Step by step: Testing the staircase model of intergroup apologies

Despite the assumption that intergroup apologies should lead to forgiveness in the same way that interpersonal ones do, research suggests that this is not the case. We report two studies exploring the application of the Staircase Model of Intergroup Apologies (Wohl et al.), in which an intergroup apology is embedded in a broader reconciliation process. Participants read accounts of an intergroup conflict (the “Troubles” in Northern Ireland) and subsequent efforts to achieve reconciliation, focusing on an official apology issued by the IRA. The content of this apology was varied such that the steps specified by the Staircase Model were presented either sequentially (Study 1) or in a manipulated order (Study 2). Both studies yielded results that were broadly supportive of the model. There were significant effects on measures of forgiveness, perceptions of the perpetrating group, and negative emotion (in particular disgust). The implications for enhancing the efficacy of intergroup apolo-gies are discussed.


| INTRODUC TI ON
There is a widespread assumption that intergroup apologiesapologies offered by one group to another group-provide an important basis for achieving reconciliation between these groups (e.g., Branscombe & Cronin, 2010;Tavuchis, 1991).This assumption is reflected in the marked increase in large-scale apologies being delivered by countries, political parties, businesses and corporations, leading to the suggestion that we have entered an "age of apology" (e.g., Barkan, 2000;Brooks, 1999;Gibney et al., 2008).Such apologies are offered for both historical and current transgressions in an effort to improve intergroup relations.However, there is limited empirical evidence to suggest that these apologies do improve such relations (Wohl et al., 2011).In an effort to identify the conditions under which intergroup apologies are most likely to be effective, Wohl et al. (2011) proposed the Staircase Model of intergroup reconciliation.In the present article we report two studies in which we sought to test predictions derived from that model.
An intergroup apology is one that is offered in a group-to-group context and differs from an interpersonal apology in that the latter is offered from one individual to another.Research shows that interpersonal apologies are often effective in eliciting forgiveness (e.g., McCullough et al., 1997;Riek & Mania, 2011), defined by McCullough and colleagues as a reduction in motives to retaliate against or move away from the perpetrator, and an increase in benevolent feelings towards the perpetrator.The effectiveness of interpersonal apologies in promoting forgiveness has led some to assume that intergroup apologies should have similar consequences, on the basis that forgiveness of a perpetrator group would also involve a decrease in negative thoughts and feelings and an increase in positive thoughts and feelings about the perpetrator.Indeed, Tavuchis (1991) argued that apologies should be seen as a panacea for repairing relationships, regardless of whether they are interpersonal or intergroup in nature.
Indeed, there is some evidence that intergroup apologies can be effective.Leonard et al. (2011) found that an apology offered to university students from a group of university professors who had written an article in a local newspaper criticizing student lifestyle was more effective in gaining forgiveness than not offering an apology.There are also findings from research conducted in the context of more violent conflict to suggest that an intergroup apology can reduce motivations for revenge and avoidance (Brown et al., 2008).
Furthermore, there is evidence that intergroup apologies can increase perceptions of perpetrator remorsefulness and that they usually leave the victim group feeling more satisfied (Philpot & Hornsey, 2008).
However, other findings suggest that recipients of an intergroup apology feel ambivalent, and that increased perceptions of perpetrator remorsefulness often fail to translate into forgiveness (Hornsey et al., 2015).In fact, there is surprisingly little evidence that intergroup apologies lead to true intergroup forgiveness.For example, Bombay et al. (2013) found that although intergroup apologies were regarded positively, victims were generally pessimistic about genuine improvements in intergroup relations.Thus intergroup apologies often seem to fail to achieve what they are presumably intended to achieve.
It is therefore important to understand what it is about intergroup apologies that stops them from achieving forgiveness.There have been suggestions that intergroup apologies cannot be trusted, perhaps because intergroup situations are characterized by greater competition and fear, leading to mistrust (Halabi et al., 2012;Insko et al., 1988).By definition, in the case of an interpersonal apology the recipient needs to put his or her trust in another individual; in the case of intergroup apology one or more persons have to trust many other individuals.This provides a relatively straightforward reason why intergroup apologies are less likely to succeed: It is more difficult to trust a group than it is to trust individuals (Balliet & Van Lange, 2013).A further point is that it is commonly thought that intergroup apologies are self-serving and insincere, and that there is no genuine concern for the victim group (Blatz et al., 2009); rather, the apology is offered to make the perpetrating group feel better about the situation.It follows that increasing the perceived trustworthiness and sincerity of an intergroup apology is likely to be pivotal to its success.
Potential reasons why intergroup apologies are problematic with respect to sincerity and trust arise not from the apology itself, but from the actions surrounding it.Perpetrator groups may believe (or be thought to believe) that the offering of an apology can "close the book" on the past, leaving the wrongdoing(s) forgotten (Corntassel & Holder, 2008).This creates the impression that there will be no further actions based on what is said in the apology.Such an impression would presumably lead to these apologies being regarded as untrustworthy and insincere.One way of alleviating this concern would be to make concrete promises about changes in behavior.The effectiveness of an apology based on promised behavioral changes should be enhanced by trust-building interactions between the perpetrator or victim groups before the actual apology is delivered.This assertion is supported by Nadler (2012, p. 294), who describes the outcome of positive intergroup reconciliation as "Trustworthy positive relations between former adversaries who enjoy secure social identities and interact in an equality-based social environment."Previous accounts of intergroup reconciliation have tended to treat it as an outcome to be sought, rather than as part of a process.
Treating reparative intergroup interactions as a multi-stage process, in which the apology is just one component, could help us to understand how impressions that intergroup apologies are self-serving and insincere can be avoided.Given the apparently low efficacy of intergroup apologies when taken in isolation, it makes sense to support them with other actions, both before and after the apology is delivered.

