Amends, Forgiveness, and the Resolution of Betrayal

In the wake of betrayal: Amends, forgiveness, and the resolution of betrayal

a b c PEGGY A. HANNON, CARYL E. RUSBULT, ELI J. FINKEL, AND MADOKA KAMASHIROd aUniversity of Washington; bVrije Universiteit, Amsterdam; cNorthwestern University; d Goldsmiths, University of London, United Kingdom

Abstract The present work advances a dyadic model of victim and perpetrator interactions following betrayals, and the effect of their interactions on betrayal resolution and relationship quality. The authors propose that perpetrator amends promotes victim forgiveness and that both amends and forgiveness contribute to betrayal resolution. In Study 1, married couples discussed unresolved betrayal incidents, and their behavior was rated by partners and trained observers. In Study 2, dating individuals used interaction records to describe betrayal incidents perpetrated by themselves or the partner over a 2-week period. In Study 3, dating partners both provided retrospective descriptions of prior betrayal incidents. All studies yielded good support for model predictions, revealing parallel ﬁndings from the point of view of victims, perpetrators, and external observers.

Close relationships are the origin of many most relationships, one partner will betray the positive experiences. We feel exhilarated other, violating relationship-relevant norms by when we develop new interdependencies, cel- engaging in acts of deception or disloyalty. ebrate positive interpersonal events in our How do we manage to sustain relationships own and others’ lives, and are deeply grat- inwhichwesuffersuchharm,andwhatare ified by relationships that are healthy, vital, the consequences of forgiving versus failing and enduring. Paradoxically, it is in roman- to forgive a partner’s act of betrayal? tic relationships that we also find ourselves The past two decades have witnessed a most vulnerable to suffering. At some point in dramatic increase in the study of forgiveness. Early research in this tradition was largely victim centered, examining the personality traits Peggy A. Hannon, Department of Health Services, Uni- that are associated with forgiveness, the cog- versity of Washington; Caryl E. Rusbult, Department of Social Psychology, Vrije Universiteit, Amsterdam, The nitive and affective events that predict for- Netherlands; Eli J. Finkel, Department of Psychology, giveness, and the personal values that promote Northwestern University; Madoka Kamashiro, Depart- forgiveness (for a review, see Worthington, ment of Psychology, Goldsmiths, University of London, London, United Kingdom. 2005). In recent years, scientists have turned This research was supported by a grant to the sec- their attention to interpersonal elements of ond author from the Templeton Foundation (Grant 5158). the forgiveness process, examining proper- Manuscript preparation was facilitated by National Can- cer Institute Training Grant 5R25CA092408. Study 3 was ties of partners and relationships that make part of Peggy A. Hannon’s doctoral dissertation, super- forgiveness more versus less probable. The vised by Caryl E. Rusbult. We would like to thank the present work contributes to this emerging lit- Marriage Project Team at the University of North Car- olina at Chapel Hill for all of their contributions to the erature in two respects. First, we advance a studies presented in this article. dyadic model of forgiveness in the context Correspondence should be addressed to Peggy A. of close relationships, exploring the interplay Hannon, University of Washington, Department of Health Services, 1107 NE 45th Street, Suite 200, Seattle, WA between perpetrator amends and victim for- 98105, e-mail: [email protected]. giveness and the associations of each variable 253 254 P. A. Hannon et al. with betrayal resolution and its relational becomes decreasingly motivated to retaliate consequences. Second, we test our model by against an [offender], decreasingly motivated examining both victim and perpetrator per- to maintain estrangement from the offender, spectives in real couples confronted with real and increasingly motivated by conciliation betrayal incidents. Few studies have examined and goodwill for the offender” (McCullough, perpetrators’ perceptions of the forgiveness Worthington, & Rachal, 1997, pp. 321–322). process, and fewer still have explored per- Our definition is consistent with this tradi- petrator perspectives in the context of actual tion yet represents forgiveness as both a psy- betrayals (as opposed to hypothetical betray- chological and behavioral event. We define als or stranger betrayals). forgiveness as the victim’s willingness to Our dyadic model rests on the principles of (a) forego vengeance and demands for ret- interdependence theory (Kelley et al., 2003; ribution and (b) react to the betrayal in a Rusbult & Van Lange, 2003) and suggests that constructive, less judgmental manner (Finkel perpetrator amends enhances victim motiva- et al., 2002). tion to forgive. Moreover, we propose that Given that victims’ immediate, gut-level both perpetrator amends and victim forgive- impulses frequently include desire for grudge ness play key roles in the successful reso- or vengeance, how do victims find their lution of betrayal incidents and that betrayal way to forgiveness? Interdependence the- resolution is beneficial to relationships from ory describes immediate, gut-level reactions the perspective of both victims and perpe- as given preferences, in that they are self- trators. As initial tests of our model, we oriented, asocial, and focus on the here report the results of three studies of betrayal, and now (Kelley et al., 2003). People depart amends, and forgiveness in ongoing roman- from self-oriented, given preferences as a tic relationships. The studies employ diverse result of transformation of motivation,a methods, including the observation of couple psychological process whereby victims take interactions regarding unresolved betrayals, into account considerations extending beyond interaction diary reports of relational betray- direct self-interest, including long-term goals, als during a 2-week period, and retrospective social dispositions and values, or concern for accounts of betrayals. a partner’s well-being. The modified preferences resulting from transformation are Perpetrator betrayal and victim forgiveness termed effective preferences; these prefer- We define betrayal as the perceived violation ences guide behavior. of an implicit or explicit relationship-relevant In betrayal situations, the victims’ trans- norm (Finkel, Rusbult, Kumashiro, & Han- formation from vengeful impulses to pro- non, 2002). Individuals experience betrayal relationship motives may not be effortless, when they believe that a partner has know- uncomplicated, or automatic. The impulse ingly departed from the norms of decency and toward negative reciprocity is strong—people fairness that are assumed to govern a rela- are inclined to fight fire with fire, respond- tionship, thereby causing harm. Given that ing in kind to a partner’s real or imagined betrayals are harmful to victims and violate negativity (Gottman, 1998; Rusbult, Verette, morality-based expectations, victims typically Whitney, Slovik, & Lipkus, 1991; Yovetich & experience righteous indignation, believe that Rusbult, 1994). Moreover, betrayal-inspired the perpetrator has incurred an interpersonal moral outrage and perceived debt may keep debt, and perceive that such incidents bode such incidents alive and feed the flames— poorly for their relationships (Leary, Springer, victims may ruminate about a betrayal and Negel, Ansell, & Evans, 1998). Whether a its implications, such that vengeful impulses couple can survive and recover from such an may linger for a considerable period of time incident rests on how betrayals are resolved. (Fincham, 2000; McCullough, Bono, & Root, Previous work has defined forgiveness as 2007). How are the flames of righteous “the set of motivational changes whereby one indignation quelled? Amends, forgiveness, and betrayal resolution 255

