in the wake of betrayal: amends, forgiveness, and the...

Personal Relationships, 17 (2010), 253–278. Printed in the United States of America.Copyright © 2010 IARR

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

PEGGY A. HANNON,a CARYL E. RUSBULT,b ELI J. FINKEL,c AND MADOKAKAMASHIROd

aUniversity of Washington; bVrije Universiteit, Amsterdam; cNorthwestern University;dGoldsmiths, University of London, United Kingdom

AbstractThe present work advances a dyadic model of victim and perpetrator interactions following betrayals, and the effectof their interactions on betrayal resolution and relationship quality. The authors propose that perpetrator amendspromotes 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 trainedobservers. In Study 2, dating individuals used interaction records to describe betrayal incidents perpetrated bythemselves or the partner over a 2-week period. In Study 3, dating partners both provided retrospective descriptionsof prior betrayal incidents. All studies yielded good support for model predictions, revealing parallel findings fromthe point of view of victims, perpetrators, and external observers.

Close relationships are the origin of manypositive experiences. We feel exhilaratedwhen we develop new interdependencies, cel-ebrate positive interpersonal events in ourown and others’ lives, and are deeply grat-ified by relationships that are healthy, vital,and enduring. Paradoxically, it is in roman-tic relationships that we also find ourselvesmost vulnerable to suffering. At some point in

Peggy A. Hannon, Department of Health Services, Uni-versity of Washington; Caryl E. Rusbult, Department ofSocial Psychology, Vrije Universiteit, Amsterdam, TheNetherlands; Eli J. Finkel, Department of Psychology,Northwestern University; Madoka Kamashiro, Depart-ment of Psychology, Goldsmiths, University of London,London, United Kingdom.

This research was supported by a grant to the sec-ond author from the Templeton Foundation (Grant 5158).Manuscript preparation was facilitated by National Can-cer Institute Training Grant 5R25CA092408. Study 3 waspart of Peggy A. Hannon’s doctoral dissertation, super-vised by Caryl E. Rusbult. We would like to thank theMarriage Project Team at the University of North Car-olina at Chapel Hill for all of their contributions to thestudies presented in this article.

Correspondence should be addressed to Peggy A.Hannon, University of Washington, Department of HealthServices, 1107 NE 45th Street, Suite 200, Seattle, WA98105, e-mail: [email protected].

most relationships, one partner will betray theother, violating relationship-relevant norms byengaging in acts of deception or disloyalty.How do we manage to sustain relationshipsin which we suffer such harm, and what arethe consequences of forgiving versus failingto forgive a partner’s act of betrayal?

The past two decades have witnessed adramatic increase in the study of forgiveness.Early research in this tradition was largely vic-tim centered, examining the personality traitsthat are associated with forgiveness, the cog-nitive and affective events that predict for-giveness, and the personal values that promoteforgiveness (for a review, see Worthington,2005). In recent years, scientists have turnedtheir attention to interpersonal elements ofthe forgiveness process, examining proper-ties of partners and relationships that makeforgiveness more versus less probable. Thepresent work contributes to this emerging lit-erature in two respects. First, we advance adyadic model of forgiveness in the contextof close relationships, exploring the interplaybetween perpetrator amends and victim for-giveness and the associations of each variable

253

254 P. A. Hannon et al.

with betrayal resolution and its relationalconsequences. Second, we test our model byexamining both victim and perpetrator per-spectives in real couples confronted with realbetrayal incidents. Few studies have examinedperpetrators’ perceptions of the forgivenessprocess, and fewer still have explored per-petrator perspectives in the context of actualbetrayals (as opposed to hypothetical betray-als or stranger betrayals).

Our dyadic model rests on the principles ofinterdependence theory (Kelley et al., 2003;Rusbult & Van Lange, 2003) and suggests thatperpetrator amends enhances victim motiva-tion to forgive. Moreover, we propose thatboth perpetrator amends and victim forgive-ness play key roles in the successful reso-lution of betrayal incidents and that betrayalresolution is beneficial to relationships fromthe perspective of both victims and perpe-trators. As initial tests of our model, wereport the results of three studies of betrayal,amends, and forgiveness in ongoing roman-tic relationships. The studies employ diversemethods, including the observation of coupleinteractions regarding unresolved betrayals,interaction diary reports of relational betray-als during a 2-week period, and retrospectiveaccounts of betrayals.

Perpetrator betrayal and victim forgiveness

We define betrayal as the perceived violationof an implicit or explicit relationship-relevantnorm (Finkel, Rusbult, Kumashiro, & Han-non, 2002). Individuals experience betrayalwhen they believe that a partner has know-ingly departed from the norms of decency andfairness that are assumed to govern a rela-tionship, thereby causing harm. Given thatbetrayals are harmful to victims and violatemorality-based expectations, victims typicallyexperience righteous indignation, believe thatthe perpetrator has incurred an interpersonaldebt, and perceive that such incidents bodepoorly for their relationships (Leary, Springer,Negel, Ansell, & Evans, 1998). Whether acouple can survive and recover from such anincident rests on how betrayals are resolved.

Previous work has defined forgiveness as“the set of motivational changes whereby one

becomes decreasingly motivated to retaliateagainst an [offender], decreasingly motivatedto maintain estrangement from the offender,and increasingly motivated by conciliationand goodwill for the offender” (McCullough,Worthington, & Rachal, 1997, pp. 321–322).Our definition is consistent with this tradi-tion yet represents forgiveness as both a psy-chological and behavioral event. We defineforgiveness as the victim’s willingness to(a) forego vengeance and demands for ret-ribution and (b) react to the betrayal in aconstructive, less judgmental manner (Finkelet al., 2002).

Given that victims’ immediate, gut-levelimpulses frequently include desire for grudgeor vengeance, how do victims find theirway to forgiveness? Interdependence the-ory describes immediate, gut-level reactionsas given preferences, in that they are self-oriented, asocial, and focus on the hereand now (Kelley et al., 2003). People departfrom self-oriented, given preferences as aresult of transformation of motivation, apsychological process whereby victims takeinto account considerations extending beyonddirect self-interest, including long-term goals,social dispositions and values, or concern fora partner’s well-being. The modified pref-erences resulting from transformation aretermed effective preferences ; these prefer-ences guide behavior.

In betrayal situations, the victims’ trans-formation from vengeful impulses to pro-relationship motives may not be effortless,uncomplicated, or automatic. The impulsetoward negative reciprocity is strong—peopleare inclined to fight fire with fire, respond-ing in kind to a partner’s real or imaginednegativity (Gottman, 1998; Rusbult, Verette,Whitney, Slovik, & Lipkus, 1991; Yovetich &Rusbult, 1994). Moreover, betrayal-inspiredmoral outrage and perceived debt may keepsuch incidents alive and feed the flames—victims may ruminate about a betrayal andits implications, such that vengeful impulsesmay linger for a considerable period of time(Fincham, 2000; McCullough, Bono, & Root,2007). How are the flames of righteousindignation quelled?

Amends, forgiveness, and betrayal resolution 255

Perpetrator amends and victim forgiveness

We suggest that although forgiveness ulti-mately rests in the hands of victims, perpe-trator actions affect both the probability offorgiveness and the likelihood that a betrayalwill be successfully resolved (Human Devel-opment Study Group, 1991). Interdependencetheory proposes that to understand the resolu-tion of interdependence dilemmas, interactionis the name of the game; that is, both partners’actions matter (Kelley et al., 2003; Rusbult& Van Lange, 2003). Given that most extantwork regarding forgiveness has focused onvictims, we know relatively little about theprocesses by which couples achieve forgive-ness and resolve betrayals. We know even lessabout how perpetrators in romantic relation-ships perceive this process and their role in it.

We define perpetrator amends as acceptingresponsibility for an act of betrayal, offer-ing genuine atonement for one’s actions.We employ this broad definition of amendsto acknowledge our belief that in the con-text of close relationships, perpetrator amendsmay not always include an explicit, verbalapology. Importantly, amends must be sin-cere—perpetrator acts that are perceived asinsincere tend to backfire, such that disingen-uous amends is likely to inhibit forgivenessand betrayal resolution (e.g., Exline, DeShea,& Holeman, 2007; Risen & Gilovich, 2007;Zechmeister, Garcia, Romero, & Vas, 2004).

How does amends work its magic? Ininterdependence terms, amends serves as aform of situation selection, moving a coupletoward interaction opportunities with supe-rior behavioral options and outcomes (Kelley,1984; Kelley et al., 2003). When John com-municates that he feels betrayed, Mary maycalmly discuss the incident with him. Recog-nizing that her actions hurt him, she may offeramends, apologizing for her actions or mak-ing things right by atoning for the harm shehas caused. In turn, John may find it easier tolet go of his hurt and anger. John’s gradualmovement toward forgiveness may also makeit easier for Mary to offer further amends, con-tinuing to respond in a loving and benevolentmanner. Thus, perpetrator amends and victim

forgiveness may be mutually reinforcing overthe course of extended interaction.

Of course, perpetrators may not find it easyto offer amends. In the wake of betrayal, Marymay suffer sadness, shame, or guilt (Baumeis-ter, Stillwell, & Heatherton, 1995). If Johndwells on the incident or is reproachful, Marymay exhibit defensive cognitive maneuvers,seeking to justify her behavior not only toJohn but also to herself (Baumeister, Stillwell,& Wotman, 1990). And perpetrators may notbe willing to suffer extended blame—if Johnis persistently hostile or vengeful, Mary mayreact with reciprocal hostility, refusing tomake amends (Hodgins & Liebeskind, 2003;Hodgins, Liebeskind, & Schwartz, 1996).Thus, just as it is difficult for victims to findtheir way to forgiveness, it is also difficultfor perpetrators to find their way to amends.If Mary reacts defensively to John’s anger,claiming that she committed no offense andthat he has no right to feel upset, forgivenessand betrayal resolution become unlikely.

The extant literature offers some supportfor our claims, demonstrating that apology ispositively associated with forgiveness (mostprior work has studied explicit apology ratherthan overall amends; Frantz & Bennigson,2005; McCullough et al., 1997; McCulloughet al., 1998; Mullet, Houdbine, Laumonier, &Girard, 1998; Zechmeister et al., 2004). How-ever, recent studies have revealed that theassociation between apology and forgivenesscan also be negative. For example, in situa-tions where the perpetrator’s offense was per-ceived as intentional (Struthers, Eaton, San-telli, Uchiyama, & Shirvani, 2008) or wherethe victim has low implicit self-esteem (Eaton,Struthers, Shomrony, & Santelli, 2007), apol-ogy is associated with decreased forgiveness.Especially in the context of close, ongo-ing relationships, there is more to be dis-covered regarding the association of amendswith forgiveness:1 First, it is unclear whether

1. In a recent PsychInfo (an online database of articles,book chapters, and other scholarly publications inpsychology and related disciplines) search wherein wecombined all variants of forgiveness with all variantsof amends, apology, and atonement, we uncovered31 studies of forgiveness in dyads. Of these, only 5studies examined perpetrator perspectives on betrayal


perpetrators perceive the positive effects ofamends on victim forgiveness, at least tothe same extent that victims do. If thesepositive effects are not apparent to perpe-trators, perpetrators may not be motivatedto make amends—even in the context ofa committed relationship—especially giventhat amends may not be an automatic or effort-less response. Second, most research has rep-resented apology as a discrete event, examin-ing a causal sequence wherein apology pro-motes victim forgiveness, and forgiveness isthe endpoint—most work has not examinedwhat transpires following victim forgiveness.It is important to demonstrate not only thatamends promotes forgiveness but also thatboth variables—from the perspective of bothpartners—contribute to successful betrayalresolution.

