Expressing forgiveness after interpersonal mistreatment: Power and status of forgivers influence transgressors' relationship restoration efforts

Summary One adverse consequence of interpersonal mistreatment is that it damages the relationship between the victim and the transgressor. Scholars have promoted forgiveness of such mistreatment as a victim response that can motivate transgressors to work towards relationship restoration. Building on social exchange theory and the social perception literature, we provide an account of when transgressors are less (vs. more) willing to restore their relationship with the victim in response to forgiveness. Specifically, we argue that transgressors perceive forgiveness from a victim who has high (vs. low) power, relative to the transgressor, as insincere, making transgressors less willing to restore the relationship. We further argue that this effect of high (vs. low) victim power is pronounced especially when the victim also has low (vs. high) status. Two experiments and two field studies support these predictions. These findings highlight the relevance of studying how contextual conditions color transgressors' perceptions of victims' behavior to understand relationship restoration after interpersonal mistreatment.


Summary
One adverse consequence of interpersonal mistreatment is that it damages the relationship between the victim and the transgressor. Scholars have promoted forgiveness of such mistreatment as a victim response that can motivate transgressors to work towards relationship restoration. Building on social exchange theory and the social perception literature, we provide an account of when transgressors are less (vs. more) willing to restore their relationship with the victim in response to forgiveness. Specifically, we argue that transgressors perceive forgiveness from a victim who has high (vs. low) power, relative to the transgressor, as insincere, making transgressors less willing to restore the relationship. We further argue that this effect of high (vs. low) victim power is pronounced especially when the victim also has low (vs. high) status. Two experiments and two field studies support these predictions.
These findings highlight the relevance of studying how contextual conditions color transgressors' perceptions of victims' behavior to understand relationship restoration after interpersonal mistreatment.

K E Y W O R D S
forgiveness, hierarchy, interpersonal mistreatment, power, relationship restoration, sincerity, status
The current research is concerned with the effects of expressed forgiveness on subsequent transgressors' relationship restoration efforts. Expressing forgiveness is often promoted as the victim stimulating the transgressor to recommit to the broken rule and to be willing to interact with the victim again; that is, to commit to relationship restoration (Fehr & Gelfand, 2012;Ren & Gray, 2009). However, research that examined transgressor responses to expressed forgiveness has revealed divergent findings. Some studies found that transgressors respond with restorative actions such as complying with victim requests (Kelln & Ellard, 1999), making amends (Hannon, Rusbult, Finkel, & Kamashiro, 2010;Leunissen, De Cremer, & Folmer, 2012), exhibiting prosocial intentions (Mooney, Strelan, & McKee, 2015), and refraining from repeating transgressions (Wallace, Exline, & Baumeister, 2008). Yet other studies found that forgiveness makes transgressors avoid the victim (Adams et al., 2015) and repeat the transgression (McNulty, 2011;McNulty & Russell, 2016). These conflicting findings point to a need to identify moderators that determine when forgiveness is more (vs. less) likely to promote transgressors' restorative efforts. Indeed, forgiveness scholars have suggested that forgiveness does not occur in a social vacuum, and its effectiveness largely depends on organizational contexts such as hierarchy (Bies et al., 2016).
In the current paper, we build on social exchange theory (Blau, 1964;Lovaglia, 1995;Thye, 2000) and the social perception literature (Fragale, Overbeck, & Neale, 2011;Kramer, 1994) to propose that the hierarchy within the victim-transgressor relationship may explain when and why expressed forgiveness promotes transgressors' restorative actions. As a fundamental part of social exchange theory, the norm of reciprocity dictates that recipients of a beneficial act (e.g., receiving forgiveness) should reciprocate with an equally beneficial act (e.g., restorative actions; Gouldner, 1960). However, social exchange theory also poses structural constraints on the reciprocity norm: The hierarchical position of actors dictates how recipients perceive their interaction partner's actions and, thus, their reciprocating actions (Blau, 1964;Lovaglia, 1995;Thye, 2000). Important in this respect, the social perception literature has revealed that low-power actors question the sincerity of their high-power interaction partner's actions (Farrell, 2004;Fiske & Durante, 2014;Kramer, 1994;Zheng, Van Dijke, Leunissen, Giurge, & De Cremer, 2016). Because recipients of a beneficial gesture reciprocate gestures they perceive as insincere less (Belmi & Pfeffer, 2015;Eilam & Suleiman, 2004;Flynn, 2006), forgiveness from a victim who has higher (vs. lower) power may make transgressors less willing to restore the relationship.
However, hierarchies are based on differentiations in power and status. Although correlated, these two variables are conceptually distinct and can vary orthogonally (Anderson & Brown, 2010;. In other words, some people with high power may have low status (e.g., security guards and reimbursement clerks), whereas other people with low power may have high status (e.g., emeritus professors and Olympic athletes; Fragale et al., 2011).
The social perception literature has shown that power and status of actors interact to shape people's perceptions of these actors (Fragale et al., 2011). Building on this, we propose that power and status of the forgiving victim should not only be distinguished but that they should be considered in interaction to understand how they shape transgressor restorative actions following forgiveness. Specifically, we will argue that the transgressor is least likely to exhibit restorative behaviors in response to a forgiveness gesture from a victim with high power and low status. We expect this because the transgressor will perceive forgiveness from such a victim as insincere. Figure 1 visually depicts our proposed model.

