Introduction

One painful workplace reality is that organizations sometimes break promises to employees. Broken promises often constitute a moral injustice that inflicts lasting harm upon persons and relationships (Kickul 2001; O’Donohue and Nelson 2009). The perception that an organization has broken their promise is known as a psychological contract breach (PCB). Psychological contract breach is associated with a host of negative employee emotions, attitudes, and behaviors (Conway and Briner 2005). In the aftermath of promise-breaking, the ethical and prudent organization may seek forgiveness by apologizing, giving restitution, or performing both actions (Boyd 2011; Tucker et al. 2006). Using such reparative actions to obtain forgiveness, the organization may address injustice, forestall PCB sequela, and restore damaged relationships. They may also accomplish moral repair, which is the restoration of emotional bonds and shared ethical expectations that bind communities together (Walker 2006).

Knowledge of how apology and restitution work to bring about forgiveness, therefore, has high ethical, managerial, and communal value and is sorely needed because of the ubiquity of moral failure (Kurzynski 1998). Despite this, business ethics literature investigating the repair and recovery process is minimal (Grover et al. 2017; Schminke et al. 2014). Similarly, relationship repair between organizations and individuals is “surprisingly understudied” (Dirks et al. 2009, p. 82). Especially scant is research about post-breach emotional processes and effective employer responses to psychological contract breach (De Ruiter et al. 2016; Tomprou et al. 2015). The dearth of research on workplace forgiveness (Bright and Exline 2012), factors involved in employee choices to forgive rather than avenge (Bradfield and Aquino 1999; Palanski 2012), and whether, when and how apology and restitution are effective in repairing relationships (Byrne et al. 2014; Tomlinson et al. 2004; Zheng et al. 2015) has also been noted.

Much of what we do know about how forgiveness is elicited using apology or restitution comes from the interpersonal relationship and clinical literature. One dominant theory of interpersonal forgiveness is the Emotion Replacement Model (ERM). ERM posits that apologies and reparation elicit forgiveness by increasing positive emotions that displace negative emotions in the offended party (Worthington 2006). Primarily focusing on interpersonal relationships, the model has never been used to understand how apology or reparation elicit forgiveness of organizations for their broken promises, or indeed for any organizational wrongdoing (for a notable exception, see Bisel and Messersmith 2012). Using ERM to frame our approach to recovery and repair after psychological contract breach suggests the following question: After broken organizational promises, do apology and restitution lead to negative-to-positive emotion replacement and thereby elicit forgiveness?

The current research sought to address this question. We tested the proposals that in the face of employee perceptions of a broken transactional promise, organizational apology and restitution will lead to forgiveness by displacing offender-oriented negative emotions with positive ones. In three experiments, participants envisioned a broken promise involving significant monetary aid from their school, then received an organizational response consisting of apology and restitution. This research adds to what we know about how transgressing organizations may recover and obtain forgiveness after ethical misdeeds, particularly for broken promises to their members. More broadly, it adds to what we know about repairing damaged emotional bonds within an organizational community. Concerning forgiveness theory, it adds to what we know about the specific emotional processes by which apology and restitution elicit forgiveness.

To develop our hypotheses, we review the literature on psychological contract breach, forgiveness and the emotion replacement model, apology, and restitution.

Psychological Contract Breach (PCB)

A psychological contract is an employee’s subjective and promissory expectation based on their belief that the employer made implicit, explicit, and even written promises (Conway and Briner 2005). Typically, an employee perceives a promise of organizational rewards in exchange for employee performance and behaviors (Rousseau 2011). Psychological contracts have both transactional and relational forms (Montes and Irving 2008). Transactional contracts occur in relatively short-term relationships and concern highly specified, publicly observable, and monetizable performance-reward contingencies (e.g., “A fair day’s work for a fair day’s pay,” Rousseau and Wade-Benzoni 1994, p. 466). Relational contracts occur in long-term relationships, and concern affect-laden, less-tangible, and more open-ended exchanges (e.g., “If I work hard the organization is obligated to continue employing me”) (Rousseau 1995). Trust is the basis for both transactional and relational psychological contracts; even specific expectations of monetary exchange are part of a broader trusting relationship with the organization (Rousseau 1989).

Psychological contract breach occurs when employer promises are perceived as broken. It is likely that a majority (Robinson and Rousseau 1994), and possibly a large majority (Conway and Briner 2002), of workers, have experienced PCB (Coyle-Shapiro and Kessler 2003). PCB may lead to an employee’s sense of violation, (i.e., feelings of anger and betrayal, Morrison and Robinson 1997), depression, anxiety (Conway and Briner 2002), emotional exhaustion (Chih et al. 2016), erosion of trust (Robinson 1996), perceptions of injustice, and a sense of relational trauma and powerlessness (Rousseau 1989, 1995). PCB is correlated with a number of negative workplace attitudes (i.e., lower employee job satisfaction and organizational commitment, and higher turnover intentions), and negative workplace behaviors (i.e., increased counterproductive work behaviors, turnover, complaints, workplace deviance, displaced aggression, rumor spreading, and decreased organizational citizenship behaviors and job performance) (see reviews in Conway and Briner 2005; Zhao et al. 2007; see also Bordia et al. 2014; Jensen et al. 2010; Restubog et al. 2007; Wei and Si 2013).

If an organization breaks a promise and thereby initiates a damaging psychological contract breach in employees, it may wish to forestall the harms associated with PCB, repair relationships, and restore communal norms of morality by seeking forgiveness.

Forgiveness and Emotion Replacement

After a transgression harms the relationship between two parties, forgiveness is the offended party’s voluntary internal decision to forego antipathy toward the offender (Enright and Fitzgibbons 2015a). Definitions of forgiveness highlight pro-social changes in motivation (McCullough et al. 1998), cognition (Flanigan 1998), and emotion (Worthington and Scherer 2004). Researchers have been careful to specify that forgiveness is not reconciliation (restoration of a relationship), condoning (failing to see an act as transgression), excusing (failure to hold transgressor responsible), justifying (defending as right), pardoning (release from the penalty), or forgetting (removal of transgression from conscious awareness) (Finch 2006; Worthington 2005).

