How do I pardon thee?: The effects of relationship type, account type, and gender on offence-specific forgiveness

Abstract We employed a Bayesian analysis to compare offence-specific forgiveness in supportive versus ambivalent relationships. We also investigated offender accounts to assess their effect on forgiveness. Participants (283 total, 171 female) read a hypothetical scenario wherein an offender from a supportive or ambivalent relationship transgressed against them. The offender then offered a mitigating (i.e., concession or excuse) or an aggravating (i.e., justification or refusal) account for their behaviour. As predicted, an ambivalent offender received less forgiveness than a supportive offender, and mitigating accounts produced more forgiveness than aggravating accounts. These results suggest that the positive aspects of an ambivalent relationship are not substantial enough to negate the negative aspects of the relationship, which results in less forgiveness being offered to an offender, independent of the type of account offered for the offense.

that it brings both spiritual and emotional benefits that can transform one's life (Rye et al., 2000).The importance of forgiveness is also a central theme that pervades secular writings such as the poetic works of Pope (1711) and Victor Hugo's Les Misérables (1862Misérables ( /1938)).In fact, numerous religious and secular texts contend that forgiveness is an essential dynamic in human relations.Although some psychologists believe that the study of forgiveness should be left to religious scholars (for a discussion see McCullough et al., 2000), the psychological study of forgiveness has enjoyed a long and productive history in the discipline.The importance of the topic is reflected in a PsychINFO search for published articles in peer reviewed journals since 1965.This search reveals that more than 3,100 publications include forgiveness as part of their focus.

Forgiveness
Forgiveness has been conceptualized as a prosocial motivational change on the part of the offended individual (Tsang et al., 2006).After an offence occurs, many experience negative emotions that centre on avoidant or vengeful emotions or behaviours.Forgiveness occurs when these negative emotions and behaviours are replaced with conciliatory ones (Worthington, 2020).This is why it is distinctly different from pardoning, condoning, excusing, reconciling, or forgetting, all of which focus only on moving past the negative emotions and behaviours of an offensive act while ignoring the positive conciliatory ones (Körner et al., 2022;McCullough, 2001).
Forgiveness can be assessed at the generalized (or dispositional) level, the dyadic level, or an offence-specific level (McCullough, 2000).The first of these assesses a person's general disposition to forgive others, whereas the second measures forgiveness across multiple occasions within a single relationship.Offence-specific forgiveness assesses forgiveness toward a particular offender for a specific offence.Studies of offence-specific forgiveness typically induce one participant to offend against a second participant (e.g., Zheng & van Dijke, 2020) or present participants with hypothetical scenarios and ask them to report their willingness to forgive an offender for a specific misdeed (e.g., Furman et al., 2017).These studies suggest that forgiveness is more or less likely depending on a number of mediating and moderating factors, such as attachment style, personality (e.g., agreeableness, emotional stability), religiosity and spirituality, empathy towards the aggressor, rumination about the transgressor, and offender account type (Körner et al., 2022;McCullough, 2001;Schumann, 2014).This research provides important insight into factors that affect the ability to forgive in a wide variety of situations.
Forgiving another has been found to have a profound positive impact on relationship quality, mental health, and social well-being (Körner et al., 2022).Forgiving individuals experience fewer negative emotions (Akhtar & Barlow, 2016), greater positive emotions (Gao et al., 2022), higher levels of social support (Lawler-Row & Piferi, 2006), and greater prosocial intentions and behaviours towards larger groups of people (Karremans & Van Lange, 2008).This research thus supports assertions made in religious and secular literatures that forgiveness is an important process that contributes favourably to a person's overall personal and social welfare.
Given the benefits associated with forgiving an offender, research has explored several factors that affect a person's ability to forgive (see above).One clear finding is that the type of relationship that exists between the victim and the offender matters (McCullough et al., 2010).For example, people are more willing to forgive those with whom they have a secure, close, and supportive (i.e., positive) relationship than those with whom they have a negative relationship (e.g., Fincham et al., 2006).Moreover, relationship-specific attachment research shows that individuals in securely attached relationships are more forgiving of their relationship partner than those in insecure relationships (Körner et al., 2022).Considerable attention has been paid to attachment styles as one way of characterizing relationships between offenders and victims (Hirst et al., 2019), and a reasonable conclusion that may be drawn from the literature is that, generally, forgiveness is more likely in positive relationships than in negative relationships and can serve an important restorative function: it's easier to forgive one whom one likes than one whom one dislikes.
But what if the relationship with another person is not uniformly positive or negative as is often conceptualized?What if the offender is someone about whom a victim has decidedly mixed feelings, some positive and some negative?How easy is it to forgive someone with whom a person has a truly ambivalent relationship?If it is the case that people's social relationships fall neatly into positive and negative categories, then previous research maps onto a social reality that is adequately covered by the existing literature.If not, however, then there is a gap in the literature.Research and theorizing suggest that investigations of positive and negative relationships fail to consider a third relationship type that is much more common and important in people's social networks: ambivalent relationships (Uchino et al., 2004).