| The Staircase Model of Intergroup Apologies
The Staircase Model of Intergroup Apologies (Wohl et al., 2011) is a framework that seeks to identify the context in which intergroup apologies are effective.It sets out a series of steps (or "stairs"), starting with the perpetrating group's acceptance of collective guilt and its willingness to set history records straight, and is structured in such a way that each successive step should bring about an improvement in intergroup relations, thereby gaining enough momentum to proceed to the next step.Wohl and colleagues claim that each step provides a foundation for subsequent steps, creating genuine intergroup communication and trust-building.The five steps are as follows: accepting collective guilt, setting straight the records of history, discussing reparations, offering an intergroup apology, and post-apology engagement.To our knowledge, the model has not yet been tested empirically for its effectiveness in promoting reconciliation.Despite the original article suggesting that the Staircase Model is a framework that may not need to be tested, we argue that the model offers a novel and interesting process perspective on intergroup reconciliation.The present research was designed to provide a test of the model and to investigate the effectiveness of viewing an apology as a part of a broader reconciliatory process, as opposed to as a "stand-alone" tool to achieve an outcome.
Accepting collective guilt entails the perpetrator group accepting its responsibility for what has happened.As Wohl et al. (2011, p. 89) put it, "If collective guilt is not accepted, an apology is unlikely (or would be perceived as insincere), and so the group will remain on the ground floor."Setting straight the records of history allows the two groups to arrive at a shared interpretation of events, and also enables victims to be heard, understood, and validated.Wohl et al. (2011, p. 90) argue that "if a mutually agreed upon history is not achieved, there is risk that members of the perpetrator group will descend the staircase.At worst, collective guilt itself might be undermined."Discussing reparations marks the beginnings of repair and establishes a shared understanding of what resources are likely to be needed to put matters right.
In the words of Wohl et al. (2011, p. 91), "When consensus is reached on the historical record, the most appropriate means of repair should be discussed.This is because repair is dependent on a mutual understanding of what needs to be mended."Intergroup apology is the key communicative step; it involves the expressions of regret and provides a validation for the victims.It might seem surprising that the apology is the fourth step in this sequence, but Wohl et al.'s (2011) point is that an intergroup apology that is not built on the previous steps is likely to be regarded as insincere.
As they put it, "To be successful, a full intergroup apology must be offered.This encompasses an explicit expression of collective guilt.In doing so, the perpetrator should acknowledge the human dignity of the victimized group and the fact that this dignity was disregarded in the past.The perpetrator group should outline the mutually understood account of history that squarely places collective guilt on the shoulders of perpetrator groups" (p.92).Postapology engagement is designed to promote genuine reconciliation and harmony between the groups.The idea here is that promises of reparation are not the same as reparation; for genuine forgiveness, the perpetrators need to demonstrate through actions that they have changed.As Wohl et al. (2011, p. 94) argue, "the apology cannot be allowed to stand as the endgame for the perpetrator group.Following the apology, the aim must be the restoration and rebuilding of relationships in novel and context-sensitive ways that promote the needs of both groups."