Betrayal resolution and postbetrayalrelationship quality

Thus, rather than representing forgiveness asthe endpoint of betrayal incidents, we adopta broadened time frame, examining not onlyamends and forgiveness but also reconcilia-tion (cf. Freedman, 1998). Why so? Thereis no guarantee that amends and forgivenesswill necessarily yield successful betrayal res-olution and the recovery of couple function-ing. Even when a perpetrator offers sincereamends and a victim genuinely forgives, part-ners may find that they cannot forget theincident or fully relegate it to the past. Forexample, even though Mary offers heart-felt amends and John genuinely forgivesher for her betrayal, they may continue toreview betrayal-relevant events (“whose faultwas that, really?”), surreptitiously monitorone another’s actions (“is she sufficientlyrepentant?” “does he still feel hurt?”), suf-fer reduced trust (“could that happen again?”),or otherwise interact in an unnatural manner.

incidents; 4 of these examined hypothetical or stagedbetrayals in nonintimate relationships and 1 examinedreal betrayals in relationships of unspecified types.We located no studies that examined victim andperpetrator perspectives on specific, shared betrayalincidents, and we located no studies that examinedperspectives on betrayal resolution.

Thus, a betrayal may be forgiven yet still bevery much alive.

To further our understanding of the after-math of betrayal, it is important to exam-ine betrayal resolution, or the perception—byeach partner—that a betrayal incident hasbeen successfully closed, such that it nolonger colors interaction. Moreover, it isimportant to recognize that broader relation-ship quality may be influenced by betrayalresolution: Navigating a potentially harm-ful betrayal may affect partners’ relationshipexpectations, influence their sense of commit-ment, or yield other cognitive or affective con-sequences. In the present work, we examinehow partners experience relationship qualityin light of a specific betrayal incident and itsresolution, thereby extending prior researchthat has examined the association of globalforgiveness tendencies with global satisfac-tion or adjustment (e.g., Paleari, Regalia, &Fincham, 2005; Thompson et al., 2005).

Victim versus perpetrator perspectives

Our work is also guided by a final goal: Manyextant studies of forgiveness have examinedevents solely from the victim’s viewpoint.Studies examining both victim and perpetra-tor perspectives reveal that partner percep-tions are not necessarily parallel; for example,victims may be less likely to perceive that abetrayal is resolved, perpetrators may mini-mize the impact of a betrayal, and both partiesmay perceive circumstances in a self-servingmanner (Baumeister et al., 1990; Hodgins &Liebeskind, 2003; Kearns & Fincham, 2005;Zechmeister & Romero, 2002). Thus, it wouldbe easy to imagine that victims and perpe-trators experience circumstances quite differ-ently and that their differing perceptions affectthe impact of each person’s actions on theperceptions and actions of the other.

We suggest that the extent and conse-quences of such bias need further exploration.Partners may shade their construals in sucha manner as to protect the self, but this isnot to say that there is no reality underly-ing perception; such bias may well operateat the margins. Indeed, many prior studiesof role-based bias have employed procedures


that may yield exaggerated evidence of role-based bias; for example, many prior studiesrest on participant-selected incidents from thepoint of view of one party to a betrayal,such that role may affect incident selection(e.g., victims and perpetrators recall differ-ent incidents, describe incidents differently,or select incidents that reflect favorably onthemselves). By examining real betrayal inci-dents through the eyes of both parties, wehope to determine whether amends and for-giveness are phenomena that are experiencedsimilarly by partners, and that exert paralleleffects on betrayal resolution and relationshipquality.

Hypotheses and research overview

As displayed in Figure 1, our dyadic modelsuggests that to the extent that perpetratorsoffer amends for their actions, victim for-giveness is more probable. In turn, perpe-trator amends and victim forgiveness exertindependent effects on betrayal resolution,and betrayal resolution promotes enhancedrelationship quality. Some extant studies ofconfession or apology have revealed find-ings consistent with one or more links in thismodel (e.g., Gonzales, Manning, & Haugen,1992; Weiner, Graham, Peter, & Zmuidinas,

1991). However, most such work has studiedhypothetical betrayals in nonintimate relation-ships and has ended with whether forgivenesswas given or denied. We suggest that it isimportant to examine real betrayals in real,ongoing relationships, in that (a) althoughsignificant hurtful events may occur duringstranger interactions (e.g., in a violent crime),it is in the context of ongoing relation-ships that meaningful, relationship-relevanttransgressions transpire; (b) the question ofwhether amends and forgiveness promotebetrayal resolution and relationship quality ismost relevant for ongoing relationships, inwhich there is not only a history but alsoa (potential) future; and (c) as noted earlier,from a methodological point of view it is cru-cial to examine both partners’ perceptions ofreal betrayal incidents.

We conducted three studies to examine theassociations among two or more model vari-ables. In Study 1, married couples discussedan unresolved betrayal incident. Immediatelyfollowing the discussion, participants vieweda video recording of their interaction andrated their own and the partner’s behavior;trained observers provided parallel ratings.In Study 2, participants completed interactionrecords to describe all betrayal incidents thattranspired in their dating relationships during

Perpetrator Amends

Victim Forgiveness

Betrayal Resolution

Relationship Quality

Tested in Tested in

Studies 2 and 3Studies 1,2, and 3

Tested in Study 3

+

+

+

+

+

Figure 1. The dyadic model of betrayal resolution.


a 2-week period, whether they were perpetra-tors or victims in each incident. In Study 3,both partners in dating relationships identifiedspecific prior betrayal incidents and describedtheir own and the partner’s behavior duringeach incident.

In Study 1, we assessed amends and for-giveness at multiple times over the courseof interaction to assess whether (a) earlieramends predicts increases over time in for-giveness. We also assessed whether (b) earlierforgiveness predicts increases over time inamends to examine the possibility that for-giveness and amends may operate in a mutu-ally reinforcing manner. In Studies 2 and3, we not only evaluated whether perpetra-tor amends and victim forgiveness are pos-itively associated but also whether (c) bothamends and forgiveness contribute uniquevariance to predicting successful betrayal res-olution. In Study 3, we also assessed whether(d) betrayal resolution predicts relationshipquality. In addition to testing these core pre-dictions of the dyadic model, we examinedvictim–perpetrator role effects. We antici-pated that although we might observe rolemain effects in levels of variables (e.g., vic-tims might perceive greater forgiveness thanperpetrators), the hypothesized associationsamong amends, forgiveness, and betrayal res-olution would be evident using both victimand perpetrator accounts of events.

Study 1

In Study 1, we asked married partners todiscuss an unresolved betrayal incident. Fol-lowing the interaction, each partner viewed avideo recording of the discussion and ratedhis or her own and the partner’s behavior. Inaddition, trained observers later viewed thevideotaped interaction and rated both part-ners’ behavior. By asking couples to dis-cuss an unresolved incident, we were ableto observe perpetrator and victim behav-iors as they unfolded during the course ofbetrayal-relevant interaction. We asked cou-ples to rate positive behavior rather thanexplicitly asking them to rate amends andforgiveness; trained observers rated positivebehaviors and perpetrator amends and victim

forgiveness. By examining behavior through-out the discussion, we were able to testwhether perpetrator positive behavior predictsincreases over time in victim positive behav-ior, as well as whether victim positive behav-ior predicts increases over time in perpetratorpositive behavior. Given that both partnersparticipated in the discussion and rated eachperson’s behavior, we were also able to exam-ine victim and perpetrator accounts of eachincident. In Study 1, we examined betrayalincidents that were not resolved at the timeof the study, so we were unable to assess theassociations of victim and perpetrator positivebehavior with betrayal resolution.

Method

Participants and recruitment

Seventy-nine married couples participated in astudy of marital processes. The data for 4 cou-ples were deleted from the analyses (2 for fail-ing to follow instructions, 1 because they werenot married, and 1 due to videotaping tech-nical difficulties), leaving a total of 75 cou-ples. Participants were recruited via noticesposted on the campus and in the communityof the University of North Carolina, as wellas through advertisements in local newspa-pers. All announcements briefly described theproject, indicated that the study involved threeresearch sessions over an 8-month period,noted that couples would be paid $50.00 fortaking part in each research session, and pro-vided contact information. When couples con-tacted us, we provided more detailed infor-mation about project activities and scheduledappointments. Data from the initial researchsession are presented here; data collected dur-ing the second and third sessions are not rel-evant to the present work.

Participants were 34 years old on average(SD = 11 years), and most were Caucasian(80% Caucasian, 11% African American,4% Hispanic, 2% Asian American, and 3%Other). Most participants had at least 4 yearsof college education (44% obtained advancedor professional degrees, 38% completed 4years of college, 11% completed 2 years ofcollege, and 7% completed high school only).


Their median personal income was $20,000–$30,000 per year. Participants had been mar-ried to one another for 6 years on average(SD = 9 years), and most did not have chil-dren (74% no children, 10% one child, 8%two children, and 8% three or more children).

Procedure

Upon arrival at the laboratory, we asked eachparticipant to complete a brief questionnairethat was later used to identify a suitable recentbetrayal incident for a videotaped conversa-tion. We did not use the word betrayal in ourinstructions because this word may connotesexual forms of disloyalty or infidelity, andwe did not want to limit participants to thesetypes of betrayals. Instead, instructions to par-ticipants described such incidents as follows:

All of us have expectations about how ourpartners should treat us. No matter howwell-behaved your partner may be in gen-eral, from time to time he or she is likely to“break the rules.” For example, your part-ner may tell a friend something that youthink should have remained private; yourpartner may do something that is hurtfulbehind your back; your partner may flirtwith another person; or your partner mayotherwise violate the rules that govern yourmarriage.

Each partner was asked to describe threesuch incidents from the past 4 months, pro-viding simple ratings of each incident on 9-point scales (e.g., “How upsetting was it?”0 = not upsetting at all, 8 = very upsetting).To identify an incident for discussion, werandomly determined whether to use an inci-dent described by the husband or the wife,and selected an incident that was moder-ately upsetting, that was not totally resolved,and that the partners were willing to discuss.While the experimenter selected an incidentfor couples to discuss, participants completedother activities that are not relevant to thepresent work.

During the videotaped interaction, part-ners were seated at adjacent sides of a tablewith a microphone positioned in front of each

person. A camera was oriented to videotapeboth partners. Following a 2-min warm-upconversation (discussing the events of the pre-vious day), the experimenter explained thatwe had randomly determined which part-ner’s incident would be discussed and selectedone of that person’s incidents as the discus-sion topic. The experimenter read the inci-dent description aloud; then partners weregiven 1 min to describe the incident, as ameans of helping them bring the incident tomind. After the experimenter left the labora-tory, the couple spent 8 min discussing theincident. Immediately following the interac-tion, partners individually reviewed and ratedthe videotaped interaction. At the end of thesession, couples were paid, thanked for theirassistance, and partially debriefed.