Ambivalent relationships and forgiveness
Relationships have frequently been categorized on a single positive/negative dimension.Uchino et al. (2004), however, have expanded the traditional positive/negative conceptualization in response to their findings that these dimensions are statistically independent.Thus, relationships can exhibit high levels of both positive and negative qualities.This conceptualization produces a new relationship category that is referred to as ambivalent relationships, whereas relationships that are characterized by high positivity and low negativity are referred to as supportive relationships.Uchino et al. found that, on average, people have 9.40 supportive and 9.39 ambivalent relationships in their social networks, as opposed to only 0.38 negative (i.e., aversive) relationships.Clearly, the number of ambivalent relationships vastly outnumbers negative ones for most people.Moreover, ambivalent relationships do not differ from positive ones in amount of relational contact or importance (Uno et al., 2002).It is not as if a person has an important group of supportive network ties and then a group of unimportant ambivalent ties: they are all important.
Not only do people have a high number of ambivalent relationships in their social networks, but the consequences of these relationships are substantial when compared to aversive network ties.Research shows that ambivalent relationships may lead to interpersonal stress above and beyond that found in aversive relationships (Birmingham et al., 2009;Uchino, 2004) and to increases in mental and emotional health outcomes that are conducive to many significant health problems including cardiovascular issues (Uchino et al., 2016).Yet the effects of these important and plentiful relationships on forgiveness have not been empirically investigated.The question yet to be answered is how forgiving a person can be toward an offender about whom they have equally positive and negative feelings.To the extent that forgiveness might improve not only the affect experienced in ambivalent relationships but might also improve the physical and emotional health of those involved, it is critical to investigate this question.

Account types
In addition to the type of relationship between two people, a second factor that affects forgiveness is the type of account an offender gives for a transgression.There is a substantial body of research that shows that when an offender offers an apology for a specific offence (i.e., says "I'm sorry"), forgiveness is more likely to occur (Green et al., 2008;Schumann, 2014), but psychological research on how the components of the apology affect forgiveness is scarce (Fehr & Gelfand, 2010).Accounts researchers contend that not all acknowledgements of an offence are equal in intent and effect.As people attempt to reconcile with an offended individual, they offer sentiments that incorporate different ways to account for their behaviours, and these sentiments may influence how an acknowledgement is received (Schmitt et al., 2004).For example, Schönbach (1990) has developed a taxonomy of four different account types that explain why every acknowledgement is not the same.He defines these as concessions, excuses, justifications, and refusals.These acknowledgements are proposed to reflect points on a continuum that represent accounts for offences that can range from mitigating to aggravating (McLaughlin et al., 1983).At one extreme, the offender concedes complete and total responsibility for the offence (offers a concession).At the other, the offender denies that an offence has occurred or that they are in any way responsible for it and refuses to acknowledge the offence (asserts a refusal; Gonzales et al., 1990).Between concessions and refusals are excuses (an admission of fault, but with mitigating circumstances) and justifications (an acceptance of responsibility, but for good reason).Accounts can thus be thought of as sub-types of a superordinate offence response.Concessions and excuses are considered mitigating types of responses, and justifications and refusals are considered aggravating types of responses.Mitigating accounts tend to focus upon the needs and concerns of the offended individual, whereas aggravating accounts are focused upon the offender.

