Content uploaded by Peggy A Hannon
Author content
All content in this area was uploaded by Peggy A Hannon on Sep 01, 2014
Content may be subject to copyright.
Dealing With Betrayal in Close Relationships:
Does Commitment Promote Forgiveness?
Eli J. Finkel
Carnegie Mellon University
Caryl E. Rusbult, Madoka Kumashiro,
and Peggy A. Hannon
University of North Carolina at Chapel Hill
This work complements existing research regarding the forgiveness process by highlighting the role of
commitment in motivating forgiveness. On the basis of an interdependence–theoretic analysis, the
authors suggest that (a) victims’ self-oriented reactions to betrayal are antithetical to forgiveness,
favoring impulses such as grudge and vengeance, and (b) forgiveness rests on prorelationship motivation,
one cause of which is strong commitment. A priming experiment, a cross-sectional survey study, and an
interaction record study revealed evidence of associations (or causal effects) of commitment with
forgiveness. The commitment–forgiveness association appearred to rest on intent to persist rather than
long-term orientation or psychological attachment. In addition, the commitment–forgiveness association
was mediated by cognitive interpretations of betrayal incidents; evidence for mediation by emotional
reactions was inconsistent.
Although close partners often treat one another in a positive and
considerate manner, negative interactions are nearly inevitable.
Such interactions result from a variety of causes, including incom-
patible preferences, external sources of stress, and extrarelation-
ship temptation. We suggest that the violation of a relationship-
relevant norm—or betrayal of one’s partner—constitutes one of
the more serious threats to a relationship, and we propose that the
resolution of betrayal incidents is not easy. Indeed, forgiveness of
betrayal arguably is one of the more difficult tasks in an ongoing
relationship.
The film The War of the Roses darkly and humorously illustrates
the complexities of betrayal and forgiveness: The marriage of
Oliver and Barbara Rose is marked by repeated betrayal. Oliver
belittles Barbara’s career as a chef. Barbara fails to support Oliver
during a frightening health crisis. Each humiliates the other, de-
livering impossible-to-forget attacks on the other’s tastes and
habits. Cumulatively, such incidents reduce their once-loving mar-
riage to a sequence of vicious, increasingly deadly battles. During
their marital Armageddon, the two find themselves entangled in a
chandelier suspended above a hallway. The mechanism supporting
the chandelier gives way, and—embraced in the arms of the
chandelier—the two crash to the unyielding terazzo floor 30 feet
below. With his dying breath, Oliver reaches out to touch Bar-
bara’s shoulder, offering amends and seeking forgiveness. Bar-
bara’s hand slowly rises to meet Oliver’s (perhaps, one imagines,
to reciprocate Oliver’s act), and with her dying breath, Barbara
flings Oliver’s hand away from her. Again, forgiveness of betrayal
arguably is one of the more difficult tasks in an ongoing
relationship.
At present, psychologists’ understanding of betrayal and for-
giveness is somewhat limited. Although the concept of forgiveness
has received considerable attention in the fields of philosophy and
theology (e.g., Dorff, 1992; Nietzsche, 1887), until recently this
phenomenon was largely ignored in the social sciences (for a
review, see McCullough, Sandage, & Worthington, 1997). Most
empirical work has sought to portray the process by which people
forgive, examining the manner in which individuals perceive and
explain acts of betrayal (e.g., Baumeister, Stillwell, & Wotman,
1990; Boon & Sulsky, 1997; Gonzales, Haugen, & Manning,
1994), the emotional reactions that accompany betrayal (e.g.,
Ohbuchi, Kameda, & Agarie, 1989; Tangney, Wagner, Hill-
Barlow, Marschall, & Gramzow, 1996), or the role of interaction
processes in promoting the resolution of betrayal incidents (e.g.,
McCullough, Worthington, & Rachal, 1997; Weiner, Graham,
Peter, & Zmuidinas, 1991). Also, some work has examined the
efficacy of clinical interventions designed to encourage forgive-
ness (e.g., Freedman & Enright, 1996; McCullough & Worthing-
ton, 1995).
Thus, the existing literature helps illuminate our understanding
of how people forgive, identifying the cognitive, affective, and
interactional concomitants of this process. Unfortunately, few
prior studies have sought to explain why people forgive, identify-
ing the motivational underpinnings of this phenomenon. The
present work uses the principles of interdependence theory (Kelley
& Thibaut, 1978) to analyze betrayal and forgiveness, emphasizing
the role of commitment in motivating interpersonal forgiveness.
We present the results of three studies designed to provide
evidence regarding the commitment–forgiveness link. In addi-
Eli J. Finkel, Department of Psychology, Carnegie Mellon University;
Caryl E. Rusbult, Madoka Kumashiro, and Peggy A. Hannon, Department
of Psychology, University of North Carolina at Chapel Hill.
This research was supported by Templeton Foundation Grant 5158.
Final manuscript preparation was facilitated by National Institute of Mental
Health Training Grant T32 MH19953.
Correspondence concerning this article should be addressed to Eli J.
Finkel, Department of Psychology, Baker Hall 436-C, Carnegie Mellon
University, Pittsburgh, Pennsylvania 15213. E-mail: eli@andrew.cmu.edu
Journal of Personality and Social Psychology Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 82, No. 6, 956–974 0022-3514/02/$5.00 DOI: 10.1037//0022-3514.82.6.956
956
tion, in two studies, we extend prior research regarding the process
of forgiveness by examining the role of mental events (i.e., cog-
nition and emotion) in mediating the commitment–forgiveness
association.
Norm Violations and the Experience of Betrayal
Betrayal typically is defined as “to be unfaithful or disloyal,”“to
reveal something meant to be hidden,” or “to seduce and desert”
(Steinmetz, 1993, p. 63). In the context of close relationships, we
define betrayal as the perceived violation by a partner of an
implicit or explicit relationship-relevant norm. Betrayal may be
said to have occurred when the victim believes that the perpetrator
has knowingly departed from the norms that are assumed to govern
their relationship, thereby causing harm to the victim. Betrayal
may involve minor or major normative infractions. Toward the
mild end of the betrayal continuum, Oliver might embarrass Bar-
bara during a dinner party, telling a story that makes her appear
ignorant. Toward the more serious end of the continuum, Barbara
might attempt to seduce Oliver’s best friend.
In previous research, we have examined a variety of relationship
maintenance acts, including (a) accommodative behavior—the ten-
dency, when a partner enacts rude or inconsiderate behaviors, to
inhibit destructive impulses and instead react in a constructive
manner (e.g., Rusbult, Verette, Whitney, Slovik, & Lipkus, 1991);
(b) willingness to sacrifice—the tendency, when partners’ interests
conflict, to forego otherwise desirable behaviors or engage in
otherwise undesirable behaviors (e.g., Van Lange et al., 1997); and
(c) derogation of alternatives—the inclination, when confronted
with an attractive alternative, to cognitively derogate that person
(e.g., Johnson & Rusbult, 1989). All of these behaviors arise in
response to interdependence situations involving the potential for
harm. What distinguishes betrayal incidents from other sorts of
negative interaction incidents?
Unlike the maintenance acts examined in previous work, be-
trayal incidents involve norm violations.
1
Norms are rule-based
inclinations to respond to particular interdependence situations in
a specified manner; norms constitute the rules by which interaction
is governed, whether the rules are relationship specific or cultur-
ally shared (Rusbult & Van Lange, 1996; Thibaut & Kelley, 1959).
For example, interaction partners may implicitly or explicitly agree
that some courses of action are forbidden (and that other courses of
action are mandated)—they agree that some interaction sequences
are not to be initiated (e.g., never humiliate the partner), that some
interaction contingencies are not to transpire (e.g., never fight with
the partner the night before important work-related events), and
that some interaction sequences are not to take place with partic-
ular sorts of partner (e.g., never become sexually intimate with an
extrarelationship partner). Norms may initially be established as a
simple matter of convenience. However, over time, such rules
frequently “take on the characteristics of a moral obligation”
(Thibaut & Kelley, 1959, p. 128).
Interaction partners develop characteristic patterns of response
to recognizable interdependence situations, including habitual
emotional, cognitive, and behavioral impulses (cf. Rusbult & Van
Lange, 1996). Humans count on adherence to rules, so adaptation
to betrayal incidents is likely to include the impulse to punish
transgressors—an impulse embodied in righteous indignation and
hostile behavioral tendencies. Because betrayals violate moral
obligations, such incidents create an interpersonal debt. Thus, the
impulse toward vengeance and other forms of debt reduction can
be seen to be functionally adaptive (at least in the short run), in that
the inclination to punish transgressors is a mechanism for enforc-
ing relationship-relevant norms. Indeed, it has been argued that
reactions such as victim vengeance and perpetrator guilt may have
an evolutionary basis, resting on the functional value to social
animals of mutual cooperation and rule adherence (cf. Ridley,
1998).
What effects do betrayal incidents exert on victims? In the
aftermath of betrayal, the victim may find it difficult to depart from
the negative affect associated with the incident—for instance,
Oliver may experience persistent and debilitating sadness or anger
(e.g., Ohbuchi et al., 1989; Rosenzweig-Smith, 1988). The victim
may develop negative patterns of cognition—Oliver may feel
confused by the event and its implications, may obsessively review
events surrounding the betrayal, or may reinterpret prebetrayal
behavior, questioning whether earlier interpretations of Barbara’s
behavior were correct (e.g., Baumeister et al., 1990; Boon &
Sulsky, 1997; Stillwell & Baumeister, 1997). In addition, the
victim may adopt negative behavioral tendencies in interaction
with the perpetrator—Oliver may rather persistently seek ven-
geance or demand retribution and atonement (e.g., Fagenson &
Cooper, 1987; Kremer & Stephens, 1983; Zillman, Bryant, Cantor,
& Day, 1975).
Perpetrators may develop patterns of negative affect, cognition,
and behavior that parallel the patterns evident among victims. For
example, perpetrators may experience persistent negative emotions
such as sadness, shame, or guilt (e.g., Baumeister, Stillwell, &
Heatherton, 1995; Tangney et al., 1996). Also, victims’ preoccu-
pation and negative attributions may be met with defensive cog-
nition on the part of perpetrators, who may seek to explain their
1
In work regarding sacrifice arising from partners’ noncorrespondent
preferences, we examine noncorrespondence that arises through no fault of
either partner; thus, noncorrespondent situations are not against the rules.
In work regarding derogation of alternatives, it is not against the rules to be
confronted with an attractive alternative; indeed, derogation of alternatives
may serve to prevent situations in which norm violations might otherwise
transpire. When partners create accommodative dilemmas by enacting rude
or inconsiderate behaviors, such acts are unpleasant and irritating, but they
are not necessarily against the rules. Of course, the boundary separating
accommodation from forgiveness may be a fuzzy one, in that both accom-
modative dilemmas and betrayals are initiated by a partner’s potentially
destructive act. Potentially destructive acts such as inconsiderate behaviors
may be interpreted as merely unpleasant (i.e., as accommodative dilem-
mas) or as rule violations (i.e., as betrayals; in fact, some norm violations
studied in the present work are properly categorized as violations of
decency and etiquette norms). Indeed, dictionaries describe forgiveness
using norm-relevant terms such as pardon and absolve, whereas accom-
modation is described using norm-irrelevant terms such as adjust, adapt,
and modify. Thus, we believe that the boundary separating accommodation
from forgiveness rests on the individual’s interpretation of a partner’s
potentially destructive act: When negative partner acts are interpreted as
annoying (or even infuriating) but not necessarily against the rules, such
acts are construed as accommodative dilemmas; the positive, relationship-
restoring act that resolves such dilemmas is accommodation. When nega-
tive partner acts are interpreted as rule violations, they arouse righteous
indignation and are construed as betrayals; the positive, relationship-
restoring act that resolves such dilemmas is forgiveness.
957
COMMITMENT AND FORGIVENESS
actions in such a manner as to justify their behavior (e.g., Gonza-
les, Manning, & Haugen, 1992; Stillwell & Baumeister, 1997).
Moreover, victims’ desire for revenge and demands for atonement
may be met with reciprocal behavioral negativity on the part of
perpetrators, in that few perpetrators suffer endless payback and
offer bottomless amends (e.g., Hodgins, Liebeskind, & Schwartz,
1996; Ohbuchi et al., 1989).
Forgiveness of Betrayal
Moving beyond this constellation of negative affect, cognition,
and behavior rests on victim forgiveness. Forgiveness typically is
defined as “to grant pardon” or “to cancel a debt or payment”
(Steinmetz, 1993, p. 259). Previous work has adopted related
definitions, including (a) “a willingness to abandon one’s right to
resentment, condemnation, and subtle revenge toward an offender