Perpetrator amends and victim forgiveness forgiveness may be mutually reinforcing over the course of extended interaction. We suggest that although forgiveness ulti- Of course, perpetrators may not find it easy mately rests in the hands of victims, perpe- to offer amends. In the wake of betrayal, Mary trator actions affect both the probability of may suffer sadness, shame, or guilt (Baumeis- forgiveness and the likelihood that a betrayal ter, Stillwell, & Heatherton, 1995). If John will be successfully resolved (Human Devel- dwells on the incident or is reproachful, Mary opment Study Group, 1991). Interdependence may exhibit defensive cognitive maneuvers, theory proposes that to understand the resolu- seeking to justify her behavior not only to tion of interdependence dilemmas, interaction John but also to herself (Baumeister, Stillwell, is the name of the game; that is, both partners’ & Wotman, 1990). And perpetrators may not actions matter (Kelley et al., 2003; Rusbult be willing to suffer extended blame—if John & Van Lange, 2003). Given that most extant is persistently hostile or vengeful, Mary may work regarding forgiveness has focused on react with reciprocal hostility, refusing to victims, we know relatively little about the make amends (Hodgins & Liebeskind, 2003; processes by which couples achieve forgive- Hodgins, Liebeskind, & Schwartz, 1996). ness and resolve betrayals. We know even less Thus, just as it is difficult for victims to find about how perpetrators in romantic relation- their way to forgiveness, it is also difficult ships perceive this process and their role in it. for perpetrators to find their way to amends. We define perpetrator amends as accepting If Mary reacts defensively to John’s anger, responsibility for an act of betrayal, offer- claiming that she committed no offense and ing genuine atonement for one’s actions. that he has no right to feel upset, forgiveness We employ this broad definition of amends and betrayal resolution become unlikely. to acknowledge our belief that in the con- The extant literature offers some support text of close relationships, perpetrator amends for our claims, demonstrating that apology is may not always include an explicit, verbal positively associated with forgiveness (most apology. Importantly, amends must be sin- prior work has studied explicit apology rather than overall amends; Frantz & Bennigson, cere—perpetrator acts that are perceived as 2005; McCullough et al., 1997; McCullough insincere tend to backfire, such that disingen- et al., 1998; Mullet, Houdbine, Laumonier, & uous amends is likely to inhibit forgiveness Girard, 1998; Zechmeister et al., 2004). How- and betrayal resolution (e.g., Exline, DeShea, ever, recent studies have revealed that the & Holeman, 2007; Risen & Gilovich, 2007; association between apology and forgiveness Zechmeister, Garcia, Romero, & Vas, 2004). can also be negative. For example, in situa- How does amends work its magic? In tions where the perpetrator’s offense was per- interdependence terms, amends serves as a ceived as intentional (Struthers, Eaton, San- form of situation selection, moving a couple telli, Uchiyama, & Shirvani, 2008) or where toward interaction opportunities with supe- the victim has low implicit self-esteem (Eaton, rior behavioral options and outcomes (Kelley, Struthers, Shomrony, & Santelli, 2007), apol- 1984; Kelley et al., 2003). When John com- ogy is associated with decreased forgiveness. municates that he feels betrayed, Mary may Especially in the context of close, ongo- calmly discuss the incident with him. Recog- ing relationships, there is more to be dis- nizing that her actions hurt him, she may offer covered regarding the association of amends amends, apologizing for her actions or mak- with forgiveness:1 First, it is unclear whether ing things right by atoning for the harm she has caused. In turn, John may find it easier to let go of his hurt and anger. John’s gradual 1. In a recent PsychInfo (an online database of articles, book chapters, and other scholarly publications in movement toward forgiveness may also make psychology and related disciplines) search wherein we it easier for Mary to offer further amends, con- combined all variants of forgiveness with all variants of amends, apology,andatonement, we uncovered tinuing to respond in a loving and benevolent 31 studies of forgiveness in dyads. Of these, only 5 manner. Thus, perpetrator amends and victim studies examined perpetrator perspectives on betrayal 256 P. A. Hannon et al. perpetrators perceive the positive effects of Thus, a betrayal may be forgiven yet still be amends on victim forgiveness, at least to very much alive. the same extent that victims do. If these To further our understanding of the after- positive effects are not apparent to perpe- math of betrayal, it is important to exam- trators, perpetrators may not be motivated ine betrayal resolution, or the perception—by to make amends—even in the context of each partner—that a betrayal incident has a committed relationship—especially given been successfully closed, such that it no that amends may not be an automatic or effort- longer colors interaction. Moreover, it is less response. Second, most research has rep- important to recognize that broader relation- resented apology as a discrete event, examin- ship quality may be influenced by betrayal ing a causal sequence wherein apology pro- resolution: Navigating a potentially harm- motes victim forgiveness, and forgiveness is ful betrayal may affect partners’ relationship the endpoint—most work has not examined expectations, influence their sense of commit- what transpires following victim forgiveness. ment, or yield other cognitive or affective con- It is important to demonstrate not only that sequences. In the present work, we examine amends promotes forgiveness but also that how partners experience relationship quality both variables—from the perspective of both in light of a specific betrayal incident and its partners—contribute to successful betrayal resolution, thereby extending prior research resolution. that has examined the association of global forgiveness tendencies with global satisfac- Betrayal resolution and postbetrayal tion or adjustment (e.g., Paleari, Regalia, & relationship quality Fincham, 2005; Thompson et al., 2005). Thus, rather than representing forgiveness as the endpoint of betrayal incidents, we adopt Victim versus perpetrator perspectives a broadened time frame, examining not only Our work is also guided by a final goal: Many amends and forgiveness but also reconcilia- extant studies of forgiveness have examined tion (cf. Freedman, 1998). Why so? There events solely from the victim’s viewpoint. is no guarantee that amends and forgiveness Studies examining both victim and perpetra- will necessarily yield successful betrayal res- tor perspectives reveal that partner percep- olution and the recovery of couple function- tions are not necessarily parallel; for example, ing. Even when a perpetrator offers sincere victims may be less likely to perceive that a amends and a victim genuinely forgives, part- betrayal is resolved, perpetrators may mini- ners may find that they cannot forget the mize the impact of a betrayal, and both parties incident or fully relegate it to the past. For may perceive circumstances in a self-serving example, even though Mary offers heart- manner (Baumeister et al., 1990; Hodgins & felt amends and John genuinely forgives Liebeskind, 2003; Kearns & Fincham, 2005; her for her betrayal, they may continue to Zechmeister & Romero, 2002). Thus, it would review betrayal-relevant events (“whose fault be easy to imagine that victims and perpe- was that, really?”), surreptitiously monitor trators experience circumstances quite differ- one another’s actions (“is she sufficiently ently and that their differing perceptions affect repentant?” “does he still feel hurt?”), suf- the impact of each person’s actions on the fer reduced trust (“could that happen again?”), perceptions and actions of the other. or otherwise interact in an unnatural manner. We suggest that the extent and consequences of such bias need further exploration. Partners may shade their construals in such incidents; 4 of these examined hypothetical or staged betrayals in nonintimate relationships and 1 examined a manner as to protect the self, but this is real betrayals in relationships of unspecified types. not to say that there is no reality underly- We located no studies that examined victim and ing perception; such bias may well operate perpetrator perspectives on specific, shared betrayal incidents, and we located no studies that examined at the margins. Indeed, many prior studies perspectives on betrayal resolution. of role-based bias have employed procedures Amends, forgiveness, and betrayal resolution 257 that may yield exaggerated evidence of role- 1991). However, most such work has studied based bias; for example, many prior studies hypothetical betrayals in nonintimate relation- rest on participant-selected incidents from the ships and has ended with whether forgiveness point of view of one party to a betrayal, was given or denied. We suggest that it is such that role may affect incident selection important to examine real betrayals in real, (e.g., victims and perpetrators recall differ- ongoing relationships, in that (a) although ent incidents, describe incidents differently, significant hurtful events may occur during or select incidents that reflect favorably on stranger interactions (e.g., in a violent crime), themselves). By examining real betrayal inci- it is in the context of ongoing relation- dents through the eyes of both parties, we ships that meaningful, relationship-relevant hope to determine whether amends and for- transgressions transpire; (b) the question of giveness are phenomena that are experienced whether amends and forgiveness promote similarly by partners, and that exert parallel betrayal resolution and relationship quality is effects on betrayal resolution and relationship most relevant for ongoing relationships, in quality. which there is not only a history but also a (potential) future; and (c) as noted earlier, from a methodological point of view it is cru- Hypotheses and research overview cial to examine both partners’ perceptions of As displayed in Figure 1, our dyadic model real betrayal incidents. suggests that to the extent that perpetrators We conducted three studies to examine the offer amends for their actions, victim for- associations among two or more model vari- giveness is more probable. In turn, perpe- ables. In Study 1, married couples discussed trator amends and victim forgiveness exert an unresolved betrayal incident. Immediately independent effects on betrayal resolution, following the discussion, participants viewed and betrayal resolution promotes enhanced a video recording of their interaction and relationship quality. Some extant studies of rated their own and the partner’s behavior; confession or apology have revealed find- trained observers provided parallel ratings. ings consistent with one or more links in this In Study 2, participants completed interaction model (e.g., Gonzales, Manning, & Haugen, records to describe all betrayal incidents that 1992; Weiner, Graham, Peter, & Zmuidinas, transpired in their dating relationships during

Perpetrator Amends

+ + Betrayal Relationship + Resolution Quality

+ Victim Forgiveness

Tested in Tested in Tested in Studies 1,2, and 3 Studies 2 and 3 Study 3

Figure 1. The dyadic model of betrayal resolution. 258 P. A. Hannon et al. a 2-week period, whether they were perpetra- forgiveness. By examining behavior through- tors or victims in each incident. In Study 3, out the discussion, we were able to test both partners in dating relationships identified whether perpetrator positive behavior predicts specific prior betrayal incidents and described increases over time in victim positive behav- their own and the partner’s behavior during ior, as well as whether victim positive behav- each incident. ior predicts increases over time in perpetrator In Study 1, we assessed amends and for- positive behavior. Given that both partners giveness at multiple times over the course participated in the discussion and rated each of interaction to assess whether (a) earlier person’s behavior, we were also able to exam- amends predicts increases over time in for- ine victim and perpetrator accounts of each giveness. We also assessed whether (b) earlier incident. In Study 1, we examined betrayal forgiveness predicts increases over time in incidents that were not resolved at the time amends to examine the possibility that for- of the study, so we were unable to assess the giveness and amends may operate in a mutu- associations of victim and perpetrator positive ally reinforcing manner. In Studies 2 and behavior with betrayal resolution. 3, we not only evaluated whether perpetrator amends and victim forgiveness are positively associated but also whether (c) both Method amends and forgiveness contribute unique Participants and recruitment variance to predicting successful betrayal resolution. In Study 3, we also assessed whether Seventy-nine married couples participated in a (d) betrayal resolution predicts relationship study of marital processes. The data for 4 cou- quality. In addition to testing these core pre- ples were deleted from the analyses (2 for fail- dictions of the dyadic model, we examined ing to follow instructions, 1 because they were victim–perpetrator role effects. We antici- not married, and 1 due to videotaping tech- pated that although we might observe role nical difficulties), leaving a total of 75 cou- main effects in levels of variables (e.g., vic- ples. Participants were recruited via notices tims might perceive greater forgiveness than posted on the campus and in the community perpetrators), the hypothesized associations of the University of North Carolina, as well among amends, forgiveness, and betrayal res- as through advertisements in local newspa- olution would be evident using both victim pers. All announcements briefly described the and perpetrator accounts of events. project, indicated that the study involved three research sessions over an 8-month period, noted that couples would be paid $50.00 for Study 1 taking part in each research session, and pro- In Study 1, we asked married partners to vided contact information. When couples con- discuss an unresolved betrayal incident. Fol- tacted us, we provided more detailed infor- lowing the interaction, each partner viewed a mation about project activities and scheduled video recording of the discussion and rated appointments. Data from the initial research his or her own and the partner’s behavior. In session are presented here; data collected dur- addition, trained observers later viewed the ing the second and third sessions are not rel- videotaped interaction and rated both part- evant to the present work. ners’ behavior. By asking couples to dis- Participants were 34 years old on average cuss an unresolved incident, we were able (SD = 11 years), and most were Caucasian to observe perpetrator and victim behav- (80% Caucasian, 11% African American, iors as they unfolded during the course of 4% Hispanic, 2% Asian American, and 3% betrayal-relevant interaction. We asked cou- Other). Most participants had at least 4 years ples to rate positive behavior rather than of college education (44% obtained advanced explicitly asking them to rate amends and or professional degrees, 38% completed 4 forgiveness; trained observers rated positive years of college, 11% completed 2 years of behaviors and perpetrator amends and victim college, and 7% completed high school only). Amends, forgiveness, and betrayal resolution 259