| Overview of the studies
The aim of the studies reported below was to apply the Staircase Model to examples of historical conflicts.In both studies the Staircase Model was applied to The "Troubles" in Northern Ireland that took place between the 1970s and 1990s and involved sustained conflict between the Irish Republican Army (IRA) and the British government and army.The British (Study 1) or Northern Irish (Study 2) participants were persons who were adults at the time of the Troubles.They therefore had an involvement in the events being described in the sample.
Both studies used a variation of an additive procedure to introduce the different steps of the Staircase Model.In Step 1 participants learned that the perpetrating group recognized that they were to blame and accepted collective guilt.Step 2 described an agreement between the two groups about the documenting of the events, as well as members of both groups participating in the drawing up of this agreement.Step 3 provided information that the perpetrating group promised to disarm, promised compensation to those affected, and promised to ensure the safety of the victim group.Step 4 included an intergroup apology, in the form of a public statement, the structure and content of this apology being modeled on a real apology given by an IRA spokesperson (Cowan & Watt, 2002).Step 5 included details of reparations, such as compensation being delivered, and the sending of flowers and representatives to a service of remembrance.In both studies, there was also a control condition in which participants were simply told that there had been no contact between the perpetrating and victim groups.
As already noted, the Staircase Model was not used by its authors to arrive at testable predictions.Given this, and in the absence of previous empirical tests of the model, our predictions were derived from the structure of the model.The model's logic suggests that with each successive step in the model, there should be an increase in positive perceptions of the perpetrator group, and a decrease in negative emotions felt towards the perpetrator group.
It was predicted that the intergroup apology, because it is not presented in isolation, but rather is part of a broader reconciliatory process, would have a significant positive effect on forgiveness.
One way to test these predictions is to examine the significance of linear trends across the steps of the model.Such analyses are reported for Study 1 (the design of Study 2 did not lend itself to such analyses).However, given that each step involves different content, we also anticipated that the steps might have differential effects on outcome variables, such that some steps might have a greater influence on certain outcomes than those that precede or follow them.For example, feelings of anger towards the perpetrator group might not dissipate in a gradual way across the five steps, but rather in a stepwise fashion once the group has acknowledged its role in causing the suffering of the victim group by agreeing on what has taken place (Step 2), because this should attenuate perceptions of "other blame", a core relational theme (see Smith & Lazarus, 1993) of anger.By contrast, feelings of fear might reduce only when the perpetrator group starts to discuss reparations (Step 3), because this should lessen perceptions of threat, a core relational theme of fear (see Smith & Lazarus, 1993).Thus for both studies we also report Bonferroni-corrected comparisons between the different model step conditions.Below we report all measures, manipulations, and exclusions used in the studies.
Both studies followed the same general format, in which there are six conditions (one control condition, and five conditions relating to the Staircase Model).To estimate the required sample size, a power analysis was conducted.We based our estimates on the effect sizes reported in Brown et al. (2008).There, in Study 1 the effect size for the Apology manipulation on the Avoidance and Revenge sub scales of TRIM-18 ranged from .71 to 1.99.These are large or very large effect sizes on the same dependent variables as those used in the current studies.The nature of the offence (a friendly fire incident) in the Brown et al. study might have made the apology they used particularly effective, so we downscaled our expected effect size to medium-to-large.G*Power 3 (Faul et al., 2007) showed that to have an 80% chance of detecting a large effect size (F = .40)a sample size of 90 would be required (15 per cell), and that to have an 80% chance of detecting a medium effect size (F = .25)a sample size of 216 would be required (36 per cell).The cell sizes of the present studies ranged between 36 and 40 and were not increased after any data analysis took place.The studies reported here were approved by the relevant institutional ethics committee.Datasets for both studies, together with the Supplementary Materials, can be found at https://osf.io/4z539/?view_only=dfd38 26862 19420 48f0c 24035 063aaeb.

| S TUDY 1
Study 1 used the context of the Troubles in Northern Ireland, and more specifically the role played by the Irish Republican Army (IRA) in that conflict.Because participants in this study were adults living in mainland Britain, IRA attacks on Britain and on British identity were emphasized in the research materials describing the Troubles.Participants in this study had to be over the age of 35, meaning that they would have been at least 18 years old when the "Good Friday Agreement", which brought an end to the intergroup conflict in Northern Ireland was signed, in 1998.This measure was taken with the aim of ensuring that participants would have first-hand memories of at least some of the events referred to in the study.After reading a description of the Troubles, participants were told that an inquiry had established that the IRA is still in existence today, although its members maintain that they are committed to peaceful protest only.This was followed by the Staircase Model manipulation, in which participants were exposed to statements made by current IRA members about the group's past.

| Participants and design
Two hundred and thirty-five participants (115 males and 120 females; mean age of 52.32) completed this study.Participants were recruited via the research company Pureprofile (www.pureprofile.com).This enabled a sample of participants to be chosen based on age (>35 years) and location (mainland Britain).The study had a fully between-subjects design comprising six conditions, with participants randomly allocated to one of them.
Although 280 participants started the survey, some were excluded before finishing because they did not provide consent (n = 8), failed an attention check (n = 20), or simply did not finish the study (n = 17).The attention check was included to ensure that participants paid careful attention to information written in the transcripts they were given.The check itself consisted of a paragraph of text regarding the Troubles which included the following sentence: "This is an attention check, please skip this question and move on to the next page."

| Staircase model manipulation
Participants were randomly allocated to conditions that corresponded to a step number in the Staircase Model.The transcripts for these different steps are shown in the Supplementary Materials.

| Forgiveness
This construct was assessed using two measures.A binary measure consisted of the item, "After reading this, do you think the IRA should be forgiven?"with "Yes" or "No" response options.
A multi-item measure was an adapted version of the Trim-18 Scale (McCullough et al., 2006).Trim-18 was designed as an interpersonal forgiveness measure, and consists of three subscales (Avoidance, Revenge, and Benevolence); for the current research the items were adapted to be group related.Example items are "People who suffered from the Troubles should keep as much distance from the IRA as possible" (Avoidance); "People who suffered from the Troubles will get even one day" (Revenge), and ""The hatchet should be buried and people who suffered from the Troubles should move forward" (Benevolence).The full version of the adapted measure can be found in the Supplementary Materials.Responses to the items were made using 5-point response scales (1 = Strongly Disagree to 5 = Strongly Agree).The Cronbach's alphas for the three subscales were high: Avoidance (α = .89),Revenge (α = .86),and Benevolence (α = .91).