The present study
In this study, we sought to understand if forgiveness is dependent on the more positive or negative aspects of a relationship, and because truly negative relationships are rare in people's social networks (Uchino et al., 2004), we focused on supportive and ambivalent relationships and did not investigate negative relationships.In addition, we examined a wider array of accounts for offensive behaviour (see Gonzales et al., 1994) to investigate their effects on offence-specific forgiveness.
Research shows that relationship ambivalence is a significant source of stress above and beyond that found in negative relationships, causing higher blood pressure, cardiovascular reactivity, and depression (Birmingham et al., 2009;Uchino et al., 2016).To the extent that stress in an ambivalent relationship exceeds that found in supportive and negative relationships because ambivalence is subjectively more aversive than both, we would expect it to be harder to forgive a person with whom one has an ambivalent, as opposed to a supportive, relationship.This leads us to propose the following hypothesis: H 1 : Victims will be less willing to forgive an offender who commits a specific offence and with whom they have an ambivalent relationship than an offender with whom they have a supportive relationship.
Although it may seem self-evident that one might be more willing to forgive when a concession or excuse (i.e., a mitigating account) is offered than when a justification or refusal (i.e., an aggravating account) is offered, there are circumstances when mitigating and aggravating accounts result in offence-specific forgiveness (see Folger & Cropanzano, 2001).Nevertheless, previous research shows that mitigating accounts are perceived to be more polite than aggravating accounts (Hocker & Wilmot, 1995), possibly because offenders assume more responsibility for a transgression when they offer a concession or an excuse than when they offer a justification or a refusal (Schönbach, 1990), thus focusing more on preserving the recipient's self-respect than the actor's self-respect (Hocker & Wilmot, 1995).This leads us to our second hypothesis: H 2 : Offence-specific forgiveness will be afforded with greater willingness to forgive in response to a mitigating account (i.e., a concession and an excuse) than an aggravating account (i.e., a justification and a refusal).
In addition to these hypotheses, we also explore the gender of the victim.Some meta-analyses suggest that females are more forgiving than males (Miller et al., 2008), whereas others suggest that there are no gender differences (Fehr & Gelfand, 2010).Interestingly, Miller et al. (2008) found that when forgiveness was defined as the abandoning of the desire for revenge (or vengeance), women were much more forgiving than are men.Given the lack of consistent findings regarding gender differences in forgiveness, we nevertheless wonder if females will desire less revenge for a specific offense than will males.This leads to the following exploratory question: R 1 : Will females desire less revenge from an offender than males?

Participants
Participants were recruited from undergraduate psychology classes at a large private university through classroom announcements and via an online recruitment management system.They were told that the study involved an examination of how people interact with each other and were invited to participate in various classrooms on campus.Two hundred eighty-three undergraduates (61% female) participated for extra course credit.The sample was 85% Caucasian, 4% multi-racial, 3% Hispanic, 3% Asian and 5% other ethnic identities.Ages ranged from 17 to 62, but most participants were in their early twenties (M = 21.40,SD = 4.51).The study was approved by the university Institutional Review Board.

Procedure and materials
Upon arriving at their assigned classrooms, participants were greeted by a researcher who obtained informed consent.Participants then began working through a series of questionnaires, among which was the Social Relationships Index (SRI; Campo et al., 2009), a widely used instrument designed to classify relationships within a person's social network as either supportive or ambivalent, and a hypothetical offence scenario that described another person's transgression perpetrated upon the participant and requiring the participant to indicate how forgiving they could be toward the offender in this offence-specific scenario.Because participants completed the SRI at the time of the experiment, the researcher retrieved it while participants completed filler questionnaires and scored it before they responded to the transgression scenario.Upon scoring the SRI, the researcher identified both supportive and ambivalent relationships within each participant's social network and then randomly assigned each participant to think about either a supportive (n = 139) or an ambivalent (n = 142) other when contemplating the scenario.This was accomplished by instructing the participants to write the initials of the person in several strategically located areas throughout the scenario to remind them of the perpetrator whom they had been assigned to think about.
After reading the offence scenario, participants completed an adjusted Transgression-Related Interpersonal Motivations Inventory (TRIM; McCullough et al., 1998), an instrument designed to measure a person's desire to engage in revenge toward and avoidance of someone who has offended them.In addition, participants answered questions regarding their willingness to forgive the supportive or ambivalent other, as well as various manipulation check questions.Upon completing all these measures, participants were debriefed and dismissed.

