This is an open access article distributed under the terms of the Creative Commons
Attribution License (In the aftermath of intergroup harms, perpetrator groups may seek to make amends for past violations through efforts such as apologies. But what purpose do these restorative justice efforts serve? Forgiveness is frequently examined in past research (Blatz & Philpot, 2010), but we believe other outcomes are also important. One such outcome is affirmation of fairness and justice in one’s current society (Wenzel, Okimoto, Feather, & Platow, 2008). We believe victim group members’ faith in societal norms of just behavior is important because past injustices can call these norms into question, leading victims to generalize to societal institutions and disengage from them (Katz et al., 2008). In this paper, we empirically examine how intergroup apologies affect perceived remorsefulness, beliefs about norms of justice, trust in the government, and intergroup forgiveness, and we make the case for why beliefs about norms of justice in society may be important for the well-being of the victims. In addition to examining new and relevant outcomes in intergroup apology research, we also examine how the degree of elaboration in the presented apology affects these outcomes. This study is the first to our knowledge to examine whether and how people react to an increasingly complex intergroup apology that was actually given for a historic injustice (see Kirchhoff, 2013 for past research on how apology elaboration affects responses in the context of a hypothetical apology).
Intergroup Forgiveness
Most research on the effects of intergroup apologies on victim groups focuses on whether or not they elicit forgiveness as the primary outcome (Brown, Wohl, & Exline, 2008; Leonard, Mackie, & Smith, 2011; Philpot & Hornsey, 2008, 2011); however, some research has examined how apologies affect attitudes towards the other groups in the conflict (Blatz, Day, & Schryer, 2014). The idea that intergroup apologies should lead to forgiveness has been widely discussed (Lazare, 2004; Minow, 2002; Tavuchis, 1991), but the evidence is mixed (Hornsey & Wohl, 2013; Iyer & Blatz, 2012). Some studies establish an intergroup apology-forgiveness link. For example, an apology offered by the U.S. government to Canadian participants for a friendly fire incident (Brown et al., 2008) and by faculty at a university for insulting the student body (Leonard et al., 2011) led to greater forgiveness toward the offenders. Other studies have failed to find an apology-forgiveness link. Across four experimental tests, Australian participants expressed an equal level of forgiveness when they read that the perpetrator group had or had not apologized for five different historical injustices (Philpot & Hornsey, 2008; see also Philpot & Hornsey, 2011). Moreover, Wohl, Matheson, Branscombe, and Anisman (2013) did not find longitudinal effects of the Canadian apology for the Chinese head tax on forgiveness among Chinese Canadians.
These mixed results might be explained by recent theorizing suggesting that an intergroup apology offered in the context of appropriate pre-conditions may be more likely to lead to forgiveness (Blatz & Philpot, 2010). Wohl, Hornsey, and Philpot (2011) propose a staircase model of intergroup apologies which outlines actions, such as admitting collective guilt, setting the historical records straight, and repairing harm, that must take place prior to an apology, and subsequent actions, such as positive intergroup contact, that should follow an apology in order for it to effectively promote forgiveness. Relatedly, multiple studies suggest that apologies are more likely to lessen the desire for revenge (Giner-Sorolla, Castano, Espinosa, & Brown, 2008) and to promote forgiveness (Wohl, Hornsey, & Bennett, 2012) when the wording of the apology is carefully tailored to express emotions in a way that meets the expectations of the victimized group (see also Fehr & Gelfand, 2010).
Forgiveness is a critical outcome to intergroup apologies, and can be an important step in the quest to improve intergroup relations. However, there are limitations of examining intergroup forgiveness as the primary outcome of intergroup apologies. Forgiveness might be difficult to achieve because victim group members may feel less accepting of the original perpetrators who committed the harm than they do of their present day representatives who did not. Moreover, members of the present victim group may feel that they lack the standing to offer forgiveness for a historical injustice. As such, they may feel that it is not their place to grant forgiveness if they were not direct victims (Iyer & Blatz, 2012).
Some recent research has begun to look beyond intergroup forgiveness. Blatz et al. (2014) found that intergroup apologies can improve how positively members of the historically victimized group evaluate the group who committed the injustice. Similarly, Bombay, Matheson, and Anisman (2013) found within a group of Aboriginal Canadians who personally had been or were related to someone who had been victimized by the residential schools incident in Canada that the government apology for this incident led to optimism that intergroup relations and the marginalized status of Aboriginal people in society would improve. This link was mediated by perceived discrimination, trust in the government, and forgiveness of the government. With the current paper, we hope to expand the list of outcomes typically examined in intergroup apology research. The present study examines how intergroup apologies affect the perceptions that governing authorities respect norms of just behavior.