Forgiveness in workplace contexts is an intentional non-retaliatory and possibly positive response to an offense, distinct from relationship and trust repair, involving parties at the individual, organizational, or collective levels (Bies et al. 2016). In the context of PCB, forgiveness is an intentional non-retaliatory response to organizational promise-breaking, distinct from reconciliation, trust repair, forgetting, condoning, or excusing. Conversely, unforgiveness is the sense of violation that develops after PCB (Morrison and Robinson 1997).

After a broken organizational promise and amidst feelings of anger and betrayal, the employee’s “voluntary internal decision to forego antipathy toward the offender” may be quite challenging. How do people become willing and able to forgive others their misdeeds?

Worthington (2006) and colleagues (Worthington and Wade 1999) set forth a stress and coping theory of forgiveness in which emotion replacement facilitates forgiveness. That is, supplanting negative emotions of unforgiveness (i.e., anger, resentment, hatred) with positive emotions (i.e., empathy, compassion, sympathy, and altruistic love) toward the offender facilitates forgiveness. This is the Emotion Replacement Model (ERM). Drawing on neuropsychological, emotion, and motivation literature, ERM posits that when experienced simultaneously, positive and negative emotions create emotional dissonance and displace one another. Similarly, McCullough et al. (1997) argued that affective empathy, conceptualized as the positive other-oriented emotions of sympathy, tenderness, and compassion for the offender, facilitates forgiveness by shifting the victim’s focus from self to offender needs. Consistent with ERM, Fehr et al. (2010) meta-analyzed 175 studies of correlates to interpersonal forgiveness and found that forgiveness was strongly positively related to victim’s state empathy (a positive offender-oriented emotion), and moderately negatively related to state anger (a negative offender-oriented emotion).

If an organization broke a promise, and now faces PCB and members’ negative unforgiving emotions, two potential ways of accomplishing negative-to-positive emotion replacement and eliciting forgiveness (and thereby laying the groundwork for relationship and moral repair) are to apologize and give restitution.

Apology and Restitution

Apology and restitution are reparative actions embedded within cultural rituals aimed at reestablishing disrupted social equilibria (Ren and Gray 2009) or mending poor impressions (Gold and Davis 2005) caused by an offense. After apology and restitution, the offended party may accept these actions by extending forgiveness (Tavuchis 1991). At a minimum, an apology is the acknowledgment of responsibility for an offense and the expression of remorse (Lewis et al. 2015). Restitution is “reparation made by giving an equivalent or compensation for loss, damage, or injury caused; … restoration to the former or original state or position” (“restitution,” n.d.); restitution has also been called corrective action (Benoit 2014). Restitution is a form of reparation: compensatory action seeking to mend damage caused by a past offense (Lazare 2005). After PCB, a minimal apology would consist of a speech act by an organizational representative acknowledging responsibility and expressing remorse for the broken promise; restitution would be equivalent to fulfilling the broken promise.

Although apologies may (and perhaps should) contain the offer of reparations (Bisel and Messersmith 2012; Boyd 2011), they are conceptually distinct (Lazare 2005). Giving an apology can be independent of giving reparation, and vice-versa. In this paper, an apology is a speech act taking responsibility and expressing remorse, while restitution is restorative reparation. To accentuate this distinction, we sometimes refer to an apology as “apology-alone,” “a simple apology,” or a “cheap” (i.e., without restitution, see Bottom et al. 2002) apology. Similarly, we sometimes refer to restitution as “restitution-alone.” When combined, we refer to “apology-with-restitution” or “costly” apology.

Offering apology or giving restitution increases the likelihood of being forgiven in interpersonal (Blatt and Wertheim 2015) and organizational contexts (Walfisch et al. 2013), although the comparative effectiveness of these tactics is still inconclusive (Carlisle et al. 2012). Apology, versus no-apology, is linked to forgiveness (Exline et al. 2007; Fehr et al. 2010) and improved perceptions of the transgressor (Fukuno and Ohbuchi 1998). It is also linked to reduced negative repercussions (Darby and Schlenker 1982) and aggression toward the transgressor (Ohbuchi et al. 1989). Other work suggests that restitution is also predictive of forgiveness. Post-offense compensatory responses such as making amends (Hannon et al. 2010), self-punishment (Bottom et al. 2002), and offense-removal (Zechmeister et al. 2004) are at least partially linked to forgiveness or trust repair. We therefore predicted apology and restitution effects:

Hypothesis 1A (apology effect)

Apology-alone versus no-apology from the offender increases the likelihood of forgiveness by the victim.

Hypothesis 1B (restitution effect)

Restitution-alone versus no-restitution from the offender increases the likelihood of forgiveness by the victim.

Theorists have proposed that the nature of the transgression moderates the strength of the apology or restitution effect. Specifically, restitution may be more effective after a transactional breach as compared with after a relational breach. In the final stage of Morrison and Robinson’s (1997) Model of Breach Development, a sense of violation (feelings of anger and betrayal) develops after PCB and is moderated by judgments about breach outcomes and process. They proposed that outcome concerns should be of great importance in transactional psychological contracts. Therefore, because giving restitution addresses outcome concerns and a simple apology does not, giving restitution after a transactional breach may be more effective than a simple apology for eliciting forgiveness.

Apology combined with restitution may be an especially effective remedial response, that is, “costly” apologies (i.e., accompanied by restitution) may be more effective than “cheap” apologies (i.e., no-restitution). Findings that offers of repair typically accompany effective apologies (Lewicki and Polin 2012; Scher and Darley 1997) suggest this. In experimental research, receiving a costly versus cheap apology boosted forgiveness (under high arousal conditions) and a cheap apology was actually worse than no-apology (Zechmeister et al. 2004). Similarly, many public intergroup apologies have failed because appropriate reparations were not also offered (Wohl et al. 2011). Apology coupled with reparations should therefore be more effective than apology-alone; in this article, we refer to this as the “boost” effect:

Hypothesis 1C (boost effect)

Apology-with-restitution versus apology-alone from the offender increases the likelihood of forgiveness by the victim.