Their median personal income was $20,000– person. A camera was oriented to videotape $30,000 per year. Participants had been mar- both partners. Following a 2-min warm-up ried to one another for 6 years on average conversation (discussing the events of the pre- (SD = 9 years), and most did not have chil- vious day), the experimenter explained that dren (74% no children, 10% one child, 8% we had randomly determined which part- two children, and 8% three or more children). ner’s incident would be discussed and selected one of that person’s incidents as the discus- Procedure sion topic. The experimenter read the incident description aloud; then partners were Upon arrival at the laboratory, we asked each given 1 min to describe the incident, as a participant to complete a brief questionnaire means of helping them bring the incident to that was later used to identify a suitable recent mind. After the experimenter left the labora- betrayal incident for a videotaped conversa- tory, the couple spent 8 min discussing the tion. We did not use the word betrayal in our incident. Immediately following the interac- instructions because this word may connote tion, partners individually reviewed and rated sexual forms of disloyalty or infidelity, and the videotaped interaction. At the end of the we did not want to limit participants to these session, couples were paid, thanked for their types of betrayals. Instead, instructions to par- assistance, and partially debriefed. ticipants described such incidents as follows: Perpetrator and victim ratings of All of us have expectations about how our interactions partners should treat us. No matter how well-behaved your partner may be in gen- Following their interaction, partners were led eral, from time to time he or she is likely to to separate video monitors and individu- “break the rules.” For example, your part- ally reviewed the videotaped interaction. The ner may tell a friend something that you experimenter stopped the videotape at the end think should have remained private; your of each 2-min segment of the interaction, ask- partner may do something that is hurtful ing participants to rate their own and the behind your back; your partner may flirt partner’s behavior during that segment. Each with another person; or your partner may participant first rated his or her own behav- otherwise violate the rules that govern your ior during a given segment and then rated the marriage. partner’s behavior. To avoid communicating our hypotheses, we used parallel scales for Each partner was asked to describe three ratings of victim and perpetrator behav- such incidents from the past 4 months, pro- ior, employing concrete descriptors of spe- viding simple ratings of each incident on 9- cific interaction behaviors (with appropriate point scales (e.g., “How upsetting was it?” changes in item wording to reflect own vs. 0 = not upsetting at all,8= very upsetting). partner behavior). Six items measured posi- To identify an incident for discussion, we tive behavior during the interaction (e.g., “My randomly determined whether to use an inci- partner tried to comfort me,” “I raised my dent described by the husband or the wife, voice toward my partner” [reverse-scored], and selected an incident that was moder- “My partner wanted to cut off the inter- ately upsetting, that was not totally resolved, action” [reverse-scored]; for all items, 0 = and that the partners were willing to discuss. do not agree at all,8= agree completely). While the experimenter selected an incident Ratings of positive behavior exhibited good for couples to discuss, participants completed reliability as rated by both perpetrators (for other activities that are not relevant to the ratings of perpetrator positive behavior, αsfor present work. the four 2-min segments ranged from .80 to During the videotaped interaction, part- .83; for ratings of victim positive behavior, αs ners were seated at adjacent sides of a table ranged from .77 to .81) and victims (for per- with a microphone positioned in front of each petrator positive behavior, αs ranged from .77 260 P. A. Hannon et al. to .82; for victim positive behavior, αs ranged critical manner” [reverse-scored], “exhibited from .72 to .76). We calculated measures of hostility” [reverse-scored]) and then com- perpetrator and victim positive behavior by pleted two additional one-item global ratings averaging participants’ ratings of each per- (“perpetrator offered amends” and “victim son’s behavior during each 2-min segment. was forgiving”; for all items, 1 = no evidence of this behavior,5= very strong evidence of Measuring possible confounds this behavior). We calculated measures of perpetrator amends and victim forgiveness Participants also completed additional instru- by averaging the two observers’ global rat- ments that were used in analyses we per- ings of the perpetrator’s amends and the vic- formed to control for possible confounds. tim’s forgiveness during each 2-min segment. Given that reports of positive behavior may (Parallel findings were evident in analyses be vulnerable to socially desirable response using observers’ concrete ratings of specific tendencies, we asked participants to complete interaction behaviors; for the analyses dis- a 20-item version of the Balanced Inven- played in Table 1, the same six residualized tory of Desirable Responding that included lagged associations were significant.) Interob- the 10 most reliable items from the self- server agreement was acceptable for ratings of deception and impression management sub- both amends and forgiveness (for amends, αs scales (Paulhus, 1984; e.g., “I always obey across the four 2-min segments ranged from laws, even if I’m unlikely to get caught”; 1 = .69 to .77; for forgiveness, αs ranged from .54 do not agree at all,7= agree completely; to .72). αs = .69 and .60). In the present sample of married couples, we were concerned that com- Character of betrayal incidents mitment (rather than partner behavior) might powerfully drive behavior; that is, the rela- Consistent with our goals for the betrayal tionship with the spouse might be so impor- interaction, the betrayal incidents partners tant that maintaining the relationship would discussed were described as moderately to take precedence over all else. To explore this severely upsetting (M = 5.17, SD = 1.68) possibility, we measured participants’ com- and were described as not yet fully resolved mitment level using a 15-item version of (M = 4.76, SD = 2.57). The incidents inclu- the Investment Model Scale (Rusbult, Martz, ded violations of dependence norms (e.g., & Agnew, 1998; e.g., “I would feel very overspending after agreeing to save money), upset if our relationship were to end in monogamy norms (e.g., not trusting the part- the near future”; 0 = do not agree at all, 8 = ner with people of the opposite gender), pri- agree completely; α = .92). vacy norms (e.g., discussing embarrassing topics in front of family members or neigh- Observer ratings of interactions bors), and decency/etiquette norms (e.g., vol- unteering the partner for something without In addition to obtaining participants’ rat- asking). ings of their interaction, we also asked two trained observers to rate each couple’s interaction. Observers received extensive train- Results ing, including studying and mastering a cod- Analysis strategy ing manual and participating in a 4-hr training session. Observers viewed the same Each participant provided four sets of rat- 2-min interaction segments as participants. ings of the videotaped betrayal interaction— Observer ratings of participant interactions ratings of the victim and ratings of the per- were structured in such a manner as to prompt petrator for each 2-min segment. Our trained global, abstract ratings of amends and for- observers provided parallel ratings. Multi- giveness: Observers first provided ratings of ple ratings from a given individual are not concrete interaction behaviors (e.g., “behaved independent, nor are the ratings of the two in a warm/friendly manner,” “behaved in a partners in a given couple. Therefore, we Amends, forgiveness, and betrayal resolution 261

Table 1. Residualized lagged analyses: Predicting later victim behavior and later amends behavior from earlier victim and perpetrator behavior: Study 1

Victim data Perpetrator data Across-partner data Observer data Later victim positive behavior from: Earlier perpetrator .18∗∗ .16∗∗ .18∗∗ .16∗∗ positive behavior Earlier victim positive .71∗∗ .80∗∗ .71∗∗ .67∗∗ behavior Later perpetrator positive behavior from: Earlier victim positive .12∗∗ .10∗∗ .05 .03 behavior Earlier perpetrator .83∗∗ .85∗∗ .87∗∗ .68∗∗ positive behavior