The Role of Justice Norms in Society
As a result of living in social groups, people need to believe that there are shared values of acceptable behavior that they are expected to follow and that they can expect others to follow (Durkheim, 1893/1964; Lerner, 1980; Wenzel et al., 2008). Group norms help people anticipate and judge the behavior of others, and trust that they will be treated fairly and well. According to Justice Restoration Theory (JRT), when someone violates a rule or commits a transgression, he or she violates these expectations. One symbolic implication of a transgression is that in violating the victim’s right to fair treatment, the transgressor implies that shared values of what is right and wrong are not important and may not be honored (Wenzel, Okimoto, & Cameron, 2012; Wenzel et al., 2008). Although people can respond with one of two kinds of justice (retributive and restorative), restorative justice addresses justice norms directly. Restorative justice often involves bringing the perpetrator, victim, and possibly third parties together to discuss the transgression, establish how it violated norms of just behavior, and have the offender make amends through apology or some other conciliatory action (Wenzel et al., 2008). In this regard, JRT is echoed by past research on the structure and function of apology. By labeling past actions as wrong and communicating that violating shared norms creates suffering in the transgressor (Blum-Kulka & Olshtain, 1984; Iyer & Blatz, 2012; Lazare, 2004), an apology affirms expectations of how people should act (Goffman, 1971; Govier & Verwoerd, 2002), and presumably promotes faith that justice norms will be upheld in the future (Okimoto & Wenzel, 2009).
JRT makes a thorough and persuasive argument that restorative justice efforts, such as apologies, restore the belief in justice norms after interpersonal transgressions (Okimoto & Wenzel, 2009; Wenzel & Okimoto, 2010, 2012). However, the belief in justice norms also matters at the societal level. Research has demonstrated that it is important to believe that powerful figures in our lives act benevolently and adhere to standards of acceptable behavior because people have a fundamental need to feel that their lives are predictable (Janoff-Bulman, 1992; Lerner, 1980). Moreover, people desire to believe that governmental agents are both predictable and just (Lerner, 1980; van der Toorn, Tyler, & Jost, 2011). According to these various theories, people want to believe that the authorities who govern them act justly. People may even go to creative lengths to maintain these beliefs in the face of strong contradictory evidence (e.g., Jost & Banaji, 1994; Lerner, 1980). For instance, people may justify inequalities created by and misdeeds committed by the government even when they suffer because of these inequalities and misdeeds (van der Toorn et al., in press).
But such legitimization may not be possible when the government’s actions are grossly unjust by today’s standards (e.g., slavery, forced removal of children from their homes). Unjustifiable acts committed by the government question the belief that the government will uphold norms of just behavior in the future. Transgressions against members of one’s group in particular suggest that the government is capable of mistreating people similar to oneself (Tajfel & Turner, 1986). In such cases, authority figures might be able to help satisfy a desire for justice by apologizing. When government representatives label their past actions as unjust and express emotional suffering in an apology, they may communicate to the historically victimized group that the present authority, and presumably the majority group on whose behalf they offer the apology, affirms these values in the present (Blatz, Schumann, & Ross, 2009).
Expressing remorse and labelling the actions as unjust are central parts of the message when governments apologize (Barkan, 2000; Tavuchis, 1991). By labelling the past actions as wrong, the perpetrator directly communicates that he or she values this norm in the present (Blum-Kulka & Olshtain, 1984; Iyer & Blatz, 2012; Lazare, 2004). By expressing remorse, the perpetrator affirms the questioned values of just behavior by communicating that violating these values caused emotional distress. As such, an apology will be effective to the extent that the apologizer is seen as genuinely remorseful (Blatz et al., 2009; Goffman, 1971; Govier & Verwoerd, 2002; Lazare, 2004; Tavuchis, 1991). Thus, we hypothesize that victim group members will believe that apologies affirm justice norms within the government and in society to the extent that these apologies are perceived as remorseful, meaning that perceived remorsefulness will mediate the effect of apology on the belief in justice norms.