Using the emotion replacement framework, we can now understand the role of changes in offender-oriented emotion when a promise-breaking organization gives apology and restitution to an employee experiencing PCB.

Emotion Replacement After Apology and Restitution

Apology and restitution, alone or combined, may elicit forgiveness in part by facilitating emotion replacement. Since apologizing and giving restitution are recognized as aversive and humbling experiences, they evoke positive emotions (e.g., empathy, compassion) toward the offender, displace negative emotions (e.g., anger, resentment), and motivate a voluntary relinquishing of the right to retribution (i.e., forgiveness). The evidence is generally consistent with this mediation. The apology effect on eliciting forgiveness was mediated by anger reduction (Kirchhoff et al. 2012), feeling valued and likeable (De Cremer and Schouten 2008), positive emotions (Takaku 2001), and empathy (Lewis et al. 2015; McCullough et al. 1997, 1998; Zechmeister and Romero 2002, but see Carlisle et al. 2012 for a null finding using a behavioral forgiveness measure). Empathy mediated the restitution effect on eliciting forgiveness in one study (Carlisle et al. 2012) but not another (Zechmeister et al. 2004). Restitution has been studied less frequently than apology and researchers have called for investigations of their comparative effects and mechanisms (De Cremer et al. 2010). We predicted:

Hypothesis 2 (mediation)

The replacement of offender-oriented negative emotions with positive ones mediates apology, restitution, and boost effects.

Hypothesis 3 (displacement)

Increases in positive emotions toward the offender correlate with reductions in negative ones.

We tested these hypotheses over three pre-post experiments where participants read narratives in which a department broke a promise to reduce tuition costs.

Overview of Experiments

In Experiment 1, we tested the effect of a simple apology, alone and unaccompanied by restitution, on forgiveness (Hypothesis 1A) by using a one-way apology (present/absent) design and measuring the likelihood of forgiveness before and after the departmental response. We compared changes in the likelihood of forgiveness between conditions. At the same time, we sought to observe two critical claims of ERM, mediation and displacement, by measuring changes in emotions toward the offending department. First, we gaged the mediation of forgiveness via increases in positive and decreases in negative emotions (Hypothesis 2). Second, we assessed the correlational pattern of change posited by ERM, namely, displacement of negative with positive emotions (Hypothesis 3).

In Experiment 2, we used the same methods but added a restitution variable in a two-way apology (present/absent) × restitution (present/absent) design. By doing so, we tested the effects of apology-alone (Hypothesis 1A), restitution-alone (Hypothesis 1B), and the relative boost in forgiveness afforded by including restitution with the apology (Hypothesis 1C). We again sought to gage mediation via changes in positive and negative emotions (Hypothesis 2) and the pattern of negative-to-positive emotion displacement (Hypothesis 3).

In Experiment 3, we attempted to manipulate the proposed mediating variable of emotion change more directly. If we could boost the effect of apology-alone, this time by manipulating emotion changes instead of by giving restitution, we could more directly test the emotion replacement claim posited by ERM. To accomplish this, we used the strategy of third-party blame. One of several possible image repair discourse strategies used in organizational crisis communication, third-party blame (or “shifting blame”), attempts to repair the organization’s image and regain trust by arguing that someone else committed the violation (Conrad 2011; Rowland and Jerome 2004). Shifting blame (also called defeasibility, Benoit 2014) is a form of denying responsibility (Benoit 1997). By comparing the boost to forgiveness likelihood after an apology accompanied by third-party blame (apology-with-3rd-party blame) versus the boost after an apology accompanied by restitution (apology-with-restitution), we sought to more directly test the mediational role of emotion change (Hypothesis 2). That is, if a cheap apology (i.e., one without restitution) becomes as effective as a costly one (i.e., with restitution) using a blame-shifting story to change emotions, then we can be more confident that changes in emotion are a more proximal cause of forgiveness than actual restitution.

Experiment 1: Apology

Experiment 1 investigated the role of negative-to-positive emotion replacement in eliciting forgiveness using a simple apology in the aftermath of a broken transactional organizational promise.

Methods

Sample, Design, and Procedures

Undergraduate students (N = 61) participated for course credit and were placed alternately into apology-alone (n = 30) and no-apology (n = 31) conditions. Apology (apology-alone/no-apology) was the independent variable and forgiveness likelihood the dependent variable; increases in positive, and decreases in negative, emotions about the offender were the proposed mediating variables. Participants read online a narrative of an undergraduate’s experience of a broken promise of financial aid and imagined themselves as the student depicted in the story. The story included an organizational promise (to produce a psychological contract) broken by the department (to produce PCB), followed by an organizational response. Table 1 lists the chronology of story events and experimental measures.

Table 1 Narrative chronology and scale reliabilities
Full size table

Materials

The story depicted a student who was given the standard $3000 first year financial aid scholarship and was verbally promised by the Chair (on behalf of the Department) an additional $7000 for the second year, upon the condition of achieving an “A” average over freshman year. Highly motivated, the student diligently attended class, worked hard, handed in assignments early, forewent partying and procrastination, and thereby earned the “A” average. Despite this, the student received notice that the second-year scholarship amount would remain unchanged ($3000). Surprised and perplexed, the student met the Chair, who denied ever making the promise; the student then realized that the Department gets more money from the College if student grades are high. After complaining politely and asking for the promised aid in a letter to the Department, the Department met to discuss the matter and responded to the student in a letter. The response letter stated:

Thank you for your recent letter concerning your financial aid. We, the department, have considered it carefully and discussed it at length at a department faculty meeting. Everyone in our department was present, and we unanimously voted to...

[no-apology:] not comment on this matter.

[apology-alone:] apologize to you. We broke our promise and we are sorry.

In both conditions, the student’s financial aid remained the same. The manipulation was successful: The apology manipulation check ratings (see below) for no-apology (mean = − 2.87, SD = 2.22,) differed from those for the apology-alone condition (mean = 1.60, SD = 2.53, p < 0.001). (All materials and data are available at http://farstudy.x10host.com/).