who acts unjustly, while fostering the undeserved qualities of
compassion, generosity, and even love toward him/her” (Enright &
Human Development Study Group, 1996, p. 108); and (b) “the set
of motivational changes whereby one becomes decreasingly mo-
tivated to retaliate against an offending partner, decreasingly mo-
tivated to maintain estrangement from the offender, and increas-
ingly motivated by conciliation and goodwill toward the offender,
despite the offender’s hurtful actions” (McCullough et al., 1997,
pp. 321–322). Distinguishing between intrapsychic and interper-
sonal events, forgiveness has also been defined in terms of (c) “[a]
the inner, intrapsychic dimension involving the victim’s emotional
state (and the cognitive and behavioral accompaniments), and [b]
the interpersonal dimension involving the ongoing relationship
within which forgiveness takes place or fails to do so” (Baumeis-
ter, Exline, & Sommer, 1998, p. 80).
The extant definitions differ in the degree to which they char-
acterize forgiveness as intrapersonal (i.e., as a within-victim men-
tal phenomenon) versus interpersonal (i.e., as a victim–perpetrator
interaction phenomenon). Indeed, lay construals of this construct
seem to be rather multifaceted, including both internal qualities
(e.g., mentally forgiving another’s transgression) and interpersonal
qualities (e.g., forgiving another by resuming prebetrayal patterns
of interaction). Our work concerns forgiveness in ongoing rela-
tionships, so we emphasize the interpersonal character of this
phenomenon, defining forgiveness as the victim’s resumption of
prebetrayal behavioral tendencies—as the tendency to forego ven-
geance and other destructive patterns of interaction, instead be-
having toward the perpetrator in a positive and constructive man-
ner. In short, the victim cancels the interpersonal debt created by
the act of betrayal.
2
As noted earlier, the victim’s immediate reaction to betrayal
often is antithetical to forgiveness (e.g., Barbara may want to crush
Oliver’s Morgan with her sports utility vehicle). Interdependence
theory describes self-oriented, gut-level impulses as given prefer-
ences (Kelley & Thibaut, 1978). Of course, the degree to which a
victim is inclined toward vengeance rather than forgiveness varies
across interactions and may be moderated by the severity of
betrayal, the importance of the domain in which betrayal occurs,
and the specific emotions and cognitions that accompany a given
act. But given that betrayals cause harm, violate moral obligations,
and create an interpersonal debt, we suggest that betrayal typically
engenders impulses such as vengeance and demands for
retribution.
However, impulsive given preferences do not necessarily guide
behavior. In reacting to an incident, individuals explicitly or im-
plicitly take account of broader considerations such as personal
values, long-term goals, and concern for a partner’s well-being.
This process is termed transformation of motivation, and the
preferences resulting from this process are termed effective pref-
erences (Kelley & Thibaut, 1978). The transformation process
leads individuals to forego behavioral impulses based on direct
self-interest and instead act on the basis of broader goals, as
embodied in reconceptualized, effective preferences.
3
Sometimes
the transformation process is automatic and habit driven, resting on
well-established patterns of interaction; sometimes the process is
mentally mediated, resting on the emergence of relatively less
blameful, more benevolent emotional reactions (e.g., reduced
anger) or cognitive interpretations (e.g., discounting of internal
causes; cf. Rusbult & Van Lange, 1996).
It is interesting to note that there is controversy regarding the
role of mental events in the forgiveness process. Why is this so?
First, the philosophical literature tends to characterize intraper-
sonal and interpersonal events as dichotomous, all-or-nothing
propositions—a victim either absolves the perpetrator of blame or
does not, either forgives a perpetrator or does not (cf. North, 1987).
Second, the Christian theological literature tends to regard inter-
personal change in the absence of intrapersonal change as the
prototype of forgiveness (cf. Marty, 1998). A rather high value is
placed on saintly forgiveness, whereby the victim recognizes the
full extent of a perpetrator’s sin and in no way absolves blame yet
nevertheless forgives. As a result of these traditions, there is a
tendency to assume that forgiveness is not—or should not be—
accompanied by changes in betrayal-relevant mental events. If
victims achieve forgiveness because they come to understand acts
of betrayal—for example, if they identify extenuating circum-
stances or discount the role of internal causes—such forgiveness
does not count. The victim has nullified the betrayal or reinter-
preted the incident in nonbetrayal terms; no real transgression is
perceived to have transpired, so there is nothing to forgive.
2
In the case of complete forgiveness, the victim no longer exhibits
interpersonal negativity in reaction to the betrayal incident. Although
complete forgiveness defines the logical endpoint of the forgiveness di-
mension, the extent to which complete forgiveness is a practical possibility
remains unclear. Thus, it is important to characterize forgiveness as a
continuum rather than an all-or-nothing proposition. Indeed, we suggest
that in understanding the forgiveness process, magnitude of forgiveness
should be evaluated in light of the magnitude of a given betrayal. Assuming
that it is more difficult to forgive relatively more serious norm violations,
it may be as impressive to exhibit forgiveness of a magnitude of x ⫺ 1 for
a serious betrayal as it is to exhibit forgiveness of a magnitude of x for a
minor betrayal. Accordingly, in studies of the motives underlying forgive-
ness, it is important to take into consideration the severity of the betrayal
for which forgiveness is relevant.
3
Recent research supports this characterization of interpersonal forgive
-
ness (Rusbult, Davis, Finkel, Hannon, & Olsen, 2001). For example, in
responding to a series of hypothetical betrayal incidents, the behaviors that
individuals endorse under conditions of limited reaction time (given pref-
erences) are substantially more destructive and less constructive than are
the behaviors they endorse under conditions of plentiful reaction time
(effective preferences). Such findings support the assertions that (a) im-
pulsive reactions to betrayal are considerably less forgiving than are actual
reactions and (b) forgiveness rests on transformation of motivation.
958
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
We suggest that neither of the aforementioned assumptions is
entirely valid. First, neither mental construal nor forgiveness is an
all-or-nothing proposition. Individuals may come to partially un-
derstand the circumstances surrounding a betrayal; given that
victim and perpetrator construals initially may differ (Stillwell &
Baumeister, 1997), understanding may simply entail achieving an
unbiased interpretation. Also, individuals may partially forgive—a
possibility that seems particularly plausible if forgiveness unfolds
over the course of extended interaction. Second, we acknowledge
that victims may sometimes exhibit interpersonal forgiveness
without modifying their mental construals (e.g., “I cannot find my
way to anything short of full and complete blame, yet I forgive
you”). At the same time, we suspect that for most people in the
context of most betrayals, some degree of understanding facilitates
some degree of forgiveness. Accordingly, we suggest that coming
to mentally understand a betrayal incident—as evidenced by re-
duced negative affect and cognition—partially mediates forgive-
ness. Given that it may be difficult for victims to develop less
blameful, more benevolent understandings, victims may rather
persistently act on the basis of vengeful preferences (cf. Enright &
Human Development Study Group, 1996; Gordon & Baucom,
1998). Therefore, it becomes important to ask, What inspires
positive mental events, prorelationship motives, and interpersonal
forgiveness?
Commitment and Forgiveness
We suggest that commitment is a fundamental property of
relationships and propose that strong commitment promotes pos-
itive mental events, prorelationship motives, and forgiveness.
Commitment is defined in terms of three components—intent to
persist, long-term orientation, and psychological attachment. Com-
mitment develops as a result of (a) increasing satisfaction (i.e., a
relationship gratifies important needs, e.g., the needs for intimacy
or security), (b) declining alternatives (i.e., important needs could
not effectively be gratified by alternative partners, friends, or kin),
and (c) increasing investments (i.e., resources such as personal
identity, effort, or material possessions become linked to a rela-
tionship; Rusbult, 1983; Rusbult, Martz, & Agnew, 1998).
Why should commitment promote positive mental events, pro-
relationship motives, and forgiveness? We suggest three lines of
reasoning in support of this prediction, identifying how the above-
noted components of commitment might—individually or collec-
tively—account for such associations. Our logic is based on the
assumption that commitment-relevant interests may be rather im-
mediate and direct or may be broader in either of two respects.
First, individuals may be concerned about their temporally ex-
tended interests or their long-term self-interest. Second, individu-
als may be concerned about their interpersonally extended interests
or the interests of the partners with whom they are interdependent.
The most primitive component of commitment is simple intent
to persist, or the decision to remain dependent on a partner. Intent
to persist is primitive because it does not in any direct manner
(theoretically or operationally) involve either broadened temporal
interests or broadened interpersonal interests. Because Oliver is
dependent, he needs to persist—his relationship provides him with
desirable outcomes, he has invested a good deal, and his alterna-
tives are poor. Because committed individuals need their relation-
ship and therefore intend to persist with their partner, they should
be more willing to forgive a partner’s transgressions: Quite simply,
the more one has to lose, the more one should be willing to forego
grudge and vengeance to hold on to what one has.
4
A second component of commitment involves broadened tem-
poral interests, or long-term orientation. Individuals with short-
term orientation may achieve relatively good outcomes by behav-
ing in accord with direct self-interest. Given long-term orientation,
it behooves partners to develop patterns of reciprocal cooperation,
in that Oliver’s long-term well-being may be enhanced if he
forgives Barbara’s transgressions today so that Barbara will for-
give his transgressions next month (cf. Axelrod, 1984; Wu &
Axelrod, 1995). Also, with long-term orientation, the costs of
forgiveness are aggregated over a longer time perspective and in
light of the partner’s reciprocal beneficence (cf. Kelley, 1983).
Thus, forgiveness may be a conscious or unconscious means of
maximizing long-term self-interest.
A third component of commitment involves broadened interper-
sonal interests, or psychological attachment, resting on the percep-
tion that one’s own well-being and the partner’s well-being are
linked. In committed relationships, the self and partner may be-
come merged to the extent that departures from self-interest ben-
efitting the partner are not experienced as antithetical to self-
interest (cf. Agnew, Van Lange, Rusbult, & Langston, 1998; Aron
& Aron, 1997). Also, commitment may yield communal orienta-
tion, including tendencies to respond to a partner’s needs in a
rather unconditional manner. Committed individuals may exert
effort without counting what they receive in return, without cal-
culating whether their beneficence will be reciprocated (cf. Clark
& Mills, 1979). Thus, commitment may inspire rather thoroughly
other-oriented actions.
The empirical literature provides indirect support for this pre-
diction, in that commitment has been shown to be associated with
prorelationship maintenance acts such as derogation of alterna-
tives, accommodative behavior, and willingness to sacrifice (e.g.,
Johnson & Rusbult, 1989; Rusbult et al., 1991; Van Lange et al.,
1997; Wieselquist, Rusbult, Foster, & Agnew, 1999). Relatively
more direct support for this proposition was reported by McCul-
lough et al. (1998, Study 3), who found that a composite measure
of commitment (Stanley & Markman, 1992) and dyadic adjust-
ment (Spanier, 1976) was positively associated with forgiveness.
However, examining the commitment–forgiveness link was a sub-
sidiary goal in the McCullough et al. (1998) research, so the study
did not provide particularly definitive evidence regarding the ef-
fects of commitment: (a) The commitment–forgiveness association
was examined in only one of four studies; (b) the authors used a
composite measure, examining variance attributable to the combi-
nation of commitment and adjustment; (c) retrospective reports of
forgiveness were examined, such that the measure of forgiveness
4
It might be argued that intent to persist is not so simple as we suggest.
Perhaps intent to persist involves more than immediate, dependence-based
need to remain in a relationship. Perhaps intent to persist, in itself, implies
temporally or interpersonally extended concerns. But given that these
broader concerns are directly embodied in the other two components of
commitment (long-term orientation and psychological attachment), if such
broad interests indeed are central to understanding why commitment
promotes interpersonal forgiveness, then we should find that one or both
of these components dominates intent to persist in explaining the
commitment–forgiveness association.
959
COMMITMENT AND FORGIVENESS