| FORGIVER POWER, FORGIVENESS SINCERITY PERCEPTIONS, AND TRANSGRESSOR RELATIONSHIP RESTORATION EFFORTS
Power is commonly defined as asymmetric control over valued resources (Galinsky, Gruenfeld, & Magee, 2003;Keltner, Gruenfeld, & Anderson, 2003). Thus, having power provides the capacity to impose one's will over others (Ng, 1980;Sturm & Antonakis, 2015). The possibility of high-power actors imposing their will motivates low-power actors to make sense of the situation, which produces a hypervigilant mode of information processing (Kramer, 1994;van Dijke, De Cremer, & Mayer, 2010). This, we argue, results in suspicion of being manipulated by one's high-power interaction partner. For example, Kramer (1994) found that first-year (relative to second-year) MBA students perceived their more (vs. less) senior classmates' actions as being driven by malicious intentions. Zheng et al. (2016) showed that low-power (vs. high-power) victims see high-power (vs. low-power) transgressors' apology as being less sincere. Hommelhoff and Richter (2017) found that individuals in nonmanagerial (vs. managerial) positions exhibit more distrust. Other studies have revealed that people see high-power (relative to low-power) actors as being interpersonally cold (Fragale et al., 2011) and dishonest (Fiske & Durante, 2014). In sum, low-power actors may question the sincerity of their high-power interaction partners' actions.
Expressing forgiveness can result from a true internal change towards the transgressor or from self-serving motivations such as attempts to enhance status, assert moral superiority, or manage impressions (Adams et al., 2015;Baumeister, Exline, & Sommer, 1998;Enright & the Human Development Study Group, 1991;Wallace et al., 2008). In the wake of interpersonal mistreatment, low-power transgressors' suspicion of being manipulated may be heightened.
They may question whether felt and expressed forgiveness are actually aligned; that is, they may question the sincerity of forgiveness expressed by their high-power interaction partner (Baumeister et al., 1998;Enright & the Human Development Study Group, 1991). When high-power victims express forgiveness, low-power transgressors may view high-power victims as simply putting on a "show" by expressing forgiveness in order to serve their self-interests. As noted, a core element of social exchange theory is that recipients of a beneficial gesture are less willing to reciprocate the gesture when they perceive it as insincere (Belmi & Pfeffer, 2015;Eilam & Suleiman, 2004;Flynn, 2006). Indeed, studies show that in conflict situations, perceived sincerity of social accounts such as transgressors' explanations for transgressions (Shapiro, 1991) and employers' explanations for layoffs (Skarlicki, Barclay, & Pugh, 2008) influence whether social accounts are effective in mitigating negative recipient reactions. Thus, we reason that low-power (vs. high-power) transgressors perceive forgiveness from forgivers with high-power (vs. low-poer) as less sincere, making transgressors less willing to reciprocate with restorative behaviors.