Note. The values are standardized regression coefficients from hierarchical linear modeling analyses. Analyses are based on data from 75 couples (150 individuals). Victim data = victim and perpetrator positive behavior as reported by victims; perpetrator data = victim and perpetrator positive behavior as reported by perpetrators; across-partner data = victim positive behavior as reported by victims and perpetrator positive behavior as reported by perpetrators; and observer data = forgiveness and amends as rated by trained observers. ∗∗p<.01. used hierarchical linear modeling to analyze victim ratings of partner behavior and their our data (cf. Kenny, Kashy, & Bolger, 1998; own behavior; (b) perpetrator data analy- Raudenbush & Bryk, 2002). The multilevel ses, employing perpetrator ratings of partner data structure includes measures assessed dur- behavior and their own behavior; (c) across- ing each 2-min segment of the interaction partner data analyses, employing perpetrator (Level 1), represented as nested within partic- ratings of their own behavior and victim rat- ipant (Level 2 in within-participant analyses), ings of their own behavior; and (d) observer and nested within couple (Level 3 in across- data analyses, employing observer ratings of partner analyses and analyses of observer rat- perpetrator amends and victim forgiveness. ings). Hierarchical linear modeling accounts We standardized all variables prior to conduct- for the nonindependence of observations by ing hierarchical linear modeling. We initially simultaneously examining variance associated performed all univariate analyses including with each level of nesting, thereby provid- main effects and interactions for participant ing unbiased hypothesis tests. Following rec- gender (male vs. female). No main effects or ommended procedures for couples research, interactions involving participant gender were we represented intercept terms as random significant, so this variable was dropped from effects and represented slope terms as fixed the analyses. effects (Kenny, Mannetti, Pierro, Livi, & Kashy, 2002). Dyadic model hypotheses All analyses reported below employed a lagged data structure, wherein we predicted a To determine whether perpetrator amends pro- later criterion from an earlier predictor. That motes victim forgiveness and whether vic- is, we simultaneously predicted (a) Time 2 tim forgiveness promotes perpetrator amends, criteria from Time 1 predictors, (b) Time 3 we first examined the simple lagged asso- criteria from Time 2 predictors, and (c) Time ciations of (a) earlier perpetrator positive 4 criteria from Time 3 predictors. In residual- behavior with later victim positive behav- ized lagged analyses, we included the earlier ior and (b) earlier victim positive behavior measure of the criterion as a control vari- with later perpetrator positive behavior. All able. We performed four types of analysis (see simple associations of earlier predictors with Table 1): (a) victim data analyses, employing later criteria were significant (for analyses 262 P. A. Hannon et al. employing victim, perpetrator, across-partner, perpetrator data. To further explore the extent and observer ratings, βs ranged from .17 to to which amends and forgiveness are phe- .35, all ps <.05). Analyses predicting change nomena that are perceived in parallel man- over time in each criterion are presented in ner by partners in a given relationship, we Table 1. In these residualized lagged analyses examined whether victims and perpetrators (controlling for earlier levels of the criterion), exhibited rough agreement in their reports of across all four types of analysis, earlier per- one another’s positive behavior. To explore petrator positive behavior predicted increases simple across-partner agreement in reports over time in victim positive behavior (beyond of behavior during the videotaped interac- variance attributable to earlier victim positive tion, we (a) regressed victims’ ratings of behavior; under Later victim positive behav- their own positive behaviors during inter- ior, see row labeled Earlier perpetrator pos- action onto perpetrators’ ratings of victims’ itive behavior, βs ranged from .16 to .18, positive behaviors and (b) regressed perpe- all ps <.01). Earlier victim positive behavior trators’ ratings of their own positive behav- predicted increases over time in perpetrator iors onto victims’ ratings of perpetrators’ positive behavior in the analyses employing positive behaviors. Both across-partner asso- victim data and perpetrator data but not in ciations were significant (for victim posi- the analyses employing across-partner data or tive behaviors β = .59, for perpetrator behav- observer data (under Later perpetrator posi- iors β = .45, both ps <.01). Thus, partners tive behavior, see row labeled Earlier victim exhibited good agreement in their ratings positive behavior). of one another’s behavior during the course of the interaction. In addition, we examined Potential confounds whether there were mean differences in lev- Are the present findings attributable to unin- els of perceived perpetrator positive behav- tended confounds, such as tendencies toward ior and perceived victim positive behavior as socially desirable responding or commitment a function of perpetrator versus victim role. to the marriage? We replicated the Table 1 Analyses in which we regressed ratings of analyses employing victim data and perpe- victim and perpetrator behavior onto role (per- trator data (for which data sets it is suitable petrator vs. victim) revealed no significant to explore potential confounds assessed using differences in victims’ and perpetrators’ rat- self-report measures) including as covari- ings of either victim positive behavior (Ms = ates measures of self-deception, impression 5.85 and 5.81), t(74) = 0.55, ns, or perpetra- management, and commitment. In analy- tor positive behavior (Ms = 6.08 and 6.00), ses that simultaneously controlled for the t(74) = 1.01, ns. Thus, victims’ and perpe- three potential confounds, as well as con- trators’ perceptions of one another’s behavior trolling for earlier levels of each criterion: were significantly associated and there were (a) in predicting later victim positive behavior no role main effects in mean levels of vari- from earlier perpetrator positive behavior, ables. Moreover, the hypothesized associa- coefficients remained significant in analyses tions between victim and perpetrator positive employing both victim and perpetrator data behavior typically were evident using both (βs = .14 and .17, both ps <.01), and (b) in victim and perpetrator accounts of events (see predicting later perpetrator positive behav- Table 1). ior from earlier victim positive behavior, the coefficient remained significant in analyses employing victim data (β = .13, p<.01) Discussion but declined to nonsignificance in analyses In Study 1, we tested the hypothesis that β = employing perpetrator data ( .08, ns). (a) perpetrator amends promotes victim forgiveness; we also examined whether (b) vic- Victim and perpetrator perspectives tim forgiveness promotes perpetrator amends, Our Table 1 findings are roughly parallel in using ratings of each partner’s positive behav- analyses employing victim data and ior during a discussion of the betrayal as a Amends, forgiveness, and betrayal resolution 263 proxy for amends and forgiveness. In the con- An important limitation of Study 1 is that text of an interaction regarding an unresolved the laboratory setting in which the interactions betrayal incident, earlier perpetrator positive took place could be experienced as artifi- behavior reliably predicted increases over cial. A second limitation of Study 1 is that time in victim positive behavior. This finding although it allowed for a focused exploration emerged whether the analyses were based on of the reciprocal, mutually reinforcing asso- victim ratings of the interaction, perpetrator ciation of amends with forgiveness, an 8-min ratings of the interaction, across-partner rat- interaction does not lend itself well to study- ings, or independent observers’ ratings. How- ing the actual resolution of betrayal incidents. ever, earlier victim positive behavior was Finally, we did not explicitly ask participants a less robust predictor of increases over to rate each other on forgiveness or amends time in perpetrator positive behavior—this but rather to rate each other on positive, sup- association was significant in only two of portive behaviors during the interaction. Thus, the four analyses displayed in Table 1, and our findings point to the significance of perpe- one of these two significant associations fell trators enacting positive behaviors following to nonsignificance when controlling for rel- betrayal but cannot be interpreted as explic- evant confounds. Thus, Study 1 provides itly supporting the association of amends with strong evidence that perpetrator amends pro- forgiveness (with the exception of the ratings motes victim forgiveness but weaker evidence provided by trained observers, who were rat- that victim forgiveness promotes perpetrator ing participants’ amends and forgiveness). In amends. This finding suggests that the bene- Studies 2 and 3, we employed complementary fits of amends are not attributable to a simple methods to examine the associations among positivity effect (i.e., the positivity of part- amends, forgiveness, and betrayal resolution, ners’ behaviors toward one another). Rather, using measures that explicitly ask participants it is the existence of positive behavior in the to rate their and their partner’s amends and context of a particular role—to wit, perpetra- forgiveness following betrayal. tor amends—that appears to be crucial to the forgiveness process. One strength of Study 1 is the control Study 2 afforded by the laboratory setting. Partici- In Study 2, we examined not only perpetra- pants were focused on a specific unresolved tor amends and victim forgiveness but also betrayal incident in their relationship and pro- betrayal resolution. In addition, we tested vided concrete ratings of each person’s behav- key hypotheses of the dyadic model by ior during each 2-min segment of their inter- examining betrayals that transpired in the action. Indeed, partners exhibited very good context of everyday interaction. Over the agreement in their ratings of one another’s course of a 2-week period, people who behavior. Moreover, this method allowed were involved in ongoing dating relationships us to obtain parallel ratings from trained described each incident in which either they observers; analyses employing observers’ or the partner violated relationship-relevant ratings confirmed findings based on victims’ norms. Interaction records completed soon and perpetrators’ ratings. Thus, we can feel after each betrayal incident included mea- relatively confident that participants’ ratings sures of perpetrator amends, victim forgive- of each other’s behavior were not unduly col- ness, and betrayal resolution. This method has ored by availability effects, role-based con- the advantage of tapping online reports of key sistency, or self-serving bias. On the other variables, thus minimizing the possibilities for hand, the observer ratings for victim forgive- biased recall. It also has the advantage of tap- ness exhibited relatively low reliability, rais- ping perceptions of incidents in which the ing some questions about the extent to which participant was the victim as well as incidents outside observers can accurately perceive and in which he or she was the perpetrator, thus rate forgiving behavior. allowing us to examine associations among 264 P. A. Hannon et al. amends, forgiveness, and betrayal resolution would.” We asked participants to complete a from both victim and perpetrator perspectives. victim diary record “to record each incident in which your partner made you feel upset, angry, or hurt, ” and to complete a perpetrator Method diary record “to record each incident in which Participants you made your partner feel upset, angry, or hurt” (we describe data from these two types Participants were 78 undergraduates (20 men of record in terms of betrayal role, victim and 58 women) who volunteered to take vs. perpetrator). Participants were asked to part in the study in partial fulfillment of record all such incidents, even if (a) the inci- the requirements for introductory psychol- dent was quite brief, (b) the participant had ogy courses at the University of North Car- already recorded a similar incident in an ear- olina. The prerequisites for participation were lier record, or (c) the incident did not involve involvement in a dating relationship of at talking. Participants were instructed to com- least 1 month in duration, in which part- plete the diary records without input from ners interacted with one another nearly every their dating partners. day, either face to face or in telephone conversations. Participants were 19 years old We asked participants to complete diary on average (SD = 1 year), most were fresh- records as soon as possible following each men or sophomores (41% freshmen, 44% incident, and to turn in booklets every Mon- sophomores, 5% juniors, and 9% seniors), day, Wednesday, and Friday. To maximize and most were Caucasian (81% Caucasian, timely reporting, we stressed the importance 13% African American, 3% Asian Ameri- of carrying diary record booklets at all times, can, 1% Hispanic, and 3% Other). Partici- and we telephoned participants Sunday, Tues- pants had been involved with their partners day, and Thursday evenings to remind them for 16 months on average (SD = 14 months), to turn in their booklets the following day. For most indicated that they dated their partners each betrayal incident that transpired, partici- steadily (83% dating steadily, 8% engaged, pants (a) recorded the date and time at which 5% dating casually, and 4% dating regu- the incident occurred, (b) recorded the date larly), and most described their relationships and time at which the interaction record was as exclusive (96%). completed, (c) recorded the duration of the incident, (d) provided a brief description of Procedure the incident, and (e) answered several questions about the incident (described below). Our procedure was modeled after previ- During Time 1 and Time 2 laboratory ses- ous work using the Rochester Interaction sions, participants also completed a variety Record (cf. Reis & Gable, 2000). Participants of questionnaires, two of which (described attended two laboratory sessions—one at the below) are relevant to the goals of the present start of the 2-week study (Time 1) and a sec- research. At the end of Time 2 sessions, par- ond at the end of the study (Time 2). During ticipants were fully debriefed and thanked for Time 1 sessions, we explained that the study their assistance. concerned negative incidents in dating relationships. As in Study 1, we did not use the Measures word betrayal. Instead, we described betrayal incidents in terms of “breaking the rules”: We Participants completed a diary record for asked participants to complete a diary record each betrayal incident that transpired over the for each incident in which “you upset your course of a 2-week period, using one type of partner or your partner upsets you .... For record for victim incidents and a second type example, your partner may do something that of record for perpetrator incidents. Item word- is hurtful behind your back, you may flirt ing for victim and perpetrator diary records with someone else at a party, or you may for- were identical except for differences reflecting get to call your partner when you said you betrayal role (e.g., “I forgave my partner” Amends, forgiveness, and betrayal resolution 265

Table 2. Predicting victim forgiveness and betrayal resolution: Study 2

Victim data Perpetrator data Combined data Victim forgiveness from: Perpetrator amends .35∗∗ .41∗ .34∗∗ Betrayal resolution from: Victim forgiveness .54∗∗ .56∗∗ .52∗∗ Perpetrator amends .27∗∗ .23∗ .24∗∗