| Positive perceptions
Four single-item questions were used to assess positive perceptions of the perpetrator group.These items assessed the perceived sincerity, remorsefulness, trustworthiness, and believability of the group.
Responses were made using a 5-point response scale from "Strongly Disagree" to "Strongly Agree."

| Emotion measures
Single-item questions were used to assess how participants felt towards the perpetrator group after reading the transcript ("When I think about the IRA, I feel …").These items related to feelings of anger, fear, sadness, and disgust.Again, they were responded to using a 5-point scale from "Strongly Disagree" to "Strongly Agree."These questions were asked twice: once after the description of the IRA's actions during the Troubles, but before the manipulation; and a second time, after the condition manipulation.

| Demographics
To control for the possible influence of confounding variables, participants were asked their religion, their knowledge of the Troubles and the IRA, whether they had any Irish relatives, and whether they or their family had been affected by either the Troubles or the IRA, either directly or indirectly.

| Procedure
Participants were first given a brief description of the study and asked to sign an on-screen consent form.Next, they completed demographic measures.The structure of the main questionnaire was as follows.First came the description of the Troubles and role of the IRA, then the attention check, followed by the Staircase Model manipulation, the discrete measure of forgiveness, and then a random ordering of the Trim-18, positive perception items, and emotion items (with items within each set also presented in a random order).
Participants were then thanked and debriefed.

| Results
The association between condition and responses to the binary forgiveness measure was analyzed using chi-square.The effect of the manipulation on the Trim-18 measure, including its subscales, and positive perceptions was analyzed using a series of one-way ANOVAs.To protect against alpha inflation, Bonferroni-corrected post-hoc tests were used to follow up significant omnibus effects.
The emotion variables were analyzed with a repeated-measures ANOVA, with the Staircase condition as the between-subjects factor and time as the within-subjects factor.
There were no significant associations with demographic variables, including the variables regarding religion, knowledge of the Troubles, whether or not respondents had Irish relatives, and whether or not participants or their families had been affected by the Troubles or the IRA.

| Forgiveness
For the binary forgiveness item, the overall frequency of forgiveness rating (i.e., "yes" responses) was 47.66%.The overall chi-square analysis showed that there was a significant association between step of the Staircase Model and how people responded to this measure, χ 2 (5) = 12.35, p = .030.Table 1 shows a clear trend for forgiveness rates to increase with increasing step numbers, rising from 28.21% in Step 0 to 64.10% in Step 5.
Bonferroni-corrected post-hoc tests showed that avoidance scores in Step 0 were significantly higher than for all of the other steps.

| Emotion variables
These were analyzed in two ways.First, a series of 2 (time of measurement; within-subjects) × 6 (condition; between-subjects) ANOVAs assessed the effects of time and condition on each emotion.Second, a series of one-way ANOVAs assessed the influence of condition on the time 2 measurements.Means and standard deviations for each emotion variable can be seen in Table 3.

Anger
There was a significant main effect of time on anger, F( 1 steps, but no other differences were significant.

Fear
The main effect of time was not significant, F(1,229) < .01,p = .948,η p 2 < .01.There was also only a marginally significant main effect of Staircase condition, F(5,229) = 2.13, p = .063,η p 2 = .04,but the interaction was significant, F(5,229) = 12.72, p < .01,η p 2 = .22.Simple effects analysis showed that this interaction effect was driven by step 5, where fear was significantly reduced at time 2. Contrast analyses revealed that the linear component of the model was significant for fear at time 2, F(5, 229) = 6.37, p < .001,η p 2 = .12.Bonferronicorrected post-hoc tests showed that fear scores for Steps 3, 4, and 5 were significantly lower than for Step 0.

Disgust
There was a significant main effect of time,  participants saw the group as sincere, remorseful, trustworthy, and believable.