| FORGIVER STATUS AND TRANSGRESSOR RELATIONSHIP RESTORATION EFFORTS
Status refers to the respect, admiration, and regard an individual has in the eyes of others (Blader & Chen, 2012;Fragale et al., 2011;. It is willingly bestowed by others on those who contribute most to the collective's success and functioning (Kemper, 2006;Van Vugt, Hogan, & Kaiser, 2008). Status serves as a "parameter" of this person's social value in the eyes of others (Chen, Peterson, Phillips, Podolny, & Ridgeway, 2012). High-status (vs. lowstatus) actors act more prosocially (Blader & Chen, 2012) and are expected to continue to display prosocial behaviors that helped them to achieve their high status in the first place (Fragale et al., 2011).
Because status reflects one's orientation towards the collective, we argue that status moderates the relationship between forgiver power and transgressor restorative efforts. Specifically, we argue that the combination of high forgiver power and low forgiver status makes transgressors least likely to respond in restorative ways to forgiveness. Transgressors likely view high-status forgivers (regardless of their power) as being oriented towards the collective, suggesting that their forgiveness gestures are sincere. In contrast, low forgiver status arguably amplifies the effect of high (vs. low) forgiver power on forgiveness sincerity perceptions and subsequent transgressor restorative actions. This is because low-status forgivers do not possess any attributes that signal they will use their power to benefit others, rather than their own self-interest. This argument culminates in our hypotheses: High (vs. low) forgiver power, relative to the transgressor, makes transgressors less willing to restore the relationship. However, this effect is pronounced particularly when the forgiver simultaneously has low (vs. high) status (H1).
The interaction effect of forgiver power and forgiver status on the transgressor's willingness to restore the relationship is mediated by the transgressor's perceptions of forgiveness sincerity (H2).
We tested our hypotheses in two laboratory experiments (Studies 1-2) and two field studies conducted among employees of various organizations (Studies 3-4).

| Participants and design
One hundred and twenty European undergraduate business students participated in exchange for course credit. We randomly assigned them to one of four conditions that resulted from orthogonally manipulating forgiver power (low vs. high) and forgiver status (low vs. high).
Based on criteria explained below (see Section 4.1.2), we included 88 participants in the analyses (48 women; M age = 21.66, SD age = 2.86). There were 22 participants in the low forgiver power/high forgiver status condition, 19 participants in the low power/low status condition, 22 participants in the high power/high status condition, and 25 participants in the high power/low status condition.