Social Relationships Index (SRI)
The SRI is a self-report version of the social support interview (Uchino et al., 2001) and is used to classify relationships as either supportive or ambivalent (see Campo et al., 2009 for psychometric information).Participants first list the initials of several individuals within their social network and then rate these individuals as to how helpful and upsetting they are on a scale from 1 (not at all) to 6 (extremely) when providing informational, emotional, and tangible support.Participants' relationships with network members are classified as supportive if the average score on all three types of support is above one for helpful and only one for upsetting.Their relationships are classified as ambivalent if their average ratings are above one on both the helpful and the upsetting dimensions.This method of classifying relationships has been employed successfully in several studies (Holt-Lunstad & Clark, 2014;Reblin et al., 2020;Uchino et al., 2012).Test-retest correlations for the SRI are r = 0.69 for supportive relationships and r = 0.51 for ambivalent relationships (Uchino et al., 2001), and prior research confirms that it is both reliable and valid (Campo et al., 2009).Consistent with previous research (Holt-Lunstad et al., 2007) participants rated same-sex others from their social networks.

Hypothetical offence scenario
We created a scenario in which participants imagined themselves in a situation where another person borrowed their class notes prior to an important exam.The person returned the notes much later than promised and in damaged condition with missing pages.This made it impossible for the participant to adequately prepare for the exam.The person then offered one of two accounts to represent the mitigating account dimension and one of two to represent the aggravating account dimension (see stimulus materials at Open Science Framework https://osf.io/84jru/?view_only= 56f4d66479c640fcb3e2a08fa38ca168).In the mitigating concession condition, the perpetrator apologized for the offence and took complete responsibility for his or her behaviour, offering no excuses.In the mitigating excuse condition, the perpetrator apologized and took responsibility for the offence while also providing mitigating circumstances that contributed to the incident.In the aggravating justification condition, the perpetrator apologized, but argued that the offence wasn't serious and that the participant shared responsibility for not waiting long enough to receive the notes.In the aggravating refusal condition, the perpetrator did not apologize and refused to accept fault or responsibility for the incident, arguing it was a professor's fault for delaying her or him and the participant's fault for not waiting long enough to get the notes.Previous research has validated thinking about a hypothetical offence scenario as a suitable substitute for experiencing or thinking about an actual offence (Green et al., 2008;Kang et al., 2011)

Adjusted Transgression-Related Interpersonal Motivations Inventory (TRIM; McCullough, 1998)
Forgiveness has been conceptualized as a prosocial motivational change whereby a victim becomes less vengeful and avoidant towards a transgressor (Tsang et al., 2006).The TRIM has been successfully used as an index of offence-specific forgiveness for actual transgressions.It consists of a 12-item selfreport measure that asks respondents to consider a person who has committed an offence against them and indicate their current thoughts and feelings about the person.Five items on the TRIM reflect a revenge motivation (e.g., I am going to get even) and seven items assess an avoidance motivation (e.g., I avoid him/her), and both are assessed using 5-point Likert scales (1 = strongly disagree, 5 = strongly agree; revenge subscale range 5-25, avoidance subscale range 7-35).The TRIM has been shown to have a high degree of internal consistency (Card, 2018).Both subscales are reliable ( ∝ s = .85)and demonstrate moderate test-retest stability (rs = approximately .50;Tsang et al., 2006) and convergent and discriminant validity (McCullough et al., 1998).
Because participants in this study were asked to think about a hypothetical situation rather than an autobiographical scenario, items on the TRIM were adjusted to reflect this difference.Specifically, the instructions were modified to ask participants to think of the hypothetical scenario instead of an actual transgression.In addition, the items were worded to reflect the hypothetical nature of the scenario instead of an actual event (e.g., I would want to get even; I would want to avoid him/her).Alphas were .87 for the revenge subscale and .91 for the avoidance subscale.The correlation between the two measures was r = .57.