Measures

Checks and Demographics

To check that a psychological contract had been created, participants rated as true or false: “In the scenario, you expect to receive $3000 in financial aid during your first year” and “...you expect to receive $10,000 in financial aid during your second year, if you earn an “A” average throughout your first year.” PCB was checked using agreement with three items (e.g., “The department broke their promise to me”). The apology manipulation was checked using “The department apologized.” Post-manipulation, perceived sorrow was used as an additional manipulation check and was measured by having participants rate their agreement with “The department members genuinely felt regret” and “…sorrow.” Participant conscientiousness was checked with “As you think about this questionnaire so far, please indicate honestly how conscientiously you imagined yourself as the student in the scenario and answered the questions to the best of your ability.” Participants also reported their age, academic major, sex, race/ethnicity, and academic year. In each experiment, data were discarded from participants incorrectly answering check questions, sample demographics did not differ across conditions, and scale reliabilities were high (see Table 1 notes).

Pre- and Post-manipulation Forgiveness Likelihood

Participants responded to “At this point, how likely is it that you would forgive the department for having broken their promise?” on a 9-point scale anchored with “Not at all,” “Slightly,” “Moderately,” “Mostly,” and “Completely.”

Pre- and Post-manipulation Negative and Positive Emotion Valence

Participants rated their agreement with eight statements about positive emotions toward the department (i.e., empathy, admiration, gratitude, warmth, anticipation, friendliness, liking, and overall positive feeling), and eight about negative emotions toward the department (i.e., anger, disapproval, fear, disgust, bitterness, resentment, upset, and overall negative feeling). We generated these items using the strategy found in the Affect Subscale of the Enright Forgiveness Inventory (EFI; Enright and Rique 2004). The EFI asks participants to think of a person who hurt them recently and to rate agreement with items such as “I feel warm toward him/her” (Enright and Fitzgibbons 2015b, pp. 255–256).

Results

Hypothesis 1A: Apology Effect

As predicted (Hypothesis 1A), an effect for apology occurred: respondents were less unwilling to forgive after a simple apology than after a no-apology (“no comment”) response. As shown in Fig. 1, forgiveness likelihood means and CIs never exceeded zero, indicating that after the breach, forgiveness was improbable, even with an apology. However, observation suggests that forgiveness likelihood increased slightly with an apology, and decreased slightly without it. Planned contrasts to investigate the effects of apology were calculated using post-manipulation (while controlling for pre-manipulation) forgiveness likelihood ratings. As predicted, apology-alone, compared to no-apology, increased forgiveness likelihood, t(58) = 2.90, p = 0.005, r = 0.37. This result supported Hypothesis 1A.

Fig. 1
figure 1

Experiment 1 means and 95% CIs for pre and post ratings of forgiveness likelihood by apology (present/absent) condition

Full size image

Hypothesis 2: Emotion Valence Mediation

We used regression analysis to investigate Hypothesis 2 that pre-post changes in positive and negative offender-oriented emotion would each mediate the apology effect. Table 2 presents standardized beta coefficients and p values for these. As expected, results indicated that positive and negative emotion change mediated the apology effect: apology predicted change scores, change scores predicted forgiveness likelihood, and beta weights for the apology-forgiveness regression became non-significant after adding change scores (p values rose from 0.06 to at least 0.28). This result supported Hypothesis 2.

Table 2 Experiment 1 standardized betas for mediation analyses
Full size table

Hypothesis 3: Displacement

Hypothesis 3 predicted that increases in positive emotions would correlate with decreases in negative emotions, that is, that positive and negative offender-oriented emotion changes would be negatively correlated. Consistent with this, changes in positive and negative emotions were negatively correlated, r = − 0.71, p < 0.001. This result supported Hypothesis 3. In addition, positive and negative emotions were also negatively correlated at both pre- (r = − 0.57, p < 0.001) and post-manipulation points (r = − 0.71, p < 0.001).

Discussion

Concerning response effects, apology-alone, versus a no-apology response from an organization that had broken a promise, made participants less unlikely to forgive after PCB. Given that the breach incurred was transactional and involved a relatively substantial financial promise, it is noteworthy that a simple apology reduced unwillingness to forgive, increased positive, and reduced negative emotions. This finding is indicative of the power of even a minimal reparative speech act. This result replicates previous research finding an effect for an apology on forgiveness and extends it to organizational wrongdoing involving PCB.

Concerning mediation and displacement, increases in positive and decreases in negative emotions about the offending organization mediated the apology effect. This result is consistent with ERM and replicates previous research showing emotion replacement in interpersonal contexts (e.g., McCullough et al. 1997). Also consistent with ERM, changes in emotions were strongly negatively correlated. To our knowledge, these findings constitute the first test of ERM emotion change mediation and displacement in an organizational PCB context. They point toward the central role of emotion in the apology ritual.

Experiment 2: Apology and Restitution

As in Experiment 1, Experiment 2 investigated the role of negative-to-positive emotion replacement in eliciting forgiveness using a simple apology, but also used restitution.

Methods

Sample, Design, and Procedures

Undergraduate students (N = 114) participated for course credit and were placed alternately into no-apology-no-restitution, apology-alone, restitution-alone, and apology-with-restitution conditions (n = 32, 29, 27, and 26, respectively). Apology (present/absent) and restitution (present/absent) were the independent variables. The dependent variable, proposed mediating variables, and procedures were identical to those in Experiment 1.

Materials and Measures

Materials and measures were identical to those used in Experiment 1 except for alterations in the manipulation (and manipulation checks) as follows. First, manipulations were accomplished via the department response letter, which contained the same introduction followed by the apology manipulation:

Thank you for your recent letter concerning your financial aid. We, the department, have considered it carefully and discussed it at length at a department faculty meeting. Everyone in our department was present, and we unanimously voted to...

[no-apology:] not comment on this matter.

[apology:] apologize to you. We broke our promise and we are sorry.

The restitution manipulation followed:

You also notice that they make no mention of reinstating your promised financial aid, but when you call Financial Aid, they confirm that

[no-restitution:] nothing has changed: you will still receive only $3,000 next year.