| Procedure
We induced participants to transgress against their interaction partner in a trust game (Berg, Dickhaut, & McCabe, 1995), which is often used to study interpersonal transgressions (Desmet & Leunissen, 2014;Leunissen et al., 2012;Zheng et al., 2016). In the trust game, two individuals are randomly assigned to either the trustor (Player 1) or the trustee (Player 2) role. Player 1 starts with a sum of initial endowments (i.e., 10 valuable chips in our study) and decides how many of these chips to send to Player 2. The number of transferred chips is tripled; thus, Player 2 receives three times the number of chips that Player 1 transferred. Player 2 then decides how many chips to return to Player 1. By sending chips to Player 2 (i.e., trusting Player 2), Player 1 can increase his/her own outcomes and those of Player 2. For instance, if Player 1 transfers all 10 chips, Player 2 receives 30 chips and could subsequently divide these chips equally, ensuring that both players end up with 15 chips. However, Player 1 is vulnerable to Player 2's willingness to return a fair number of chips. If Player 2 returns a number of chips that makes Player 1 end up with fewer chips than Player 2, Player 2 commits a transgression by violating Player 1's trust (Leunissen et al., 2012).
Prior research shows that in a trust game, Player 2 is likely to ensure that both players end up with equal outcomes when Player 2 feels fully trusted by Player 1 (i.e., when Player 1 transfers all his/her chips); when Player 2 feels not fully trusted, he/she is likely to reciprocate by ensuring that Player 1 ends up with fewer chips than Player 2 (Pillutla, Malhotra, & Murnighan, 2003). The procedure that we used to induce participants to commit a trust-violating mistreatment (taken from Leunissen et al., 2012;Desmet & Leunissen, 2014; see also Zheng et al., 2016) builds on this prior work by raising participants' uncertainty regarding Player 1's initial endowment. Specifically, Player 2 learns that Player 1's initial endowment could be anything from 10 to 30 chips. Because the exact initial endowment is unknown to participants, most participants assume that Player 1's original endowment is larger than 10 chips. As a result, when receiving 10 chips from Player 1, Player 2 is likely to keep more chips for him/herself than to send back to Player 1.
In our study, all participants were seated in separate cubicles and received all information via a computer. We informed participants that the research assistant they met at the beginning of the study was in one of the other cubicles and would interact with them via the computer network in an exchange exercise. We then explained the trust game and informed participants that they would be Player 2; the research assistant would be Player 1. In reality, all actions from Player 1 were preprogramed.
Before the game started, participants read a message that introduced Player 1 to them. We used this to manipulate the power and status of Player 1. We framed Player 1's power relative to Player 2 in terms of asymmetric control over valuable resources (i.e., chips in this case) without actually changing the game reward structure (taken from Zheng et al., 2016). We operationalized status as the prestige, respect, and esteem the forgiver has in the eyes of others (Anicich, Fast, Halevy, & Galinsky, 2016;Blader & Chen, 2012;Fragale et al., 2011). Specifically, in the low/high forgiver power conditions, participants read: Player 1 depends on you to receive chips. Since you divide the chips that are tripled, Player 1 has little power to influence the final division in this game. /Player 1 divides the initial chips. Thus, Player 1 has a lot of power to influence the final division.
Subsequently, in the low/high forgiver status conditions, participants read: In addition/however, based on the evaluation from many participants who interacted with Player 1 in the game before, they don't have much respect and admiration for her. /They respect and admire her.
Following the definition of interpersonal mistreatment as taking counternormative negative actions or termination of normative positive actions (Cortina & Magley, 2003), we manipulated interpersonal mistreatment as a trust violation. To induce mistreatment, we informed participants that Player 1 had received between 10 and 30 chips and had decided to transfer 10 chips to them. As we expected, most participants (N = 88, 77%) subsequently committed a trust-violating transgression by returning a number of chips that made Player 1 end up with fewer chips than Player 2 (this percentage is similar to Leunissen et al., 2012, in which 74% committed a transgression). Twenty-six participants returned a number of chips that made Player 1 end up with the same number or more chips than Player 2 and thus did not commit mistreatment; Six participants indicated they have participated in a similar trust game before; we thus included 88 participants in our hypotheses tests.
Because we manipulated interpersonal mistreatment as a trust violation, to check whether participants felt they mistreated their interaction partner, we asked them to indicate "To what extent do you think you violated Player 1's trust in the first round?" (1 = not at all; 2 = to a small extent; 3 = to some extent; 4 = to a moderate extent; 5 = to a considerable extent, 6 = to a great extent; and 7 = completely).
After the first round of the trust game, we presented participants a chart with the outcome for both players. This chart indicated that Player 1 had 10 chips and sent out all. Player 1 thus ended up with fewer chips than Player 2 for 88 participants. These participants then received an email message from Player 1 that expressed forgiveness (adopted from Leunissen et al., 2012; see also Wallace et al., 2008): "Hey! I have fewer chips than you! That is too bad. But I will give you the benefit of the doubt for now. I will forgive you but please be cooperative in the future." After having received the forgiveness message from Player 1, a second round of the trust game commenced. We informed participants that Player 1's endowment was in this round again between 10 and 30 chips and that Player 1 sent them 10 chips. Participants then decided how many chips to return in round 2.

| Measures
After participants read the power and status manipulation instructions, they rated Player 1's power with two items from previous studies (Blader & Chen, 2012;Zheng et al., 2016): "In this game, Player 1 has a lot of power over me" and "In this game, Player 1 has a big influence on the outcomes of the game" (1 = strongly disagree; 7 = strongly agree; M = 5.07, SD = 1.51; α = .86). Participants rated Player 1's status with two items from Blader and Chen (2012): "In this game, Player 1 is respected by other participants" and "In this game, Player 1 is held in high regard" (M = 4.22, SD = 1.75; α = .90).
A core element of relationship restoration is the transgressor's recommitment to the norm that was broken (i.e., trust in this case; Ren & Gray, 2009). In light of this, we operationalized transgressor relationship restoration as the increase in the number of chips returned to Player 1 in round 2, relative to round 1 (M = 2.94, SD = 3.65). This increase indicates how much transgressors want to make up for their violation in round 1 (see Wallace et al., 2008 and Desmet & Leunissen, 2010 for a similar approach). Table S1 shows the correlations between the study variables.