Perpetrator and victim ratings ofinteractions

Following their interaction, partners were ledto separate video monitors and individu-ally reviewed the videotaped interaction. Theexperimenter stopped the videotape at the endof each 2-min segment of the interaction, ask-ing participants to rate their own and thepartner’s behavior during that segment. Eachparticipant first rated his or her own behav-ior during a given segment and then rated thepartner’s behavior. To avoid communicatingour hypotheses, we used parallel scales forratings of victim and perpetrator behav-ior, employing concrete descriptors of spe-cific interaction behaviors (with appropriatechanges in item wording to reflect own vs.partner behavior). Six items measured posi-tive behavior during the interaction (e.g., “Mypartner tried to comfort me,” “I raised myvoice toward my partner” [reverse-scored],“My partner wanted to cut off the inter-action” [reverse-scored]; for all items, 0 =do not agree at all, 8 = agree completely).Ratings of positive behavior exhibited goodreliability as rated by both perpetrators (forratings of perpetrator positive behavior, αs forthe four 2-min segments ranged from .80 to.83; for ratings of victim positive behavior, αsranged from .77 to .81) and victims (for per-petrator positive behavior, αs ranged from .77


to .82; for victim positive behavior, αs rangedfrom .72 to .76). We calculated measures ofperpetrator and victim positive behavior byaveraging participants’ ratings of each per-son’s behavior during each 2-min segment.

Measuring possible confounds

Participants also completed additional instru-ments that were used in analyses we per-formed to control for possible confounds.Given that reports of positive behavior maybe vulnerable to socially desirable responsetendencies, we asked participants to completea 20-item version of the Balanced Inven-tory of Desirable Responding that includedthe 10 most reliable items from the self-deception and impression management sub-scales (Paulhus, 1984; e.g., “I always obeylaws, even if I’m unlikely to get caught”; 1 =do not agree at all, 7 = agree completely;αs = .69 and .60). In the present sample ofmarried couples, we were concerned that com-mitment (rather than partner behavior) mightpowerfully drive behavior; that is, the rela-tionship with the spouse might be so impor-tant that maintaining the relationship wouldtake precedence over all else. To explore thispossibility, we measured participants’ com-mitment level using a 15-item version ofthe Investment Model Scale (Rusbult, Martz,& Agnew, 1998; e.g., “I would feel veryupset if our relationship were to end inthe near future”; 0 = do not agree at all, 8 =agree completely; α = .92).

Observer ratings of interactions

In addition to obtaining participants’ rat-ings of their interaction, we also asked twotrained observers to rate each couple’s inter-action. Observers received extensive train-ing, including studying and mastering a cod-ing manual and participating in a 4-hr train-ing session. Observers viewed the same2-min interaction segments as participants.Observer ratings of participant interactionswere structured in such a manner as to promptglobal, abstract ratings of amends and for-giveness: Observers first provided ratings ofconcrete interaction behaviors (e.g., “behavedin a warm/friendly manner,” “behaved in a

critical manner” [reverse-scored], “exhibitedhostility” [reverse-scored]) and then com-pleted two additional one-item global ratings(“perpetrator offered amends” and “victimwas forgiving”; for all items, 1 = no evidenceof this behavior, 5 = very strong evidence ofthis behavior). We calculated measures ofperpetrator amends and victim forgivenessby averaging the two observers’ global rat-ings of the perpetrator’s amends and the vic-tim’s forgiveness during each 2-min segment.(Parallel findings were evident in analysesusing observers’ concrete ratings of specificinteraction behaviors; for the analyses dis-played in Table 1, the same six residualizedlagged associations were significant.) Interob-server agreement was acceptable for ratings ofboth amends and forgiveness (for amends, αsacross the four 2-min segments ranged from.69 to .77; for forgiveness, αs ranged from .54to .72).

Character of betrayal incidents

Consistent with our goals for the betrayalinteraction, the betrayal incidents partnersdiscussed were described as moderately toseverely upsetting (M = 5.17, SD = 1.68)and were described as not yet fully resolved(M = 4.76, SD = 2.57). The incidents inclu-ded violations of dependence norms (e.g.,overspending after agreeing to save money),monogamy norms (e.g., not trusting the part-ner with people of the opposite gender), pri-vacy norms (e.g., discussing embarrassingtopics in front of family members or neigh-bors), and decency/etiquette norms (e.g., vol-unteering the partner for something withoutasking).

Results

Analysis strategy

Each participant provided four sets of rat-ings of the videotaped betrayal interaction—ratings of the victim and ratings of the per-petrator for each 2-min segment. Our trainedobservers provided parallel ratings. Multi-ple ratings from a given individual are notindependent, nor are the ratings of the twopartners in a given couple. Therefore, we


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

Victim data Perpetrator data Across-partner data Observer data

Later victim positive behavior from:Earlier perpetrator

positive behavior.18∗∗ .16∗∗ .18∗∗ .16∗∗

Earlier victim positivebehavior

.71∗∗ .80∗∗ .71∗∗ .67∗∗

Later perpetrator positive behavior from:Earlier victim positive

behavior.12∗∗ .10∗∗ .05 .03

Earlier perpetratorpositive behavior

.83∗∗ .85∗∗ .87∗∗ .68∗∗

Note. The values are standardized regression coefficients from hierarchical linear modeling analyses. Analyses arebased on data from 75 couples (150 individuals). Victim data = victim and perpetrator positive behavior as reportedby victims; perpetrator data = victim and perpetrator positive behavior as reported by perpetrators; across-partnerdata = 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 analyzeour data (cf. Kenny, Kashy, & Bolger, 1998;Raudenbush & Bryk, 2002). The multileveldata structure includes measures assessed dur-ing each 2-min segment of the interaction(Level 1), represented as nested within partic-ipant (Level 2 in within-participant analyses),and nested within couple (Level 3 in across-partner analyses and analyses of observer rat-ings). Hierarchical linear modeling accountsfor the nonindependence of observations bysimultaneously examining variance associatedwith each level of nesting, thereby provid-ing unbiased hypothesis tests. Following rec-ommended procedures for couples research,we represented intercept terms as randomeffects and represented slope terms as fixedeffects (Kenny, Mannetti, Pierro, Livi, &Kashy, 2002).

All analyses reported below employed alagged data structure, wherein we predicted alater criterion from an earlier predictor. Thatis, we simultaneously predicted (a) Time 2criteria from Time 1 predictors, (b) Time 3criteria from Time 2 predictors, and (c) Time4 criteria from Time 3 predictors. In residual-ized lagged analyses, we included the earliermeasure of the criterion as a control vari-able. We performed four types of analysis (seeTable 1): (a) victim data analyses, employing

victim ratings of partner behavior and theirown behavior; (b) perpetrator data analy-ses, employing perpetrator ratings of partnerbehavior and their own behavior; (c) across-partner data analyses, employing perpetratorratings of their own behavior and victim rat-ings of their own behavior; and (d) observerdata analyses, employing observer ratings ofperpetrator amends and victim forgiveness.We standardized all variables prior to conduct-ing hierarchical linear modeling. We initiallyperformed all univariate analyses includingmain effects and interactions for participantgender (male vs. female). No main effects orinteractions involving participant gender weresignificant, so this variable was dropped fromthe analyses.

Dyadic model hypotheses

To determine whether perpetrator amends pro-motes victim forgiveness and whether vic-tim forgiveness promotes perpetrator amends,we first examined the simple lagged asso-ciations of (a) earlier perpetrator positivebehavior with later victim positive behav-ior and (b) earlier victim positive behaviorwith later perpetrator positive behavior. Allsimple associations of earlier predictors withlater criteria were significant (for analyses


employing victim, perpetrator, across-partner,and observer ratings, βs ranged from .17 to.35, all ps < .05). Analyses predicting changeover time in each criterion are presented inTable 1. In these residualized lagged analyses(controlling for earlier levels of the criterion),across all four types of analysis, earlier per-petrator positive behavior predicted increasesover time in victim positive behavior (beyondvariance attributable to earlier victim positivebehavior; under Later victim positive behav-ior, see row labeled Earlier perpetrator pos-itive behavior, βs ranged from .16 to .18,all ps < .01). Earlier victim positive behaviorpredicted increases over time in perpetratorpositive behavior in the analyses employingvictim data and perpetrator data but not inthe analyses employing across-partner data orobserver data (under Later perpetrator posi-tive behavior, see row labeled Earlier victimpositive behavior).

Potential confounds

Are the present findings attributable to unin-tended confounds, such as tendencies towardsocially desirable responding or commitmentto the marriage? We replicated the Table 1analyses employing victim data and perpe-trator data (for which data sets it is suitableto explore potential confounds assessed usingself-report measures) including as covari-ates measures of self-deception, impressionmanagement, and commitment. In analy-ses that simultaneously controlled for thethree potential confounds, as well as con-trolling for earlier levels of each criterion:(a) in predicting later victim positive behaviorfrom earlier perpetrator positive behavior,coefficients remained significant in analysesemploying both victim and perpetrator data(βs = .14 and .17, both ps < .01), and (b) inpredicting later perpetrator positive behav-ior from earlier victim positive behavior,the coefficient remained significant in analy-ses employing victim data (β = .13, p < .01)but declined to nonsignificance in analysesemploying perpetrator data (β = .08, ns).

Victim and perpetrator perspectives

Our Table 1 findings are roughly parallel inanalyses employing victim data and

perpetrator data. To further explore the extentto which amends and forgiveness are phe-nomena that are perceived in parallel man-ner by partners in a given relationship, weexamined whether victims and perpetratorsexhibited rough agreement in their reports ofone another’s positive behavior. To exploresimple across-partner agreement in reportsof behavior during the videotaped interac-tion, we (a) regressed victims’ ratings oftheir own positive behaviors during inter-action onto perpetrators’ ratings of victims’positive behaviors and (b) regressed perpe-trators’ ratings of their own positive behav-iors onto victims’ ratings of perpetrators’positive behaviors. Both across-partner asso-ciations were significant (for victim posi-tive behaviors β = .59, for perpetrator behav-iors β = .45, both ps < .01). Thus, partnersexhibited good agreement in their ratingsof one another’s behavior during the courseof the interaction. In addition, we examinedwhether there were mean differences in lev-els of perceived perpetrator positive behav-ior and perceived victim positive behavior asa function of perpetrator versus victim role.Analyses in which we regressed ratings ofvictim and perpetrator behavior onto role (per-petrator vs. victim) revealed no significantdifferences in victims’ and perpetrators’ rat-ings of either victim positive behavior (Ms =5.85 and 5.81), t (74) = 0.55, ns, or perpetra-tor positive behavior (Ms = 6.08 and 6.00),t (74) = 1.01, ns. Thus, victims’ and perpe-trators’ perceptions of one another’s behaviorwere significantly associated and there wereno role main effects in mean levels of vari-ables. Moreover, the hypothesized associa-tions between victim and perpetrator positivebehavior typically were evident using bothvictim and perpetrator accounts of events (seeTable 1).