may have been colored by self-report bias or motivated memory;
and (d) the study was correlational and did not seek to rule out
possible third-variable confounds, such that cause-and-effect rela-
tions remain ambiguous.
Research Overview
We conducted three studies to test the prediction that commit-
ment motivates forgiveness, using both experimental and nonex-
perimental methods to obtain converging hypothesis-relevant
evidence. Study 1 uses a priming procedure to manipulate com-
mitment and assess reactions to hypothetical betrayals. Study 2 is
a cross-sectional survey study in which individuals described
previous betrayal incidents in their ongoing relationship. Study 3 is
an interaction record study in which individuals provided in-the-
moment reports of betrayal incidents over the course of a 2-week
period. All three studies test the hypothesis that strong commit-
ment is associated with (or causes) increased forgiveness. To
enrich our understanding of the process by which forgiveness
comes about, in Studies 2 and 3 we examined both (a) interper-
sonal forgiveness, or behavioral tendencies in response to betrayal,
and (b) mental events, or the emotional and cognitive concomitants
of such responses. To understand how commitment promotes
forgiveness, in Studies 2 and 3 we assessed whether forgiving
behavior was mediated by cognition and emotion. Because the
transformation process by which commitment shapes motivation
and behavior can be rather automatic and habit driven, we antic-
ipated that mediation by mental events would be partial rather than
complete. We also addressed two subsidiary issues: To explore lay
construals of forgiveness, in Studies 2 and 3 we assessed whether
answers to the statement “I forgive my partner” align with behav-
ioral tendencies and mental events. Finally, to explore precisely
why commitment promotes forgiveness, in Study 3 we adminis-
tered an instrument to measure the three components of commit-
ment—intent to persist, long-term orientation, and psychological
attachment.
Study 1
In Study 1 we developed an experimental procedure to deter-
mine whether commitment causes interpersonal forgiveness in
ongoing relationships. The Study 1 procedure is predicated on the
assumption that individuals experience some day-to-day and
moment-to-moment variation in commitment level. Although a
given individual may exhibit strong commitment on average, the
strength of that commitment presumably varies somewhat in re-
sponse to temporal changes in interaction quality or the salience of
dependence on a relationship. On the basis of this assumption, we
developed a priming procedure to momentarily activate low versus
high commitment to an ongoing relationship. To our knowledge,
this is the first attempt to experimentally manipulate commitment
level in ongoing relationships.
Following the low versus high commitment prime, participants
indicated how they would react to each of several hypothetical acts
of betrayal. We wished to determine whether commitment exerts
differential effects on constructive versus destructive and active
versus passive forgiveness-relevant reactions. Therefore, for each
betrayal, we measured forgiveness with items designed to assess
exit, voice, loyalty, and neglect reactions (Rusbult, 1993). Exit
reactions are actively destructive (e.g., seeking vengeance), ne-
glect reactions are passively destructive (e.g., giving the partner
the cold shoulder), voice reactions are actively constructive (e.g.,
suggesting that the partners discuss the incident), and loyalty
reactions are passively constructive (e.g., continuing to support the
partner despite dissatisfaction). We predicted that in comparison
with participants exposed to the low commitment prime, those
exposed to the high commitment prime would exhibit greater
forgiveness, reacting to betrayal with lesser exit and neglect along
with greater voice and loyalty.
Method
Participants. Participants were 89 undergraduates (22 men, 67 women)
who volunteered to take part in partial fulfillment of the requirements for
introductory psychology at the University of North Carolina at Chapel Hill.
Sign-up sheets listed the following requirement: “To participate you must
currently be involved in a dating relationship of at least one month in
duration.” Individuals participated in groups ranging in size from 2 to 15
persons. Within each session, participants were randomly assigned to one
of two experimental conditions (low vs. high commitment prime), with
about equal proportions of men and women in the two conditions. Partic-
ipants had been involved with their partner for 19.81 months on average.
Most described their relationships as steady dating relationships (7% dating
casually, 13% dating regularly, 75% dating steadily, 3% engaged or mar-
ried), and most indicated that they dated their partner exclusively (91%
reported that neither partner dated others, 7% reported that one partner
dated others, 2% reported that both dated others). About 44% indicated that
they were involved in long-distance relationships.
Procedure. In priming research, it is desirable that participants remain
unaware of any link between the priming manipulation and key dependent
variables. Accordingly, participants were informed that they would be
asked to take part in two separate studies during the session. The first of
these was described as a pilot study for work in an extended research
program concerning relationships, and the second was described as a study
of the positive and negative events that routinely occur in relationships. To
reinforce the impression that the two portions of the study were indepen-
dent, we ensured that the name of the experimenter conducting the osten-
sible Study 1 was different from the name of the experimenter conducting
the ostensible Study 2, and we printed research materials using different
fonts and print sizes.
In Study 1, participants completed a one-page questionnaire including
five open-ended questions. Participants in the high commitment prime
condition answered questions designed to activate thoughts regarding
dependence and commitment (e.g., “If your relationship were to end in the
near future, what would upset you the most about not being with your
partner anymore?”; “Describe two ways in which you feel that your life has
become “linked to’ your partner”). Participants in the low commitment
prime condition answered questions designed to activate thoughts regard-
ing independence and lack of commitment (e.g., “Describe an activity that
you enjoy engaging in when your partner is not around”; “Describe two
ways in which you are independent of your partner”). When all participants
had completed activities for Study 1, Study 2 commenced.
We explained that the goal of Study 2 was to determine how dating
partners react to a variety of events. Participants read descriptions of 12
hypothetical acts of betrayal (e.g., “Your partner lies to you about some-
thing important”; “Your partner flirts with a classmate”) and responded to
four items regarding each betrayal. In keeping with previous research
regarding responses to dissatisfaction, the four items for each betrayal
assessed tendencies toward exit, voice, loyalty, and neglect (Rusbult &
Zembrodt, 1983; one measure of each response category for each of 12
acts; e.g., “I would suggest that we go out to dinner and have a constructive
talk about flirting”;0⫽ not at all likely to react this way,8⫽ extremely
likely to react this way).
960
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
To assess the effectiveness of the commitment prime, we then measured
commitment level using the Investment Model Scale (Rusbult et al., 1998;
seven items; e.g., “I would feel very upset if our relationship were to end
in the near future”;0⫽ do not agree at all,8⫽ agree completely). We also
measured self-deception and impression management using the Balanced
Inventory of Desirable Responding (Paulhus, 1984; 40 items; e.g., “I
always obey laws, even if I’m unlikely to get caught”;1⫽ do not agree at
all,7⫽ agree completely). At the end of the session, participants were
debriefed and thanked for their assistance. During the debriefing, the
experimenter carefully probed for suspicion. No participants indicated
awareness of a link between Study 1 and Study 2. Also, no participants
knew that whereas some people answered Study 1 questions regarding
dependence and commitment, others answered questions regarding inde-
pendence and lack of commitment.
Reliability and validity of measures. Reliability analyses revealed ac-
ceptable coefficients for items designed to measure exit, voice, loyalty, and
neglect (
␣
s ⫽ .78, .84, .79, and .72, respectively), commitment level (
␣
⫽
.86), and self-deception and impression management (
␣
s ⫽ .64 and .78,
respectively). Therefore, we developed a single measure of each vari-
able—a count of extreme scores for self-deception and impression man-
agement (following Paulhus’, 1984, procedure), and item averages for
other variables. We performed correlational analyses to evaluate the va-
lidity of our measures. Consistent with expectations, destructive exit and
neglect were strongly positively correlated with one another, r(89) ⫽ .75,
p ⬍ .01, and were moderately negatively correlated with measures of voice
and loyalty (exit, average r ⫽⫺.30; neglect, average r ⫽⫺.26); also,
constructive voice and loyalty were strongly positively correlated with one
another, r(89) ⫽ .58, p ⬍ .01, and were moderately negatively correlated
with measures of exit and neglect (voice, average r ⫽⫺.37; loyalty,
average r ⫽⫺.19).
Results and Discussion
Manipulation check. To determine whether the priming ma-
nipulation influenced momentary feelings of commitment, we
performed a two-factor analysis of variance on the commitment
prime manipulation check. In this analysis, commitment prime
(low vs. high) and participant sex (male vs. female) were between-
subjects variables. The analysis revealed a significant main effect
of the commitment prime, F(1, 85) ⫽ 11.79, p ⬍ .01. Participants
reported stronger subjective commitment in the high commitment
prime condition than in the low commitment prime condition
(Ms ⫽ 7.06 and 5.73, respectively). In addition, the main effect
of sex was significant, F(1, 85) ⫽ 3.91, p ⬍ .05. Compared
with women, men reported stronger subjective commitment in
both the low and the high commitment prime condition (means for
low and high commitment: men ⫽ 6.14 and 7.13, respectively;
women ⫽ 4.63 and 6.81).
Effects of commitment prime and participant sex. To evaluate
the overall effects of the commitment prime and participant sex on
exit, voice, loyalty, and neglect, we first analyzed all four mea-
sures simultaneously. The criteria in this analysis were measures of
the four response tendencies; the independent variables were con-
structiveness of response (neglect and exit vs. loyalty and voice; a
within-subject factor), activity of response (neglect and loyalty vs.
exit and voice; a within-subject factor), commitment prime (low
vs. high; a between-subjects factor), and participant sex (male vs.
female; a between-subjects factor). To interpret the overall analy-
sis, we also analyzed each dependent variable individually—sep-
arately for each response, we performed analyses that included
as independent variables commitment prime and participant sex.
Table 1 presents mean scores for each dependent variable as a
function of the commitment prime along with summary statistics
for the commitment prime main effect.
As can be seen in Table 1, compared with participants in the low
commitment prime condition, those in the high commitment prime
condition exhibited descriptively lower exit and neglect as well as
descriptively greater voice and loyalty. The overall analysis re-
vealed a significant interaction of Commitment Prime ⫻ Construc-
tiveness of Response, F(1, 85) ⫽ 6.78, p ⬍ .01. Consistent with
the claim that strong commitment causes forgiveness, univariate
analyses revealed that the commitment main effect was significant
for exit and neglect tendencies (see Table 1). However, the com-
mitment effect was nonsignificant for voice and loyalty. Thus, the
priming of commitment more powerfully influenced destructive
reactions (exit and neglect) than constructive reactions (voice and
loyalty).
The overall analysis also revealed a significant interaction of
Participant Sex ⫻ Constructiveness of Response ⫻ Activity of
Response, F(1, 85) ⫽ 10.44, p ⬍ .01. Univariate analyses revealed
that the main effect of participant sex was significant for voice
tendencies (see Table 1). In comparison with men, women exhib-
ited somewhat greater forgiveness, reporting greater inclinations to
react to betrayal with voice (Ms ⫽ 4.42 and 5.42, respectively).
Finally, the overall analysis revealed several significant effects
involving type of response—a main effect of constructiveness of
response, F(1, 85) ⫽ 49.70, p ⬍ .01, a main effect of activity of
response, F(1, 85) ⫽ 6.06, p ⬍ .02, and an interaction of Con-
structiveness ⫻ Activity, F(1, 85) ⫽ 221.19, p ⬍ .01. The main
effects reflect the fact that participants exhibited higher scores on
average for constructive than destructive responses and exhibited
higher scores on average for passive than active responses. The
interaction effect reflects the fact that for constructive responses,
scores were higher for active voice than for passive loyalty,
whereas for destructive responses, scores were higher for passive
neglect than for active exit. Does this mean that in everyday life,
individuals are most likely to react to betrayal with actively for-
giving behaviors and are least likely to react with actively vengeful
behaviors? We think not, in that absolute levels of agreement with
each exit, voice, loyalty, and neglect item rest on the precise
Table 1
Forgiveness of Partner Betrayal as a Function of Commitment
Level: Study 1
Tendency
Low
commitment
prime
High
commitment
prime
Commitment
main effect