Trust
In addition to examining how a government apology may restore faith that justice norms are important to the government, we were interested to see if people would be more willing to trust that the government would not exploit their vulnerabilities in the future (Kramer & Carnevale, 2001). An apology may affirm that the compromised values are important to the current government, thereby suggesting that their future actions will adhere to these values (Goffman, 1971). Alternatively, if someone believes that the government does not regret violating an important value, he or she is unlikely to trust it in the future. While there are good reasons to believe that apologizing may facilitate trust, there are also reasons to be skeptical that such an association will always occur. Consistent with past theorizing on when apologies may lead to forgiveness, trust may only be developed after continued adherence to shared norms, ongoing positive contact between groups, and further restorative justice efforts (Wohl et al., 2011; Wohl et al., 2013). Moreover, unlike trust in one individual, trust in the government may be particularly difficult to achieve because it is often in flux. As such, it is unclear whether the people who run the next government will share the same values (Iyer & Blatz, 2012; Smith, 2008). Alternatively, even if a victimized group member believes that the apologizer is genuine, they may legitimately question whether others in the government or the dominant group in society supports the apology. Thus, while there are good reasons to think an intergroup apology may lead to trust, there are also reasons to be skeptical.
The Role of Intergroup Apology Elaboration
We propose that government apologies to members of a minority group may help them believe that the present government upholds standards of just behavior because the apology communicates remorse for violating these standards. But what makes victim group members perceive an apology as remorseful? By definition, an apology involves labelling past actions as unjust and expressing that the actions have caused regret and remorse in the transgressors (Barkan, 2000; Tavuchis, 1991). Moreover, past research has identified multiple speech elements that can elaborate on the core message of the apology to make it seem more sincerely remorseful such as statements that accept responsibility or acknowledge harm (Blatz et al., 2009; Bono, 2005; Lazare, 2004; Scher & Darley, 1997; Tavuchis, 1991). In interpersonal apology research, including more of these elements enhances the effectiveness of the apology. For instance, Scher and Darley (1997) asked participants to read about a hypothetical situation in which someone wronged a friend and then apologized. Then they varied the degree of elaboration in the apology through the number of additional speech elements it included (e.g., statements that expressed remorse, accepted responsibility, promised forbearance, and promised repair). They found both that even the most basic apology led participants to think that the perpetrator was more conscientious and apologetic, and a more reliable friend than no apology, and that these ratings increased with the number of apology elements included. Similar effects have been replicated in other interpersonal studies in which people were asked to imagine a transgressor apologizing for a hypothetical harm (Darby & Schlenker, 1982; Kirchhoff, Wagner, & Strack, 2012) or in which the offender apologized after an interpersonal offense staged in the lab (Anderson, Linden, & Habra, 2006; Bono, 2005).
Building off of this research and theorizing, we believe that the elaborateness of an apology is important to intergroup apologies because more elaboration will increase perceived sincerity and remorsefulness. When a representative of the perpetrator group offers a public apology to an entire victim group, many victim group members may only hear about the apology from secondhand sources, so the expression of remorse must extend beyond non-verbal cues (Tavuchis, 1991). Simple apologies may suffice for minor harms; however, the increased severity of intergroup harms may require a symmetrical increase in the elaborateness of the apologies (Hornsey & Wohl, 2013). For example, in addition to the function that an interpersonal apology primarily plays to restore a relationship, intergroup apologies attempt to establish a public record about the wrong (Tavuchis, 1991) and rebuild a belief that institutions will be fair in the future (de Greiff, 2008; Goffman, 1971). Consistent with the notion that intergroup harms need to be addressed by elaborate apologies, Blatz et al. (2009) analyzed the content of thirteen actual intergroup apologies offered by governments (e.g., apologies for the internment of Japanese Americans), and found that they typically contain many elements (see Table 2, pp. 227 of Blatz et al., 2009). Little experimental research tests whether more elaborate apologies are more effective in intergroup contexts (Blatz & Philpot, 2010), with the exception of work by Kirchhoff (2013; for examples of tests of apology vs. no apology see Blatz, 2008; Philpot & Hornsey, 2008; see also Wohl, Hornsey, & Bennett, 2012).
In a doctoral dissertation, Kirchhoff (2013) asked participants to imagine a harm (sexual discrimination by male professors) or to remember past armed conflict in Bosnia and Herzegovina and then to imagine that the offending group offered one of five apologies varying in elaborateness. In the sexual discrimination study, the first apology had a statement that expressed remorse. In the next three apologies, statements were added that accepted responsibility, promised forbearance, and conveyed emotion. Six additional elements were added to the final apology including a statement that acknowledged harm and a statement that admitted a norm violation (somewhat analogous to admitting injustice). Participants reported greater acceptance of the apology with four elements compared to those that were less elaborate. All other contrasts were non-significant. The apologies for past armed conflict were similar, but the elements were presented in a different order, such that the first apology condition only expressed remorse, and then additional elements were added to each subsequent apology condition that admitted norm violations, acknowledged harm, and praised the victim group. A final condition included all of the previous elements as well as ten additional elements. In this study, results found little effect of elaboration on how much people accepted the apology. Thus, these two studies, examining hypothetical apologies, showed conflicting effects of elaboration on how much people accepted the apology. We add to these studies by examining the role of elaboration in a different context, involving a real harm and a real apology, and, more importantly, by examining how apology affects a different outcome – belief in justice norms.