Willingness to forgive
Whereas the TRIM assesses revenge and avoidance motivations as an index of offence-specific forgiveness, we sought to supplement this assessment with additional questions.Specifically, we wanted to know how much participants thought they could forgive the other person.Responses to three questions, How difficult would it be for you to forgive this person if this situation had really occurred?, How willing would you be to forgive this person if this situation had really occurred?, and How much do you think you could forgive this person if this situation had really occurred?, were combined to measure participants' willingness to forgive.The individual items for this index were rated on 5-point Likert-type scales from 1 (easily forgive) to 5 (could never forgive).The responses were reverse scored so that a higher score represented a greater willingness to forgive, and the index proved to be reliable (∝ = .81).The correlation between this measure and the desire for revenge was r = −.50 and r = −.56 for the desire to avoid.

Results
For the analysis, we performed a generalized linear regression model through Bayesian inference as opposed to the typical ANOVA approach.By working under the Bayesian paradigm, we could introduce more flexibility into the model, which is especially important when the dependent variables are truncated.We used Bayesian inference here due to the violation of assumptions to an ANOVA test, namely the nonnormality of the dependent variables.Further, Bayesian concepts tend to be more flexible and independent of specific stopping rules.The Bayesian theorem can also take both prior knowledge and the strength of the evidence for a hypothesis into account, which is theoretically closer to scientific reasoning than frequentist statistics (Morey et al., 2016).Additionally, we can combine information from previous sources, aiding the field in terms of reproducibility.Finally, the results from the Bayesian approach are easier to interpret.That is, the results can be understood without the convoluted confidence intervals and arbitrary p-values.

Hypothesis tests
Table 1 presents descriptive statistics for all dependent variables in the study.To investigate the effects of relationship type, account type, and participant gender on our dependent variables, we performed a Bayesian generalized linear model on our measures of revenge, avoidance, and willingness to forgive.The set of regressors included a dichotomous variable of account type, grouped as either aggravating or mitigating, gender, and relationship type.We also conducted an analysis including all four account types (concession, excuse, justification, and refusal) and found the dichotomous account type had better fit.Additionally, the four-account-types model showed a grouping of concession with excuse and justification with refusal.Therefore, only the model with the dichotomous account type is analysed.
Because each measure is in a bounded region, we transformed each score y to be y � ¼ yÀ a bÀ a where b is the maximum score and a is the minimum score given the measure.When the extreme scores (either a 0 or the maximum) were reached, the transformation for all scores was then =n as suggested by Smithson and Verkuilen (2006).After the transformation, we have where μ is the mean and ϕ is the precision parameter.Additionally, for the generalized linear model, we used the linear predictor η ¼ Xβ where β are the coefficients for the independent and control variables.Finally, we used logit link function g To finish specifying the Bayesian model, we put a normal prior on the β coefficients and an inverse gamma prior on the precision parameter ϕ.In particular, we had We then used the Metropolis-Hastings algorithm to sample from the joint posterior (Gelman et al., 2013).Both trace plots and the stationarity test based on Heidelberger and Welch (1983) indicated convergence of the simulation.