[restitution:] you will receive the full $10,000 you expected next year.

The manipulation checks consisted of two ratings, one for apology “The department apologized,” and one for restitution “The department gave me the scholarship they owe me.” Both manipulations were successful (p < 0.001 for all comparisons): Apology check ratings in no-apology (mean = − 3.49, SD = 0.95) conditions were lower than ratings in apology conditions (mean = 1.59, SD = 2.43). Restitution check ratings in no-restitution (mean = − 3.21, SD = 1.75) conditions were lower than ratings in restitution conditions (mean = 1.19, SD = 3.07).

Results

Hypothesis 1A, 1B, and 1C: Apology, Restitution, and Boost Effects

We predicted apology (Hypothesis 1A), restitution (1B), and boost effects (1C); all occurred, and restitution was more effective than an apology. Observing Fig. 2, we note first that all forgiveness likelihood means and CIs never exceeded + 1, indicating that at best forgiveness was only moderately likely. However, pre-post trends differed markedly, suggesting a strong restitution effect, a moderate apology effect, and a marked boost effect when the two were combined. Figure 2 shows that forgiveness likelihood means increased pre to post for the two conditions with restitution, but either gave no evidence of change or decreased for the two conditions without restitution. Inferential analysis confirmed these observations: Controlling for pre-manipulation forgiveness likelihood ratings, a 2 (apology/no-apology) × 2 (restitution/no-restitution) ANCOVA on post-manipulation forgiveness likelihood ratings found a strong main effect for restitution, F(1, 108) = 44.07, p < 0.001, r = 0.51, a weak effect for apology, F(1, 108) = 4.09, p = 0.04, r = 0.15, and no interaction (p = 0.69).

Fig. 2
figure 2

Experiment 2 means and 95% CIs for pre and post ratings of forgiveness likelihood by apology × restitution condition

Full size image

Pairwise comparison between the apology-alone versus no-apology-no-restitution conditions once again showed an apology effect, t(58) = 2.02, p = 0.046, r = 0.27, replicating the Experiment 1 result for Hypothesis 1A. Pairwise comparison between restitution-alone and no-apology-no-restitution conditions showed an even stronger restitution effect, t(56) = 2.83, p = 0.006, r = 0.52. However, pairwise comparison failed to show differences between restitution-alone and apology-with-restitution, t(51) = 1.11, p = 0.27, r = 0.18, indicating that once restitution was given, apology added little. These results supported an apology effect (Hypothesis 1A), and especially a restitution effect (Hypothesis 1B). Pairwise comparisons between apology-with-restitution and apology-alone showed a strong boost effect, t(53) = 4.35, p = < 0.001, r = 0.48. These results support the boost effect (that restitution boosts the effectiveness of apology, Hypothesis 1C). When it comes to forgiving, words (of apology) count, but actions (restitution) count more.

Hypothesis 2: Emotion Valence Mediation

Regression analysis was used to investigate Hypothesis 2 that pre-post changes in positive and negative offender-oriented emotion mediated the apology, restitution, and boost effects on forgiveness likelihood (see Table 3). First, and contrary to prediction, neither positive nor negative emotion changes mediated the apology effect, because as Table 3 shows, the independent variable (apology) failed to predict the proposed mediators reliably (p = 0.11 for positive and p = 0.19 for negative emotion change). Sobel tests (see Table 3 note a) also show no mediation effects for apology. However, meta-analytically averaging Experiments 1 and 2 apology effect mediation results using n-weighted z-scores (Rosenthal and Rosnow 2007, p. 496) does show the predicted mediation effects for positive (z = 2.94, p = 0.002) and negative emotion changes (z = 1.76, p = 0.04). Second, as predicted, positive and negative emotion changes partially mediated the restitution effect and the boost effect. In sum, emotion changes did not mediate the apology effect reliably in Experiment 2, though they did partially mediate it for Experiments 1 and 2 combined, and emotion changes partially mediated the restitution and boost effects. These results supported Hypothesis 2 for apology (overall), restitution, and boost effects.

Table 3 Experiment 2 standardized betas for mediation analyses
Full size table

Hypothesis 3: Displacement

Hypothesis 3 predicted that positive and negative offender-oriented emotion changes would be negatively correlated. Consistent with this, changes in positive and negative emotions were negatively correlated, r = − 0.83, p < 0.001. The results supported Hypothesis 3. In addition, positive and negative emotions were also strongly negatively correlated at both pre- (r = − 0.65, p < 0.001) and post-manipulation points (r = − 0.85, p < 0.001).

Discussion

Concerning organizational response effects, restitution versus no-restitution (strongly), and apology versus no-apology (weakly), each made participants less unlikely to forgive after PCB; when combined, participants became moderately likely to forgive. These results were as predicted, again replicate the apology effect, and also replicate a very small number of previous study findings showing restitution effects (e.g., Carlisle et al. 2012). Also, by operationally distinguishing apology-alone from restitution-alone, our findings add new empirical knowledge about their unique, comparative, and combined effects in eliciting forgiveness in organizational contexts.

Concerning mediation, Experiment 2 did not replicate apology effect mediation by offender-oriented emotion changes, contrary to our prediction. Apology-alone, compared with controls, did yield positive and negative emotion changes in the predicted directions, but these changes were not reliable. However, when combined with Experiment 1 results, the predicted outcomes occurred. Restitution and boost effects also were each partially mediated by positive and negative emotion changes. Thus, the results (overall) were as predicted and support the ERM proposition that offender-oriented emotion changes mediate the effects of apology and restitution on forgiveness. Concerning displacement effects, positive displaced negative offender-oriented emotions.

Experiment 3: Emotion-Reversal

Whereas Experiments 1 and 2 manipulated the independent variables (apology and restitution), Experiment 3 sought to manipulate the mediating variable (negative-to-positive emotion valence change). Specifically, we sought to reverse the net negative emotions caused by PCB; hence we titled Experiment 3 emotion-reversal.