Current Study
We manipulated the elaborateness of an actual government apology by varying the number of elements (identified by Blatz et al., 2009) it contained compared to no apology. African-American participants recruited on-line from around the United States read a description of the Tuskegee Syphilis Study – a harm perpetrated by the U.S. government in which African-American men who had contracted syphilis went untreated for decades after a cure was discovered (Katz et al., 2008). Years after the study ended, former U.S. President Bill Clinton apologized to the surviving victims and their families (The White House, Office of the Press Secretary, 1997). Memory of this study resonates with many African-Americans today. In a study assessing U.S. residents’ knowledge of the Tuskegee study, African-Americans were more likely to know about the study than other groups, and of the African-Americans who knew about this study, hearing about President Clinton’s apology made them more trusting of biomedical research (Katz et al., 2008). We re-wrote the apology in a way that we could manipulate the apology elements participants read. The final apologies presented in each condition all used direct quotes from Clinton’s apology, but the individual sentences were presented in a different sequence from what Clinton originally said to meet the goals of the present study. We included six conditions: a no apology condition and five conditions presenting an apology with varying levels of elaboration (see Appendix A). All conditions mentioned that there was an event at the White House and a brief statement by Clinton honoring the victims. Thus, even in the no apology condition some action was taken. After the no apology condition, the most basic apology condition only included a statement of remorse, conceptually similar to a simple “I’m sorry” for an interpersonal offense (Meier, 1997; Scher & Darley, 1997), which we predict would be perceived as empty or only an expression of empathy (Hornsey & Wohl, 2013; Lazare, 2004; Schlenker & Darby, 1981). Because we were interested primarily in justice norms, our second apology condition included a statement that admitted the Tuskegee study was unjust (Lazare, 2004; Tavuchis, 1991). Beyond this, we predicted that as we added elements designed to communicate an apology’s sincere remorse (Barkan, 2000; Tavuchis, 1991), the successive apologies would be seen as more remorseful and more affirming of justice norms, and we hypothesized that perceived remorsefulness would mediate the effect of the apology on justice norms. Therefore, we examined a mediation model that tested whether each apology, compared to no apology, would lead to increased affirmation of justice norms mediated by perceived remorsefulness.
Trust in the government may become more likely when an apology is offered. However, we expected effects on trust to be weaker than effects for perceived remorsefulness and justice norms. Finally, the evidence for an intergroup apology-forgiveness link in past research was mixed, so we did not make predictions about forgiveness. We did not preclude the possibility of forgiveness, but we had a number of reasons to expect little or no change in forgiveness as a result of the presented apologies. For example, the apology did not include a statement of empowerment to victims that has been found to facilitate readiness for reconciliation (Shnabel & Nadler, 2008; Wenzel et al., 2008), and it did not express emotions know to facilitate forgiveness (Wohl et al., 2012) and lessen the desire for revenge (Giner-Sorolla et al., 2008). Although our primary concern in this study was the effect of the various apology conditions on perceived remorsefulness, justice norms, and trust, we included a widely used measure of intergroup forgiveness (Philpot & Hornsey, 2008) to remain consistent with the past literature.
Method
Participants
We recruited 145 African-Americans online through Amazon’s Mechanical Turk. Ninety-two were female and had a mean age of 31.52 (SD = 19.74). Thirty-nine of the participants had completed some college, 56 had a college degree, and 15 had a masters or doctoral degree. From August 2011 to January 2012, 43 people completed a brief survey of demographic items and then were invited to participant in our study without explicitly being told that they were recruited because they identified as African-American. Participation tapered off, thus we began explicitly recruiting people who identified as Black or African-American, and 102 more people participated between February and April 2012. The procedure during both forms of recruitment was identical. There were no main effects of recruitment method on any of the outcome variables, Fs < 1.75, ps > .189, or interactions between recruitment method and condition, Fs < 2.00, ps > .082. To boost participation, we also progressively increased the compensation given to participants over the course of the study ($0.30, n = 40; $0.40, n = 19, $0.45, n = 22, $0.50, n = 56). There were no differences on our outcome variables across the different levels of compensation, Fs < 1.68, ps > .176, and there were no interactions between compensation and condition, Fs < 0.88, ps > .588.