Note. The values are standardized regression coefficients from hierarchical linear modeling analyses. Analyses are based on data from 78 individuals. Victim data = predictors and criteria as reported in victim diary records; perpetrator data = predictors and criteria as reported in perpetrator diary records; combined data = predictors and criteria as reported in victim and perpetrator diary records combined. ∗p<.05. ∗∗p<.01. vs. “My partner forgave me”). Record items sessions, participants completed the 40-item were structured in such a manner as to prompt Balanced Inventory of Desirable Respond- global, abstract ratings of amends or forgive- ing, which assesses both self-deception and ness: Participants not only provided ratings impression management (Paulhus, 1984; e.g., of concrete interaction behaviors involving “I always obey laws, even if I’m unlikely amends (e.g., “I showed real remorse about to get caught”; 1 = do not agree at all,7= the incident”; as discussed in the introductory agree completely; respective αs = .54 and text, our goal with these items was to cap- .80). Finally, during Time 2 sessions, partici- ture amends-related behaviors in addition to pants answered questions about the validity of explicit apologies) or forgiveness (e.g., “I tried their data, indicating that they recorded 94% to get even with my partner” [reverse-scored]) of the betrayal incidents that transpired dur- but also completed additional one-item global ing the 2-week period and completed diary ratings (for amends, “I tried to make it up to records about 90 min following the inci- my partner”; for forgiveness, “I forgave my dents; 92% reported that the records accu- partner”; for all items, 1 = do not agree at all, rately reflected their experiences during the 7 = agree completely). The analyses reported 2-week period,2 and 79% reported that this below are based on participants’ one-item period was typical of the sorts of incidents global ratings. (Parallel findings were evident they experienced with their partners. in analyses using participants’ concrete ratings; for the analyses displayed in Table 2, eight of nine associations were significant.) Character of betrayal incidents Records also included a one-item measure of Sixty-four participants (18 men and 46 betrayal resolution (“By the end of the inter- women) described one or more betrayal inci- action, the incident was resolved”). dents in the victim role and 54 participants (16 Additional items were also included for men and 38 women) described one or more use in analyses we performed to control for incidents in the perpetrator role. As would possible confounds: One item assessed sever- be expected for the sort of day-to-day betray- ity of incident (“When this incident occurred, als that emerge over the course of a 2-week I thought it had the potential to seriously period, the incidents were described as mildly harm our relationship”) and one item assessed magnitude of betrayal (“When this incident occurred, I thought my partner had ‘bro- 2. All Study 2 analyses were replicated excluding data ken the rules’ of our relationship”). Both for the 8% of participants who indicated that the diary of these items were measured using 7-point records did not accurately reflect their experiences = = during the 2-week period. As expected (given the scales, where 1 do not agree at all and 7 small percentage of participants in this group), these agree completely. In addition, during Time 1 analyses revealed identical conclusions. 266 P. A. Hannon et al. to moderately severe betrayals (victim role Dyadic model hypotheses M = 3.23, perpetrator role M = 2.44) and as Table 2 summarizes results relevant to the having mild to moderate potential for harm dyadic model. In all three analyses, amends (victim role M = 2.51, perpetrator role M = was significantly positively associated with 2.39). forgiveness (under Victim forgiveness,see The incidents included violations of depen- row labeled Perpetrator amends): That is, dence norms (e.g., failing to provide assis- in everyday betrayal incidents, the tendency tance when it was needed), monogamy norms toward victim forgiveness is greater to the (e.g., disappearing into a bedroom for an extent that perpetrators offer greater amends hour with an ex-partner), privacy norms (e.g., for their acts of betrayal. Moreover, when telling a friend an important secret), and we regressed betrayal resolution simultane- decency/etiquette norms (e.g., playing a trick ously onto amends and forgiveness, both vari- on the partner and then lying about it). ables accounted for unique variance (under Betrayal resolution, see rows labeled Perpe- Results trator amends and Victim forgiveness); that is, everyday betrayals were more likely to be Analysis strategy successfully resolved to the extent that per- Multiple diary records from a given partic- petrators offered greater amends and victims ipant are not independent, so we used hier- found their way to greater forgiveness. archical linear modeling to analyze our data, employing strategies that are suitable for the Potential confounds analysis of diary data (cf. Bolger, Davis, & Are these findings attributable to unintended Rafaeli, 2003; Nezlek, 2001; for an empir- confounds, such as severity of incident? ical illustration, see Bolger, Zuckerman, & We replicated the Table 2 combined anal- Kessler, 2000). The two-level data struc- yses including as covariates measures of ture includes measures assessed in each diary severity of incident, magnitude of betrayal, record (Level 1), represented as nested within self-deception, and impression management. participant (Level 2). For example, a partici- In analyses that simultaneously controlled pant who experienced three betrayal incidents for the four potential confounds: (a) the during the 2-week period provided reports of association of amends with forgiveness was three separate betrayals (for each incident, significant (β = .36, p<.01) and (b) in pre- reports of amends, forgiveness, and betrayal dicting betrayal resolution, both perpetra- resolution). tor amends (β = .26, p<.01) and victim We performed three types of analysis (see forgiveness accounted for unique variance Table 2): (a) victim data analyses, employing (β = .44, p<.01). participant ratings of the partner’s amends and their own forgiveness; (b) perpetrator Victim and perpetrator perspectives data analyses, employing participant ratings of their own amends and the partner’s forgive- Our Table 2 findings are roughly parallel in ness; and (c) combined data analyses, includ- analyses employing victim data and perpetra- ing both victim data and perpetrator data. We tor data. In addition, we examined whether standardized all variables prior to conducting there were mean differences in levels of per- hierarchical linear modeling. We initially perceived perpetrator amends, victim forgive- formed all combined data analyses including ness, and betrayal resolution as a function main effects and interactions for participant of perpetrator versus victim role. Analyses gender (male vs. female) and betrayal role in which we regressed each variable onto (victim vs. perpetrator). No main effects or role (perpetrator vs. victim) revealed a signif- interactions were significant for either vari- icant effect of role on perpetrator amends (for able, so these factors were dropped from the perpetrator and victim diary records, Ms = analyses. 4.37 and 3.66), t(166) = 2.58, p<.02; a Amends, forgiveness, and betrayal resolution 267 nonsignificant effect of role on victim forgive- Study 3 ness (for perpetrator and victim diary records, We used a retrospective method in Study 3, Ms = 5.57 and 5.39), t(168) = 0.93, ns;and testing our hypotheses by asking dating part- a significant effect of role on perceived ners to recall recent betrayal incidents in their betrayal resolution (for perpetrator and victim relationship. Partners in ongoing dating rela- diary records, Ms = 5.06 and 4.49), t(168) = tionships identified an occasion on which the 2.69, p<.01. Thus, although we observed partner violated a relationship-relevant norm. some role main effects in mean levels of Each person then provided information about variables (e.g., compared with victims, per- perpetrator amends, victim forgiveness, and petrators reported offering higher levels of betrayal resolution, not only for the incident amends and betrayal resolution), the hypothe- he or she identified but also for the inci- sized associations between amends and for- dent identified by the partner. This method giveness typically were evident using both has the advantage of tapping betrayal inci- victim and perpetrator accounts of events. dents of greater severity than those examined in Study 2. In addition, Study 3 included a Discussion new measure of relationship quality so that we could assess whether and how partici- Study 2 revealed strong support for our pants felt that these specific betrayal inci- hypotheses. Examining the sorts of betray- dents had influenced their relationship. And als that transpire in the context of everyday finally, by obtaining victim and perpetrator interactions over a 2-week period, perpetrator accounts of the same incident, we were able amends was positively associated with victim to assess the extent to which partners agree in forgiveness, and both amends and forgive- their descriptions of one another’s actions, as ness contributed positively to the prediction well as to examine the associations between of successful betrayal resolution. These asso- victim and perpetrator reports while control- ciations did not differ significantly for betrayal ling for the severity of a given betrayal. As incidents described by participants when they displayed in Figure 1, we predicted that per- were in the role of victims versus perpetra- petrator amends would predict victim forgive- tors, nor were these associations attributable ness, that both amends and forgiveness would to potential confounds such as severity of inci- predict betrayal resolution, and that betrayal dent or tendencies toward socially desirable resolution would predict relationship quality. responding. At the same time, these findings are limited in several respects. First, Study 2 exam- Method ined betrayal incidents that spanned a nar- Participants row range of severity, representing relatively minor, “everyday” forms of betrayal. Based Participants were 70 heterosexual dating cou- on these findings, it is unclear whether perpe- ples who responded to notices posted on the trator amends plays a significant role in rel- campus and in the community of the Uni- atively more serious betrayal incidents. Sec- versity of North Carolina, as well as through ond, in Study 2, participants completed diary advertisements in local newspapers. The data records for all betrayal incidents that they for 2 couples were deleted from the analy- experienced during the course of the 2-week ses (1 because one partner did not follow study. These findings would be augmented by instructions, 1 due to experimenter error). hypothesis tests in which we examine both Couples were paid $40.00 for taking part in partners’ descriptions of the same betrayal the study. Participants were 22 years old on incidents and their aftermath. And third, in average (SD = 3 years), and most were Cau- order to test our full dyadic model, it is impor- casian (86% Caucasian, 9% African Ameri- tant to examine how the resolution of betrayal can, 2% Hispanic, 1% Asian American, and incidents relates to the current state of the 2% Other). Eighty-eight percent were enrolled relationship (see Figure 1). in college or had completed undergraduate 268 P. A. Hannon et al. degrees; an additional 11% were enrolled in Participants then completed questionnaires (or had completed) graduate or professional in which they provided measures of vic- degrees. Partners had been involved with tim behavior, perpetrator behavior, betrayal one another for 20 months on average (SD = resolution, and relationship quality for the 18 months), most indicated that they dated assigned incident. Following a 5-min break,3 one another steadily (81% dating steadily, 7% the experimenter distributed the description dating regularly, 7% engaged, and 2% dating of the second incident selected for each cou- casually), and most described their relation- ple, and participants completed questionnaires ships as exclusive (96%). describing that incident. At the end of the session, couples were paid, thanked for their Procedure assistance, and debriefed. Partners were seated in individual cubicles Measures so that they could neither see nor interact with one another and were informed that The questionnaires participants completed to the study concerned negative incidents in describe their own and the partner’s behav- dating relationships. In an initial question- ior during each incident were nearly identi- naire, each partner was asked to identify a cal except for differences reflecting betrayal betrayal incident that transpired within the role (e.g., “I forgave my partner” vs. “My past 4 months. The instructions were worded partner forgave me”). Victim forgiveness was as in Study 1, such that we avoided using assessed with a 12-item instrument (e.g., “I tried to work things out with my partner”; the word betrayal, instead describing betrayal for all items, 0 = do not agree at all,8= incidents in terms of “breaking the rules.” We agree completely; for ratings by victims and asked participants to describe two such inci- perpetrators αs = .84 and .87). Perpetrator dents, indicating that later in the session we amends was assessed with a 24-item instru- would identify one incident for each partner, ment designed to tap our broad, atonement- and ask both partners to provide informa- based definition of amends (e.g., “My part- tion about each incident. We asked partici- ner confessed that he/she had done wrong,” pants to report on nontrivial incidents and to “My partner attempted to make up for his/her identify incidents that they (and the partner) behavior”; for ratings by victims and perpetra- would feel comfortable describing in greater tors, αs = .90 and .86).4 Betrayal resolution detail. Participants wrote brief descriptions of was measured with a single item, “My partner two incidents and rated the extent to which each incident had the potential to harm the = = relationship (0 no potential for harm,8 3. Participants were allowed to interact during the break strong potential for harm). For each partner, but were instructed not to discuss the study. Many we identified an incident that had moderate to participants remained in the laboratory during their break, where the experimenter could verify that they severe potential for harm. While the experi- followed these instructions. Most participants who left menter identified an incident for each partner, the laboratory did so to use the restroom, and did not participants completed other questionnaires interact with their partners during this time. 4. For each incident, participants provided information describing themselves and their relationship. regarding both immediate and delayed victim and The experimenter prepared photocopies of perpetrator behavior. Items concerning “immediate” the selected incidents and randomly deter- behavior described actions immediately following the betrayal incident; items concerning “delayed” behav- mined which partner’s incident would be ior described actions at present (or the last time the addressed first. The experimenter asked each couple engaged in incident-relevant interaction). Mea- partner to read the incident description, and sures of immediate and delayed behavior were significantly correlated with one another for both victims asked the perpetrator whether he or she rec- and perpetrators (all rs ≥ .60, all ps <.01), and anal- ognized the incident described by the partner. yses performed separately for immediate and delayed One participant did not recognize the incident; measures revealed parallel findings. Accordingly, the Study 3 analyses are based on averaged measures of this couple was allowed to converse briefly immediate and delayed victim forgiveness and perpe- so that the partner could clarify the incident. trator amends. Amends, forgiveness, and betrayal resolution 269 and I have completely resolved this incident.” each betrayal incident (Level 2) and in turn Relationship quality was assessed with a new nested within each couple (Level 3). We per- 6-item instrument designed to measure the formed four types of analysis (see Tables 3 impact of a specific betrayal on the present and 4): (a) victim analyses, employing vic- quality of the relationship (the items for victim ratings of partner amends and their own tims were as follows [wording was altered as forgiveness; (b) perpetrator analyses, employ- appropriate for perpetrators]; “Our relation- ing perpetrator ratings of partner forgiveness ship is ruined” [reverse scored], “Our rela- and their own amends; (c) combined analyses, tionship can still work, but it will never be as including both victim ratings and perpetrator good as it was before the incident” [reverse ratings; and (d) across-partner analyses, scored], “I think my partner and I learned employing perpetrator ratings of their own something valuable from the incident,” “I amends and victim ratings of their own for- believe my partner will not behave this way giveness (as well as, in each analysis, rat- again,” “My partner has a better understand- ings of betrayal resolution and relationship ing of my expectations for this relationship quality). We standardized all variables prior now,” “Our relationship is better now than to conducting hierarchical linear modeling. ever before”; for ratings by victims and per- We initially performed all univariate analy- petrators, αs = .77 and .77). For the purpose ses including main effects and interactions of controlling for possible confounds, partici- for participant gender (male vs. female), pants also completed the Study 1 measures of betrayal role (victim vs. perpetrator), and self-deception, impression management,and incident order (victim vs. perpetrator inci- commitment level (αs = .72,.73, and .96, dent described first). Two betrayal role main respectively). effects were significant (described below), but no interactions were significant. Given that Character of betrayal incidents our findings were not reliably moderated by participant gender, betrayal role, or incident The betrayal incidents partners recounted order, these variables were dropped from the were described as having moderate to high analyses. potential for harm (M = 5.71, SD = 1.76). The incidents included violations of depen- Dyadic model hypotheses dence norms (e.g., breaking a promise to quit smoking, then lying about the lapse), Table 3 summarizes results relevant to the monogamy norms (e.g., becoming sexually dyadic model, presenting standardized coef- intimate with an extrarelationship partner), ficients for analyses performed separately as privacy norms (e.g., speaking to a third party a function of betrayal role as well as for about matters the individual regarded as pri- the sample as a whole. As hypothesized, per- vate), and decency/etiquette norms (e.g., sur- petrator amends was significantly associated reptitiously “borrowing” the partner’s debit with victim forgiveness (under Victim forgiveness card to pay for a drinking spree). , see row labeled Perpetrator amends). Analyses regressing betrayal resolution onto victim forgiveness and perpetrator amends Results revealed that betrayal resolution is significantly predicted by: (a) victim forgiveness in Analysis strategy the perpetrator data analyses and the com- Data from the two betrayal incidents described bined analyses (under Betrayal resolution,see by each participant are not independent, nor single row labeled Victim forgiveness) and are data from the two partners in a given (b) perpetrator amends in all three analyses couple. Therefore, as described in Studies (see single row labeled Perpetrator amends). 1 and 2, we used hierarchical linear mod- In addition, when we regressed betrayal res- eling to analyze our data. The three-level olution simultaneously onto forgiveness and data structure includes measures provided amends, forgiveness accounted for signifi- by each participant (Level 1), nested within cant unique variance in two of the three 270 P. A. Hannon et al.