Methods

Sample, Design, and Procedures

Undergraduate students (N = 95) participated for course credit and were placed alternately into apology-alone, apology-with-restitution, and apology-with-3rd-party blame conditions (n = 31, 30, and 34, respectively). The first two of these conditions constituted the restitution manipulation; the first and the third conditions constituted the third-party blame manipulation. Thus, restitution (present/absent) and third-party blame (present/absent) were the independent variables. The dependent variable, proposed mediating variables, and procedures were as in Experiments 1 and 2.

Materials and Measures

Materials and measures were identical to those used in Experiments 1 and 2 except as follows. All conditions conveyed the standard apology embedded in the department response letter:

Thank you for your recent letter concerning your financial aid. We, the department, have considered it carefully and discussed it at length at a department faculty meeting. Everyone in our department was present, and we unanimously voted to apologize to you. We broke our promise and we are sorry.

One of the three conditions then followed. The first two of these conditions were identical to the apology-alone (“You also notice that they make no mention of reinstating your promised financial aid, but when you call Financial Aid, they confirm that you will still receive only $3000 next year”) and the apology-with-restitution (“…you will receive the full $10,000 you expected next year”) conditions in Experiment 2.

The third condition, apology-with-3rd-party blame, was identical to apology-alone except that the following paragraph was inserted in the department letter after the apology:

This situation occurred because an employee in the university’s accounting office embezzled large amounts of money. This employee was fired, but unfortunately, the amount stolen was so great that the university was unable to increase anyone’s merit scholarship this year; indeed, it was not even able to honor faculty cost-of-living salary increases (this has been a hardship for some). Everyone seems to have been affected by the embezzlement, but again, we are sorry.

The apology check ratings in all three conditions did not differ (mean = 1.67, SD = 2.21, p = 0.14) and were similar to apology condition means in Experiment 2. The restitution check ratings in no-restitution conditions (mean = − 2.68, SD = 2.10) also did not differ (p = 0.32) and were lower than ratings in restitution condition (mean = 0.77, SD = 3.38) (p < 0.001); the restitution manipulation was therefore considered successful. Comparing emotion valences of apology-with-3rd-party blame versus apology-alone conditions, positive emotion was higher (mean = − 1.42, SD = 1.62, vs. mean = − 3.03, SD = 1.25) and negative emotion was lower (mean = 0.82, SD = 1.86, vs. mean = 2.27, SD = 2.02) (all p < 0.01). The third-party blame manipulation of the mediating variables, positive and negative emotion valence, was therefore considered successful.

Results

Boost (Hypothesis 1C) and Emotion-Reversal Effects

We predicted a boost effect (Hypothesis 1C) and to test emotion-reversal, we expected a third-party blame effect; both occurred. Figure 3 shows that, as with the two previous experiments, forgiveness likelihood never recovered fully from the breach, but pre-post trends differed markedly by condition. Observation suggests, and pairwise comparisons show, that compared with apology-alone, forgiveness likelihood means increased when restitution was added (t(59) = 4.66, p = < 0.001, r = 0.55, boost effect, Hypothesis 1C), and when breach was blamed on a third-party (t(63) = 4.79, p = < 0.001, r = 0.54, third-party blame effect). Indeed, blaming a third-party elicited the same likelihood of forgiveness as giving restitution. We also note that because the apology-alone and apology-with-restitution conditions were identical to the apology conditions in Experiment 2, their comparison constituted a replication of the boost effect. This result again supported Hypothesis 1C.

Fig. 3
figure 3

Experiment 3 means and 95% CIs for pre and post ratings of forgiveness likelihood by response condition

Full size image

Hypothesis 2: Emotion Valence Mediation

Table 4 reports regression analyses showing that pre-post changes in positive and negative emotion (Hypothesis 2) each partially mediated the boost effect, thus replicating the boost effect mediation in Experiment 2. Emotion changes also partially mediated the third-party blame effect. These results supported Hypothesis 2.

Table 4 Experiment 3 standardized betas for mediation analyses
Full size table

Hypothesis 3: Displacement

Changes in positive and negative emotions were again negatively correlated, r = − 0.65, p < 0.001, supporting Hypothesis 3. Positive and negative emotions were also again negatively correlated at both pre- (r = − 0.61, p < 0.001) and post-manipulation points (r = − 0.73, p < 0.001).

Discussion

Concerning organizational response effects, a message blaming a third-party was as effective as compensatory restitution in eliciting forgiveness. The boost effect for restitution results replicated Experiment 2, giving additional evidence of the effectiveness of making one’s apology costly by adding restitution. The boost from restitution is as would be expected, especially with a transactional breach.

The third-party blame effect raises an ethical issue. That organizations may blame shift for broken promises in no way implies that they should do so. Or that it would be wise; groups can be quite effective at ferreting out the facts over time (DiFonzo 2010) and a history of leader wrongdoing is strongly negatively correlated with employees’ tendency to forgive them (Basford et al. 2014, see Table 2, p. 109). It may be that participants were willing to believe the apology-with-3rd-party blame response because it was the first time they received it, was more detailed than the brief apology-alone response, and indicated that department members had shared in the negative consequences of the embezzlement transgression.

Concerning mediation by emotion changes, boost and third-party blame effects were each partially mediated by positive and negative offender-oriented emotion changes. Further, we replicated boost effect mediation seen in Experiment 2. These predicted results are consistent with the ERM idea that emotion changes play a more proximal mediational role in eliciting forgiveness. Concerning displacement effects, positive and negative emotion changes were for a third time strongly negatively correlated as predicted; this result accords with the idea that positive offender-oriented emotions displace negative ones.

General Discussion

Summary

Over three experiments using a vignette methodology, after transactional PCB a simple organizational apology and compensatory restitution each elicited greater likelihoods of forgiveness as compared to control conditions in which no-apology or reparation was offered. Restitution exceeded apology effects, and when combined, restitution boosted apology effects. Simultaneous decreases in negative and increases in positive offender-oriented emotions partially mediated these effects, and these emotion valence changes were strongly correlated with each other.