F(1, 85)
Exit 2.51 1.85 ⫺.42 13.15**
Voice 4.95 5.39 .17 2.10
Loyalty 3.80 4.10 .05 0.20
Neglect 3.92 3.50 ⫺.32 6.94**
Note. Values in the Low commitment prime and High commitment prime
columns are means for each experimental condition. Higher values reflect
greater levels of each construct; the possible range for each variable is
from 0 to 8. Each row presents findings regarding the impact of commit-
ment on a single criterion (e.g., on exit tendencies). The

column presents
standardized coefficients for the effect, and the F column presents F
statistics for the effect.
** p ⬍ .01.
961
COMMITMENT AND FORGIVENESS
wording of the items used in the present work. Thus, readers
should exercise caution in interpreting these findings.
5
Are the effects of the commitment prime mediated by subjective
commitment? Our priming manipulation was intended to activate
thoughts regarding dependence and commitment, thereby modify-
ing participants’ subjective commitment to their relationship. Are
the observed effects of the prime indeed attributable to across-
conditions differences in subjective commitment? To address this
issue, we performed mediation analyses; examining the role of
subjective commitment in mediating the impact of the prime on
exit and neglect tendencies (cf. Baron & Kenny, 1986). Consistent
with the requirements for assessing mediation, the measure of
subjective commitment (a) differed significantly for the low and
high commitment prime conditions (see earlier analyses) and (b)
was significantly correlated with measures of exit and neglect,
rs(89) ⫽⫺.30 and ⫺.29, respectively, both ps ⬍ .01.
To evaluate whether the effects of the commitment prime are
attributable to differences in subjective commitment, we examined
the impact of the commitment prime in analyses in which we
included subjective commitment as a covariate. In the analysis for
exit, the effect of subjective commitment was significant,