| Discussion of Study 1 and introduction to Study 2
In showing that forgiveness from a victim who has high power and simultaneously low status is relatively unlikely to stimulate the transgressor to restore the relationship, the results of Study 1 support H1.
However, a potential limitation of Study 1 was that we tested H1 in the stylized trust game context. In this context, it may not have been clear how previous players developed the "respect and admiration" they purportedly communicated to the participant as part of our status manipulation. Furthermore, we had to operationalize interpersonal mistreatment narrowly as a trust violation. In Study 2, we therefore used a more realistic workplace setting, that is, an in-basket task 1 .
Such a task delivers findings with high internal validity but also ecological validity for workplace experiences (Treviño, 1992). This allowed operationalizing status as respect and admiration from colleagues and interpersonal mistreatment more broadly than a trust violation. Furthermore, in Study 2, we introduced diversity in power operationalizations using a structural power manipulation in which we emphasized asymmetrical outcome dependence: The forgiver was either the transgressor's leader or his/her subordinate who held a formal position that gave him/her a great deal of (vs. very little) control over valued resources in the company. Finally, we tested the process that purportedly drives the effect of forgiver power on transgressor F I G U R E 2 The interactive effect of forgiver power and forgiver status on relationship restoration efforts in Study 1 1 The in-basket task is a workplace simulation that is often used in personnel selection.
Participants are presented with various materials such as memos and phone/email messages and make decisions based on the available information. Experimental manipulations can be embedded in the materials (Treviño, 1992). relationship restoration efforts, as a function of forgiver status; that is, forgiveness sincerity perceptions.

| Participants and design
In total, 117 European undergraduate business students participated in exchange for course credit (64 women; M age = 20.30, SD age = 3.37).
We randomly assigned them to one of four conditions that resulted from orthogonally manipulating forgiver power (low vs. high) and forgiver status (low vs. high).

| Procedure
We seated each participant in a separate soundproof cubicle. All instructions were communicated via a computer. Participants learned that they worked in an organization called Duron Paints, a multinational manufacturer of paint products with approximately 3,000 employees. They would either be a leader or a subordinate in the organization, ostensibly based on their responses to items that measured leadership skills. This was to ensure that participants believed that their role in the organization was appointed in a legitimate manner (see Galinsky et al., 2003;Hoogervorst, De Cremer, & Van Dijke, 2013). In reality, we assigned participants randomly to the high or low forgiver power condition. We manipulated status as the prestige, respect, and esteem the victim had in the eyes of colleagues (Blader & Chen, 2012 In addition (however), Andrew is not very respected or admired (highly respected and admired) by other members of the company. As a result, Andrew does not have (has) influence over others, because these individuals do not value (value) Andrew's opinion.
After reading the role descriptions, participants responded to manipulation checks. Consistent with Study 1 and following the definition of interpersonal mistreatment (Cortina & Magley, 2003), we manipulated mistreatment as taking counternormative actionsparticipants read a workplace mistreatment scenario in which they presented Andrew's contribution to a project as their own contribution during a presentation in a meeting. Participants then received an email, ostensibly from Andrew. We adapted the forgiveness message from previous studies (Adams et al., 2015;Wallace et al., 2008), such that it would be suitable in the workplace context. Specifically, the email indicated: Hi (participant's name), I am writing this email to tell you not to worry about what just happened. I forgive you and I hope we are good now.

| Measures
Except when indicated otherwise, participants responded on 7-point scales (1 = not at all; 7 = completely). Forgiver power was rated with "To what extent do you feel Andrew is in charge in the company?" (Galinsky et al., 2003). Forgiver status was rated with "To what extent do you feel Andrew is respected in the company?" (Blader & Chen, 2012).
We measured forgiveness sincerity perceptions with a 4-item scale.
Because no existing scale measures forgiveness sincerity perceptions, we developed this scale based on items from previous studies (Mooney et al., 2015;Strelan, McKee, Calic, Cook, & Shaw, 2013). We specified items to the current situation. The items are "Andrew's expression of forgiveness reveals his true forgiving self," "Andrew shows consistency between his forgiving intention and the expression of forgiveness," "Andrew's expression of forgiveness is guided by a genuine forgiving intention," and "I perceive his forgiveness as sincere." We averaged these items into a reliable index (M = 4.70, SD = 1.15, α = .81).
Consistent with Study 1, we operationalized relationship restoration as the extent to which the transgressor recommits to the broken rule. Specifically, we adapted McNulty and Russell's (2016) 1-item measure of motivation to refrain from transgressions: "After seeing his forgiveness, I will try very hard not to offend him again in the future" (1 = strongly disagree; 7 = strongly agree; M = 5.62, SD = 1.57). Table S2 shows the correlations between the study variables.