Theoretical Implications

Broadly, this research contributes to that currently minimal (Grover et al. 2017) portion of the business ethics literature pertaining to the repair of damaged norms and relationships within an organizational community. Repair and ethics are intimately connected within community. Foundational moral principles such as “Do to others as you would have them do to you” (Luke 6:31, New International Version) are communal in nature; they involve “others” and “you.” Questions concerning the violation of moral norms in organizations (e.g., promise-breaking) have an intrinsic communal character and are a proper focus of business ethics. However, community is an ongoing and living entity, and therefore what happens after the violation also possesses an intrinsically moral and communal character. Thus, questions concerning the thinking, feelings, and actions surrounding the repair of damaged organizational norms and relationships are of central interest to business ethics research.

Specifically, our results contribute to the knowledge of the repair process in three main ways. First, they add to our understanding of reparative effects of organizational apology and restitution by means of procuring forgiveness. Second, they add to a dual-track conceptualization of repair processes, and this has implications for the repair of damaged relationships after psychological contract breach. Third, they point to the centrality of emotion in reparative actions. They do this by supporting the ERM itself and by suggesting a way of conceptualizing forgiveness of an organization. We take each in turn.

Reparative Effects of Apology and Restitution

Business ethics is interested in the effectiveness of initiatives that seek to redress moral violations and rebuild moral expectations. Our findings point toward the value of an apology, the greater value of reparation, and the greater-still value of their combination in organizational responses intended to foster forgiveness, restore trust, reconcile relationships, and repair community moral expectations. Simple words of contrition buffered the net negative emotions and greater unwillingness to forgive following a non-apology but were most effective when buttressed with restitution. These results support Walker’s (2006) assertion that making amends is fundamental to moral repair efforts. The relatively smaller effectiveness of apology-alone accords with other recent research indicating that apologies are sometimes not effective in facilitating reconciliation (De Cremer and Schouten 2008; Skarlicki et al. 2004). Apology-alone may be a weak treatment for serious injury.

Dual Processes in Repair

Business ethics is also interested in the nature of moral violations (e.g., in what ways do violations disrupt moral expectations and relational equity?) and the nature of the repair efforts involved in redressing moral violations (e.g., how do certain types of repair initiatives rebuild moral expectations and restore relational equity?). To these points, in addition to being a weak treatment for serious injury, apology-alone may be the wrong treatment for the injury. That is, the reparation effort may not match the nature of the violation.

In line with this idea, Ren and Gray (2009) set forth two types of violations in an organizational context that trigger relationship conflict: those that threaten identity needs (e.g., respect and inclusion) and control of desired outcomes (e.g., fair access to resources). The restoration process for each of these violations will differ: identity-violation restoration goals include face-saving and recommitment to social order, whereas control violations include compensation for loss and the restoration of the sense of procedural justice. These theorists proposed that appropriate restoration behaviors for violation of control include penance and reframing accounts, but not an apology. Our results support this formula. After a control violation (promised tuition aid unfairly withheld), forgiveness likelihood increased with restitution (penance), or with reframing the violation as the fault of a third-party, regardless of the presence of apology (see also Kim et al. 2009). Reb et al. (2006) advanced a comparable understanding of how different offenses (i.e., those that violate procedural vs. interactional justice) obstruct different needs (i.e., control vs. belonging needs). Different offenses thus require different organizational remedies to redress those needs (i.e., compensation to address control needs vs. apology to address belonging needs).

Similarly, our findings likely generalize more to transactional than relational forms of PCB. Although the perceived promise depicted in our vignette was embedded within a college student–department relationship, often regarded as a long-term and almost familial association (e.g., alma mater), the salient aspect of the relationship was a transaction: a specified monetary reward in exchange for a specified level of academic performance. As noted in the introduction, Morrison and Robinson (1997) proposed that after a transactional breach, outcome concerns should be of great importance in transactional psychological contracts. We reasoned that because giving restitution addresses outcome concerns and a simple apology does not, giving restitution after a transactional breach should be more effective than a simple apology for eliciting forgiveness. Our results support this idea, that is, for transactional psychological contract breach, restitution was superior to apology in eliciting forgiveness. This finding adds to the dearth of research on management and repair of PCB, particularly transactional PCB.

However, even in transactional breach, judgments framed by the salient social contract context may be more determinative than outcome concerns. Social contract context is the framework of norms and expectations used to judge conduct. Morrison and Robinson’s (1997) model posited that social contract context also moderates violation development (see also Aycan and Kabasakal 2006; Rousseau 2011). Our results are consistent with this model in that even without reparation, the knowledge that another party caused the transgression absolved the organization of wrongdoing and brought about forgiveness. The embezzlement scenario successfully relieved the department of responsibility for fulfilling its promise. Indeed, our results may add to the Morrison and Robinson model in suggesting that even in transactional relationships, social contract context was a more powerful factor than outcomes in the development of post-PCB violation. Speech acts (Worthington and Scherer 2004) that successfully shift the social contract context may thus override the necessity of reparation in procuring forgiveness.

Emotion in the Repair Process

Finally, business ethics is interested in the psychological mechanisms by which the repair of communal moral norms and the restoration of relational equity are accomplished. To this point, at least two implications of our results are noteworthy and expand on the role of emotion in the repair through forgiveness process. First, the results of this investigation lend additional support to ERM itself. Reductions in negative emotions were strongly correlated with increases in positive emotions, and these changes mediated apology (overall), restitution, boost, and third-party blame effects on forgiveness. Our study constitutes a stronger test of ERM than has heretofore been demonstrated. Despite the model’s general acceptance, little research has specifically investigated its central tenets, namely, negative and positive offender-oriented emotion change mediation of forgiveness, and negative-to-positive offender-oriented emotion displacement. In particular, we are not aware of any previous ERM research explicitly gaging the hypothesized pattern of displacement, that is, the association between gains in positive, and losses in negative, offender-oriented emotions. Our research addresses this gap.