Discussion

In Study 1, we tested the hypothesis that(a) perpetrator amends promotes victim for-giveness; we also examined whether (b) vic-tim forgiveness promotes perpetrator amends,using ratings of each partner’s positive behav-ior during a discussion of the betrayal as a


proxy for amends and forgiveness. In the con-text of an interaction regarding an unresolvedbetrayal incident, earlier perpetrator positivebehavior reliably predicted increases overtime in victim positive behavior. This findingemerged whether the analyses were based onvictim ratings of the interaction, perpetratorratings of the interaction, across-partner rat-ings, or independent observers’ ratings. How-ever, earlier victim positive behavior wasa less robust predictor of increases overtime in perpetrator positive behavior—thisassociation was significant in only two ofthe four analyses displayed in Table 1, andone of these two significant associations fellto nonsignificance when controlling for rel-evant confounds. Thus, Study 1 providesstrong evidence that perpetrator amends pro-motes victim forgiveness but weaker evidencethat victim forgiveness promotes perpetratoramends. This finding suggests that the bene-fits of amends are not attributable to a simplepositivity effect (i.e., the positivity of part-ners’ behaviors toward one another). Rather,it is the existence of positive behavior in thecontext of a particular role—to wit, perpetra-tor amends—that appears to be crucial to theforgiveness process.

One strength of Study 1 is the controlafforded by the laboratory setting. Partici-pants were focused on a specific unresolvedbetrayal incident in their relationship and pro-vided concrete ratings of each person’s behav-ior during each 2-min segment of their inter-action. Indeed, partners exhibited very goodagreement in their ratings of one another’sbehavior. Moreover, this method allowedus to obtain parallel ratings from trainedobservers; analyses employing observers’ratings confirmed findings based on victims’and perpetrators’ ratings. Thus, we can feelrelatively confident that participants’ ratingsof each other’s behavior were not unduly col-ored by availability effects, role-based con-sistency, or self-serving bias. On the otherhand, the observer ratings for victim forgive-ness exhibited relatively low reliability, rais-ing some questions about the extent to whichoutside observers can accurately perceive andrate forgiving behavior.

An important limitation of Study 1 is thatthe laboratory setting in which the interactionstook place could be experienced as artifi-cial. A second limitation of Study 1 is thatalthough it allowed for a focused explorationof the reciprocal, mutually reinforcing asso-ciation of amends with forgiveness, an 8-mininteraction does not lend itself well to study-ing the actual resolution of betrayal incidents.Finally, we did not explicitly ask participantsto rate each other on forgiveness or amendsbut rather to rate each other on positive, sup-portive behaviors during the interaction. Thus,our findings point to the significance of perpe-trators enacting positive behaviors followingbetrayal but cannot be interpreted as explic-itly supporting the association of amends withforgiveness (with the exception of the ratingsprovided by trained observers, who were rat-ing participants’ amends and forgiveness). InStudies 2 and 3, we employed complementarymethods to examine the associations amongamends, forgiveness, and betrayal resolution,using measures that explicitly ask participantsto rate their and their partner’s amends andforgiveness following betrayal.

Study 2

In Study 2, we examined not only perpetra-tor amends and victim forgiveness but alsobetrayal resolution. In addition, we testedkey hypotheses of the dyadic model byexamining betrayals that transpired in thecontext of everyday interaction. Over thecourse of a 2-week period, people whowere involved in ongoing dating relationshipsdescribed each incident in which either theyor the partner violated relationship-relevantnorms. Interaction records completed soonafter each betrayal incident included mea-sures of perpetrator amends, victim forgive-ness, and betrayal resolution. This method hasthe advantage of tapping online reports of keyvariables, thus minimizing the possibilities forbiased recall. It also has the advantage of tap-ping perceptions of incidents in which theparticipant was the victim as well as incidentsin which he or she was the perpetrator, thusallowing us to examine associations among


amends, forgiveness, and betrayal resolutionfrom both victim and perpetrator perspectives.

Method

Participants

Participants were 78 undergraduates (20 menand 58 women) who volunteered to takepart in the study in partial fulfillment ofthe requirements for introductory psychol-ogy courses at the University of North Car-olina. The prerequisites for participation wereinvolvement in a dating relationship of atleast 1 month in duration, in which part-ners interacted with one another nearly everyday, either face to face or in telephoneconversations. Participants were 19 years oldon average (SD = 1 year), most were fresh-men or sophomores (41% freshmen, 44%sophomores, 5% juniors, and 9% seniors),and most were Caucasian (81% Caucasian,13% African American, 3% Asian Ameri-can, 1% Hispanic, and 3% Other). Partici-pants had been involved with their partnersfor 16 months on average (SD = 14 months),most indicated that they dated their partnerssteadily (83% dating steadily, 8% engaged,5% dating casually, and 4% dating regu-larly), and most described their relationshipsas exclusive (96%).

Procedure

Our procedure was modeled after previ-ous work using the Rochester InteractionRecord (cf. Reis & Gable, 2000). Participantsattended two laboratory sessions—one at thestart of the 2-week study (Time 1) and a sec-ond at the end of the study (Time 2). DuringTime 1 sessions, we explained that the studyconcerned negative incidents in dating rela-tionships. As in Study 1, we did not use theword betrayal. Instead, we described betrayalincidents in terms of “breaking the rules”: Weasked participants to complete a diary recordfor each incident in which “you upset yourpartner or your partner upsets you . . . . Forexample, your partner may do something thatis hurtful behind your back, you may flirtwith someone else at a party, or you may for-get to call your partner when you said you

would.” We asked participants to complete avictim diary record “to record each incidentin which your partner made you feel upset,angry, or hurt, ” and to complete a perpetratordiary record “to record each incident in whichyou made your partner feel upset, angry, orhurt” (we describe data from these two typesof record in terms of betrayal role, victimvs. perpetrator). Participants were asked torecord all such incidents, even if (a) the inci-dent was quite brief, (b) the participant hadalready recorded a similar incident in an ear-lier record, or (c) the incident did not involvetalking. Participants were instructed to com-plete the diary records without input fromtheir dating partners.

We asked participants to complete diaryrecords as soon as possible following eachincident, and to turn in booklets every Mon-day, Wednesday, and Friday. To maximizetimely reporting, we stressed the importanceof carrying diary record booklets at all times,and we telephoned participants Sunday, Tues-day, and Thursday evenings to remind themto turn in their booklets the following day. Foreach betrayal incident that transpired, partici-pants (a) recorded the date and time at whichthe incident occurred, (b) recorded the dateand time at which the interaction record wascompleted, (c) recorded the duration of theincident, (d) provided a brief description ofthe incident, and (e) answered several ques-tions about the incident (described below).During Time 1 and Time 2 laboratory ses-sions, participants also completed a varietyof questionnaires, two of which (describedbelow) are relevant to the goals of the presentresearch. At the end of Time 2 sessions, par-ticipants were fully debriefed and thanked fortheir assistance.

Measures

Participants completed a diary record foreach betrayal incident that transpired over thecourse of a 2-week period, using one type ofrecord for victim incidents and a second typeof record for perpetrator incidents. Item word-ing for victim and perpetrator diary recordswere identical except for differences reflectingbetrayal role (e.g., “I forgave my partner”


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 basedon 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 reportedin victim and perpetrator diary records combined.∗p < .05. ∗∗p < .01.

vs. “My partner forgave me”). Record itemswere structured in such a manner as to promptglobal, abstract ratings of amends or forgive-ness: Participants not only provided ratingsof concrete interaction behaviors involvingamends (e.g., “I showed real remorse aboutthe incident”; as discussed in the introductorytext, our goal with these items was to cap-ture amends-related behaviors in addition toexplicit apologies) or forgiveness (e.g., “I triedto get even with my partner” [reverse-scored])but also completed additional one-item globalratings (for amends, “I tried to make it up tomy partner”; for forgiveness, “I forgave mypartner”; for all items, 1 = do not agree at all,7 = agree completely). The analyses reportedbelow are based on participants’ one-itemglobal ratings. (Parallel findings were evidentin analyses using participants’ concrete rat-ings; for the analyses displayed in Table 2,eight of nine associations were significant.)Records also included a one-item measure ofbetrayal resolution (“By the end of the inter-action, the incident was resolved”).

Additional items were also included foruse in analyses we performed to control forpossible confounds: One item assessed sever-ity of incident (“When this incident occurred,I thought it had the potential to seriouslyharm our relationship”) and one item assessedmagnitude of betrayal (“When this incidentoccurred, I thought my partner had ‘bro-ken the rules’ of our relationship”). Bothof these items were measured using 7-pointscales, where 1 = do not agree at all and 7 =agree completely. In addition, during Time 1

sessions, participants completed the 40-itemBalanced Inventory of Desirable Respond-ing, which assesses both self-deception andimpression management (Paulhus, 1984; e.g.,“I always obey laws, even if I’m unlikelyto get caught”; 1 = do not agree at all, 7 =agree completely; respective αs = .54 and.80). Finally, during Time 2 sessions, partici-pants answered questions about the validity oftheir data, indicating that they recorded 94%of the betrayal incidents that transpired dur-ing the 2-week period and completed diaryrecords about 90 min following the inci-dents; 92% reported that the records accu-rately reflected their experiences during the2-week period,2 and 79% reported that thisperiod was typical of the sorts of incidentsthey experienced with their partners.


Sixty-four participants (18 men and 46women) described one or more betrayal inci-dents in the victim role and 54 participants (16men and 38 women) described one or moreincidents in the perpetrator role. As wouldbe expected for the sort of day-to-day betray-als that emerge over the course of a 2-weekperiod, the incidents were described as mildly

2. All Study 2 analyses were replicated excluding datafor the 8% of participants who indicated that the diaryrecords did not accurately reflect their experiencesduring the 2-week period. As expected (given thesmall percentage of participants in this group), theseanalyses revealed identical conclusions.


to moderately severe betrayals (victim roleM = 3.23, perpetrator role M = 2.44) and ashaving mild to moderate potential for harm(victim role M = 2.51, perpetrator role M =2.39).

The incidents included violations of depen-dence norms (e.g., failing to provide assis-tance when it was needed), monogamy norms(e.g., disappearing into a bedroom for anhour with an ex-partner), privacy norms (e.g.,telling a friend an important secret), anddecency/etiquette norms (e.g., playing a trickon the partner and then lying about it).

Results

Analysis strategy

Multiple diary records from a given partic-ipant are not independent, so we used hier-archical linear modeling to analyze our data,employing strategies that are suitable for theanalysis of diary data (cf. Bolger, Davis, &Rafaeli, 2003; Nezlek, 2001; for an empir-ical illustration, see Bolger, Zuckerman, &Kessler, 2000). The two-level data struc-ture includes measures assessed in each diaryrecord (Level 1), represented as nested withinparticipant (Level 2). For example, a partici-pant who experienced three betrayal incidentsduring the 2-week period provided reports ofthree separate betrayals (for each incident,reports of amends, forgiveness, and betrayalresolution).