(86) ⫽
.23, p ⬍ .03, and the main effect of the commitment prime
declined somewhat: excluding subjective commitment,

(87) ⫽
.30, p ⬍ .01; including subjective commitment,

(86) ⫽ .22, p ⬍
.04. A test of the significance of mediation (Kenny, Kashy, &
Bolger, 1998) revealed that subjective commitment marginally
mediated the impact of the prime on exit tendencies (z ⫽ 1.86, p ⬍
.06). In the analysis for neglect, the effect of subjective commit-
ment was significant,

(86) ⫽ .25, p ⬍ .02, and the main effect of
the commitment prime declined to nonsignificance: excluding
subjective commitment,

(87) ⫽ .22, p ⬍ .04; including subjective
commitment,

(86) ⫽ .14, ns. Subjective commitment marginally
mediated the impact of the prime on neglect (z ⫽ 1.94, p ⬍ .06).
Thus, the effects of the manipulation were at least partially attrib-
utable to the effects of the prime on participants’ subjective com-
mitment. (Of course, subjective commitment was assessed follow-
ing measurement of exit, voice, loyalty, and neglect. Given that
subjective commitment may have been influenced not only by the
commitment prime but also by questions regarding betrayal and
forgiveness, conclusions regarding the relationship between sub-
jective commitment and interpersonal forgiveness should be re-
garded as tentative.)
Ruling out alternative explanations. Is it possible that the
observed effects of the commitment prime are spurious, resulting
from inclinations toward socially desirable responding? To explore
this possibility, we first calculated the correlations of our exit,
voice, loyalty, and neglect measures with self-deception and im-
pression management. Self-deception was not significantly corre-
lated with any of the measures, rs(89) ⫽ .05, .05, ⫺.06, and ⫺.04,
respectively, all ns. However, impression management was signif-
icantly or marginally correlated with exit, voice, loyalty, and
neglect, rs(89) ⫽ .32, ⫺.29, ⫺.20, and .20, respectively, all ps ⬍
.06. Subjective commitment—the priming manipulation check—
was not significantly correlated with self-deception or impression
management, rs(89) ⫽ .02 and .11, respectively, both ns.
To ensure that the earlier reported effects of the commitment
prime were evident even when we controlled for socially desirable
responding, we performed auxiliary analyses of covariance, exam-
ining the impact of the commitment prime and participant sex on
each dependent variable; we included self-deception and impres-
sion management, in turn, as covariates. The earlier reported
analyses revealed significant effects of the commitment prime for
exit and neglect; in auxiliary analyses, the commitment effect
remained significant in analyses that included as covariates both
self-deception, Fs(1, 84) ⫽ 13.10 and 6.49, respectively, both ps ⬍
.01, and impression management, Fs(1, 84) ⫽ 9.27 and 4.84, both
ps ⬍ .03. Also, the earlier reported analyses revealed a main effect
of participant sex for voice; in auxiliary analyses, the sex main
effect remained significant, Fs(1, 84) ⫽ 9.38 and 7.17, respec-
tively, both ps ⬍ .01. Thus, earlier reported findings do not appear
to be attributable to socially desirable responding.
Study 2
Study 1 reveals evidence that in comparison with less commit-
ted individuals, highly committed individuals are more likely to
forgive partners’ acts of betrayal. Given that Study 1 effected an
experimental manipulation of commitment level, these findings
provide good support for the assumed causal role of commitment
in encouraging interpersonal forgiveness. However, Study 1 ex-
amined relatively artificial norm violations, exploring reactions to
hypothetical acts of partner betrayal.
In Study 2, we used a nonexperimental method to examine
associations with commitment in real betrayal incidents. Partici-
pants recalled an incident in which the dating partner violated a
5
Are the effects of commitment evident for diverse types of betrayal?
The partner acts to which participants responded involved four categories
of betrayal—violations of monogamy norms (e.g., “You find out that your
partner kissed someone else at a party”), dependence norms (e.g., “In a
disagreement with a third person, your partner takes the other person’s
side”), privacy norms (e.g., “Your partner tells friends about an embar-
rassing secret from your past”), and decency and etiquette norms (e.g.,
“Your partner forgets your birthday”). We calculated measures of exit,
voice, loyalty, and neglect separately for each type of betrayal and per-
formed exploratory analyses to assess the effects of betrayal type (monog-
amy vs. dependence vs. privacy vs. decency and etiquette) in conjunction
with constructiveness of response, activity of response, commitment prime,
and participant sex.
In addition to replicating earlier reported findings, this analysis reveals
that the interaction of Commitment Prime ⫻ Betrayal Type was nonsig-
nificant—the impact of commitment did not differ as a function of betrayal
type. Consistent with earlier analyses, the commitment effect was signif-
icant or marginal for exit in reaction to all four types of betrayal, Fs(1,
85) ⫽ 4.81, 7.95, 8.42, 11.90, respectively, all ps ⬍ .03, for neglect in
reaction to privacy and monogamy betrayals, Fs(1, 85) ⫽ 3.92 and 4.11,
both ps ⬍ .05, and for voice in reaction to privacy betrayals, F(1,
85) ⫽ 3.67, p ⬍ .06. Also, the sex effect was significant or marginal for
voice in reaction to privacy betrayals, F(1, 85) ⫽ 10.69, p ⬍ .01, and for
exit and voice in reaction to monogamy betrayals, respective Fs(1,
85) ⫽ 2.91 and 8.42, both ps ⬍ .09. The analysis also revealed several
effects involving betrayal type: For dependence and privacy violations,
participants were somewhat more constructive than destructive—they were
most likely to react with voice, followed by loyalty, followed by neglect
and exit. In contrast, participants were somewhat more destructive than
constructive in reacting to decency and etiquette violations (e.g., lying,
deliberately hurtful behavior) and were considerably more destructive than
constructive in reacting to monogamy violations (e.g., flirting, sexual
infidelity).
962
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
relationship-relevant norm and described their immediate and de-
layed reactions to the incident. We examined both immediate
and delayed reactions because, unlike other sorts of interdepen-
dence dilemmas, incidents involving betrayal may not be resolved
in the course of a single interaction—in reacting to real betrayals,
the process by which individuals achieve forgiveness may unfold
over time. In addition to assessing (a) immediate and delayed
behavioral tendencies (e.g., vengeance, stated forgiveness), we
also examined the mental events accompanying forgiveness, in-
cluding (b) immediate and delayed cognitive interpretations (e.g.,
attributing the partner’s actions to internal vs. external causes) and
(c) immediate and delayed emotional reactions (e.g., sadness,
anger).
Assuming that forgiveness is not easy and that the forgiveness
process unfolds over time, we anticipated that individuals would
exhibit less forgiveness immediately following betrayal than at a
later time. We anticipated that we might observe one of two
patterns for commitment. First, we might observe a main effect of
commitment, such that strong commitment promotes forgiveness
both immediately following betrayal and at a later time. Second,
we might observe an interaction of commitment with time, such
that highly committed individuals exhibit greater movement over
time toward forgiveness. The latter possibility seems plausible in
that the motives underlying forgiveness may exert their effects
over the course of extended interaction—immediate, gut-level
impulses may be rather negative irrespective of commitment level
(immediate reactions may reflect pretransformation impulses),
whereas over time, relatively committed individuals may find their
way to prorelationship motives and forgiveness (delayed reactions
may reflect posttransformation tendencies). Indeed, the strength of
immediate and delayed associations with commitment may differ
for emotion, cognition, and behavioral tendencies. We did not
advance a priori predictions about whether we would observe main
effects of commitment or interactions of commitment with time,
nor did we specify whether patterns of association with commit-
ment would differ for the three types of variable we examined.
Given that Study 2 uses a nonexperimental method, it is impor-
tant to demonstrate—insofar as it is possible to do so—that any
observed association of commitment with forgiveness is not at-
tributable to variables that may be confounded with these vari-
ables. Accordingly, we examined the associations of duration of
relationship and recency of betrayal with commitment and forgive-
ness, reasoning that less committed individuals might be involved
in briefer relationships, might describe more recent betrayals, and
might exhibit weaker forgiveness not because they are less com-
mitted but because they experienced more recent betrayal (i.e.,
they had less time to work through the forgiveness process). We
also examined the association of severity of betrayal with com-
mitment and forgiveness, reasoning that less committed individu-
als might be less committed precisely because they experienced
more severe betrayals or might exhibit weaker forgiveness because
they experienced more severe betrayals (rendering it more difficult
to forgive). We examined six possible confounds, including (a)
properties of betrayal incidents (severity of betrayal, time since
betrayal), (b) features of relationships (duration of relationship,
long-distance involvement), and (c) tendencies toward socially
desirable responding (self-deception, impression management).
Method
Participants. Participants were 155 undergraduates (50 men, 104
women, 1 person who did not specify sex) who volunteered to take
part in partial fulfillment of the requirements for introductory psychol-
ogy at the University of North Carolina at Chapel Hill. As in Study 1,
sign-up sheets indicated that to take part, participants must be involved
in a dating relationship of at least 1 month in duration. Individuals
participated in groups ranging in size from 2 to 20 persons. Participants
had been involved with their dating partner for 16.56 months on
average. Most described their relationships as steady dating relation-
ships (24% dating casually, 13% dating regularly, 57% dating steadily,
6% engaged or married), and most indicated that they dated their
partner exclusively (80% reported that neither partner dated others, 4%
reported that one partner dated others, 16% reported that both dated
others). About 29% indicated that they were involved in long-distance
relationships.
Procedure. Participants were told that the goal of the study was to
explore the manner in which partners react to the positive and negative
events that routinely occur in dating relationships. In one portion of the
questionnaire, participants were asked to describe a partner act of
betrayal. We avoided using the word betrayal in the questionnaire, in
that betrayal may connote exclusively sexual norm violations or may
arouse anxiety or desire to present the self or the relationship in a
socially desirable manner. Therefore, incidents of betrayal were de-
scribed in the following manner:
All of us have expectations about how our partners should treat us. No
matter how well-behaved a partner may be in general, from time to
time he or she is likely to violate those expectations; that is, your
partner is likely to “break the rules” of your relationship. For example:
your partner may talk to a friend about something that you believe
should have remained private; your partner may do something that is
hurtful to you behind your back; your partner may bring up sensitive
issues from your past that you believe should have been forgotten; or
your partner may otherwise behave in a way that violates your
expectations about how your partner should behave.
Participants were asked to (a) write a paragraph description of what the
partner did that violated their expectations, (b) write a paragraph descrip-
tion of how they reacted to the partner’s behavior, and (c) indicate when the
incident occurred (i.e., number of weeks/months/years ago).
Then participants completed a questionnaire designed to assess im-
mediate reactions to betrayal (reactions at the time the incident oc-
curred; i.e., “What were your immediate thoughts and feelings about the
way your partner behaved; what was your initial reaction to your