| Discussion
The aim of Study 1 was to show the positive effects of applying the Staircase Model in a context relevant to the participants.It was predicted that each step of the model would provide additive benefits, but also that particular steps might have differential effects on outcome variables.Significant effects of applying the model were observed for the binary forgiveness measure, perceptions of the perpetrator, and for emotions.There were also significant effects on the avoidance motivation subscale of the multi-item measure of forgiveness.
A key finding of this study is that applying the Staircase Model to a setting in which participants had a reasonably close relation to the historical events showed that the model had a positive impact on forgiveness rates.This provides good initial support for the model, especially given the general pessimism in the literature about the effectiveness of intergroup apologies and their ability to promote forgiveness following a conflict.The linear contrasts were significant for the majority of the outcome variables, consistent with the notion that there is an additive benefit of each step for the model.In addition, different variables were differentially affected at certain steps.Notably, ratings of fear reduced significantly at Steps 4 and 5, and the post-manipulation fear scores were lower at Steps 3, 4, and 5 than in the control condition.
Furthermore, disgust ratings reduced significantly at Step 4, and post-manipulation disgust scores were also lower at Steps 3, 4, and 5 than in the control condition.This suggests that the offering of an intergroup apology (Step 4) that is preceded by acknowledgement of guilt, setting straight the record of events, and discussion of compensation can reduce fear and disgust felt towards the perpetrator group.These results can also be regarded as support for the stepwise structure of the model.
Although these findings are consistent with the view that proceeding through the steps of the model will have a beneficial effect on intergroup perceptions, and thereby promote reconciliation, the evidence does not unequivocally support the sequencing of the steps as proposed in the model.It may be that the steps themselves have independent effects and that presenting them in a different sequence would result in the same effect, which would call into question the "staircase" notion that is inherent to the model.This possibility was tested in Study 2.

| S TUDY 2
Again, the context of the Troubles and the IRA was chosen, but this time there was an attempt to make the involved status of the participants even more salient than was the case in Study 1.This was achieved by recruiting participants from Northern Ireland.This should result in still greater personal relevance of the transgression and the subsequent reconciliation efforts.This change necessitated some changes in the materials used in Study 2. Rather than the intergroup transgression being "IRA atrocities in Britain", in Study 2 the intergroup transgression was simply "IRA atrocities", intended to include IRA actions in Northern Ireland, and the British government was described as acknowledging its role and shared responsibility for the events of the Troubles.
With regard to the Staircase Model, we focused on step 4, the intergroup apology, and varied where in the sequence of steps this step was located.This was again done in a cumulative manner, with the apology appearing (a) on its own, (b) after step 1, (c) after steps 1 and 2, or (d) after steps 1, 2, and 3-the last of these being its "correct" location in the Staircase Model.In addition, a control condition was included, in which no apology was offered, and an "alternative model" condition was also added, in which all of the information preceding the apology remained the same but was presented in the reverse order (i.e., discussing reparations, setting history records straight, accepting guilt, intergroup apology).The idea here was to test the importance of the sequencing of the steps within the model.If the sequence proposed by the Staircase Model is important, perceptions of the perpetrator group, emotions felt towards this group, and willingness to forgive the group should all be greater when the apology follows step 3, by comparison with the alternative, reordered condition.
An important finding in Study 1 was that disgust felt towards the perpetrator group only reduced after the intergroup apology was offered.To explore this further, there was an attempt to distinguish between different aspects of disgust.It has been argued that disgust has two facets, one more physical in nature and associated with avoidance of physical contamination, the other being socio-moral in nature and associated with avoidance of social or cultural contamination (Russell & Giner-Sorolla, 2013;Tyber et al., 2009).We aimed to investigate whether the intergroup apology would affect measures of these different facets of disgust in different ways.We also wanted to examine whether feelings of disgust would be reduced by the intergroup apology, regardless of the context in which it is given, or whether this effect depended on other steps in the model.

Staircase model manipulation
The location of the intergroup apology in the Staircase Model was manipulated.In condition 1, participants only read the intergroup apology; in condition 2 they read the first step of the Staircase Model, followed by the apology; in condition 3 they read the first two steps of the model, followed by the apology; in condition 4, the apology followed steps 1, 2, and 3, as in the original model.The fifth condition was a reordering of the model, with the intergroup apology still being the final step, but the preceding steps were 3, 2, 1 (discussing reparations, documenting history, accepting collective guilt).There was also a control condition (condition 0) in which participants were told there had not been any contact from the perpetrator group.The transcripts relating to the different conditions can be found in the Supplementary Materials.
It is worth noting that the apology used in the original model included dates relating to some of the individual steps, e.g."In late 2005" in step 1, and "In early 2006" in step 3. When reordering the apology for condition 5 (the alternative sequence), this chronology was retained, such that the content of step 1 always referred to an earlier date than the content of subsequent steps.Similarly, where there was a qualifying introduction to a step, such as "in addition to accepting guilt", the wording was changed to ensure that it was consistent with the preceding step(s).

Forgiveness
The forgiveness measures were the same as those used the previous study.Thus, there was one binary measure, reading "After reading this, do you think the IRA should be forgiven?"with "Yes" or "No" response options; and the adapted multi-item Trim-18 measure (McCullough et al., 2006).

Positive perceptions
The single-item questions and response scales assessing perceptions of the perpetrating group were the same as those used Study 1.

Disgust measures
To examine more closely the effect of intergroup apology on reducing disgust, observed in Study 1, three items were used to measure physical disgust ("When I think about the IRA … I feel physically sick," "… I feel my stomach turning," and "… my stomach is quivering;" α = .80)and another three items to measure socio-moral disgust ("The IRA are bad people," "The IRA are morally wrong," and "The IRA are evil;" α = .84).