We performed three types of analysis (seeTable 2): (a) victim data analyses, employingparticipant ratings of the partner’s amendsand their own forgiveness; (b) perpetratordata analyses, employing participant ratingsof their own amends and the partner’s forgive-ness; and (c) combined data analyses, includ-ing both victim data and perpetrator data. Westandardized all variables prior to conductinghierarchical linear modeling. We initially per-formed all combined data analyses includingmain effects and interactions for participantgender (male vs. female) and betrayal role(victim vs. perpetrator). No main effects orinteractions were significant for either vari-able, so these factors were dropped from theanalyses.


Table 2 summarizes results relevant to thedyadic model. In all three analyses, amendswas significantly positively associated withforgiveness (under Victim forgiveness, seerow labeled Perpetrator amends): That is,in everyday betrayal incidents, the tendencytoward victim forgiveness is greater to theextent that perpetrators offer greater amendsfor their acts of betrayal. Moreover, whenwe regressed betrayal resolution simultane-ously onto amends and forgiveness, both vari-ables accounted for unique variance (underBetrayal resolution, see rows labeled Perpe-trator amends and Victim forgiveness); thatis, everyday betrayals were more likely to besuccessfully resolved to the extent that per-petrators offered greater amends and victimsfound their way to greater forgiveness.

Potential confounds

Are these findings attributable to unintendedconfounds, such as severity of incident?We replicated the Table 2 combined anal-yses including as covariates measures ofseverity of incident, magnitude of betrayal,self-deception, and impression management.In analyses that simultaneously controlledfor the four potential confounds: (a) theassociation of amends with forgiveness wassignificant (β = .36, p < .01) and (b) in pre-dicting betrayal resolution, both perpetra-tor amends (β = .26, p < .01) and victimforgiveness accounted for unique variance(β = .44, p < .01).


Our Table 2 findings are roughly parallel inanalyses employing victim data and perpetra-tor data. In addition, we examined whetherthere were mean differences in levels of per-ceived perpetrator amends, victim forgive-ness, and betrayal resolution as a functionof perpetrator versus victim role. Analysesin which we regressed each variable ontorole (perpetrator vs. victim) revealed a signif-icant effect of role on perpetrator amends (forperpetrator and victim diary records, Ms =4.37 and 3.66), t (166) = 2.58, p < .02; a


nonsignificant effect of role on victim forgive-ness (for perpetrator and victim diary records,Ms = 5.57 and 5.39), t (168) = 0.93, ns; anda significant effect of role on perceivedbetrayal resolution (for perpetrator and victimdiary records, Ms = 5.06 and 4.49), t (168) =2.69, p < .01. Thus, although we observedsome role main effects in mean levels ofvariables (e.g., compared with victims, per-petrators reported offering higher levels ofamends and betrayal resolution), the hypothe-sized associations between amends and for-giveness typically were evident using bothvictim and perpetrator accounts of events.

Discussion

Study 2 revealed strong support for ourhypotheses. Examining the sorts of betray-als that transpire in the context of everydayinteractions over a 2-week period, perpetratoramends was positively associated with victimforgiveness, and both amends and forgive-ness contributed positively to the predictionof successful betrayal resolution. These asso-ciations did not differ significantly for betrayalincidents described by participants when theywere in the role of victims versus perpetra-tors, nor were these associations attributableto potential confounds such as severity of inci-dent or tendencies toward socially desirableresponding.

At the same time, these findings are limitedin several respects. First, Study 2 exam-ined betrayal incidents that spanned a nar-row range of severity, representing relativelyminor, “everyday” forms of betrayal. Basedon these findings, it is unclear whether perpe-trator amends plays a significant role in rel-atively more serious betrayal incidents. Sec-ond, in Study 2, participants completed diaryrecords for all betrayal incidents that theyexperienced during the course of the 2-weekstudy. These findings would be augmented byhypothesis tests in which we examine bothpartners’ descriptions of the same betrayalincidents and their aftermath. And third, inorder to test our full dyadic model, it is impor-tant to examine how the resolution of betrayalincidents relates to the current state of therelationship (see Figure 1).

Study 3

We used a retrospective method in Study 3,testing our hypotheses by asking dating part-ners to recall recent betrayal incidents in theirrelationship. Partners in ongoing dating rela-tionships identified an occasion on which thepartner violated a relationship-relevant norm.Each person then provided information aboutperpetrator amends, victim forgiveness, andbetrayal resolution, not only for the incidenthe or she identified but also for the inci-dent identified by the partner. This methodhas the advantage of tapping betrayal inci-dents of greater severity than those examinedin Study 2. In addition, Study 3 included anew measure of relationship quality so thatwe could assess whether and how partici-pants felt that these specific betrayal inci-dents had influenced their relationship. Andfinally, by obtaining victim and perpetratoraccounts of the same incident, we were ableto assess the extent to which partners agree intheir descriptions of one another’s actions, aswell as to examine the associations betweenvictim and perpetrator reports while control-ling for the severity of a given betrayal. Asdisplayed in Figure 1, we predicted that per-petrator amends would predict victim forgive-ness, that both amends and forgiveness wouldpredict betrayal resolution, and that betrayalresolution would predict relationship quality.

Method

Participants

Participants were 70 heterosexual dating cou-ples who responded to notices posted on thecampus and in the community of the Uni-versity of North Carolina, as well as throughadvertisements in local newspapers. The datafor 2 couples were deleted from the analy-ses (1 because one partner did not followinstructions, 1 due to experimenter error).Couples were paid $40.00 for taking part inthe study. Participants were 22 years old onaverage (SD = 3 years), and most were Cau-casian (86% Caucasian, 9% African Ameri-can, 2% Hispanic, 1% Asian American, and2% Other). Eighty-eight percent were enrolledin college or had completed undergraduate


degrees; an additional 11% were enrolled in(or had completed) graduate or professionaldegrees. Partners had been involved withone another for 20 months on average (SD =18 months), most indicated that they datedone another steadily (81% dating steadily, 7%dating regularly, 7% engaged, and 2% datingcasually), and most described their relation-ships as exclusive (96%).

Procedure

Partners were seated in individual cubiclesso that they could neither see nor interactwith one another and were informed thatthe study concerned negative incidents indating relationships. In an initial question-naire, each partner was asked to identify abetrayal incident that transpired within thepast 4 months. The instructions were wordedas in Study 1, such that we avoided usingthe word betrayal, instead describing betrayalincidents in terms of “breaking the rules.” Weasked participants to describe two such inci-dents, indicating that later in the session wewould identify one incident for each partner,and ask both partners to provide informa-tion about each incident. We asked partici-pants to report on nontrivial incidents and toidentify incidents that they (and the partner)would feel comfortable describing in greaterdetail. Participants wrote brief descriptions oftwo incidents and rated the extent to whicheach incident had the potential to harm therelationship (0 = no potential for harm, 8 =strong potential for harm). For each partner,we identified an incident that had moderate tosevere potential for harm. While the experi-menter identified an incident for each partner,participants completed other questionnairesdescribing themselves and their relationship.

The experimenter prepared photocopies ofthe selected incidents and randomly deter-mined which partner’s incident would beaddressed first. The experimenter asked eachpartner to read the incident description, andasked the perpetrator whether he or she rec-ognized the incident described by the partner.One participant did not recognize the incident;this couple was allowed to converse brieflyso that the partner could clarify the incident.

Participants then completed questionnairesin which they provided measures of vic-tim behavior, perpetrator behavior, betrayalresolution, and relationship quality for theassigned incident. Following a 5-min break,3

the experimenter distributed the descriptionof the second incident selected for each cou-ple, and participants completed questionnairesdescribing that incident. At the end of thesession, couples were paid, thanked for theirassistance, and debriefed.

Measures

The questionnaires participants completed todescribe their own and the partner’s behav-ior during each incident were nearly identi-cal except for differences reflecting betrayalrole (e.g., “I forgave my partner” vs. “Mypartner forgave me”). Victim forgiveness wasassessed with a 12-item instrument (e.g., “Itried to work things out with my partner”;for all items, 0 = do not agree at all, 8 =agree completely; for ratings by victims andperpetrators αs = .84 and .87). Perpetratoramends was assessed with a 24-item instru-ment designed to tap our broad, atonement-based definition of amends (e.g., “My part-ner confessed that he/she had done wrong,”“My partner attempted to make up for his/herbehavior”; for ratings by victims and perpetra-tors, αs = .90 and .86).4 Betrayal resolutionwas measured with a single item, “My partner

3. Participants were allowed to interact during the breakbut were instructed not to discuss the study. Manyparticipants remained in the laboratory during theirbreak, where the experimenter could verify that theyfollowed these instructions. Most participants who leftthe laboratory did so to use the restroom, and did notinteract with their partners during this time.

4. For each incident, participants provided informationregarding both immediate and delayed victim andperpetrator behavior. Items concerning “immediate”behavior described actions immediately following thebetrayal incident; items concerning “delayed” behav-ior described actions at present (or the last time thecouple engaged in incident-relevant interaction). Mea-sures of immediate and delayed behavior were signif-icantly correlated with one another for both victimsand perpetrators (all rs ≥ .60, all ps < .01), and anal-yses performed separately for immediate and delayedmeasures revealed parallel findings. Accordingly, theStudy 3 analyses are based on averaged measures ofimmediate and delayed victim forgiveness and perpe-trator amends.


and I have completely resolved this incident.”Relationship quality was assessed with a new6-item instrument designed to measure theimpact of a specific betrayal on the presentquality of the relationship (the items for vic-tims were as follows [wording was altered asappropriate for perpetrators]; “Our relation-ship is ruined” [reverse scored], “Our rela-tionship can still work, but it will never be asgood as it was before the incident” [reversescored], “I think my partner and I learnedsomething valuable from the incident,” “Ibelieve my partner will not behave this wayagain,” “My partner has a better understand-ing of my expectations for this relationshipnow,” “Our relationship is better now thanever before”; for ratings by victims and per-petrators, αs = .77 and .77). For the purposeof controlling for possible confounds, partici-pants also completed the Study 1 measures ofself-deception, impression management, andcommitment level (αs = .72, .73, and .96,respectively).


The betrayal incidents partners recountedwere described as having moderate to highpotential for harm (M = 5.71, SD = 1.76).The incidents included violations of depen-dence norms (e.g., breaking a promise toquit smoking, then lying about the lapse),monogamy norms (e.g., becoming sexuallyintimate with an extrarelationship partner),privacy norms (e.g., speaking to a third partyabout matters the individual regarded as pri-vate), and decency/etiquette norms (e.g., sur-reptitiously “borrowing” the partner’s debitcard to pay for a drinking spree).