partner’s behavior?”) and delayed reactions to betrayal (reactions at
present; i.e., “What are your current thoughts and feelings about the
way your partner behaved; what is your present reaction to your
partner’s behavior?”). All participants reported immediate reactions
before reporting delayed reactions. For both immediate and delayed
reactions, we measured positive behavioral tendencies (seven items;
e.g., “I thought that I had the right to “get even’ with my partner,”
reverse scored; “I forgave my partner”), positive cognitive interpreta-
tions (four items; e.g., “I thought that my partner didn’t try hard enough
to behave in a positive manner,” reverse scored), and positive emotional
reactions (four items; e.g., “I felt very angry about the way my partner
behaved,” reverse scored; 0 ⫽ do not agree at all,8⫽ agree com-
pletely). Immediate and delayed items were identical except for changes
in verb tense (e.g., for delayed reactions, “I feel very upset about the
way my partner behaved”).
Participants also completed several other instruments. As in Study 1, we
assessed commitment level using the measure from the Investment Model
Scale (Rusbult et al., 1998). We also measured several potential confounds,
including duration of relationship (number of months), time since betrayal
963
COMMITMENT AND FORGIVENESS
(number of weeks), and long-distance involvement (“Do you and your
partner live within 60 miles of each other?”;1⫽ yes,2⫽ no). As in
Study 1, we measured self-deception and impression management using
the Balanced Inventory of Desirable Responding (Paulhus, 1984). At the
end of the session, participants were debriefed and thanked for their
assistance.
Measuring severity of betrayal. To assess severity of betrayal, we
asked trained undergraduate research assistants to rate participants’
paragraph descriptions of partner betrayal. We wished to obtain objec-
tive ratings of severity, unconfounded by the level of distress partici-
pants reported and unconfounded by the consequences of the incident.
Therefore, research assistants based their ratings on participants’ para-
graph descriptions of betrayal incidents and were unaware of partici-
pants’ reactions to the partner’s behavior (as well as other questionnaire
responses). Also, assistants adopted the perspective of an objective
observer, rating incidents in relation to absolute standards, not in
relation to the participant’s perception of incident severity. Two assis-
tants rated each description on three dimensions: (a) severity of betrayal
(i.e., “In the context of the average undergraduate dating relationship,
how serious a betrayal is this?”;0⫽ not at all serious,4⫽ extremely
serious), (b) victim distress (i.e., “How distressed would the average
person in the average undergraduate dating relationship feel about this
incident?”;0⫽ no distress at all,4⫽ extreme distress), and (c)
potential for relationship harm (i.e., “How much potential is there for an
incident of this sort to harm the average undergraduate relationship?”;
0 ⫽ no harm at all,4⫽ extreme harm).
Reliability and validity of measures. Reliability analyses revealed ac-
ceptable coefficients for items designed to measure commitment level (
␣
⫽
.96), self-deception (
␣
⫽ .54), impression management (
␣
⫽ .80), imme-
diate and delayed behavioral tendencies (
␣
s ⫽ .83 and .87, respectively),
and immediate and delayed emotional reactions (
␣
s ⫽ .82 and .90). One
item measuring cognitive interpretations exhibited weak associations with
the remaining items, so that item was dropped (alphas for three-item
measures ⫽ .68 and .64). Research assistants’ ratings of severity of
betrayal exhibited good interrater agreement: for severity, distress, and
harm, intraclass rs(155) ⫽ .52, .48, and .43, respectively, all ps ⬍ .01; for
averaged measures of the three dimensions,
␣
⫽ .95. Therefore, we
developed a single measure of each variable—a count of extreme scores for
self-deception and impression management, and item averages for other
variables.
To explore the lay meaning of forgiveness, we examined the association
of the item “I forgive my partner” with immediate and delayed behavior,
cognition, and emotion (dropping “I forgive. . .” from measures of behav-
ioral tendencies; alphas for six-item measures ⫽ .84 and .88). The mea-
sures of immediate and delayed stated forgiveness (“I forgive. . .”) were
positively correlated with immediate and delayed behavioral tendencies,
rs(153) ⫽ .40 and .52, both ps ⬍ .01, and cognitive interpretations,
rs(153) ⫽ .29 and .24, both ps ⬍ .01. Delayed stated forgiveness was
positively associated with delayed emotions, r(153) ⫽ .20, p ⬍ .01, but
immediate stated forgiveness was not significantly associated with imme-
diate emotions, r(153) ⫽ .01, ns. Thus, “I forgive my partner” tended to be
more closely aligned with behavioral tendencies than with cognition or
emotion. (In all remaining analyses, measures of behavior represent aver-
ages that include scores for stated forgiveness.)
We also examined the associations among measures of immediate and
delayed behavioral tendencies, cognitive interpretations, and emotional
reactions. The six measures were positively correlated, average r ⫽ .53;
rs(153) ranged from .33 to .77, all ps ⬍ .01. As expected, measures of
immediate and delayed constructs exhibited strong within-construct con-
vergence (average correlation for immediate and delayed measures of the
same construct ⫽ .66) and moderate across-constructs convergence (aver-
age correlation for immediate and delayed measures of different con-
structs ⫽ .48).
Content of betrayal incidents. What sorts of betrayal incidents did
participants describe? Our raters’ severity codings suggest that the average
incident was moderately serious, distressing, and potentially harmful
(M ⫽ 2.05, SD ⫽ 0.86). We examined the content of participants’ para-
graph descriptions and found that they concerned (a) violations of monog-
amy norms (22%)—the partner was emotionally or physically unfaithful or
exhibited insufficient commitment (e.g., flirted without regard to partici-
pant’s feelings), (b) violations of dependence norms (35%)—the partner
exhibited jealous or possessive behavior or let the participant down (e.g.,
insisted that they have sex, even though participant did not want to or
brought friends along on what was meant to be a romantic dinner), (c)
violations of privacy norms (15%)—the partner shared private information
with others (e.g., talked to someone else about a personal matter), and (d)
violations of decency or etiquette norms (22%)—the partner lied to or
misled the participant, insulted the participant, embarrassed the participant
in public, was judgmental, or engaged in disapproved activities (e.g.,
claimed to go home after work but instead went to a party; bragged about
sex in front of family; said loudly that the participant smelled bad). An
additional 6% of the descriptions were unclassifiable (e.g., insufficient
information to code).
Results and Discussion
Effects of commitment level, time, and participant sex. To
evaluate the overall effects of commitment, time, and partici-
pant sex across our three dependent variables, we first analyzed
all three measures simultaneously. The criteria in this analysis
were measures of behavioral tendencies, cognitive interpretations,
and emotional reactions; the independent variables were time
(immediate vs. delayed reactions; a within-subject factor), partic-
ipant sex (male vs. female; a between-subjects factor), and com-
mitment level (a continuous, between-subjects factor). To interpret
the overall analysis, we also analyzed each dependent variable
individually, including time, participant sex, and commitment
level as independent variables. Table 2 presents mean scores for
each dependent variable as a function of time and commitment:
means for low versus high commitment were conditioned at values
of one standard deviation above and below the mean of commit-
ment level (Aiken & West, 1991). Table 2 also presents statistics
for the commitment main effect and the Time ⫻ Commitment
interaction.
As can be seen in Table 2, in comparison with low commitment
participants, high commitment participants exhibited descriptively
more positive immediate and delayed behavioral tendencies, im-
mediate and delayed cognitive interpretations, and delayed emo-
tional reactions; compared with low commitment participants, high
commitment participants exhibited descriptively more negative
immediate emotional reactions. The overall analysis revealed sig-
nificant main effects of commitment, F(1, 148) ⫽ 4.42, p ⬍ .04,
and time, F(1, 148) ⫽ 28.11, p ⬍ .01, along with an interaction of
Commitment ⫻ Time, F(1, 148) ⫽ 7.65, p ⬍ .01. Consistent with
expectations, the main effect of commitment was significant for
behavioral tendencies and cognitive interpretations (see Table 2).
The Time ⫻ Commitment interaction was significant for emo-
tional reactions—commitment was marginally negatively associ-
ated with immediate emotions, r(153) ⫽⫺.16, p ⬍ .06, but was
marginally positively associated with delayed emotions, r(153) ⫽
.14, p ⬍ .09. The main effect of time was significant for behavior,
964
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
F(1, 148) ⫽ 25.63, p ⬍ .01, cognition, F(1, 148) ⫽ 4.59, p ⬍ .03,
and emotion, F(1, 148) ⫽ 15.67, p ⬍ .01.
6
The overall analysis also revealed two significant effects involv-
ing participant sex—a main effect of participant sex, F(1,
148) ⫽ 10.20, p ⬍ .01, and an interaction of Commitment ⫻ Sex,
F(1, 148) ⫽ 5.40, p ⬍ .02. Univariate analyses revealed that the
main effect of participant sex was significant for behavior, F(1,
148) ⫽ 4.49, p ⬍ .04, cognition, F(1, 148) ⫽ 5.75, p ⬍ .02, and
emotion, F(1, 148) ⫽ 10.42, p ⬍ .01—in Study 2 (inconsistent
with the results of Study 1), compared with men, women exhibited
fewer positive behavioral tendencies (Ms ⫽ 6.29 and 5.87, respec-
tively), cognitive interpretations (Ms ⫽ 5.20 and 4.76), and emo-
tional reactions (Ms ⫽ 4.26 and 3.76). The interaction of Com-
mitment ⫻ Sex was significant for emotional reactions, F(1,
148) ⫽ 7.47, p ⬍ .01—on average, commitment was more
strongly associated with immediate and delayed emotion among
women (average r ⫽ .32) than among men (average r ⫽ .13).
Thus—and consistent with the assumption that forgiveness is
not easy—individuals’ immediate reactions to betrayal were more
negative than were their delayed reactions. Also, the commitment
main effect was significant for behavior and cognition—to the
degree that individuals were committed, they formed more benign
interpretations of the partner’s acts and reported more positive
behavioral tendencies. Also—and consistent with the claim that
the motives underlying forgiveness may work their magic over
time—the Commitment ⫻ Time interaction was significant for
emotions: Compared with less committed individuals, highly com-
mitted individuals exhibited marginally more negative immediate
emotions (perhaps reflecting pretransformation impulses) and ex-
hibited marginally more positive delayed emotions (perhaps re-
flecting posttransformation tendencies). That is, committed indi-
viduals were more interpersonally forgiving despite the fact that
they initially were more hurt and angered by betrayal. (However,
note that the time perspective for reporting commitment was
concurrent with that for reporting delayed reactions.) Finally, in
Study 2 men exhibited more positive emotions, cognitions, and
behavior on average than did women, and the association of
commitment with emotions was stronger among women than men.
Are associations with commitment mediated by mental events?
Earlier, we suggested that the association of commitment with
interpersonal forgiveness may be partially mediated by mental
events. To address this issue, we performed mediation analyses,
examining the role of cognitive interpretations and emotional
reactions in mediating the association of commitment with behav-
ioral tendencies (cf. Baron & Kenny, 1986). Consistent with the
requirements for assessing mediation, our measures of cognition
and emotion were significantly associated with behavioral tenden-
cies; although cognition was significantly associated with commit-
ment, the positive association of emotion with commitment was
evident only for delayed reactions (see earlier analyses).
To evaluate whether the association of commitment with behav-
ioral tendencies is attributable to cognitive interpretations, we
regressed our measure of behavioral tendencies simultaneously
onto commitment and cognition. In this analysis, the association
with cognition was significant,

