (Oxburgh & Ost, 2011).Therefore, it is likely important to avoid making such errors in the first place, as far as possible.This may hint to the notion that interviewers may adhere moe closely to judgment error prevention (where errors should be prevented at all costs) over judgment error management (where errors are seen as inevitable and one can better learn from them) when preparing for a victim interview (Dimitrova et al., 2017;van der Byl & Vredenburg, 2023).Nonetheless, while errors should be prevented, it still seems worthwhile to train for effective error response strategies.This training may maximize the chances of restoring trust and rapport, which would, in turn, increase the chances of restoring cooperation, given the clear association between these variables in our study and others (Alison et al., 2013;Webster & Oxburgh, 2022).

Limitations
Although we provide the first direct evidence for the effects of judgment errors on victim interviewees, there are three limitations that need to be considered.The first limitation focuses on the realism of the experiments.In our study participants were asked to imagine being a victim of sexual violence, after which they were interviewed by a role-played interviewer, and these interviews were performed from home, which may raise the issue of ecological validity.In our study we tried to make the scenario as close to the perception of the participants as possible by making the story fitting to their behavior and age (psychological realism; Evans et al., 2010).Furthermore, we developed an interview script and our judgment error together with practitioners who perform these interviews in the field.Last, our findings directly corroborate earlier qualitative research in this area (Jackson et al., 2017;Patterson, 2011;Webster & Oxburgh, 2022).Yet, we do not claim that our participants directly represent genuine victims of sexual assault, and the interplay between our variables may operate differently in the high-stakes setting of a genuine sexual assault investigation.Future research may be able to test for the effects we identify here by identifying judgment errors within genuine interviews, and considering how this affected interviewee cooperation before and after errors, or in interviews with and without (or with fewer) errors present.
The second limitation focuses on the length of the interactions.The interviews entailed only 12 questions, which simulated a full victim interview, while these interviews usually last longer.We decided for this interview length, as it allowed us to make the different interviews comparable in nature, and so we were sure that any effects we found were due to the presence/absence of the judgment error.In the future we would, however, be interested to see what the effect of time is on the impact of errorsespecially as we made this error relatively early in the interview.We know from other research in the suspect and victim interviewing domain that these stages of the interview are usually crucial in building trust and rapport (Vrij et al., 2014); however, it could very well be that the impact of such an error is milder when made later in the interview, once more trust and rapport has been established (Oostinga et al., 2018a).
The third limitation relates to the representativeness of our sample.Although it is relatively unique to have a non-European or North American sample in this area of research (Henrich et al., 2010;Irvine et al., 2018), research in the police interviewing domain shows that culture and nationality can affect how participants respond to interviewer behavior (Beune et al., 2009;Giebels et al., 2017).For example, our qualitative analysis suggested that participants' responses partly reflected local beliefs about the necessity for female solidarity and challenging machista behavior.Nonetheless, the findings in our study are, grosso modo, in line with research in European samples on non-reporting of sexual violence (Ceelen et al., 2019), as well as the direct effect of errors and repair in the interviewing domain, although more pronounced (Oostinga et al., 2018b).It is therefore important to replicate our study to determine cross-cultural generalizability and to what extent our findings are dependent on the severity of our error.

Conclusion
This is the first study to experimentally test the effect of judgment errors on the victim and extends previous research in the victim interviewing and crisis negotiation domain by introducing consideration of secondary victimization.Our findings corroborate qualitative research in this area that shows the negative effect of judgment errors on trust, rapport, willingness to provide information, and an The Power of Words increase in secondary victimization.Moreover, our study shows that an apology is effective in repairing the damage done to the victim in terms of rapport and trust but is insufficient to restore victims' willingness to provide information or to prevent victims experiencing secondary victimization.Our study shows how important it is that a police interviewer, when interviewing victims of sexual violence, wears not only their crime fighter hat but also their helper hat (Zakimi et al., 2022).So, we need to create more awareness of the impact and power of the words spoken by victim interviewers, as it may directly impact a victims' state of mind, trust perceptions, and experienced relationship, all of which relate to their willingness to share what happened.

Table 2 .
Our experimental conditions statistically significantly affected all dependent variables:

Table 2 .
Means for all dependent variables across experimental conditions.
a Differs significantly from control group, p < .05. b Differs significantly from error group, p < .05.

Table 3 .
Mean dependent variables depending on experimental condition and gender.