Design and Procedure
Participants read a brief summary of the Tuskegee Syphilis Study (Katz et al., 2008):
“In 1932 the U.S. Public Health Service worked with the Tuskegee University in Montgomery County, Alabama to conduct a study on syphilis. The Tuskegee Study involved 600 Black male participants, 399 who had syphilis and 201 who did not. The participants were not given the opportunity to decide whether or not they wanted to participate in the study. For the 40 year duration of the study, none of the infected men were treated, even after a known cure for syphilis was found in 1947. The study was stopped in 1972 after an Associated Press article reported on it. The government gave the victims and families a 10-million dollar settlement outside of court and provided them with lifetime healthcare. Various groups felt that compensation was not enough and demanded that the government make a public statement about the study.”
Afterwards, they were randomly assigned to one of six conditions. In all conditions, including the no apology condition, participants read a short statement by Clinton honoring the remaining survivors. In each of the apology conditions, we included excerpts from Clinton’s apology, adding a new element to the apology from the preceding condition. Statements to represent the apology elements were based on previous coding (Blatz et al., 2009). For the text of each condition see Appendix A.
Measures
After participants read Clinton’s statement, they reported whether or not they thought an apology had been offered. Three people in the no apology condition wrongly thought that Clinton had offered an apology. Nine participants in the apology conditions said that he did not offer an apology (n = 3 in Condition 2, n = 4 in Condition 3, n = 1 in Condition 4, n = 1 in Condition 5). Participants were not excluded from the analyses as we believed that their ratings on the outcome measures were valid even if they did not define Clinton’s statement as an apology.i
All items described below were rated on a scale from 1 (strongly disagree) to 7 (strongly agree). To measure perceived remorsefulness, participants rated if “President Clinton’s response was sincere” and if “President Clinton’s response was remorseful”, r = .79, p < .001. Then participants indicated agreement with six items measuring beliefs that justice was upheld by Clinton’s response (α = .88; adapted from Okimoto & Wenzel, 2009). One item stated: “President Clinton’s response helped the government see that their past practices were wrong”. For the full scale see Appendix B. Participants responded to three items about trust in the U.S. government (α = .75), including: “I trust politicians and other prominent individuals to act fairly in the interest of those who were affected”, “The government will uphold agreed upon ethical standards of research in the future”, and “There is reason to believe the government has changed from its past behavior”. Participants completed a 30-item forgiveness scale about the current government adapted from Philpot and Hornsey (2008; α = .97; see Appendix C).
Results
For the descriptive statistics overall and by condition see Table 1. For correlations between the outcome measures see Table 2. To examine our hypotheses, we first conducted an Analysis of Variance (ANOVA) on each of the dependent measures followed by planned comparisons between each apology condition and the no apology condition. Then we conducted linear trend analyses on each outcome measure (Roberts & Russo, 1999, pp. 98-109). Assuming we considered the apologies as ordered intervals, a linear trend analysis determined if each incremental element added to the apology changed the scores on the relevant outcome variable (Roberts & Russo, 1999). This analysis essentially tested if the means on justice norms for each condition was fit well by a straight line that had a slope significantly different from zero. When we found a significant linear trend, we conducted a second test to ensure that the linear trend adequately represented the data. To do this, we conducted a test of the residual variance to be sure that after removing the variance associated with the linear trend the remaining variance only represented random error variance (Abelson & Prentice, 1997; Niedenthal, Brauer, Robin, & Innes-Ker, 2002). After the main effects analyses, we conducted a mediation analysis to test whether the effect of each apology (compared to no apology) on justice norms was mediated by perceived remorsefulness. For all analyses, we conducted case-wise deletion of missing data to maintain power.
Table 1
Descriptive Statistics
| Condition | Overall | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|---|
| N | 145 | 25 | 22 | 26 | 25 | 23 | 24 |
| Remorsefulness | |||||||
| M | 5.36 | 4.38 | 5.14 | 5.23 | 5.80 | 5.98 | 5.69 |
| SD | 1.32 | 1.67 | 1.20 | 1.26 | 0.90 | 0.85 | 1.23 |
| Justice Norms | |||||||
| M | 4.91 | 4.20 | 4.58 | 4.90 | 4.93 | 5.62 | 5.25 |
| SD | 1.19 | 1.35 | 0.86 | 1.16 | 1.11 | 1.06 | 1.08 |
| Trust | |||||||
| M | 3.88 | 3.50 | 3.82 | 4.14 | 3.60 | 4.25 | 3.97 |
| SD | 1.29 | 1.41 | 1.22 | 0.97 | 1.32 | 1.24 | 1.48 |
| Forgiveness | |||||||
| M | 4.28 | 4.14 | 4.19 | 4.32 | 4.00 | 4.49 | 4.48 |
| SD | 1.22 | 1.17 | 1.05 | 1.19 | 1.28 | 1.21 | 1.44 |
Note. Condition 1: No apology. Condition 2: statement expressing remorse. Condition 3: plus statement admitting injustice. Condition 4: plus statement accepting responsibility. Condition 5: plus statement acknowledging harm. Condition 6: plus statement promising forbearance.