(148) ⫽ .55, p ⬍ .01, and the
6
We also reasoned that if the motives underlying forgiveness exert their
effects over the course of extended interaction, then we should observe an
interaction of commitment with time since betrayal (number of weeks). To
explore this line of reasoning, we examined the effects on delayed behav-
ioral tendencies, cognitive interpretations, and emotional reactions of com-
mitment, time since betrayal, and the interaction of these variables. These
analyses revealed marginally significant Time ⫻ Commitment interactions
for emotional reactions and behavioral tendencies, Fs(1, 141) ⫽ 3.00
and 3.13, respectively, both ps ⬍ .09; the interaction was nonsignificant for
cognitive interpretations, F(1, 141) ⫽ 0.10, ns. That two of three associ-
ations with commitment were marginally stronger to the extent that more
time had elapsed since the betrayal is compatible with the claim that, over
time, committed individuals find their way to prorelationship emotional
reactions and interpersonal forgiveness.
Table 2
Forgiveness of Partner Betrayal as a Function of Time and Commitment Level: Study 2
Forgiveness component
Low
commitment
High
commitment
Commitment
main effect
Time ⫻
Commitment
interaction

F(1, 148)

F(1, 148)
Positive behavioral tendencies .20 6.01* .09 1.21
Immediate reactions 4.80 5.18
Delayed reactions 6.82 7.48
Positive cognitive interpretations .25 8.90** .09 1.07
Immediate reactions 4.14 4.86
Delayed reactions 4.95 5.89
Positive emotional reactions .00 0.00 .28 11.73**
Immediate reactions 2.64 2.12
Delayed reactions 5.16 5.68
Note. Values in the Low commitment and High commitment columns are means for each condition. Higher
values reflect greater levels of each construct; the possible range for each variable is from 0 to 8. Predicted means
were conditioned at values of one standard deviation above and below the mean of commitment level (Aiken &
West, 1991). For each dependent variable, we present findings regarding the association of commitment with that
single criterion only (e.g., with positive behavioral tendencies). The

columns present standardized coefficients
for the effect, and the F columns present F statistics for the effect.
* p ⬍ .05. ** p ⬍ .01.
965
COMMITMENT AND FORGIVENESS
association with commitment declined substantially: excluding
cognition,

(149) ⫽ .20, p ⬍ .02; including cognition,

(148) ⫽
.06, ns. A test of the significance of mediation revealed that
cognitive interpretations significantly mediated the association
with commitment (z ⫽ 2.82, p ⬍ .01). In contrast, when we
regressed behavioral tendencies onto commitment and emotional
reactions, the association with emotion was significant,

(148) ⫽
.54, p ⬍ .01, but the association with commitment remained
sizeable: including emotion,

(148) ⫽ .20, p ⬍ .01. (It is not
appropriate to test for the significance of mediation by emotional
reactions because this variable did not meet the prerequisites for
assessing mediation.) Thus, committed individuals may feel quite
hurt or angry following betrayal—indeed, high commitment par-
ticipants experienced more negative immediate emotion than did
low commitment participants—yet still find their way to forgive-
ness. In contrast, developing benevolent (or less malevolent) cog-
nitive interpretations of a partner’s behavior appears to be a key to
understanding the association of commitment with interpersonal
forgiveness.
Ruling out alternative explanations. It is possible that the
association of commitment with interpersonal forgiveness is spu-
rious, resulting from confounds such as the recency or severity
of partner betrayal? To explore this possibility, we first calculated
the correlations of commitment and both immediate and delayed
behavioral tendencies with severity of betrayal, time since be-
trayal, duration of relationship, long-distance involvement, self-
deception, and impression management; for the duration measures
(time since betrayal, duration of relationship), we examined both
simple and log-transformed measures. Commitment was nega-
tively associated with long-distance involvement, r(154) ⫽⫺.23,
p ⬍ .01, was positively associated with severity of betrayal, time
since betrayal (simple and transformed), duration of relationship
(simple and transformed), and impression management, rs(145 to
155) ranged from .18 to .47, all ps ⬍ .05, and was essentially
unrelated to self-deception, r(155) ⫽ .03. Measures of immediate
and delayed behavior were negatively associated with severity of
betrayal, rs(148) ⫽⫺.40 and ⫺.38, respectively, both ps ⬍ .01,
but were weakly or inconsistently related to time since betrayal
(simple and transformed), duration of relationship (simple and
transformed), long-distance involvement, self-deception, and im-
pression management, rs(145 to 150) ranged from ⫺.25 to .17, 10
of 14 were ns.
To ensure that the earlier reported findings are evident even
when we control for each potential confound, we performed aux-
iliary analyses of covariance, examining the associations of com-
mitment and sex with behavioral tendencies. In these analyses,
time was a within-subject variable and commitment and sex were
between-subjects variables; we included each potential confound,
in turn, as a covariate (e.g., one analysis examined the effects of
time, sex, commitment, and severity of betrayal on behavior). The
earlier reported analysis revealed a main effect of commitment on
behavioral tendencies; auxiliary analyses revealed that when we
control for each potential confound, the commitment main effect
remains significant, Fs(1, 139 to 147) ranged from 4.10 to 11.64,
all ps ⬍ .04. Thus, the commitment–forgiveness association is not
entirely attributable to any of several variables with which com-
mitment and interpersonal forgiveness were associated.
7
In addi-
tion, the earlier reported analysis revealed a main effect of time;
auxiliary analyses revealed that the time effect remains significant
in analyses in which we control for each potential confound, Fs(1,
139 to 147) ranged from 10.66 to 31.22, all ps ⬍ .01.
Finally, the earlier reported analysis revealed a main effect of
participant sex; auxiliary analyses revealed that the sex effect
remains significant in analyses in which we control for all but one
potential confound, Fs(1, 139 to 147) ranged from 4.33 to 12.97,
all ps ⬍ .04. In the analysis controlling for severity of betrayal, the
sex effect declined to nonsignificance, F(1, 147) ⫽ 2.43, ns,
suggesting that the earlier reported sex effect might be attributable
to sex differences in severity (e.g., perhaps men experienced less
severe betrayals). However, men’s and women’s betrayal descrip-
tions did not differ significantly in rated severity (Ms ⫽ 1.95
and 2.10, respectively), F(1, 147) ⫽ 1.00, ns. (Moreover, severity
did not significantly mediate the association of sex with behavioral
tendencies, z ⫽ 1.47, ns.) Given that severity did not meet the
requirements for assessing mediation, these results are
inconclusive.
Study 3
Study 1 reveals evidence that commitment is causally linked
with forgiveness, and Study 2 reveals that in people’s reactions to
real betrayal incidents in ongoing relationships, commitment is
positively associated with forgiveness. However, the results of
Study 2 rest on retrospective accounts of prior betrayal incidents,
so it is possible that individuals’ accounts are colored by motivated
memory processes. For example, it is possible that strong com-
mitment yields tendencies to reconstruct the past in a benevolent
manner, such that highly committed individuals recall their emo-
tion, cognition, and behavior as more positive than they were in
fact. (Also, given that the time perspective for reporting commit-
ment was concurrent with that for reporting delayed reactions,
reports of immediate reactions may have been colored by malev-
olent reconstruction, with less committed individuals justifying
their weak commitment by describing betrayals more negatively.)
Therefore, in Study 3 we measured reactions to betrayal using
daily interaction records, testing the hypothesis that commitment is
positively associated with forgiveness of routine (as well as not so
routine) betrayals. In Study 3 participants described all betrayals
that occurred over a 2-week period, reporting on the degree to
which they felt committed to their partners and describing their
reactions to each incident. As in Study 2, in addition to assessing
behavioral tendencies, we also examined the mental events accom-
panying forgiveness (i.e., cognitive interpretations, emotional re-
actions). Assuming that the forgiveness process may unfold over
7
Participants described betrayal incidents that varied in severity, includ
-
ing some minor norm violations and some relatively more serious viola-
tions. Does commitment exert differential effects depending on the severity
of betrayal incidents? To address this question, we analyzed immediate and
delayed behavioral tendencies, including as independent variables time,
commitment level, severity of betrayal, and participant sex; scores for
commitment and severity were centered. The interaction of Commitment ⫻
Severity was nonsignificant, F(1, 144) ⫽ 0.60, ns, suggesting that the
association of commitment with interpersonal forgiveness did not differ
substantially as a function of severity of betrayal. (Indeed, parallel analyses
examining possible interactions with commitment for the remaining seven
potential confounds revealed no significant interactions. Thus, our findings
do not appear to be moderated by variations in commitment level.)
966
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
time, why should we anticipate that commitment will be associated
with in-the-moment forgiveness, as reported in interaction
records? In light of the assumption that the process by which
commitment inspires prorelationship motivation may become rel-
atively automatic and habitual—and in light of the fact that Study
2 reveals commitment effects for both immediate and delayed
reactions to betrayal—it seemed reasonable to anticipate that, on
average, we would observe positive commitment–forgiveness as-
sociations even within the context of specific interactions.
Given that Study 3 is nonexperimental, it is important to dem-
onstrate that any observed association of commitment with for-
giveness is not attributable to factors that may be confounded with
these variables. Therefore, we sought to demonstrate that commit-
ment accounts for unique variance in forgiveness beyond proper-
ties of betrayal incidents, features of relationships, and tendencies
toward socially desirable responding. In addition, in Study 3 we
obtained exploratory data to address the question, Why does com-
mitment promote forgiveness? Specifically, we administered a
new instrument including subscales to measure the three compo-
nents of commitment (intent to persist, long-term orientation,
psychological attachment) and pitted the three components against
one another to determine which aspect of commitment most reli-
ably predicts forgiveness.
Method
Participants. Participants were 78 undergraduates (20 men, 58
women) who volunteered to take part in partial fulfillment of the require-
ments for introductory psychology at the University of North Carolina at
Chapel Hill. As in Studies 1 and 2, sign-up sheets indicated that to take
part, participants must be involved in a dating relationship of at least 1
month in duration. Sign-up sheets also indicated that it was necessary that
participants interact with their partner almost every day, either on the
telephone or in person. The analyses reported below are based on data
from 64 participants (18 men, 46 women) who reported at least one
betrayal incident over the course of the 2-week study; 14 individuals (2
men, 12 women) reported no betrayal incidents. Participants had been
involved with their partner for 17.16 months on average. Most participants
described their relationship as a steady dating relationship (6% dating
casually, 3% dating regularly, 81% dating steadily, 9% engaged or mar-
ried), and most indicated that they dated their partner exclusively (97%
reported that neither partner dated others, 3% reported that both dated
others). About 44% indicated that they were involved in long-distance
relationships.
Procedure. Our materials and procedures were modeled after those
used in research with the Rochester Interaction Record (Reis & Wheeler,
1991). Each participant attended two research sessions—one at the start of
the 2-week study (Time 1), and a second at the end of the study (Time 2).
During Time 1 sessions, we explained that the study concerned negative
incidents in dating relationships. As in Study 3, we avoided the word
betrayal, defining betrayal incidents as follows:
Please use one record sheet to record each incident in which your
partner made you feel upset, angry, or hurt...nomatter how small or
big the incident is. . . For example: your partner may tell a friend
something that you believe should have remained private; your part-
ner may do something that is hurtful behind your back; your partner
may flirt with someone else at a party; or your partner may forget to
call you when your partner said he/she would.
We asked participants to record all such incidents, even if an incident was
quite brief, even if they felt fine by the end of the incident. If the same
incident came up later during the 2-week period they were to complete a
second record. We said we were interested in “incidents that involved
talking as well as incidents that didn’t necessarily involve talking (e.g.,
your partner behaves in a way that upsets you, you hear about something
your partner did).” We asked participants to complete records as soon as
possible following each incident and to turn in booklets every Monday,
Wednesday, and Friday (as a reminder, we telephoned participants Sunday,
Tuesday, and Thursday evenings).
These instructions were summarized on the cover sheet for each record
booklet. The remaining sheets were forms on which betrayal incidents were
recorded. For each incident, participants were asked to (a) record the date
and time at which the incident occurred, (b) record the date and time at
which the record sheet was completed, (c) record the duration of the
incident, (d) provide a description of the incident (i.e., “Describe the
incident in which your partner upset you [made you feel angry or hurt]”),
and (e) answer several questions about the incident, including the manner
in which they felt, thought, and behaved. We assessed positive behavioral
tendencies (three items; e.g., “My partner ought to be especially nice to me
to make up for what he/she did,” reverse scored; “I forgive my partner”),
positive cognitive interpretations (two items; e.g., “My partner upset me on
purpose,” reverse scored), and positive emotional reactions (two items;
e.g., “I feel very hurt by the way my partner behaved,” reverse scored; 1 ⫽
do not agree at all,7⫽ agree completely). Also, one item assessed
commitment level (i.e., “I feel committed to maintaining my relationship
with my partner”), one item assessed severity of incident (i.e., “When this
incident occurred, I thought it had the potential to seriously harm our
relationship”), and one item assessed strength of betrayal (i.e., “When this
incident occurred, I thought my partner had ‘broken the rules’ of our
relationship”).
At Time 1 research sessions, participants also completed several other
instruments. As in Studies 1 and 2, we measured self-deception and
impression management using the Balanced Inventory of Desirable Re-
sponding (Paulhus, 1984). We measured duration of relationship and
long-distance involvement as in Study 2. In addition, at Time 1 and Time 2
sessions, participants completed a new instrument designed to measure the
three components of commitment (Arriaga & Agnew, in press; five items
for each construct)—intent to persist (e.g., “I intend to stay in this rela-
tionship”), long-term orientation (e.g., “When I make plans about future
events in my life, I think about the impact of my decisions on our
relationship”), and psychological attachment (e.g., “In all honesty, my
family and friends are more important to me than this relationship,” reverse
scored). At the end of Time 2 sessions, participants were debriefed and
thanked for their assistance.
8
Reliability and validity of measures. Reliability analyses revealed ac-
ceptable coefficients for items designed to measure self-deception (
␣
⫽
.78) and impression management (
␣
⫽ .87), for Time 1 and Time 2 items
measuring intent to persist (
␣
s ⫽ .88 and .93, respectively), and for
interaction record items measuring behavioral tendencies (
␣
⫽ .74) and
emotional reactions (
␣
⫽ .76). One interaction record item measuring
cognitive interpretations exhibited weak associations with other positively
8
During Time 2 sessions, participants answered questions regarding the
validity of their data. Participants reported that they recorded 94% of the
betrayal incidents that occurred, completing interaction records an average
of 91 min following each incident; 92% indicated that the records accu-
rately reflected the incidents they experienced during the past 2 weeks, and
79% indicated that this period was typical of the sorts of incidents they
experienced with the partner. They reported that it was not particularly
difficult to maintain the interaction records (M ⫽ 2.42, SD ⫽ 1.43), they
did not talk to their partner about the records (M ⫽ 1.52, SD ⫽ 1.17),
participating in the study did not interfere with their everyday lives
(M ⫽ 2.08, SD ⫽ 1.38), and participating in the study did not affect the
activities in which they engaged (M ⫽ 1.44, SD ⫽ 1.11; for all items, 1 ⫽
do not agree at all,7⫽ agree completely).
967
COMMITMENT AND FORGIVENESS
toned items, so that item was dropped. One item measuring Time 1 and
Time 2 long-term orientation exhibited weak associations with the remain-
ing items, so that item was dropped (alphas for four-item measures ⫽ .75
and .86, respectively). Also, one item measuring Time 1 and Time 2
psychological attachment exhibited weak associations with the remaining
items, so that item was dropped (alphas for four-item measures ⫽ .85 and
.80). We developed a single measure of each variable—a count of extreme
scores for self-deception and impression management, and item averages
for other variables.
To explore the lay meaning of forgiveness, we examined the association
of the interaction record item “I forgive my partner” with measures of
behavior, cognition, and emotion (dropping “I forgive. . .” from the mea-
sure of behavior; alpha for two-item measure ⫽ .69). Given that the
multiple interaction records provided by a given participant are not inde-
pendent, we used hierarchical linear modeling to perform these analyses
(see below for additional information regarding this technique; cf. Bryk &
Raudenbush, 1992). The measure of stated forgiveness (“I forgive. . .”) was
positively associated with measures of behavior,