| Manipulation checks
To check whether participants felt they mistreated, we asked:
In support of H1, simple effects analyses revealed that for lowstatus forgivers, having high power significantly reduced transgres- We then tested H2. This hypothesis implies, first, that forgiveness will be perceived as relatively insincere when it is communicated by a high-power (rather than low-power) forgiver, who is at the same time low (rather than high) in status. To test this, we conducted a Forgiver

| Discussion of Study 2 and introduction to Study 3
The results of Study 2 support H1 and H2. When transgressors received a forgiveness message from a high-power victim who also F I G U R E 3 The interactive effect of forgiver power and forgiver status on (a) relationship restoration efforts and (b) forgiveness sincerity perceptions in Study 2 had low status, they were relatively unlikely to restore the relationship. This results because they perceived forgiveness from a highpower forgiver who simultaneously had low status as being less sincere. A limitation of Study 2 was that it measured transgressors' intention to restore the relationship. In addition, Study 2 and Study 1 were both conducted in a laboratory context. In Study 3, we therefore tested our hypotheses on organization members' actual behaviors in existing work relationships. 6 | STUDY 3 6.1 | Method

| Participants
We recruited participants via a professional Dutch research agency, Flycatcher. The Flycatcher panel has the ISO-26362 certification for access panels (i.e., it meets the qualitative ISO requirements for social scientific research, market research, or opinion polls) and consists of approximately 16,000 Dutch citizens. Prior research suggests that this and similar research panels (e.g., study response in the United States) are reliable methods for data collection (Hoogervorst et al., 2013). The agency contacted 350 employees who worked for at least 20 hr each week and stated that the inclusion criteria include the recollection of a specific workplace incident where they transgressed against a fellow colleague and this colleague expressed forgiveness to them. For their participation, they received credit points that allowed them to choose certain small gifts (e.g., movie tickets).
One hundred and twenty-two employees (52 females; M age = 39.36 years, SD = 11.61) indicated they had such an experience and thus completed our online questionnaire. Two independent coders evaluated the recollections in terms of whether they described an incident specified in the instructions or not and agreed that all participants followed the instructions. As to highest completed education, 23% indicated having secondary education (high school); 35% had subsequent vocational education; 24% had a bachelor's degree, and 18% had a master's degree. Of all the participants, 66% worked for more than 5 years with their current organization. In terms of hierarchical position, 53% were line managers/supervisors and 47% had a nonmanagement function.

| Procedure
We used a critical incident technique to elicit salient experiences of having enacted interpersonal mistreatment (Aquino, Tripp, & Bies, 2006;Hershcovis et al., 2018;Karremans & Smith, 2010;Tepper & Henle, 2011). Specifically, participants read: Please recall a specific incident that happened in the last six months where you did something that offended, harmed or hurt somebody in the company, and after the transgression, this person forgave you. By forgiveness, we mean this person either offered an explicit verbal statement or exhibited behaviors indicating that he/she does not have any negative emotions towards you and he/she will not cause you any harm.