Table 2
Correlations
| Justice Norms | Trust | Forgiveness | |
|---|---|---|---|
| Remorsefulness | .51*** | .07 | .05 |
| Justice Norms | .23** | .13 | |
| Trust | .54*** |
Note. N = 142, except for the correlation between remorsefulness and trust, N = 144.
**p < .01. ***p < .001, two-tailed tests.
Main Effects
Participants perceived differing levels of remorsefulness across the conditions, as indicated by a significant one-way ANOVA, F(5, 139) = 5.51, p < .001, ηp2 = .166. To test the hypothesis that an apology would lead to greater perceived remorsefulness than no apology, planned contrasts were conducted to compare each of the apology conditions to the no apology condition. African-American participants perceived the apology with a statement of remorse alone as more remorseful than no apology, F(1, 139) = 4.42, p = .037, d = 0.63.ii The apology that expressed remorse and admitted injustice was perceived as more remorseful than no apology, F(1, 139) = 6.09, p = .015, d = 0.70. When a statement that expressed responsibility was also added to the apology, participants perceived it as more remorseful compared to no apology, F(1, 139) = 16.65, p < .001, d = 1.18. The apology that acknowledged harm in addition to the other elements was perceived as more remorseful than no apology, F(1, 139) = 20.11, p < .001, d = 1.40. Compared to no apology, the apology with all five elements, including the promise of forbearance, was perceived by participants as more remorseful, F(1, 139) = 13.83, p < .001, d = 1.08. To test the apology elaboration hypothesis that each apology element increased perceived remorsefulness we conducted a linear trend analysis (Roberts & Russo, 1999). As predicted, as the apology became more elaborate, it was seen as more sincerely remorseful, F(1, 139) = 20.99, p < .001, ηp2 = .131. According to the method outlined by Abelson and Prentice (1997) and modified by Niedenthal et al. (2002), another F-test was conducted by dividing the difference in the sums of squares between the contrast and the total by the sums of squares error. No significant residual variance remained after conducting the linear trend analysis, F(1, 139) = 1.05, p = .308, ηp2 = .042, indicating that a linear trend adequately explained the data.
Our primary prediction concerned whether or not an apology would lead to an increased belief that justice norms had been affirmed. According to a one-way ANOVA, participants thought the degree to which the response from Clinton affirmed justice norms in society differed, F(5, 136) = 4.62, p = .001, ηp2 = .145. Planned contrasts were conducted to test the hypothesis that each apology condition would affirm justice more than the no apology condition. All of the apologies affirmed justice norms more than no apology, except for the simplest apology with only a statement of remorse, F(1, 136) = 1.28, p = .259, d = 0.35. This one exception is consistent with the notion that an apology with a statement of remorse alone is considered empty or as an expression of empathy rather than as an actual apology (Hornsey & Wohl, 2013). Participants thought that the apology that expressed remorse and admitted the commission of injustice affirmed justice norms more than no apology, F(1, 136) = 4.71, p = .032, d = 0.63. The apology that expressed remorse, admitted injustice, and accepted responsibility affirmed justice norms more compared to no apology, F(1, 136) = 5.16, p = .025, d = 0.65. When a statement that acknowledged harm was added to the apology it affirmed justice norms more than no apology, F(1, 136) = 18.93, p < .001, d = 1.29. Compared to no apology, the apology with all five elements, including the promise of forbearance, affirmed justice norms more, F(1, 136) = 10.52, p = .001, d = 0.95. Next we were interested if greater apology elaboration would lead to increased feelings that norms of just behavior had been affirmed. More elaborate apologies affirmed justice norms more than less elaborate apologies, as indicated by a significant linear trend analysis, F(1, 136) = 19.03, p < .001, ηp2 = .123. Per the method recommended by Abelson and Prentice (1997) and modified by Niedenthal et al. (2002), a second F-test revealed that there was not significant residual variance after conducting the linear trend analysis, F(1, 136) = 1.03, p = .312, ηp2 = .030, indicating that a linear trend adequately explained the data.
We hypothesized that an apology would lead to greater trust in the government. However, this hypothesis was not supported through the analysis of the direct effect. A one-way ANOVA revealed that the various apology conditions did not significantly differ from each other in the amount of trust expressed, F(5, 136) = 1.26, p = .285, ηp2 = .044. Moreover, the linear trend analysis did not reveal any increases in trust as the apology became more complex, F(1, 136) = 1.97, p = .163, ηp2 = .014.