Results

Analysis strategy

Data from the two betrayal incidents describedby each participant are not independent, norare data from the two partners in a givencouple. Therefore, as described in Studies1 and 2, we used hierarchical linear mod-eling to analyze our data. The three-leveldata structure includes measures providedby each participant (Level 1), nested within

each betrayal incident (Level 2) and in turnnested within each couple (Level 3). We per-formed four types of analysis (see Tables 3and 4): (a) victim analyses, employing vic-tim ratings of partner amends and their ownforgiveness; (b) perpetrator analyses, employ-ing perpetrator ratings of partner forgivenessand their own amends; (c) combined analyses,including both victim ratings and perpetratorratings; and (d) across-partner analyses,employing perpetrator ratings of their ownamends and victim ratings of their own for-giveness (as well as, in each analysis, rat-ings of betrayal resolution and relationshipquality). We standardized all variables priorto conducting hierarchical linear modeling.We initially performed all univariate analy-ses including main effects and interactionsfor participant gender (male vs. female),betrayal role (victim vs. perpetrator), andincident order (victim vs. perpetrator inci-dent described first). Two betrayal role maineffects were significant (described below), butno interactions were significant. Given thatour findings were not reliably moderated byparticipant gender, betrayal role, or incidentorder, these variables were dropped from theanalyses.


Table 3 summarizes results relevant to thedyadic model, presenting standardized coef-ficients for analyses performed separately asa function of betrayal role as well as forthe sample as a whole. As hypothesized, per-petrator amends was significantly associatedwith victim forgiveness (under Victim forgive-ness, see row labeled Perpetrator amends).Analyses regressing betrayal resolution ontovictim forgiveness and perpetrator amendsrevealed that betrayal resolution is signifi-cantly predicted by: (a) victim forgiveness inthe perpetrator data analyses and the com-bined analyses (under Betrayal resolution, seesingle row labeled Victim forgiveness) and(b) perpetrator amends in all three analyses(see single row labeled Perpetrator amends).In addition, when we regressed betrayal res-olution simultaneously onto forgiveness andamends, forgiveness accounted for signifi-cant unique variance in two of the three


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 .08Perpetrator amends .49∗∗ .25∗∗ .35∗∗

Note. The values are standardized regression coefficients from hierarchical linear modeling analyses for one-predictormodels (single-row models) and for two- or three-factor models (grouped as two- or three-row models). Analysesare 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 reportedin victim and perpetrator reports combined.∗p < .05. ∗∗p < .01.

analyses, and amends accounted for signif-icant variance in two of the three analyses(under Betrayal resolution, see two-row mod-els). Thus, (a) in resolving betrayal incidents,perpetrator amends plays an important rolebeyond the effects of victim forgiveness, and(b) irrespective of role, partners’ perceptionsof betrayal resolution are influenced by part-ners’ behavior at least as much as by theirown behavior.

As predicted, betrayal resolution was sig-nificantly associated with relationship qual-ity in all three analyses (under Relation-ship quality, see single row labeled Betrayalresolution); that is, relationship quality isgreater among couples who more success-fully resolve their betrayal incidents. Areour results consistent with the hypothesisthat betrayal resolution mediates the associ-ations of amends and forgiveness with rela-tionship quality (cf. Baron & Kenny, 1986;Kenny et al., 1998)? We addressed media-tion using the combined data. The prerequi-sites for assessing mediation were satisfied,in that both forgiveness and amends wereassociated with relationship quality (βs = .16

and .47, both ps < .01) and with betrayalresolution (βs = .28 and .20, both ps < .01),and betrayal resolution accounted for signif-icant unique variance in relationship qual-ity beyond forgiveness and amends (β = .28,p < .01). Tests of the significance of medi-ation revealed that the association of victimforgiveness with relationship quality was sig-nificantly and wholly mediated by betrayalresolution (controlling for perpetrator amends,z = 3.31, p < .01) and the association ofamends with relationship quality was signifi-cantly but partially mediated by betrayal res-olution (controlling for victim forgiveness,z = 2.62, p < .01).5

5. We were concerned with the possibility that someitems in our relationship quality measure potentiallyalso tapped betrayal resolution (e.g., “I think my part-ner and I learned something valuable from this inci-dent.”). We repeated the mediation analyses using asubset of relationship quality items that tapped thecurrent state of the relationship (e.g., “Our relation-ship is better now than ever before.”). We found thatbetrayal resolution significantly mediated the associa-tion of forgiveness and amends with relationship qual-ity using this 3-item version of the relationship qualityscale (α = .68).


Table 4. Predicting victim forgiveness,betrayal resolution, and relationship qual-ity—Across-partner analyses: Study 3

Across-partnerdata

Victim forgiveness (VR) from:Perpetrator amends (PR) .15†

Betrayal resolution (VR) from:Victim forgiveness (VR) .20∗Perpetrator amends (PR) .21∗Betrayal resolution (PR) from:Victim forgiveness (VR) .10Perpetrator amends (PR) .21∗Relationship quality (VR) from:Betrayal resolution (PR) .26∗∗Relationship quality (PR) from:Betrayal resolution (VR) .32∗∗

Note. The values are standardized regression coefficientsfrom hierarchical linear modeling analyses. Analyses arebased on data from 68 couples (136 individuals). Across-partner data = predictors and criteria as reported byindividual specified for each variable; VR = victim reportof variable; PR = perpetrator report of variable.†p < .10. ∗p < .05. ∗∗p < .01.

Across-partner analyses

Is it possible that the Table 3 findings reflectcommon method variance or tendenciestoward consistency in self-report? To addressthis question, we employed across-partneranalyses to examine the direct associationspredicted in the dyadic model, employingvictim-reports of victim behavior andperpetrator-reports of perpetrator behavior(see Table 4). Most predicted associationswere significant or marginal: (a) perpetrator-reported amends was positively associatedwith victim-reported forgiveness (β = .15,p < .10) and betrayal resolution (β = .21,p < .05), and perpetrator-reported betrayalresolution was positively associated withvictim-reported relationship quality (β = .26,p < .01), and (b) victim-reported betrayalresolution was positively associated withperpetrator-reported relationship quality (β =.32, p < .01). However, victim-reported for-giveness was not significantly associated withperpetrator-reported betrayal resolution (β =.10, ns).

Potential confounds

Are the present findings attributable to unin-tended confounds, such as commitment levelor tendencies toward socially desirableresponding? We replicated the Table 3 anal-yses (simple models) including as covariatesmeasures of self-deception, impression man-agement, and commitment level. In analysesthat simultaneously controlled for the threepotential confounds, all associations amongkey model variables remained significant (βsranged from .18 to .34, all ps < .01).


Our Table 3 findings are roughly parallel inanalyses employing victim data and perpe-trator data. To further explore the extent towhich amends and forgiveness are phenom-ena that are perceived in parallel manner bypartners in a given relationship, we exam-ined whether victims and perpetrators exhib-ited rough agreement in their reports of oneanother’s forgiveness and amends. To exploresimple across-partner agreement in reportsof behavior, we regressed victims’ reports ofeach variable onto perpetrators’ reports of par-allel variables. All four across-partner asso-ciations were significant (βs ranged from.27 to .53, all ps < .01). Thus, partnersexhibited good agreement in their ratings ofone another’s behavior during a given betrayalincident. In addition—and as reported ear-lier—although preliminary analyses revealedno significant interactions as a function ofbetrayal role, two main effects were signifi-cant: In comparison to victims, perpetratorsreported lower levels of victim forgiveness(Ms = 5.05 and 4.85), t (68) = 4.57, p < .01,as well as greater levels of relationship qual-ity (Ms = 6.01 and 6.26), t (68) = 4.24, p <

.05. Thus, although there were two role maineffects in mean levels of variables (e.g., per-petrators reported lower levels of victim for-giveness), victims’ and perpetrators’ percep-tions of one another’s behavior were sig-nificantly associated, and the hypothesizedassociations between amends and forgive-ness typically were evident using both victimand perpetrator accounts of a given betrayalincident.


Discussion

Study 3 revealed good support for ourhypotheses. Examining partners’ descriptionsof recent betrayal incidents, we obtained con-sistent support for the hypothesized asso-ciation of perpetrator amends with victimforgiveness. Moreover, the combined analy-ses revealed that both partners’ actions con-tribute to the resolution of betrayal incidents,and that betrayal resolution positively pre-dicts relationship quality. These findings gen-erally were evident for both victims and per-petrators, with one interesting exception: Inanalyses based on victim-report data, perpe-trator amends accounted for significant uniquevariance in betrayal resolution, whereas theirown forgiveness did not, and in analysesbased on perpetrator-report data the oppositewas true—victim forgiveness accounted forsignificant unique variance in betrayal reso-lution, whereas their own amends did not.This suggests that in understanding whether abetrayal is successfully resolved, the behav-ior of the partner may be somewhat moresalient than one’s own behavior. Strikingly,the dyadic model received moderately goodsupport not only in within-participant analysesbut also in across-partner analyses examiningthe associations of self-report measures withpartner-report measures.

The main limitation of these findings isthat they rest on retrospective reports of priorbetrayal incidents. It is possible that (a) recallof prior betrayal interactions might be shapedby the present state of the relationship, or that(b) different aspects of betrayal incidents maybe salient in the memories of the two part-ners. However, our confidence in the validityof participants’ descriptions is enhanced bythe fact that partners exhibited good agree-ment about the extent to which each per-son did versus did not offer amends andwas versus was not forgiving, as well as bythe fact that across-partner analyses revealedgood support for model predictions. Study3 extends Study 1 by examining the reso-lution of betrayal incidents, as well as theassociation of betrayal resolution with rela-tionship quality. And Study 3 extends Study

2 by obtaining both partners’ reports of inter-action behaviors following betrayal incidents,by examining more severe betrayal incidents,and by examining how behavior during priorbetrayal incidents relates to present relation-ship quality.

General Discussion

An observational study of married couples,an interaction record study of individuals inongoing dating relationships, and a cross-sectional survey study of dating partnersyielded findings that extend previous researchregarding forgiveness. These studies revealedthat forgiveness and betrayal resolution arethoroughly interpersonal processes—processes that are shaped by the behavior ofboth victim and perpetrator. Moreover, ourfindings regarding betrayal role are consistentwith the claim that amends and forgivenessare phenomena that are perceived in paral-lel manner by victims and perpetrators andthat exert parallel effects on betrayal reso-lution and relationship quality as perceivedby each partner. In the following paragraphs,we review findings relevant to our theoreticalmodel and discuss their implications.

Support for the dyadic model

Figure 2 presents a modified version of ourdyadic model of forgiveness based on ourfindings across Studies 1–3 and displaysmeta-analytic coefficients for each link inour model (with coefficients weighted by thenumber of participants in each study). First,we address results regarding the associationof perpetrator amends with victim forgive-ness. A meta-analytic summary of findingsfrom Studies 2 and 3 revealed a reliableconcurrent association of perpetrator amendswith victim forgiveness (see Figure 2; meta-analytic β = .27, p < .01). This associationwas evident not only in analyses employingvictim data and perpetrator data but also inacross-partner analyses and in analyses thatcontrolled for potential confounds such asseverity of betrayal, commitment level, andtendencies toward socially desirable respond-ing. Thus, and consistent with predictions,


Perpetrator Amends

Victim Forgiveness

Betrayal Resolution

Relationship Quality

Tested in

Studies 1, 2, and 3 Tested in

Studies 2 and 3

Tested in Study 3

.32**

.21*

.28**

.36**

.27** (.08) (.17**)

.07

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 indicateinsignificant model associations.*p < .05. **p < .01

when Mary offers amends for her offense,John is more forgiving. And conversely, whenMary is defensive or hostile, John is lessforgiving. In interdependence terms, Mary’samends indeed appear to facilitate John’stransformation of motivation from potentiallyvengeful given preferences to effective pref-erences that favor a forgiving response.