| Measures
To check whether participants indeed felt they committed interpersonal mistreatment, they answered one question after recalling the mistreatment: "To what extent do you think she/he is victimized by you?" (Aquino et al., 2006; 1 = not at all; 2 = to some extent; 3 = to a moderate extent; 4 = to a considerable extent; 5 = completely). A t test showed that mean perceptions (M = 2.50, SD = 1.18) were significantly higher than 2, t(121) = 4.38, p < .001, 95% CI [.26, .68], d = 0.80, indicating that, on average, participants felt they victimized the forgiver more than to some extent.
In addition, two independent coders coded the extent to which the victim felt mistreated (1 = not at all to 5 = completely). A t test showed that means (M = 2.41, SD = 0.80) were significantly higher than 2 (to some extent), t(121) = 5.64, p < .001, 95% CI [.27, .56], d = 1.03, indicating that, on average, victims felt they were at least to some extent mistreated. The coders also coded types of mistreatment as incivility, aggression, and bullying based on the classification by Yang, Caughlin, Gazica, Truxillo, and Spector (2014); 65.5% were incivility, 29.4% were aggression, and 5% were bullying.
After participants recalled the incident, we measured forgiver power with a measure developed by Aquino et al. (2006). Participants indicated whether the person whom they had transgressed against was a "subordinate," a "supervisor," a "manager," an "administrator," a "peer," or "other." Consistent with Study 2, we used structural power to capture asymmetrical outcome dependence. Participants who reported their victim's position as "other" were asked to specify their relationship with this person. Two independent coders classified participants who indicated "other" (N = 8) into one of the categories based on their specified relationship with the forgiver. There was no disagreement in terms of classification. Consistent with Aquino et al. (2006), we combined supervisor, manager, and administrator into a high-power forgiver category (N = 32); peer represents the equal power forgiver category (N = 58); subordinate represents the lowpower forgiver category (N = 32).
We measured all other items on 5-point scales (1 = strongly disagree; 5 = strongly agree). We measured forgiver status with the 8-item organizational status scale (Rogers & Ashforth, 2014;van Quaquebeke & Eckloff, 2010). Sample items are: "Other employees respect him/her at the workplace" and "Other employees hold him/her in high regard" (M = 3.28, SD = 0.71, α = .84). We measured relationship restoration using six items taken from Tabak, McCullough, Luna, Bono, and Berry's (2012) transgression reconciliation checklist. We introduced the items as follows: "After he/she expressed forgiveness to you, to what extent do the following T A B L E 1 Results of hierarchical regression analysis in Studies 3 (upper half) and 4 (lower half)

Criterion variable
Step 1 Step 2 Step 1 Step 2 Table S3 shows the correlations between the study variables.

| Hypotheses testing
We tested H1 with ordinary least squares (OLS) regression in which transgressors' relationship restoration was the criterion variable. We included forgiver power (1 = high-power forgiver; 0 = equal power forgiver, −1 = low-power forgiver), forgiver status, and their interaction (based on a mean centered version of forgiver status) as predictor variables. Table 1 shows the results. Most importantly, in step 2, the Forgiver Power × Forgiver Status interaction significantly predicted relationship restoration (Figure 4).
We used Hayes' (2013) PROCESS macro (Model 8; 5,000 bootstrap samples) to test the full model (see Table S5

| Discussion of Study 3 and introduction to Study 4
The results of Study 3 provide further support for our hypotheses.
Specifically, by operationalizing power as position power and status as organizational status, we again found that forgiveness from a victim who is high in power but low in status is perceived as less sincere, therefore facilitating less transgressor restoration effort.

| Results
We tested H1 with OLS regression analysis. The Forgiver Power × Forgiver Status interaction predicted relationship restoration efforts. We proceeded to test H2 using OLS regression analyses. Table 1 presents the results. Most importantly, in step 2, the Forgiver Power

| GENERAL DISCUSSION
Across four studies, we found that following interpersonal mistreatment, forgiveness promotes less relationship restoration when the forgiver has high (vs. low) power. Moreover, we found this effect of forgiver power on relationship restoration in particular among forgivers who have low (rather than high) status. We obtained evidence for this effect in two laboratory experiments (Studies 1-2) and among employees in organizations (Studies 3-4). Studies 2-4 also showed that the effect results because transgressors perceive a forgiveness gesture of victims with high power and low status as relatively insincere. Importantly, we found similar results in the United States (Study 4) and the Netherlands (Studies 1-3), suggesting that the effect we identified generalizes across cultures that are at least somewhat different.