Participants also did not report differences in the level of intergroup forgiveness because of President Clinton’s response, according to a one-way ANOVA, F(5, 138) = 0.57, p = .721, ηp2 = .020. Moreover, the linear trend analysis did not reveal any increases in forgiveness as the apology became more complex, F(1, 138) = 0.87, p = .353, ηp2 = .006.
Testing Mediation
We hypothesized that perceived remorsefulness would mediate the effect of each apology, compared to no apology, on the belief in justice norms. According to the method outlined by Hayes and Preacher (2014), we tested mediation with five dummy coded variables to represent the six conditions, using the no apology condition as the reference group. For the coding of the conditions see Table 3. For the mediation model see Figure 1 (PROCESS Model 4, Hayes, 2013). Using ordinary least squares regression, five separate analyses were conducted to estimate the linear model. In these five analyses, 95% confidence intervals (CI) to test the significance of mediation (of each apology condition compared to the no apology condition) were computed using bias-corrected bootstrapping with 5,000 resamples as recommended by Hayes (2013). Thus, the indirect effect (ab) through perceived remorsefulness and the direct effect (c’) of the apology on justice norms had to be interpreted in light of the coding of the dummy variables (which compared each apology vs. no apology).
Table 3
Dummy Coding
| Dummy Variables | D1 | D2 | D3 | D4 | D5 |
|---|---|---|---|---|---|
| Condition 1 | 0 | 0 | 0 | 0 | 0 |
| Condition 2 | 1 | 0 | 0 | 0 | 0 |
| Condition 3 | 0 | 1 | 0 | 0 | 0 |
| Condition 4 | 0 | 0 | 1 | 0 | 0 |
| Condition 5 | 0 | 0 | 0 | 1 | 0 |
| Condition 6 | 0 | 0 | 0 | 0 | 1 |
Note. Condition 1: No apology. Condition 2: statement expressing remorse. Condition 3: plus statement admitting injustice. Condition 4: plus statement accepting responsibility. Condition 5: plus statement acknowledging harm. Condition 6: plus statement promising forbearance.
Figure 1
Mediation model of each apology by perceived remorsefulness on justice norms.
Note. Condition 1: No apology. Condition 2: statement expressing remorse. Condition 3: plus statement admitting injustice. Condition 4: plus statement accepting responsibility. Condition 5: plus statement acknowledging harm. Condition 6: plus statement promising forbearance.
Consistent with our predictions about the simplest apology, mediation of the apology that expressed remorse alone (compared to no apology) on justice norms was not significant, abD1 = 0.30, 95% CI [-0.004, 0.70]. Also, consistent with our hypotheses, mediation of each of the more elaborate apologies compared to no apology was significant. Mediation of the apology with a statement of remorse and a statement admitting injustice was significant, abD2 = 0.37, 95% CI [0.06, 0.79]. Mediation of the apology with statements that expressed remorse, admitted injustice, and accepted responsibility was significant, abD3 = 0.55, 95% CI [0.24, 1.00]. Mediation of the apology with statements of remorse, injustice, responsibility, and acknowledgment of harm was significant, abD4 = 0.61, 95% CI [0.28, 1.08]. Mediation of the most elaborate apology that also included a statement promising forbearance was significant, abD5 = 0.51, 95% CI [0.19, 0.95].
Discussion
People are motivated to believe that others in their society (Wenzel et al., 2008) as well as governmental authorities (Jost & Banaji, 1994; Lerner, 1980) uphold values of just behavior, and can be trusted to treat them well. However, historical injustices threaten these beliefs (Starzyk, Blatz, & Ross, 2009). In this research, we examined whether an intergroup apology for one such historical injustice would help restore the belief that the government valued norms of just behavior in society (de Greiff, 2008), an effect previously only found at the interpersonal level (Okimoto & Wenzel, 2009). African-Americans read a shortened version of President Clinton’s apology for the Tuskegee Syphilis Study. We varied the degree of elaboration in each of five apology conditions. Participants then indicated how remorseful they found the statement, and whether the statement upheld norms of just behavior.
Compared to no apology, most apology conditions led to increased perceived remorsefulness and justice norms. The one exception to this was the apology with a statement expressing remorse alone. Compared to no apology, expressing remorse unsurprisingly led to more perceived remorsefulness, but it did not lead to a significant increase in justice norms. This finding is consistent with the past suggestions that a remorse-only apology would be ineffective because it is seen as shallow or incomplete (Hornsey & Wohl, 2013). However, all the other analyses testing the effects of apology elaboration revealed that as the apology became more thorough, people felt the apology was more sincerely remorseful and that the apology upheld norms of just behavior. As hypothesized, perceived remorsefulness significantly mediated the effect of the apologies with two or more elements (compared to no apology) on justice norms, meaning that the perceptions of Clinton’s intentions explained why participants thought that Clinton’s words addressed norms in the government and in society. Demonstrating this role of an intergroup apology is perhaps the most important contribution of the present work.