Is this association a unidirectional relation-ship wherein amends promotes forgiveness,or is the association of amends with forgive-ness reciprocal? In Study 1, we examined thetemporal unfolding of amends and forgive-ness in the course of betrayal-relevant inter-actions. Study 1 revealed good support forthe assertion that earlier perpetrator amendspromotes increases over time in victim for-giveness (see Figure 2; meta-analytic β = .17,p < .01). This residualized lagged associ-ation was significant not only in analysesemploying victim data and perpetrator databut also in across-partner analyses and inanalyses employing observer data, as well asin analyses that controlled for potential con-founds such as commitment level and sociallydesirable responding. Following an act of

betrayal, to the extent that Mary apologizesfor her actions or atones for the harm shehas caused, John becomes increasingly likelyto let go of his hurt and anger, graduallymoving toward forgiveness. However, earliervictim forgiveness does not reliably promoteincreases over time in perpetrator amends(meta-analytic β = .08, ns). This residualizedlagged association was significant in analysesemploying victim data and perpetrator data,but not in across-partner analyses or in anal-yses employing observer data. Moreover, thisassociation was even less reliably observedin analyses that controlled for potential con-founds. Therefore, it appears that the power ofpositive behavior in the wake of betrayal maybe dependent on the role of the actor; per-petrator amends more reliably inspires victimforgiveness than victim forgiveness inspiresperpetrator amends. Mary’s amends create apowerful form of situation selection; to theextent that amends improves John’s willing-ness to forgive, the couple will have morepositive interactions.

Next, we address findings regarding theassociations of amends and forgiveness with


betrayal resolution and relationship qual-ity. A meta-analytic summary of findingsfrom Studies 2 and 3 revealed that perpe-trator amends and victim forgiveness reli-ably account for unique variance in betrayalresolution (see Figure 2; meta-analytic βs =.21 and .32, both ps < .05). These associ-ations were evident not only in analysesemploying victim data and perpetrator databut also in across-partner analyses and inanalyses that controlled for diverse potentialconfounds. Thus, and consistent with predic-tions, forgiveness per se is no magic bulletin the resolution of betrayal dilemmas. Theprocess of betrayal resolution is thoroughlyinterpersonal, and rests on pro-relationshiptransformation on the part of both victim andperpetrator.

Finally, it is important to comment onour findings regarding relationship quality.Study 3 revealed not only that betrayal reso-lution is reliably associated with relationshipquality (see Figure 2; meta-analytic β = .28,p < .01) but also that perpetrator amendsaccounts for unique variance in relation-ship quality beyond betrayal resolution (meta-analytic β = .36, p < .01). The former find-ing is of consequence—it is important todemonstrate that successful betrayal resolu-tion is associated with healthy couple func-tioning (and that failure to resolve betrayalsis harmful). Yet the latter, unexpected find-ing is perhaps even more striking. It appearsthat perpetrator amends yields benefits thatextend beyond the fact that amends promotesbetrayal resolution; that is, amends also helpscouples move forward in a positive manner inthe aftermath of betrayal. We speculate thatthe direct association of amends with couplefunctioning may be attributable to the factthat amends promotes broader healing pro-cesses, perhaps by helping couples reestablishrelationship-relevant norms or by promotingthe recovery of trust or commitment. It is alsopossible that these findings are in part due toour exploring perpetrator amends (defined asbehaviors to atone for the betrayal) in thesestudies rather than just perpetrator apology(e.g., an explicit verbal statement about beingsorry for the betrayal). Future research shouldexplore the precise mechanisms that account

for the restorative properties of perpetratoramends.

Thus, although betrayal incidents tend tobe upsetting and potentially harmful, it isimportant to recognize that partners’ behaviorfollowing such incidents can be highly diag-nostic of the degree to which they value theirrelationship (cf. Holmes & Rempel, 1989).Their behavior provides meaningful infor-mation about (a) each person’s dispositions,values, and motives, as well as (b) each per-son’s probable future behavior. After violatingMary’s expectations about normative behav-ior, if John promises not to do so again, apol-ogizes for causing Mary pain, and behavesin such a manner as to partially “repay thedebt” he has incurred, Mary may have reasonto believe that John has developed a betterunderstanding of her expectations and that hevalues their relationship sufficiently to changehis ways. Conciliatory behavior of this sortmay do a good deal to reduce victim uncer-tainty, assuage anxiety, and increase intimacyby communicating perpetrator understanding,validation, and caring for the victim (Lau-renceau, Barrett, & Pietromonaco, 1998; Reis& Shaver, 1988). Of course, rather than offer-ing amends, John might alternatively declarethat Mary’s expectations are unreasonable orbelittle her for feeling hurt by his actions.Either way, Mary has learned a good dealabout John that she would not learn undersunnier circumstances of interdependence.

Victim versus perpetrator perspectives

All three studies allowed us to explore possi-ble differences between perpetrators and vic-tims in perceptions of betrayal incidents. InStudy 2, we “controlled for person,” asking agiven individual to report on victim and per-petrator behavior in the context of betrayalincidents in which he or she was the per-petrator as well as in betrayal incidents inwhich he or she was the victim. In Studies1 and 3, we “controlled for betrayal inci-dent,” asking two partners to report on theirown behavior and the partner’s behavior ina given betrayal incident. Consistent withprior research regarding betrayal role (e.g.,Baumeister et al., 1990; Kearns & Fincham,


2005; Zechmeister & Romero, 2002), weobserved some evidence of biased or self-serving perception. For example, perpetratorstended to perceive that they offered greateramends than victims perceived they wereoffered (Study 2), and tended to perceive thatbetrayals had more positive relational out-comes (Study 3); victims tended to perceivethat they offered greater levels of forgive-ness than perpetrators believed they received(Study 3). However, such role effects wereunreliably observed (e.g., no role effects wereevident in Study 1).

Thus, victims and perpetrators do not livein separate worlds—partners’ perceptions areconsiderably more convergent than divergent.At heart, many social psychologists are socialconstructionists; we are intrigued by evi-dence of biased perception and cognition. Thepresent research suggests that although part-ners may shade their construals in such amanner as to protect the self, this is not tosay that no shared reality underlies percep-tion. In the present research, the elephant onthe stage was the reality of betrayal-relevantinteraction, which was perceived in roughlyparallel manner not only by partners but alsoby outside observers.

Broader implications

Folk wisdom suggests that “it takes two totango”—that there must be two parties to aquarrel. The present studies suggest that theremust also be two parties to the resolution of aquarrel. An interaction-based analysis of theforgiveness process allows us to recognize thecontributions of both victims and perpetra-tors to the successful resolution of betrayal.To be sure, work regarding intra personalprocesses (cognition and emotion) informsour understanding of important social psycho-logical phenomena. But beyond such intrap-ersonal processes, an interdependence-basedanalysis illuminates our understanding of theinterpersonal character of forgiveness and rec-onciliation, highlighting the fact that behav-ior during interaction (a) has important directconsequences for individuals and (b) shapescognitive, emotional, and motivational eventsin ways that govern the course of future

interaction (Kelley et al., 2003; Rusbult &Van Lange, 2003). These broader effects arenot fully tractable in individual-focused socialpsychological analyses.

An interdependence-based analysis alsoaddresses the consequences of interaction fordyads. The present work revealed that partnerswho offer amends and forgiveness experiencegreater betrayal resolution and more positiverelational outcomes than those who do not.These findings have clear potential for appli-cation in couples counseling and family ther-apy. Moreover, this work may have impor-tant implications for understanding forgive-ness in other contexts—contexts involvingnonclose dyads and groups. It seems plausi-ble that even in settings wherein no formalrelationship exists between victim and per-petrator, perpetrator amends may yield ben-eficial consequences.6 For example, whencriminal defendants exhibit remorse for theiractions, the victims of violent crimes maymore readily recover from their maltreatment;when formal organizations apologize for theirtransgressions, they may enjoy more con-genial relations with constituents (e.g., cor-porations’ apologies for shareholder abuse,governments’ apologies for prior group-basedexploitation). Numerous interventions exist tohelp victims achieve forgiveness (Enright &the Human Development Study Group, 1996);the present work suggests the need for moreinterventions that help people make amendswhen they have hurt others (intentionallyor not).

Limitations and directions for future research

Before closing, we should address several lim-itations of the present work. Ironically, themost serious limitation rests on a feature ofthe present work that may also be regarded as

6. Interestingly, everyday experience suggests that evenin instances such as these, victims yearn for perpetratoramends. For example, during criminal trials, victimsare soothed by signs of remorse on the part of criminaldefendants; the victims of group-based exploitationwelcome apologies and confessions of wrongdoingon the part of formal “perpetrator representatives”(e.g., politicians’ apologies for reprehensible societalphenomena, such as the existence of slavery duringprior generations).


one of its greatest strengths: In all three stud-ies, participants described betrayals that tran-spired in ongoing relationships. This approachenabled us to address some important gapsin the forgiveness literature; we were ableto study betrayal interactions in real time,study both victim and perpetrator behavior(and perceptions of partner behavior) duringthe forgiveness process, and learn how bothvictims and perpetrators perceive the after-math of betrayal and forgiveness. However,it could be argued that all of the couples inStudies 1 and 3 experienced at least somedegree of forgiveness and reconciliation; oth-erwise, the couples would not have livedto tell their tales. The procedures employedin Study 2—asking individuals to completediary records of betrayals in their dating rela-tionships as they occurred—are not subject tothis critique, but the betrayals captured withthis method were comparatively mild. Due tothis limitation, our findings about the forgive-ness process may be limited to describing howthe forgiveness process unfolds in ongoingrelationships; it remains to be seen whetherthe same processes and relationships amongour model constructs apply to betrayals thatoccur in relationships that dissolve. Futurestudies should employ longitudinal methodsto explore betrayal incidents of even greaterconsequence, toward determining whether thesurvival of a relationship is influenced bydyadic forgiveness processes.

A second limitation stems from the factthat many of our findings rest on self-reportratings of one’s own or a partner’s behavior. Itis reassuring that partners exhibit moderate togood agreement in their descriptions of oneanother’s behavior. Nevertheless, given thatself-report measures are vulnerable to bias,our Study 1 analyses of observers’ ratingsare particularly compelling. However, giventhat Study 1 did not assess betrayal resolu-tion and relationship quality, future studiesmight seek to examine interaction-based ele-ments of the forgiveness process by studyinginteractions regarding both resolved and unre-solved betrayal incidents. Such studies wouldalso allow for studying additional aspects ofthe forgiveness process, such as perceived

sincerity of amends and forgiveness that wedid not include in these studies.