Second, our research has implications for emotion replacement involved in the repair of moral norms and relational equity involved in relationships between persons and organizations. That is, it has implications for the conceptualization of repair through forgiveness of an organization as compared to a person. Developed for interpersonal contexts, ERM posits that because the victim recognizes apology or reparation as a humbling and aversive experience, the victim feels empathy for the offender. Affective empathy is the “capacity to experience the emotions of another” (Joliffe and Farrington 2006, p. 589). However, organizations are not persons. It is unlikely that employees would envision, let alone experience, an impersonal entity’s “emotions,” after humbling itself or otherwise. More plausibly, apology and reparation from a promise-breaking organization simply foster positive change in the affective component of the attitude toward that organization. The affective component of attitude consists of one’s feelings toward an object (Eagly and Chaiken 1998). In response to receiving apology and restitution, the employee feels greater positive emotions toward the organization, for example, greater admiration, gratitude, and liking for the organization. Affective empathy, in the sense of experiencing the emotions of another, may therefore be unnecessary for eliciting forgiveness of organizations. If so, then a broad set of negative-to-positive organization-oriented emotions may be more germane to organizational forgiveness. This idea accords with the recent proposal from the field of organizational emotionality that “empathy may be defined and measured differently based on the level of analysis chosen by the researcher” (Burch et al. 2016, p. 172).

In addition to theoretical implications, our results suggest practical lessons for organizational leaders, to which we turn next.

Practical Implications

Transactional PCB is likely widespread (e.g., 89% of a large-scale survey of public employees in Britain, Coyle-Shapiro and Kessler 2003) and harmful (Fu and Cheng 2014; Zhao et al. 2007, see Table 4, p. 668). To forestall the development of PCB and all of its attendant negative consequences, organizations can opt to seek forgiveness through apology, restitution, or both. After breaking a transactional promise, our research suggests the organization should restore what was promised and apologize. The apology should take responsibility for the offense and sincerely express remorse (see also Basford et al. 2014). In taking these initiatives, organizations can aim to replace negative with positive emotions oriented toward the organization. Organizational “best practices” should include forgiveness-seeking after promise-breaking.

However, there are potentially “dark” implications of speaking of “eliciting forgiveness” as an organizational leadership skill or best practice. First, apology and restitution may work to dissociate the offender from the offense by differentiating the offender’s identity, which may be yet another way of evading rather than admitting, responsibility (Bentley 2015; Wittig 2009). Second, teaching organizations about the means and efficacy of apology and restitution may weaken inhibitions to commit the offense in the first place. “Do first, ask forgiveness later” has become a popular business mantra. Third, focusing on the utility of the reparative action may undermine its moral virtue (Taft 2000). An organizational mea culpa can become a utilitarian means to the end of diffusing employee resentment. Similarly, focusing on the strategic benefits of seeking forgiveness may minimize the volitional character of a remorseful apology; the offender may regard forgiveness-seeking as something they would not do unless forced (e.g., by ethical convention or social pressure). A strategic focus may also cast forgiveness-seeking as a social transaction (i.e., “an apology will purchase forgiveness”). In light of these utilitarian pitfalls, a question for the offending organization to consider is “Would we apologize if we perceived no potential subsequent benefit?” Finally, from the victim’s perspective, an apology or restitution might convey subtle pressure to forgive, also undercutting the moral roots of these reparative rituals (DiBlasio 1998). Promoting “forgiveness elicitation skills and best practices” may thus have unintended unethical consequences.

Strengths, Limitations, and Future Research

Finally, our study had several strengths and limitations. Strengths included sequential experimentation that enables opportunity for replication and is well-suited for focused and systematic testing of a theoretical model. Also, the pre-to-post design allowed measurement of within-subjects change and better control for individual variation, thus affording greater power. Further, the vignette was tailored to undergraduates’ experiences and concerns (college admissions, financial aid, and work effort vs. socializing) to maximize experimental realism and involvement. But although vignette experiments continue to be a staple of apology research (e.g., Darby and Schlenker 1982; Tucker et al. 2006; Walfisch et al. 2013), researchers should exercise caution in generalizing vignette study results (Collett and Childs 2011). A more extensive variety of research methods would afford stronger confidence in our conclusions, particularly field surveys, field experiments, and more experiential laboratory testing of our hypotheses.

Future research, especially in field settings, should also use a more complex and standardized measure of the dependent variable, forgiveness. Forgiveness measure development has typically occurred in interpersonal relationship rather than organizational contexts, but even those developed for workplace contexts pertain to interpersonal offenses (e.g., Aquino et al. 2006; Boonyarit et al. 2013). As in our study, forgiveness research sometimes uses single-item measures (e.g., “To what extent have you forgiven [the person who hurt you],” Enright and Fitzgibbons 2015b, p. 257). More complex forgiveness measures often assess revenge and avoidance motivation (McCullough et al. 1998), forgiveness stage (Enright and Fitzgibbons 2015b), trust, or wishing the offender well (McCullough et al. 1997). These multi-faceted aspects of forgiveness are inappropriate for a simple vignette methodology, but would be suitable for actual forgiveness situations in workplace field settings.

Future research should also seek to assess the mediating variable using other methods. Our study used third-party blame information to manipulate the proposed mediating variable, but this method also removed the transgressor’s responsibility for the offense. Future research should attempt a stronger test of the mediational hypothesis by altering victim offender-oriented emotion (e.g., use information not related to the violation) while leaving transgressor responsibility in place.

Conclusion

Knowledge about the processes involved in recovery after organizational misconduct is not only highly practical and useful, but also highly relevant to business ethics (Grover et al. 2017). Business ethics researchers have lately shown interest in the effects and mechanisms of moral repair (Goodstein et al. 2016). Our study adds to this nascent literature and links it with forgiveness theory. Apology and restitution, but especially the latter, may rectify the damage done to distributive justice norms and communal emotional bonds when a transactional promise is broken (Cugueró-escofet et al. 2014). Apology and restitution may also foster emotional healing from injury (Taft 2000). In short, apology and restitution may accomplish much-needed moral repair within an organizational community (Walker 2006). The psychological mechanisms involved in moral repair give central place to granting victims a voice and acknowledging their emotions after injustice. Our study substantiates and gives further texture to this idea, showing that emotion replacement is at the heart of the moral repair process. Thus, apology and restitution bring about the emotional changes necessary for moral repair to occur.