Demographics
To control for the influence of possible confounding variables, participants were asked to report their religious affiliation, political views, their knowledge of the Troubles and the IRA, and whether they or their family had been affected by either the Troubles or the IRA, either directly or indirectly.

| Procedure
Participants were first given a brief description of the study and asked to sign an on-screen consent form.Next, they completed the demographic measures.The main questionnaire began with a brief description of the Troubles and role of the IRA.Then followed the attention check, the Staircase Model manipulation, the discrete measure of forgiveness, and a random ordering of the Trim-18, positive perception questions, and disgust questions (with items within each set also presented in a random order).Participants were then thanked and debriefed.

| Results
Unsurprisingly, political views were strongly correlated with all the main dependent variables.The more that participants considered themselves to be "unionists/loyalists," the less likely they were to have positive scores on all dependent variables.Because of this, political views were controlled for in all subsequent analyses.This was done by including the political views variable as a covariate in each analysis (apart from the one involving the binary forgiveness measure, where an alternative strategy was adopted); means and standard deviations reported below are adjusted for the influence of the covariate.There were no significant associations with any other demographic variable, including variables assessing religious affiliation, knowledge of the Troubles, and whether participants or their families had been directly or indirectly affected by the Troubles or the IRA.

| Forgiveness
For the binary forgiveness item, the overall frequency of forgiveness rating (i.e., "yes" responses) was 44.14%.The overall chisquare analysis showed that there was an association between condition and responses to this question, χ 2 (5) = 23.02,p < .001.
Table 4 shows a clear trend for forgiveness rates to increase as the placement of the intergroup apology moves to a successively later step in the sequence, from 22.22% in the control condition to 65.79% in condition 4, the sequence matching the staircase model, before dropping to 32.43% in the alternative sequence condition.To account for the influence of political views, subsequent analyses were conducted with the participants split, with those scoring 1-5 defined as "Nationalist/Republican" (n = 113) and those scoring 6-10 defined as "Unionist/Loyalist" (n = 109).
The percentages of "yes" answers for these two groups are also shown in Table 4.The chi-squared analysis for those coded as "Nationalist/Republican" showed that there was a significant association between condition and responses to this question, χ 2 (5) = 13.96,p = .016.Here we see a much larger percentage of forgiveness score in condition 4 (80.95%),compared with both the control condition (28.57%) and the alternative sequence condition (50.00%).The chi-squared analysis for those coded as "Unionist/Loyalist" showed that there the association between condition and responses to this question did not reach the conventional significance threshold, χ 2 (5) = 10.19,p = .070.Here we see a larger percentage of forgiveness score in conditions 2, 3, and 4 (all 47.06%), compared with the control condition (13.33%), condition 1 (22.72%) and the alternative sequence condition (19.05%).
The means and standard deviations for the three Trim-18 subscales at each step are shown in Table 5. Condition had a significant main effect on the avoidance motivation subscale, F(5, 215) = 5.69, p < .001,η p 2 = .12.Bonferroni-corrected post-hoc tests showed that the control condition and the condition consisting of only the intergroup apology led to significantly higher avoidance scores than all other conditions, apart from the alternative sequence condition.
There was also a significant main effect on the revenge subscale, F(5, 215) = 2.25, p = .050,η p 2 = .05,although Bonferroni-corrected post-hoc tests showed that none of the conditions differed significantly from each other.Finally, there was a significant main effect on the benevolence subscale, F(5, 215) = 4.84, p < .001,η p 2 = .10,with Bonferroni-corrected post-hoc tests showing that the control condition elicited significantly lower benevolence scores than any of the other conditions.The means and standard deviations for the positive perception measures are shown in Table 5.There was a significant main effect of condition on the positive perceptions of the perpetrator group, F( 5

| Disgust
The measures of the two facets of disgust were positively and significantly correlated (r = .47,p < .001).Both facets were also strongly correlated with all of the other dependent variables (all ps < .001).Although the strength of the correlations of the two disgust measures with many of the other dependent variables is similar, the socio-moral disgust measure is much more strongly related to the revenge motivation subscale = .61)than is the physical disgust measure (r = .34).Means and standard deviations for the two disgust measures are shown in Table 6.

Socio-moral disgust
There was a significant main effect of condition, F(5, 215) = 6.62, p < .001,η p 2 = .13.Bonferroni-corrected post-hoc tests showed that all conditions that included an apology led to a significantly lower score than the control condition.
As was the case in Study 1, correlation analyses (details are reported in the Supplementary Materials) revealed strong and significant associations between positive perceptions, Trim-18 scores, and disgust ratings, consistent with the notion that forgiveness and feeling less disgust towards the perpetrator group were shaped by the extent to which participants saw the group as sincere, remorseful, trustworthy, and believable.