Table 3. Predicting victim forgiveness, betrayal resolution, and relationship quality: Study 3

Victim data Perpetrator data Combined data Victim forgiveness from: Perpetrator amends .24∗∗ .25∗ .23∗∗ Betrayal resolution from: Victim forgiveness .19 .35∗∗ .28∗∗ Perpetrator amends .29∗∗ .21∗ .20∗∗ Victim forgiveness .09 .33∗∗ .25∗∗ Perpetrator amends .26∗ .12 .16∗ Relationship quality from: Betrayal resolution .39∗∗ .46∗∗ .38∗∗ Betrayal resolution .22∗∗ .35∗∗ .28∗∗ Victim forgiveness .04 .10 .08 Perpetrator amends .49∗∗ .25∗∗ .35∗∗

Note. The values are standardized regression coefficients from hierarchical linear modeling analyses for one-predictor models (single-row models) and for two- or three-factor models (grouped as two- or three-row models). Analyses are based on data from 68 couples (136 individuals). Victim data = predictors and criteria as reported by victims; perpetrator data = predictors and criteria as reported by perpetrators; combined data = predictors and criteria as reported in victim and perpetrator reports combined. ∗p<.05. ∗∗p<.01. analyses, and amends accounted for signif- and .47, both ps <.01) and with betrayal icant variance in two of the three analyses resolution (βs = .28 and .20, both ps <.01), (under Betrayal resolution, see two-row mod- and betrayal resolution accounted for signif- els). Thus, (a) in resolving betrayal incidents, icant unique variance in relationship qual- perpetrator amends plays an important role ity beyond forgiveness and amends (β = .28, beyond the effects of victim forgiveness, and p<.01). Tests of the significance of medi- (b) irrespective of role, partners’ perceptions ation revealed that the association of victim of betrayal resolution are influenced by part- forgiveness with relationship quality was sig- ners’ behavior at least as much as by their nificantly and wholly mediated by betrayal own behavior. resolution (controlling for perpetrator amends, As predicted, betrayal resolution was sig- z = 3.31, p<.01) and the association of nificantly associated with relationship qual- amends with relationship quality was signifi- ity in all three analyses (under Relation- cantly but partially mediated by betrayal res- ship quality, see single row labeled Betrayal olution (controlling for victim forgiveness, resolution); that is, relationship quality is z = 2.62, p<.01).5 greater among couples who more successfully resolve their betrayal incidents. Are our results consistent with the hypothesis 5. We were concerned with the possibility that some items in our relationship quality measure potentially that betrayal resolution mediates the associ- also tapped betrayal resolution (e.g., “I think my part- ations of amends and forgiveness with rela- ner and I learned something valuable from this inci- tionship quality (cf. Baron & Kenny, 1986; dent.”). We repeated the mediation analyses using a subset of relationship quality items that tapped the Kenny et al., 1998)? We addressed media- current state of the relationship (e.g., “Our relation- tion using the combined data. The prerequi- ship is better now than ever before.”). We found that sites for assessing mediation were satisfied, betrayal resolution significantly mediated the association of forgiveness and amends with relationship qual- in that both forgiveness and amends were ity using this 3-item version of the relationship quality associated with relationship quality (βs = .16 scale (α = .68). Amends, forgiveness, and betrayal resolution 271

Table 4. Predicting victim forgiveness, Potential confounds betrayal resolution, and relationship qual- Are the present findings attributable to unin- ity—Across-partner analyses: Study 3 tended confounds, such as commitment level Across-partner or tendencies toward socially desirable data responding? We replicated the Table 3 analyses (simple models) including as covariates Victim forgiveness (VR) from: measures of self-deception, impression man- Perpetrator amends (PR) .15† agement, and commitment level. In analyses Betrayal resolution (VR) from: that simultaneously controlled for the three ∗ Victim forgiveness (VR) .20 potential confounds, all associations among ∗ Perpetrator amends (PR) .21 key model variables remained significant (βs Betrayal resolution (PR) from: ranged from .18 to .34, all ps <.01). Victim forgiveness (VR) .10 ∗ Perpetrator amends (PR) .21 Victim and perpetrator perspectives Relationship quality (VR) from: Betrayal resolution (PR) .26∗∗ Our Table 3 findings are roughly parallel in Relationship quality (PR) from: analyses employing victim data and perpe- Betrayal resolution (VR) .32∗∗ trator data. To further explore the extent to which amends and forgiveness are phenomena that are perceived in parallel manner by Note. The values are standardized regression coefficients from hierarchical linear modeling analyses. Analyses are partners in a given relationship, we exam- based on data from 68 couples (136 individuals). Across- ined whether victims and perpetrators exhib- partner data = predictors and criteria as reported by ited rough agreement in their reports of one individual specified for each variable; VR = victim report another’s forgiveness and amends. To explore of variable; PR = perpetrator report of variable. ∗ ∗∗ simple across-partner agreement in reports †p<.10. p<.05. p<.01. of behavior, we regressed victims’ reports of each variable onto perpetrators’ reports of par- Across-partner analyses allel variables. All four across-partner asso- Is it possible that the Table 3 findings reflect ciations were significant (βs ranged from common method variance or tendencies .27 to .53, all ps <.01). Thus, partners toward consistency in self-report? To address exhibited good agreement in their ratings of this question, we employed across-partner one another’s behavior during a given betrayal analyses to examine the direct associations incident. In addition—and as reported ear- predicted in the dyadic model, employing lier—although preliminary analyses revealed victim-reports of victim behavior and no significant interactions as a function of perpetrator-reports of perpetrator behavior betrayal role, two main effects were signifi- (see Table 4). Most predicted associations cant: In comparison to victims, perpetrators were significant or marginal: (a) perpetrator- reported lower levels of victim forgiveness reported amends was positively associated (Ms = 5.05 and 4.85), t(68) = 4.57, p<.01, with victim-reported forgiveness (β = .15, as well as greater levels of relationship qual- p<.10) and betrayal resolution (β = .21, ity (Ms = 6.01 and 6.26), t(68) = 4.24, p< p<.05), and perpetrator-reported betrayal .05. Thus, although there were two role main resolution was positively associated with effects in mean levels of variables (e.g., per- victim-reported relationship quality (β = .26, petrators reported lower levels of victim for- p<.01), and (b) victim-reported betrayal giveness), victims’ and perpetrators’ percep- resolution was positively associated with tions of one another’s behavior were sig- perpetrator-reported relationship quality (β = nificantly associated, and the hypothesized .32, p<.01). However, victim-reported for- associations between amends and forgive- giveness was not significantly associated with ness typically were evident using both victim perpetrator-reported betrayal resolution (β = and perpetrator accounts of a given betrayal .10, ns). incident. 272 P. A. Hannon et al.

Discussion 2 by obtaining both partners’ reports of interaction behaviors following betrayal incidents, Study 3 revealed good support for our by examining more severe betrayal incidents, hypotheses. Examining partners’ descriptions and by examining how behavior during prior of recent betrayal incidents, we obtained con- betrayal incidents relates to present relation- sistent support for the hypothesized asso- ship quality. ciation of perpetrator amends with victim forgiveness. Moreover, the combined analyses revealed that both partners’ actions con- General Discussion tribute to the resolution of betrayal incidents, An observational study of married couples, and that betrayal resolution positively pre- an interaction record study of individuals in dicts relationship quality. These findings gen- ongoing dating relationships, and a cross- erally were evident for both victims and per- sectional survey study of dating partners petrators, with one interesting exception: In yielded findings that extend previous research analyses based on victim-report data, perpe- regarding forgiveness. These studies revealed trator amends accounted for significant unique that forgiveness and betrayal resolution are variance in betrayal resolution, whereas their thoroughly interpersonal processes— own forgiveness did not, and in analyses processes that are shaped by the behavior of based on perpetrator-report data the opposite both victim and perpetrator. Moreover, our was true—victim forgiveness accounted for findings regarding betrayal role are consistent significant unique variance in betrayal reso- with the claim that amends and forgiveness lution, whereas their own amends did not. are phenomena that are perceived in paral- This suggests that in understanding whether a lel manner by victims and perpetrators and betrayal is successfully resolved, the behav- that exert parallel effects on betrayal reso- ior of the partner may be somewhat more lution and relationship quality as perceived salient than one’s own behavior. Strikingly, by each partner. In the following paragraphs, the dyadic model received moderately good we review findings relevant to our theoretical support not only in within-participant analyses model and discuss their implications. but also in across-partner analyses examining the associations of self-report measures with Support for the dyadic model partner-report measures. Figure 2 presents a modified version of our The main limitation of these findings is dyadic model of forgiveness based on our that they rest on retrospective reports of prior findings across Studies 1–3 and displays betrayal incidents. It is possible that (a) recall meta-analytic coefficients for each link in of prior betrayal interactions might be shaped our model (with coefficients weighted by the by the present state of the relationship, or that number of participants in each study). First, (b) different aspects of betrayal incidents may we address results regarding the association be salient in the memories of the two part- of perpetrator amends with victim forgive- ners. However, our confidence in the validity ness. A meta-analytic summary of findings of participants’ descriptions is enhanced by from Studies 2 and 3 revealed a reliable the fact that partners exhibited good agree- concurrent association of perpetrator amends ment about the extent to which each per- with victim forgiveness (see Figure 2; meta- son did versus did not offer amends and analytic β = .27, p<.01). This association was versus was not forgiving, as well as by was evident not only in analyses employing the fact that across-partner analyses revealed victim data and perpetrator data but also in good support for model predictions. Study across-partner analyses and in analyses that 3 extends Study 1 by examining the reso- controlled for potential confounds such as lution of betrayal incidents, as well as the severity of betrayal, commitment level, and association of betrayal resolution with rela- tendencies toward socially desirable respond- tionship quality. And Study 3 extends Study ing. Thus, and consistent with predictions, Amends, forgiveness, and betrayal resolution 273