Relationship type
We hypothesized that participants would seek more revenge and would try harder to avoid an offender with whom they had an ambivalent relationship than an offender with whom they had a supportive relationship.We also hypothesized that they would be less willing to forgive an ambivalent offender than a supportive offender.Results from the generalized linear model revealed a significant main effect for supportive relationship type on revenge with an estimate of −0.79, indicating that the odds a participant felt revengeful toward an ambivalent offender were 54% greater than the odds of desiring revenge in a supportive relationship (see Table 2).There was also a main effect for supportive relationship type on avoidance with an estimate of −0.59 (and an odds ratio of 0.55), indicating that participants wished to avoid an ambivalent offender more.Finally, the effect for willingness to forgive was significant with an estimate of 0.38 and an odds ratio of 1.47, suggesting that participants were more likely to forgive a supportive than an ambivalent offender.Results thus confirmed that offenders with whom participants had an ambivalent social relationship fared poorly when compared to offenders with whom they had a supportive relationship (see Table 2).The positive aspects of the ambivalent relationship were not substantial enough to negate the negative aspects of the relationship, independent of the type of account offered for an offense.

Account type
We hypothesized that the type of account offered by an offender would affect the desire for revenge, avoidance, and the willingness to forgive.Specifically, we hypothesized that a mitigating account would elicit the most favourable response, followed by an aggravating account (see Table 2).The type of account was significant in all measures, with a mitigating account type having an odds ratio of 1.76 for the willingness to forgive, meaning the odds of forgiving following a mitigating account were 76% greater than the odds of forgiving following an aggravating account.For revenge and avoidance, the odds ratios are 0.79 and 0.72, respectively.In other words, the odds of desiring revenge after a mitigating account (i.e., a concession and an excuse) were 21% less than the odds for an aggravating account (i.e., a justification and a refusal).The odds of avoidance were 28% less for a mitigating account than for an aggravating account.

Gender
Finally, in answer to our research question we found that men desired more revenge in response to an offense than did women with an odds ratio of 1.59.That is, the odds that men desired revenge were 59% greater than the odds a female desired revenge.There were no gender differences, however, in the desire to avoid the perpetrator (odds ratio = 0.95), or in the willingness to forgive her or him (odds ratio = 1.13).

Discussion
The results of our study provide helpful insights for the study of interpersonal relationships.First, we investigated offence-specific forgiveness in a relationship type that, to our knowledge, has not been studied before, yet is pervasive and consequential in people's lives: the ambivalent network tie (Uno et al., 2002).As hypothesized, we found that victims were less forgiving of people with whom they had an ambivalent relationship than those with whom they had a supportive relationship.Even though an ambivalent offender was as helpful as a supportive offender in providing informational, emotional, and tangible support to the victim, this positive aspect of the relationship was neutralized by the ambivalent offender's ability to be very upsetting when providing support.
It appears that despite the positivity that is part of an ambivalent social tie, the negative and upsetting aspects of the relationship tempered victims' ability to forgive an interpersonal transgression.A positive-negative asymmetry effect found in social psychological research may provide a reason as to why supportive transgressors were forgiven more than ambivalent ones.This asymmetry effect contends that, in general, negative information is processed more thoroughly and has more impact than positive information on social judgments (Baumeister et al., 2001).The effect is particularly robust in impression formation contexts where demonstrations of the negative bias have a long history (Rozin & Royzman, 2001).In line with this research, it is possible that the negative aspects of an ambivalent relationship were simply too salient to ignore, making it more difficult to forgive a specific offence committed by an ambivalent other.Our findings mirror the results of research investigating offence-specific forgiveness in negative relationships (Fincham et al., 2006), but in this case the findings occurred in ambivalent relationships.Because ambivalent relationships are more common and important than negative relationships and just as common and important as supportive relationships (Uno et al., 2002), forgiveness or lack of forgiveness in response to specific offences has the potential to have a stronger impact on people's everyday lives.For example, a recent meta-analysis found strong positive relationships between forgiveness and individuals' subjective well-being and life satisfaction (Gao et al., 2022).Being less willing to forgive in ambivalent relationships where interaction with people is frequent may have considerable negative consequences.This finding thus enriches the offence-specific forgiveness literature.
With respect to our second hypothesis, the results were fully supportive.Participants attributed responsibility to the offender in a manner that matched the type of account being offered for the offensive act and reported a greater willingness to forgive and less desire to avoid or seek out revenge in response to a mitigating account than an aggravating account.This finding suggests that admitting fault for an offense, but offering mitigating information or conceding that an offence has occurred but denying personal responsibility, may be insufficient for obtaining forgiveness.To truly promote offence-specific forgiveness within a relationship, accounts for the offence need to be based on mitigating principles.That is, they need to be more concerned about the feelings and needs of the offended person than about the self (Fehr & Gelfand, 2010).In addition, offenders need to be willing to accept responsibility for and attempt to correct the consequences of offensive words or deeds.Ultimately, they need to be more concerned with meeting the needs of the hurt individual than with their own needs.Such accounts should result in a greater willingness to forgive the transgressor and less desire to avoid her or him or to seek revenge.
Our research question was answered as follows: Although men were no less willing than women to forgive an offender for a specific offence and were no more likely to want to avoid the person after the event, they desired significantly more revenge, a result that is consistent with previous research (Miller et al., 2008).According to gender role and social learning theories, men are socialized to defend their higher social status through seeking vengeance against offenders, whereas women are socialized to maintain harmony in social relationships through seeking to reconcile with offenders (Wilkowski et al., 2012).Our results suggest that assessing a person's willingness to forgive an offender may not always capture a reparative attitude completely, as men may be willing to forgive, but not to forget.Given the opportunity to retaliate for a real or perceived offence, men might be more likely than women to take a shot at the offender.