| Discussion
The aims of this study were to provide a more stringent test of the staircase model and to investigate more closely the impact of intergroup apology on forgiveness, perceptions of the perpetrator group, and disgust.For the specific sequencing of the Staircase Model to be supported, the sequence proposed in the model should have had more positive effects than the alternative sequence on measures of forgiveness, positive perceptions of the perpetrator group, and disgust.This was shown to be the case for the binary measure of forgiveness, with more than two-thirds of participants in the proposed sequence condition responding "yes," compared to fewer than one-third of participants in the alternative sequence condition.The proposed sequence condition also led to more positive perceptions of the perpetrator group than did the alternative sequence condition-although here the difference fell just short of statistical significance-and to lower scores on the physical disgust measure.These findings provide good support for the sequencing of steps proposed in the Staircase Model.Also supportive of the model are the findings that the theoretically proposed sequence differed from the control condition in eliciting more positive perceptions of the perpetrator group and lower physical disgust.
Turning to the specific impact of intergroup apology, particularly interesting findings from this study relate to the measures of avoidance motivation and disgust.It is striking that the offering of the apology alone did not reduce avoidance motivation, but that when the intergroup apology was set in a broader context-even if this context simply meant that the apology followed the first step of the model-it did reduce avoidance motivation, except in the alternative sequence condition.This highlights the point that simply offering an intergroup apology is unlikely to be effective in improving intergroup relations.
Distinguishing between socio-moral and physical disgust also led to some interesting findings with regard to the impact of the intergroup apology.Although it was found that all conditions that included an apology resulted in lower socio-moral disgust scores than did the control condition, it was only the theoretically proposed sequence, with the apology following three previous steps in the model, that led to reduced physical disgust, relative to the control condition, the apology-alone condition, and the alternative sequence condition.This shows that offering an intergroup apology may be sufficient to reduce socio-moral disgust felt towards the perpetrator group, but that the apology is only effective in reducing feelings of physical disgust when it is presented at the point proposed in the Staircase Model.
It should be noted that while Study 2 offers support for the proposed Staircase Model sequence over an alternative sequence, the results observed for the theoretically proposed sequence did not differ significantly from those from sequences that remained intact but with one or two steps missing (i.e., conditions 2 and 3).
This suggests that as long as the steps remain in the sequence of the proposed model, not all of them may be needed in order to improve intergroup relations.To account for the fact that presenting the steps in the "wrong" sequence is less effective, it is worth remembering that the different steps address different issues.For example, accepting guilt acknowledges responsibility for wrongdoing, thereby reducing anger, which then enables thoughts about the events to be discussed, thereby enabling the parties to arrive at an agreed account of what happened, which in turn facilitates the discussion of reparations.This is consistent with the reasoning underlying the model.In the words of the models" authors, "Lower floors are foundations or preconditions to move to the next floor" (Wohl et al., 2011, p. 88).
It is also worth noting the effects of the different sequences on physical disgust.The full Staircase Model sequence led to significantly lower ratings of physical disgust than did the control condition, the apology-only condition, and the alternative sequence.
So, although it might be possible to remove certain steps from the Staircase Model and still achieve improved intergroup relations, we conclude that the proposed full Staircase Model sequence represents the most consistently effective way of reducing feelings of disgust towards the perpetrator group.