.36**

Perpetrator Amends

.21*

.28** Betrayal Relationship (.17** ) .27** (.08) Resolution Quality

.32**

Victim Forgiveness .07

Tested in Tested in Tested in Studies 1, 2, and 3 Studies 2 and 3 Study 3

Figure 2. The dyadic model of betrayal resolution: Associations observed in Studies 1, 2, and 3. Note. Coefficients in parentheses are residualized lagged associations. Dashed lines indicate insignificant model associations. *p<.05. **p<.01 when Mary offers amends for her offense, betrayal, to the extent that Mary apologizes John is more forgiving. And conversely, when for her actions or atones for the harm she Mary is defensive or hostile, John is less has caused, John becomes increasingly likely forgiving. In interdependence terms, Mary’s to let go of his hurt and anger, gradually amends indeed appear to facilitate John’s moving toward forgiveness. However, earlier transformation of motivation from potentially victim forgiveness does not reliably promote vengeful given preferences to effective pref- increases over time in perpetrator amends erences that favor a forgiving response. (meta-analytic β = .08, ns). This residualized Is this association a unidirectional relation- lagged association was significant in analyses ship wherein amends promotes forgiveness, employing victim data and perpetrator data, or is the association of amends with forgive- but not in across-partner analyses or in anal- ness reciprocal? In Study 1, we examined the yses employing observer data. Moreover, this temporal unfolding of amends and forgive- association was even less reliably observed ness in the course of betrayal-relevant inter- in analyses that controlled for potential con- actions. Study 1 revealed good support for founds. Therefore, it appears that the power of the assertion that earlier perpetrator amends positive behavior in the wake of betrayal may promotes increases over time in victim for- be dependent on the role of the actor; per- giveness (see Figure 2; meta-analytic β = .17, petrator amends more reliably inspires victim p<.01). This residualized lagged associ- forgiveness than victim forgiveness inspires ation was significant not only in analyses perpetrator amends. Mary’s amends create a employing victim data and perpetrator data powerful form of situation selection; to the but also in across-partner analyses and in extent that amends improves John’s willing- analyses employing observer data, as well as ness to forgive, the couple will have more in analyses that controlled for potential con- positive interactions. founds such as commitment level and socially Next, we address findings regarding the desirable responding. Following an act of associations of amends and forgiveness with 274 P. A. Hannon et al. betrayal resolution and relationship qual- for the restorative properties of perpetrator ity. A meta-analytic summary of findings amends. from Studies 2 and 3 revealed that perpe- Thus, although betrayal incidents tend to trator amends and victim forgiveness reli- be upsetting and potentially harmful, it is ably account for unique variance in betrayal important to recognize that partners’ behavior resolution (see Figure 2; meta-analytic βs = following such incidents can be highly diag- .21 and .32, both ps <.05). These associ- nostic of the degree to which they value their ations were evident not only in analyses relationship (cf. Holmes & Rempel, 1989). employing victim data and perpetrator data Their behavior provides meaningful infor- but also in across-partner analyses and in mation about (a) each person’s dispositions, analyses that controlled for diverse potential values, and motives, as well as (b) each per- confounds. Thus, and consistent with predic- son’s probable future behavior. After violating tions, forgiveness per se is no magic bullet Mary’s expectations about normative behav- in the resolution of betrayal dilemmas. The ior, if John promises not to do so again, apol- process of betrayal resolution is thoroughly ogizes for causing Mary pain, and behaves interpersonal, and rests on pro-relationship in such a manner as to partially “repay the transformation on the part of both victim and debt” he has incurred, Mary may have reason perpetrator. to believe that John has developed a better Finally, it is important to comment on understanding of her expectations and that he our findings regarding relationship quality. values their relationship sufficiently to change Study 3 revealed not only that betrayal reso- his ways. Conciliatory behavior of this sort lution is reliably associated with relationship may do a good deal to reduce victim uncer- quality (see Figure 2; meta-analytic β = .28, tainty, assuage anxiety, and increase intimacy p<.01) but also that perpetrator amends by communicating perpetrator understanding, accounts for unique variance in relation- validation, and caring for the victim (Lau- ship quality beyond betrayal resolution (meta- renceau, Barrett, & Pietromonaco, 1998; Reis analytic β = .36, p<.01). The former find- & Shaver, 1988). Of course, rather than offering is of consequence—it is important to ing amends, John might alternatively declare demonstrate that successful betrayal resolu- that Mary’s expectations are unreasonable or tion is associated with healthy couple func- belittle her for feeling hurt by his actions. tioning (and that failure to resolve betrayals Either way, Mary has learned a good deal is harmful). Yet the latter, unexpected find- about John that she would not learn under ing is perhaps even more striking. It appears sunnier circumstances of interdependence. that perpetrator amends yields benefits that extend beyond the fact that amends promotes Victim versus perpetrator perspectives betrayal resolution; that is, amends also helps couples move forward in a positive manner in All three studies allowed us to explore possi- the aftermath of betrayal. We speculate that ble differences between perpetrators and vic- the direct association of amends with couple tims in perceptions of betrayal incidents. In functioning may be attributable to the fact Study 2, we “controlled for person,” asking a that amends promotes broader healing pro- given individual to report on victim and per- cesses, perhaps by helping couples reestablish petrator behavior in the context of betrayal relationship-relevant norms or by promoting incidents in which he or she was the per- the recovery of trust or commitment. It is also petrator as well as in betrayal incidents in possible that these findings are in part due to which he or she was the victim. In Studies our exploring perpetrator amends (defined as 1 and 3, we “controlled for betrayal inci- behaviors to atone for the betrayal) in these dent,” asking two partners to report on their studies rather than just perpetrator apology own behavior and the partner’s behavior in (e.g., an explicit verbal statement about being a given betrayal incident. Consistent with sorry for the betrayal). Future research should prior research regarding betrayal role (e.g., explore the precise mechanisms that account Baumeister et al., 1990; Kearns & Fincham, Amends, forgiveness, and betrayal resolution 275