| Theoretical implications
Research has long focused on victim characteristics to explain the emergence of interpersonal mistreatment, overlooking the role of the transgressor and even running the risk of "blaming the victim" for a transgression (Cortina, Rabelo, & Holland, 2018;Dalal & Sheng, 2018 (Belmi & Pfeffer, 2015;Eilam & Suleiman, 2004;Flynn, 2006). To date, studies on the role of sincerity in social exchange between victims and transgressors have focused on victims' perceptions of apology sincerity (e.g., Zheng et al., 2016) and ignored that transgressors also need to perceive victims' gestures as sincere to decide upon relationship restoration.
Second, we provide first empirical evidence of the interactive effects of actors' power and status on the other party's perceptions and subsequent behaviors towards these actors in actual relationships. Power and status are theorized as distinct constructs that underlie social exchange processes (Blau, 1964;Lovaglia, 1995;Thye, 2000). Given that they are two coexisting hierarchical dimensions and vary orthogonally in any exchange relationships, they should interact in shaping how others perceive and behave in these exchange rela-

| Practical implications
Our finding that forgivers' power and status "color" transgressors' perceptions of forgiveness sincerity, which influence their relationship restoration efforts, implies that sincerity perceptions are to some extent independent from a person's actual sincerity. In fact, previous studies show that high-power actors act more out of their sincere intentions (Galinsky, Magee, Gruenfeld, Whitson, & Liljenquist, 2008;Hirsch, Galinsky, & Zhong, 2011;Kifer, Heller, Perunovic, & Galinsky, 2013). Thus, although high-power victims' forgiveness may be driven by sincere intentions, ironically, low-power transgressors may interpret their forgiveness as being insincere. This disconnect has implications for transgressors and victims. Transgressors should be aware that their perceptions of forgiveness sincerity may be biased. Indeed, studies have shown that perceivers can minimize bias effects in impression formation when they are made aware of such effects (Macrae & Bodenhausen, 2000).
Although victims may forgive a transgressor out of the best intentions, they should be aware that their position within the organization (i.e., a position of high power and low status) may cause their forgiveness to be perceived as insincere, thus failing to stimulate relationship

| Limitations and future directions
Despite a number of contributions to the literature, our research has limitations that future work should address. First, future research should consider the effectiveness of victim behaviors other than forgiveness, such as reconciliation (i.e., extending acts of goodwill), peaceful coexistence (i.e., bearing hostility but resuming the work relationship), and détente (i.e., reducing tension through meetings and agreements prescribing future behaviors) in resolving workplace conflicts (Bies et al., 2016), and the mediating role of sincerity perceptions in explaining the effectiveness of such behaviors. Bies et al. (2016) suggested that forgiveness is viewed as a virtue among friends and family but as less appropriate in organizations. Indeed, our and others' research revealed that forgiveness in organizations sometimes is perceived negatively (Adams et al., 2015). For high-power/low-status victims, reconciliation, peaceful coexistence, or détente may be perceived as being more professional and thus be more effective than forgiveness in restoring relationships.
Second, our findings and previous studies (Zheng et al., 2016) show that organization members perceive their interaction partner as less sincere when this partner has high (vs. low) power. This is because low-power people suspect that their behaviors may be driven by instrumental motives such as impression management. Interestingly, other research shows that in the specific context of receiving favors, people perceive low-power (vs. high-power) actors as insincere. This results because people perceive favors from low-power actors as driven by instrumental motives such as unsolicited influence attempts (Inesi, Gruenfeld, & Galinsky, 2012). Both streams of research are consistent with recent work showing that high-and low-power people can exhibit similar cognition and behaviors (which differ from those in equal power interaction partners) when instrumental goals are salient (Schaerer, Du Plessis, Yap, & Thau, 2018). In some social exchange contexts (e.g., favor exchange), high-power people see low-power interaction partners as insincere, whereas in other social exchange contexts (e.g., conflict resolution), low-power people see high-power interaction partners as insincere. To more fully understand how power shapes forgiveness sincerity perceptions, research should identify contextual factors that moderate such perceptions in unequal power relationships.
Finally, we focused on the transgressor's perspective and showed that the power and status of a victim interactively influence the sincerity of this victim's forgiveness in the eyes of the transgressor even in experiments in which transgressors always received the same forgiveness message (Studies 1-2). However, previous studies have revealed that high-power/low-status victims are more likely to actually behave in demeaning ways in conflict situations (Anicich et al., 2016;Fincham, Hall, & Beach, 2006;Kim, Smith, & Brigham, 1998).
To achieve further integration of theory on person perception and theory on the effects of having status and power, future research should use dyadic designs to simultaneously examine the victim's actual forgiveness content and the transgressors' perceptions as a function of the victim's power and status.