There was no direct effect of the apology manipulation on trust when looking at the one-way ANOVA or linear trend analysis. Neither did we find evidence that intergroup forgiveness differed based on whether participants read that President Clinton offered an apology or not in any analysis. Intergroup forgiveness also did not differ based on the degree of elaboration contained in Clinton’s apology. Past research shows that apologies elicit forgiveness in some circumstances but not in others (Blatz & Philpot, 2010), and there are critical moderators as to when intergroup apologies lead to forgiveness (Hornsey & Wohl, 2013). But in the present data the effects of an intergroup apology on forgiveness were not present. An apology can still have important benefits for the victimized group, and perhaps for society, under circumstances in which it does not elicit forgiveness (see also Blatz et al., 2014; Bombay et al., 2013).
We offer three possible explanations for the null effects on trust and on forgiveness. First, the apology did not include a direct statement of empowerment for the victims, which has been shown to enhance the effectiveness of reconciliatory statements (Shnabel & Nadler, 2008; Wenzel et al., 2008).iii Relatedly, if victim group members feel obligated to accept the apology because of expectation of the perpetrator or even objective third parties (Risen & Gilovich, 2007; Zaiser & Giner-Sorolla, 2013), their need for empowerment may be undermined. Second, the apology did not express the emotions that past work suggests make victim group members more forgiving (Wohl et al., 2012), and less wanting for revenge (Giner-Sorolla et al., 2008). If the apology more closely met the emotional needs and expectations of the victim group, it may have been more effective. Third, the historic harm to the African-American men in the study was a direct result of prejudice and discrimination. Race-based prejudice and discrimination is still a part of intergroup relations and systemic structures in the U.S. today (Hughey, 2011), and therefore, government discrimination might represent an intergroup harm with great personal relevance to the participants. As suggested by past research (Nadler & Liviatan, 2006) and theorizing (Hornsey & Wohl, 2013; Wohl et al., 2011), restorative justice efforts will likely be most effective when offered in the context of a positive intergroup relationship.
In addition to the null results for trust, this study had some other limitations. Some apology elements were impossible to separate from each other. First, in the text of Clinton’s actual apology, all of the statements that accepted responsibility also admitted injustice, meaning that we cannot make claims about how each of these elements operated independently. This was an inevitable limitation of using this actual apology that was offered as our stimulus material. Another limitation is that it may be that more complex apologies affirmed justice norms more because they were longer, rather than because they communicated additional apology elements. Once again, this limitation could not be avoided when using Clinton’s actual apology as we could not add additional elements without increasing the length.
Another limitation of the present study is that we did not include a manipulation check for each element of the apology (see Kirchhoff, 2013; Kirchhoff et al., 2012), and thus do not know for sure if people were specifically attending to these elements. Nevertheless, given that participants were randomly assigned to conditions, we do not have alternative explanations for the differences in how participants responded except for the additional statements included in the apologies. Even though number of words and additional elements cannot be separated, we believe participants attributed meaning to the increased elaboration of the apologies because they perceived more remorsefulness and affirmation of justice norms. Future research could test if any type of apology elaboration produces stronger effects, or if specific apology elements need to be added to do so.
Nevertheless, this study is the first we are aware of to demonstrate that norms of just behavior in society are partially restored following intergroup apology. It is also the first study we are aware of to examine elaboration of an actual government apology for a historical injustice against a minority group in society. Research on intergroup apology still remains a relatively under-explored area. We believe the present study contributes to a movement to create a more nuanced understanding of what effects intergroup apologies have and do not have, and the circumstances under which these apologies are more effective (Blatz & Philpot, 2010; Hornsey & Wohl, 2013; Wohl et al., 2011). This study complements recent research that examines how victim groups react to real intergroup apologies (Blatz et al., 2014; Bombay et al., 2013; Philpot & Hornsey, 2011; Wohl et al., 2013). Moreover, we believe more research on intergroup apologies should use actual content from apologies that have been offered. As such, we recruited a participant population whose group identity overlapped with the identities of the victims of an actual historic injustice, and we asked them to read an apology that had actually been given. In addition, we randomly assigned people to condition, and thus can draw stronger causal inferences. Although this ecologically valid experimental method posed some limitations, we believe it is one of this study’s greatest strengths.