Limitations
One limitation in our research is that we did not examine how ambivalent relationships compare to negative ones with respect to offence-specific forgiveness.We did so because people rarely choose to associate with or do a favour for someone they do not like.Moreover, purely negative relationships are extremely rare in people's social networks (Uchino et al., 2004).This is evident in the finding that of the nearly 1,400 relationships that study participants reported on the SRI, only 25 were classified as negative.That is just under 2%.Nevertheless, it could be informative to compare ambivalent to negative relationships.
We also placed participants in a potentially highly artificial situation in which they had to imagine that a person with whom they had a supportive relationship committed the offence described in the scenario, particularly when the person offered an aggravating account for the behaviour.Random debriefings, however, revealed that although many found such a scenario difficult to imagine, it was not impossible.Indeed, many thought more deeply about the event because it was so novel.Although the use of hypothetical scenarios is typical in offence-specific forgiveness research (Green et al., 2008;Kang et al., 2011), our results would be enhanced by examining real-life events, perhaps carefully staged in the laboratory as was done in classic research by (Gonzales et al., 1990).
Finally, our results are limited in their generalizability by our reliance on a sample from the United States.Recent research has begun to investigate peoples' conceptions of forgiveness and their motives to forgive in different cultures.For example, research representative of this literature has found that a cognitive dimension of forgiveness that involves virtues that preserve social relationships is more frequently observed in Hong Kong Chinese culture (as reflected in its idioms) than in American culture (Ho & Worthington, 2020), and that subjects from a collectivist culture endorse relationship-and offender-focused motives for forgiving more than do subjects from an individualistic culture (Huwaë & Schaafsma, 2019).This research suggests that there are potentially differences in how forgiveness is conceptualized and endorsed in supportive and ambivalent relationships in different cultures, so future research investigating cross-cultural differences (and similarities) would add nuance, breadth, and depth to this literature.

Conclusion
Occasionally within every social relationship, words are said, or events transpire that result in one person being hurt or offended.This creates a seminal moment for the relationship.The hurt or offended individual must decide how they will respond.The choice of that individual may have long lasting implications with regards to trust, confidence, and ultimately the quality of the relationship.It may even determine if the relationship will endure.This study reaffirms that there are important relational and offence-related factors that influence that choice.Future research should continue to incorporate ambivalence into a taxonomy of relationship types to better understand social relations between people.A broader taxonomy of accounts, beyond simple apologies, should also be incorporated into offence-specific forgiveness research to further explore the richness of strategies people use to address offended others.

Table 2 . Results of Bayesian generalized linear model on measures of revenge, avoidance, and forgiveness
Notes: *indicates a 95% probability that the estimate is nonzero.95% CI refers to credible intervals.Relationship type compares supportive with ambivalent relationships, account type compares aggravating with mitigating accounts, and gender compares males with females.Exponential estimates to obtain odds.