2005; Zechmeister & Romero, 2002), we interaction (Kelley et al., 2003; Rusbult & observed some evidence of biased or self- Van Lange, 2003). These broader effects are serving perception. For example, perpetrators not fully tractable in individual-focused social tended to perceive that they offered greater psychological analyses. amends than victims perceived they were An interdependence-based analysis also offered (Study 2), and tended to perceive that addresses the consequences of interaction for betrayals had more positive relational out- dyads. The present work revealed that partners comes (Study 3); victims tended to perceive who offer amends and forgiveness experience that they offered greater levels of forgive- greater betrayal resolution and more positive ness than perpetrators believed they received relational outcomes than those who do not. (Study 3). However, such role effects were These findings have clear potential for appli- unreliably observed (e.g., no role effects were cation in couples counseling and family ther- evident in Study 1). apy. Moreover, this work may have impor- Thus, victims and perpetrators do not live tant implications for understanding forgive- in separate worlds—partners’ perceptions are ness in other contexts—contexts involving considerably more convergent than divergent. nonclose dyads and groups. It seems plausi- At heart, many social psychologists are social ble that even in settings wherein no formal constructionists; we are intrigued by evi- relationship exists between victim and per- dence of biased perception and cognition. The petrator, perpetrator amends may yield ben- present research suggests that although part- eficial consequences.6 For example, when ners may shade their construals in such a criminal defendants exhibit remorse for their manner as to protect the self, this is not to actions, the victims of violent crimes may say that no shared reality underlies percep- more readily recover from their maltreatment; tion. In the present research, the elephant on when formal organizations apologize for their the stage was the reality of betrayal-relevant transgressions, they may enjoy more con- interaction, which was perceived in roughly genial relations with constituents (e.g., cor- parallel manner not only by partners but also porations’ apologies for shareholder abuse, by outside observers. governments’ apologies for prior group-based exploitation). Numerous interventions exist to help victims achieve forgiveness (Enright & Broader implications the Human Development Study Group, 1996); Folk wisdom suggests that “it takes two to the present work suggests the need for more tango”—that there must be two parties to a interventions that help people make amends quarrel. The present studies suggest that there when they have hurt others (intentionally must also be two parties to the resolution of a or not). quarrel. An interaction-based analysis of the forgiveness process allows us to recognize the Limitations and directions for future research contributions of both victims and perpetrators to the successful resolution of betrayal. Before closing, we should address several lim- To be sure, work regarding intra personal itations of the present work. Ironically, the processes (cognition and emotion) informs most serious limitation rests on a feature of our understanding of important social psycho- the present work that may also be regarded as logical phenomena. But beyond such intrap- ersonal processes, an interdependence-based 6. Interestingly, everyday experience suggests that even analysis illuminates our understanding of the in instances such as these, victims yearn for perpetrator interpersonal character of forgiveness and rec- amends. For example, during criminal trials, victims are soothed by signs of remorse on the part of criminal onciliation, highlighting the fact that behav- defendants; the victims of group-based exploitation ior during interaction (a) has important direct welcome apologies and confessions of wrongdoing consequences for individuals and (b) shapes on the part of formal “perpetrator representatives” (e.g., politicians’ apologies for reprehensible societal cognitive, emotional, and motivational events phenomena, such as the existence of slavery during in ways that govern the course of future prior generations). 276 P. A. Hannon et al. one of its greatest strengths: In all three stud- sincerity of amends and forgiveness that we ies, participants described betrayals that tran- did not include in these studies. spired in ongoing relationships. This approach A third limitation is that although we enabled us to address some important gaps examined betrayal role in Studies 1 and 3, in the forgiveness literature; we were able it must be recognized that all three studies to study betrayal interactions in real time, ultimately are nonexperimental. Therefore, we study both victim and perpetrator behavior cannot form confident conclusions regarding (and perceptions of partner behavior) during the causal relations among amends, forgive- the forgiveness process, and learn how both ness, and betrayal resolution. Although the victims and perpetrators perceive the after- residualized lagged analyses we performed in math of betrayal and forgiveness. However, Study 1 suggest that perpetrator amends may it could be argued that all of the couples in causally precede victim forgiveness, defini- Studies 1 and 3 experienced at least some tive causal evidence will require experimental degree of forgiveness and reconciliation; oth- work regarding perpetrator–victim interac- erwise, the couples would not have lived tion (e.g., via the use of priming techniques, to tell their tales. The procedures employed false feedback regarding perpetrator behav- in Study 2—asking individuals to complete ior). These methods often require a trade- diary records of betrayals in their dating rela- off between control and artificiality. In the tionships as they occurred—are not subject to case of real romantic relationships, there are this critique, but the betrayals captured with also ethical problems associated with manip- this method were comparatively mild. Due to ulating betrayal or amends, particularly when we seek to study severe betrayal incidents. this limitation, our findings about the forgive- Although we cannot draw causal conclusions ness process may be limited to describing how on the basis of the present work, we per- the forgiveness process unfolds in ongoing formed diverse confound analyses to help rule relationships; it remains to be seen whether out several potential alternative explanations, the same processes and relationships among such as socially desirable responding or com- our model constructs apply to betrayals that mitment to a relationship. Finally, the external occur in relationships that dissolve. Future validity of our findings may be questioned, studies should employ longitudinal methods in that our findings rest on evidence provided to explore betrayal incidents of even greater by North American participants. For example, consequence, toward determining whether the work employing role-playing methods sug- survival of a relationship is influenced by gests that in organizational settings, amends dyadic forgiveness processes. following promise breaking is regarded as A second limitation stems from the fact significantly more appropriate by Japanese that many of our findings rest on self-report undergraduates than by American undergrad- ratings of one’s own or a partner’s behavior. It uates (Takaku, 2000). is reassuring that partners exhibit moderate to good agreement in their descriptions of one another’s behavior. Nevertheless, given that Conclusions self-report measures are vulnerable to bias, The present research provides unique infor- our Study 1 analyses of observers’ ratings mation regarding the role of perpetrator are particularly compelling. However, given amends in the aftermath of betrayal, estab- that Study 1 did not assess betrayal resolu- lishing the importance of amends in facilitat- tion and relationship quality, future studies ing victim forgiveness and successful betrayal might seek to examine interaction-based ele- resolution. Moreover, the present research ments of the forgiveness process by studying revealed that both partners’ actions shape the interactions regarding both resolved and unre- manner in which betrayals are experienced solved betrayal incidents. Such studies would and resolved by couples. Furthermore, data also allow for studying additional aspects of obtained from both partners—irrespective the forgiveness process, such as perceived of their role in the betrayal—revealed the Amends, forgiveness, and betrayal resolution 277 importance of both victim and perpetrator Freedman, S. (1998). Forgiveness and reconciliation: The behavior in bringing about forgiveness and importance of understanding how they differ. Coun- betrayal resolution. These findings highlight seling and Values, 42, 200–217. Gonzales, M. H., Manning, D. J., & Haugen, J. A. (1992). the importance of moving beyond the tradi- Explaining our sins: Factors influencing offender tional victim-centered approach to explain- accounts and anticipated victim responses. Journal of ing forgiveness, illustrating the utility of an Personality and Social Psychology, 62, 958–971. interaction-based analysis of couple reactions Gottman, J. M. (1998). Psychology and the study of to betrayal. To the extent that both partners are marital processes. Annual Review of Psychology, 49, 169–197. able to enact constructive behaviors follow- Hodgins, H. S., & Liebeskind, E. (2003). Apology versus ing acts of betrayal, both are likely to reap defense: Antecedents and consequences. Journal of the rewards of enhanced outcomes for their Experimental Social Psychology, 39, 297–316. relationship. Hodgins, H. S., Liebeskind, E., & Schwartz, W. (1996). Getting out of hot water: Facework in social predica- ments. Journal of Personality and Social Psychology, References 71, 300–314. Holmes, J. G., & Rempel, J. K. (1989). Trust in close Baron, R. M., & Kenny, D. A. (1986). The moderator- relationships. In C. Hendrick (Ed.), Review of person- mediator variable distinction in social psychological ality and social psychology (Vol. 10, pp. 187–220). research: Conceptual, strategic, and statistical consid- London, England: Sage. erations. Journal of Personality and Social Psychol- Human Development Study Group. (1991). Five points ogy, 51, 1173–1182. on the construct of forgiveness within psychotherapy. Baumeister, R. F., Stillwell, A. M., & Heatherton, T. F. Psychotherapy: Theory, Research, Practice, Training, (1995). Personal narratives about guilt: Role in action 28, 493–496. control and interpersonal relationships. Basic and Kearns, J. N., & Fincham, F. D. (2005). Victim and per- Applied Social Psychology, 17, 173–198. petrator accounts of interpersonal transgressions: Self- Baumeister, R. F., Stillwell, A. M., & Wotman, S. R. serving or relationship-serving biases? Personality (1990). Victim and perpetrator accounts of inter- and Social Psychology Bulletin, 31, 321–333. personal conflict: Autobiographical narratives about Kelley, H. H. (1984). The theoretical description of inter- anger. Journal of Personality and Social Psychology, dependence by means of transition lists. Journal of 59, 994–1005. Personality and Social Psychology, 47, 956–982. Bolger, N., Davis, A., & Rafaeli, E. (2003). Diary meth- Kelley, H. H., Holmes, J. G., Kerr, N. L., Reis, H. T., ods: Capturing life as it is lived. Annual Review of Rusbult, C. E., & Van Lange, P. A. M. (2003). An Psychology, 54, 579–616. atlas of interpersonal situations.NewYork,NY: Bolger, N., Zuckerman, A., & Kessler, R. C. (2000). Cambridge University Press. Invisible support and adjustment to stress. Journal of Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data Personality and Social Psychology, 79, 953–961. analysis in social psychology. In D. T. Gilbert, S. T. Eaton, J., Struthers, C. W., Shomrony, A., & Santelli, Fiske, & G. Lindzey (Eds.), Handbook of social A. G. (2007). When apologies fail: The moderating psychology (4th ed., Vol. 1, pp. 233–265). New York, effect of implicit and explicit self-esteem on apology NY: McGraw-Hill. and forgiveness. Self and Identity, 6, 209–222. Kenny, D. A., Mannetti, L., Pierro, A., Livi, S., & Enright, R. D., & the Human Development Study Group. Kashy, D. A. (2002). The statistical analysis of data (1996). Counseling within the forgiveness triad: On from small groups. Journal of Personality and Social forgiving, receiving forgiveness, and self-forgiveness. Psychology, 83, 126–137. Counseling and Values, 40, 107–122. Laurenceau, J. P., Barrett, L. F., & Pietromonaco, P. R. Exline, J. J., DeShea, L., & Holeman, V. T. (2007). Is (1998). Intimacy as an interpersonal process: The apology worth the risk? Predictors, outcomes, and importance of self-disclosure, partner disclosure, and ways to avoid regret. Journal of Social and Clinical perceived partner responsiveness in interpersonal Psychology, 26, 479–504. exchanges. Journal of Personality and Social Psychol- Fincham, F. D. (2000). The kiss of the porcupine: From ogy, 74, 1238–1251. attributing responsibility to forgiving. Personal Rela- Leary, M. R., Springer, C., Negel, L., Ansell, E., & tionships, 7, 1–23. Evans, K. (1998). The causes, phenomenology, and Finkel, E. J., Rusbult, C. E., Kumashiro, M., & Hannon, consequences of hurt feelings. Journal of Personality P. A. (2002). Dealing with betrayal in close rela- and Social Psychology, 74, 1225–1237. tionships: Does commitment promote forgiveness? McCullough, M. E., Bono, G., & Root, I. M. (2007). Journal of Personality and Social Psychology, 82, Rumination, emotion, and forgiveness: Three longitu- 956–974. dinal studies. Journal of Personality and Social Psy- Frantz, C. M., & Bennigson, C. (2005). Better late than chology, 92, 490–505. early: The influence of timing on apology. Journal of McCullough, M. E., Rachal, K. C., Sandage, S. J., Experimental Social Psychology, 41, 201–207. Worthington, E. L., Brown, S. W., & Hight, T. L. 278 P. A. Hannon et al.

(1998). Interpersonal forgiving in close relationships: level, satisfaction level, quality of alternatives, and II. Theoretical elaboration and measurement. Journal investment size. Personal Relationships, 5, 357–391. of Personality and Social Psychology, 75, 1586–1603. Rusbult, C. E., & Van Lange, P. A. M. (2003). Inter- McCullough, M. E., Worthington, E. L., & Rachal, K. C. dependence, interaction, and relationships. Annual (1997). Interpersonal forgiving in close relationships. Review of Psychology, 54, 351–375. Journal of Personality and Social Psychology, 73, Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F., 321–336. & Lipkus, I. (1991). Accommodation processes in Mullet, E., Houdbine, A., Laumonier, S., & Girard, M. close relationships: Theory and preliminary empirical (1998). “Forgivingness”: Factor structure in a sample evidence. Journal of Personality and Social Psychol- of young, middle-aged, and elderly adults. European ogy, 60, 53–78. Psychologist, 3, 289–297. Struthers, C. W., Eaton, J., Santelli, A. G., Uchiyama, M., Nezlek, J. B. (2001). Multilevel random coefficient anal- & Shirvani, N. (2008). The effects of attributions of yses of event-and interval-contingent data in social intent and apology on forgiveness: When saying sorry and personality psychology research. Personality and may not help the story. Journal of Experimental Social Social Psychology Bulletin, 27, 771–785. Psychology, 44, 983–992. Paleari, F. G., Regalia, C., & Fincham, F. (2005). Marital Takaku, S. (2000). Culture and status as influences quality, forgiveness, empathy, and rumination: A lon- on account giving: A comparison between the United gitudinal analysis. Personality and Social Psychology States and Japan. Journal of Applied Social Bulletin, 31, 368–378. Psychology, 30, 371–388. Paulhus, D. L. (1984). Two-component models of Thompson, L. Y., Snyder, C. R., Hoffman, L., Michael, socially desirable responding. Journal of Personality S. T., Rasmussen, H. N., Billings, L. S., . . . , Roberts, and Social Psychology, 46, 598–609. D. E. (2005). Dispositional forgiveness of self, others, Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical and situations. Journal of Personality, 73, 313–359. linear models: Applications and data analysis meth- Weiner, B., Graham, S., Peter, O., & Zmuidinas, M. ods. London, England: Sage. (1991). Public confession and forgiveness. Journal of Reis, H. T., & Gable, S. L. (2000). Event-sampling and Personality, 59, 281–312. other methods for studying everyday experience. Worthington, E. L. (Ed.). (2005). Handbook of forgive- In H. T. Reis & C. M. Judd (Eds.), Handbook of ness. New York, NY: Brunner-Routledge. research methods in social and personality psychology Yovetich, N. A., & Rusbult, C. E. (1994). Accommoda- (pp. 190–222). Cambridge, England: Cambridge Uni- tive behavior in close relationships: Exploring trans- versity Press. formation of motivation. Journal of Experimental Reis, H. T., & Shaver, P. (1988). Intimacy as an inter- Social Psychology, 30, 138–164. personal process. In S. Duck (Ed.), Handbook of Zechmeister, J. S., Garcia, S., Romero, C., & Vas, S. N. personal relationships: Theory, relationships, and (2004). Don’t apologize unless you mean it: A interventions (pp. 367–389). Chichester, England: laboratory investigation of forgiveness and retalia- Wiley. tion. Journal of Social and Clinical Psychology, 23, Risen, J. I., & Gilovich, T. (2007). Target and observer 532–564. differences in the acceptance of questionable apologies. Zechmeister, J. S., & Romero, C. (2002). Victim and Journal of Personality and Social Psychology, 92, offender accounts of interpersonal conflict: Autobi- 418–433. ographical narratives of forgiveness and unforgive- Rusbult, C. E., Martz, J. M., & Agnew, C. R. (1998). ness. Journal of Personality and Social Psychology, The investment model scale: Measuring commitment 82, 675–686.