A third limitation is that although weexamined betrayal role in Studies 1 and 3,it must be recognized that all three studiesultimately are nonexperimental. Therefore, wecannot form confident conclusions regardingthe causal relations among amends, forgive-ness, and betrayal resolution. Although theresidualized lagged analyses we performed inStudy 1 suggest that perpetrator amends maycausally precede victim forgiveness, defini-tive causal evidence will require experimentalwork regarding perpetrator–victim interac-tion (e.g., via the use of priming techniques,false feedback regarding perpetrator behav-ior). These methods often require a trade-off between control and artificiality. In thecase of real romantic relationships, there arealso ethical problems associated with manip-ulating betrayal or amends, particularly whenwe seek to study severe betrayal incidents.Although we cannot draw causal conclusionson the basis of the present work, we per-formed diverse confound analyses to help ruleout several potential alternative explanations,such as socially desirable responding or com-mitment to a relationship. Finally, the externalvalidity of our findings may be questioned,in that our findings rest on evidence providedby North American participants. For example,work employing role-playing methods sug-gests that in organizational settings, amendsfollowing promise breaking is regarded assignificantly more appropriate by Japaneseundergraduates than by American undergrad-uates (Takaku, 2000).

Conclusions

The present research provides unique infor-mation regarding the role of perpetratoramends in the aftermath of betrayal, estab-lishing the importance of amends in facilitat-ing victim forgiveness and successful betrayalresolution. Moreover, the present researchrevealed that both partners’ actions shape themanner in which betrayals are experiencedand resolved by couples. Furthermore, dataobtained from both partners—irrespectiveof their role in the betrayal—revealed the


importance of both victim and perpetratorbehavior in bringing about forgiveness andbetrayal resolution. These findings highlightthe importance of moving beyond the tradi-tional victim-centered approach to explain-ing forgiveness, illustrating the utility of aninteraction-based analysis of couple reactionsto betrayal. To the extent that both partners areable to enact constructive behaviors follow-ing acts of betrayal, both are likely to reapthe rewards of enhanced outcomes for theirrelationship.

References

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychologicalresearch: Conceptual, strategic, and statistical consid-erations. Journal of Personality and Social Psychol-ogy, 51, 1173–1182.

Baumeister, R. F., Stillwell, A. M., & Heatherton, T. F.(1995). Personal narratives about guilt: Role in actioncontrol and interpersonal relationships. Basic andApplied Social Psychology, 17, 173–198.

Baumeister, R. F., Stillwell, A. M., & Wotman, S. R.(1990). Victim and perpetrator accounts of inter-personal conflict: Autobiographical narratives aboutanger. Journal of Personality and Social Psychology,59, 994–1005.

Bolger, N., Davis, A., & Rafaeli, E. (2003). Diary meth-ods: Capturing life as it is lived. Annual Review ofPsychology, 54, 579–616.

Bolger, N., Zuckerman, A., & Kessler, R. C. (2000).Invisible support and adjustment to stress. Journal ofPersonality and Social Psychology, 79, 953–961.

Eaton, J., Struthers, C. W., Shomrony, A., & Santelli,A. G. (2007). When apologies fail: The moderatingeffect of implicit and explicit self-esteem on apologyand forgiveness. Self and Identity, 6, 209–222.

Enright, R. D., & the Human Development Study Group.(1996). Counseling within the forgiveness triad: Onforgiving, receiving forgiveness, and self-forgiveness.Counseling and Values, 40, 107–122.

Exline, J. J., DeShea, L., & Holeman, V. T. (2007). Isapology worth the risk? Predictors, outcomes, andways to avoid regret. Journal of Social and ClinicalPsychology, 26, 479–504.

Fincham, F. D. (2000). The kiss of the porcupine: Fromattributing responsibility to forgiving. Personal Rela-tionships, 7, 1–23.

Finkel, E. J., Rusbult, C. E., Kumashiro, M., & Hannon,P. A. (2002). Dealing with betrayal in close rela-tionships: Does commitment promote forgiveness?Journal of Personality and Social Psychology, 82,956–974.

Frantz, C. M., & Bennigson, C. (2005). Better late thanearly: The influence of timing on apology. Journal ofExperimental Social Psychology, 41, 201–207.

Freedman, S. (1998). Forgiveness and reconciliation: Theimportance of understanding how they differ. Coun-seling and Values, 42, 200–217.

Gonzales, M. H., Manning, D. J., & Haugen, J. A. (1992).Explaining our sins: Factors influencing offenderaccounts and anticipated victim responses. Journal ofPersonality and Social Psychology, 62, 958–971.

Gottman, J. M. (1998). Psychology and the study ofmarital processes. Annual Review of Psychology, 49,169–197.

Hodgins, H. S., & Liebeskind, E. (2003). Apology versusdefense: Antecedents and consequences. Journal ofExperimental Social Psychology, 39, 297–316.

Hodgins, H. S., Liebeskind, E., & Schwartz, W. (1996).Getting out of hot water: Facework in social predica-ments. Journal of Personality and Social Psychology,71, 300–314.

Holmes, J. G., & Rempel, J. K. (1989). Trust in closerelationships. In C. Hendrick (Ed.), Review of person-ality and social psychology (Vol. 10, pp. 187–220).London, England: Sage.

Human Development Study Group. (1991). Five pointson the construct of forgiveness within psychotherapy.Psychotherapy: Theory, Research, Practice, Training,28, 493–496.

Kearns, J. N., & Fincham, F. D. (2005). Victim and per-petrator accounts of interpersonal transgressions: Self-serving or relationship-serving biases? Personalityand Social Psychology Bulletin, 31, 321–333.

Kelley, H. H. (1984). The theoretical description of inter-dependence by means of transition lists. Journal ofPersonality and Social Psychology, 47, 956–982.

Kelley, H. H., Holmes, J. G., Kerr, N. L., Reis, H. T.,Rusbult, C. E., & Van Lange, P. A. M. (2003). Anatlas of interpersonal situations. New York, NY:Cambridge University Press.

Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Dataanalysis in social psychology. In D. T. Gilbert, S. T.Fiske, & G. Lindzey (Eds.), Handbook of socialpsychology (4th ed., Vol. 1, pp. 233–265). New York,NY: McGraw-Hill.

Kenny, D. A., Mannetti, L., Pierro, A., Livi, S., &Kashy, D. A. (2002). The statistical analysis of datafrom small groups. Journal of Personality and SocialPsychology, 83, 126–137.

Laurenceau, J. P., Barrett, L. F., & Pietromonaco, P. R.(1998). Intimacy as an interpersonal process: Theimportance of self-disclosure, partner disclosure, andperceived partner responsiveness in interpersonalexchanges. Journal of Personality and Social Psychol-ogy, 74, 1238–1251.

Leary, M. R., Springer, C., Negel, L., Ansell, E., &Evans, K. (1998). The causes, phenomenology, andconsequences of hurt feelings. Journal of Personalityand Social Psychology, 74, 1225–1237.

McCullough, M. E., Bono, G., & Root, I. M. (2007).Rumination, emotion, and forgiveness: Three longitu-dinal studies. Journal of Personality and Social Psy-chology, 92, 490–505.

McCullough, M. E., Rachal, K. C., Sandage, S. J.,Worthington, E. L., Brown, S. W., & Hight, T. L.


(1998). Interpersonal forgiving in close relationships:II. Theoretical elaboration and measurement. Journalof Personality and Social Psychology, 75, 1586–1603.

McCullough, M. E., Worthington, E. L., & Rachal, K. C.(1997). Interpersonal forgiving in close relationships.Journal of Personality and Social Psychology, 73,321–336.

Mullet, E., Houdbine, A., Laumonier, S., & Girard, M.(1998). “Forgivingness”: Factor structure in a sampleof young, middle-aged, and elderly adults. EuropeanPsychologist, 3, 289–297.

Nezlek, J. B. (2001). Multilevel random coefficient anal-yses of event-and interval-contingent data in socialand personality psychology research. Personality andSocial Psychology Bulletin, 27, 771–785.

Paleari, F. G., Regalia, C., & Fincham, F. (2005). Maritalquality, forgiveness, empathy, and rumination: A lon-gitudinal analysis. Personality and Social PsychologyBulletin, 31, 368–378.

Paulhus, D. L. (1984). Two-component models ofsocially desirable responding. Journal of Personalityand Social Psychology, 46, 598–609.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchicallinear models: Applications and data analysis meth-ods. London, England: Sage.

Reis, H. T., & Gable, S. L. (2000). Event-sampling andother methods for studying everyday experience.In H. T. Reis & C. M. Judd (Eds.), Handbook ofresearch methods in social and personality psychology(pp. 190–222). Cambridge, England: Cambridge Uni-versity Press.

Reis, H. T., & Shaver, P. (1988). Intimacy as an inter-personal process. In S. Duck (Ed.), Handbook ofpersonal relationships: Theory, relationships, andinterventions (pp. 367–389). Chichester, England:Wiley.

Risen, J. I., & Gilovich, T. (2007). Target and observerdifferences in the acceptance of questionable apologies.Journal of Personality and Social Psychology, 92,418–433.

Rusbult, C. E., Martz, J. M., & Agnew, C. R. (1998).The investment model scale: Measuring commitment

level, satisfaction level, quality of alternatives, andinvestment size. Personal Relationships, 5, 357–391.

Rusbult, C. E., & Van Lange, P. A. M. (2003). Inter-dependence, interaction, and relationships. AnnualReview of Psychology, 54, 351–375.

Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F.,& Lipkus, I. (1991). Accommodation processes inclose relationships: Theory and preliminary empiricalevidence. Journal of Personality and Social Psychol-ogy, 60, 53–78.

Struthers, C. W., Eaton, J., Santelli, A. G., Uchiyama, M.,& Shirvani, N. (2008). The effects of attributions ofintent and apology on forgiveness: When saying sorrymay not help the story. Journal of Experimental SocialPsychology, 44, 983–992.

Takaku, S. (2000). Culture and status as influenceson account giving: A comparison between the UnitedStates and Japan. Journal of Applied SocialPsychology, 30, 371–388.

Thompson, L. Y., Snyder, C. R., Hoffman, L., Michael,S. T., Rasmussen, H. N., Billings, L. S., . . . , Roberts,D. E. (2005). Dispositional forgiveness of self, others,and situations. Journal of Personality, 73, 313–359.

Weiner, B., Graham, S., Peter, O., & Zmuidinas, M.(1991). Public confession and forgiveness. Journal ofPersonality, 59, 281–312.

Worthington, E. L. (Ed.). (2005). Handbook of forgive-ness. New York, NY: Brunner-Routledge.

Yovetich, N. A., & Rusbult, C. E. (1994). Accommoda-tive behavior in close relationships: Exploring trans-formation of motivation. Journal of ExperimentalSocial Psychology, 30, 138–164.

Zechmeister, J. S., Garcia, S., Romero, C., & Vas, S. N.(2004). Don’t apologize unless you mean it: Alaboratory investigation of forgiveness and retalia-tion. Journal of Social and Clinical Psychology, 23,532–564.

Zechmeister, J. S., & Romero, C. (2002). Victim andoffender accounts of interpersonal conflict: Autobi-ographical narratives of forgiveness and unforgive-ness. Journal of Personality and Social Psychology,82, 675–686.

in the wake of betrayal: amends, forgiveness, and the...

Documents