| G ENER AL D ISCUSS I ON
The aim of this research was to provide an initial test of the Staircase Model by applying it to real world intergroup conflict settings.It was hypothesized that exposing participants to the steps proposed by the model, and in the sequence proposed by the model, would result in increased forgiveness, more positive perceptions of the perpetrator group, and reduced negative emotions felt towards the perpetrator group.It was also predicted that these effects would vary as a function of the number of steps to which participants were exposed, and that outcome variables would be differentially affected by the different steps.
Across the two studies, a good level of support for each of these hypotheses was found.There was evidence in both studies that being exposed to all steps of the model, and in the sequence proposed by the model, led to increased forgiveness, more positive perceptions of the perpetrator group, and reduced negative emotion felt towards the perpetrator group.
To assess forgiveness, we used both a binary measure and an adapted version of the Trim-18 forgiveness scale.Evidence of an impact on this measure was found in both studies, with avoidance subscale scores being significantly reduced, and lower scores on the vengeance subscale and higher scores on the benevolence subscale also observed in Study 2, where the participants were drawn from the population most directly affected by the intergroup conflict.This suggests that changes in motivation as captured by the Trim-18 measure can be influenced by applying the Staircase Model.
The increased forgiveness scores were accompanied by more positive perceptions of the perpetrator group, including more positive ratings on items measuring trust and sincerity.If intergroup apologies fail because they are seen as insincere, and the groups offering them are regarded as untrustworthy, here is evidence suggesting that proceeding through the steps of the Staircase Model can reduce perceptions of insincerity and untrustworthiness.
Exposing participants to the steps in the Staircase Model also had consistent effects on participants" self-reported emotions.In Study 1, anger was reduced at each step of the model, with all steps resulting in significantly lower scores than the control condition.This is consistent with previous evidence that intergroup apologies reduce anger (Leonard et al., 2011;Maitner et al., 2006).Fear decreased significantly after step 5 in Study 1.
Step 5 entails concrete behaviors undertaken by the perpetrator group.As suggested earlier, it may be that the appraisal of threat that is the hallmark of fear is only attenuated by the promise of concrete actions undertaken by the perpetrator group.Disgust was the only emotion that reduced significantly after the intergroup apology, step 4 of the model.This was the case both in Study 1, where the focus was on interactions between condition and time of measurement, reflecting changes in emotion as a result of the steps taken; and in Study 2, where two types of disgust were examined.Although the bodily mutilation that resulted from some IRA actions could be seen as the basis for disgust felt towards the group, it seems more likely that these feelings of disgust stemmed from the appraisal that the group violated social norms of peaceful coexistence.The offering of an intergroup apology preceded by steps that increased its perceived sincerity appears to have reassured participants that the group will abide by these norms and thereby reduced the disgust felt towards norm violators.
The results from Study 2 shed further light on the effect of intergroup apology on feelings of disgust.Consistent with the findings of Study 1, offering an intergroup apology in isolation led to lower socio-moral disgust.However, when the apology was embedded in the broader context of the model, it also reduced physical disgust.
This suggests that the function of an apology in intergroup contexts differs from its function in interpersonal contexts.In the latter case, it is generally assumed that expressions of apology trigger forgiveness because the apology recipient accepts that the harm done was unintentional and that it will not recur.In an intergroup context, feelings of disgust towards another group are known to be related to a tendency to engage in dehumanization, and intergroup disgust predicts prejudice and discrimination (Hodson et al., 2014).There is also evidence that brain areas associated with disgust reactions are activated when individuals view targets who are considered to be "less than human" (Harris & Fiske, 2006).This suggests that reducing disgust felt towards another group should diminish any tendency to dehumanize its members.If intergroup apologies are effective in specifically reducing disgust, whether socio-moral or physical, this suggests that they can play a key role in achieving intergroup reconciliation.This also highlights the value of the Staircase Model: By decomposing the reconciliation process into different steps, it helps to identify the specific effects of each step in that process.The fact that both disgust and fear were only reduced after certain steps suggests that particular kinds of information have specific effects on the reconciliation process.
Some limitations of the present research should be acknowledged.
Both studies were scenario studies in which participants responded to information about historical transgressions.It could be argued that their responses may not reflect how those who were more directly involved in the conflicts in question would have reacted.Although it is clearly important for future research to address this issue by studying ,229) = 50.23,p < .001,η p 2 = .18,with anger scores at time 2 (M = 3.32) being significantly lower than those at time 1 (M = 3.68).The main effect of Staircase condition on anger was close to significance, F(5,229) = 2.22, p = .053,η p 2 = .05.There was also a significant interaction, F(5,229) = 12.02, p < .001,η p 2 = .21.Simple effects analysis showed that this interaction was driven by opposing patterns in the different Staircase conditions.At Step 0, anger increased at time 2, but for all other steps, anger reduced at time 2. Contrast analyses revealed that the linear component of the model was significant for anger at time 2, F(5, 229) = 7.37, p < .001,η p 2 = .14,and Bonferronicorrected comparisons showed that Step 0 differed from all other (1,229) = 14.04, p < .001,η p 2 = .06,with disgust at time 2 (M = 3.54) being significantly lower than at time 1 (M = 3.75).There was also a significant main effect of Staircase condition, F(5,229) = 2.86, p = .016,η p 2 = .06,which Bonferroni-corrected post-hoc tests showed was due to disgust at step 5 being significantly lower than after step 0. The interaction was also significant, F(5,229) = 8.33, p < .001,η p 2 = .15.Simple effects analysis showed that this was driven by steps 0, 4, and 5.At step 0 disgust scores increased at time 2, whereas at steps 4 and 5 disgust scores decreased significantly at time 2. Contrast analyses revealed that the linear component of the model was significant for disgust at time 2, F(5, 229) = 7.16, p < .001,η p 2 = .14.Bonferroni-corrected post-hoc tests showed that time 2 disgust scores for Steps 3, 4, and 5 were significantly lower than for Step 0. Disgust scores were also significantly lower for Step 5 than for Step 1. Correlation analyses (details are reported in the Supplementary Materials) revealed strong and significant associations between positive perceptions, Trim-18 scores, and emotion ratings, consistent with the notion that forgiveness and feeling less negatively towards the perpetrator group were shaped by the extent to which TA B L E 3 Mean emotion ratings (with standard deviations in parentheses) at each step of the Staircase Model at Time 1 and Time 2 (Study 1) Two hundred and twenty-two participants (110 males and 112 females; mean age of 50.49) completed this study.Participants were recruited via Pureprofile (www.pureprofile.com).Participants were all residents of Northern Ireland.Like Study 1, this study had a fully between-subjects design comprising six conditions, with participants randomly allocated to one of them.Although 260 participants started the study, some were excluded before finishing because they did not provide consent (n = 12), failed an attention check (n = 22), or simply did not finish the study (n = 4).The attention check used was the same as the one used in Study 1.