(87) ⫽ .53, p ⬍ .01,
cognition,

(87) ⫽ .26, p ⬍ .01, and emotion),

(87) ⫽ .44, p ⬍ .01. Thus,
and as expected, “I forgive my partner” tended to be somewhat more
closely aligned with behavioral tendencies than with cognition or emotion.
(In all remaining analyses, measures of behavior represent averages that
include scores for stated forgiveness.) We performed additional hierarchi-
cal analyses to examine the associations among behavioral tendencies,
cognitive interpretations, and emotional reactions. The three measures
were positively associated: average

⫽ .29;

s(87) ranged from .15 to .49,
all ps ⬍ .05.
Content of betrayal incidents. What sorts of betrayal incidents did
participants describe? Participants reported an average of 2.38 betrayal
incidents (range ⫽ 1 to 11), described incidents that had moderate potential
for harm (M ⫽ 2.51, SD ⫽ 1.70), and indicated that the incidents broke the
rules to a moderate degree (M ⫽ 3.23, SD ⫽ 1.92). We examined the
content of their descriptions and found that they generally paralleled the
Study 2 descriptions: (a) violations of monogamy norms (12%; e.g.,
disappeared into a bedroom for an hour with an ex-partner; said he or she
was bored with participant, then claimed it was a joke), (b) violations of
dependence norms (56%; e.g., accused participant of cheating; failed to
provide assistance when it was greatly needed), (c) violations of privacy
norms (2%; e.g., told a friend an important secret), and (d) violations of
decency and etiquette norms (22%; e.g., made cruel accusations; played a
trick on participant with a friend, then lied about it). An additional 7% of
the descriptions were unclassifiable (e.g., insufficient information to code).
Results and Discussion
Analysis strategy. As noted above, data from the multiple
interaction records provided by a given participant are not inde-
pendent. Therefore, we used hierarchical linear modeling to ana-
lyze our data (cf. Bryk & Raudenbush, 1992). Hierarchical linear
modeling simultaneously examines within-subject (lower level)
and between-subjects (upper level) variance, thereby modeling
each source of variation while accounting for statistical character-
istics of the other level. Our design includes two levels of variable,
in that data from one or more interaction records are nested within
participants. Variables that were measured in interaction records
are represented in our analyses as lower level variables (e.g.,
interaction record reports of commitment, behavior, cognition,
emotion). Variables that were measured during Time 1 or Time 2
research sessions are represented as upper level variables. Partic-
ipant was the upper level unit.
We initially performed key analyses allowing intercepts and
slopes to vary randomly across participants (for lower level slopes,
the analyses use data for participants with two or more lower level
observations). Tests examining the variance and covariance com-
ponents in these preliminary analyses revealed nonsignificant
across-participants differences in slopes (zs ranged from 0.18 to
1.14, all ns) and nonsignificant covariance between intercepts and
slopes (zs ranged from ⫺0.89 to 1.61, all ns). Therefore, we
allowed intercepts to vary randomly but represented slopes as
fixed effects (the analyses reported below revealed patterns of
significance vs. nonsignificance that were identical to those ob-
tained in preliminary analyses). Also, preliminary analyses re-
vealed no significant effects involving participant sex, so this
variable was dropped from the analyses.
Effects of commitment level. To examine the association of
commitment with interpersonal forgiveness and mental events, we
performed hierarchical analyses on each dependent variable. The
lower level criteria in these analyses were interaction-based reports
of behavior, cognition, and emotion; the lower level predictor was
interaction-based reports of commitment; participant was the up-
per level unit. The results of these analyses are summarized in
Table 3. Consistent with predictions, the analyses revealed signif-
icant associations of commitment with behavioral tendencies, cog-
nitive interpretations, and emotional reactions. Thus, in the course
of everyday interaction, during betrayal incidents in which indi-
viduals felt strongly committed to their relationships, they exhib-
ited more positive emotion, cognition, and behavior; during inci-
dents in which they felt less committed, they exhibited reduced
positivity and lesser tendencies toward interpersonal forgiveness.
Are associations with commitment mediated by mental events?
To determine whether the association of commitment with inter-
personal forgiveness was partially mediated by mental events, we
examined the role of cognitive interpretations and emotional reac-
tions in mediating the association of interaction-based reports of
commitment with behavioral tendencies (cf. Baron & Kenny,
1986). Consistent with the requirements for assessing mediation,
our interaction-based measures of cognition and emotion (a) were
significantly associated with behavioral tendencies and (b) were
significantly associated with interaction-based reports of commit-
ment (see earlier analyses).
To evaluate whether the association of commitment with behav-
ioral tendencies is attributable to cognitive interpretations, we
regressed behavioral tendencies simultaneously onto commitment
Table 3
Association of Commitment Level With Forgiveness of Partner
Betrayal: Study 3
Forgiveness component
Commitment main effect

F(1, 87)
Positive behavioral tendencies .48 20.59**
Positive cognitive interpretations .30 6.66**
Positive emotional reactions .29 5.17*
Note. Table values are from hierarchical linear modeling analyses includ-
ing commitment as a lower level predictor variable, with participant as the
upper level unit. Each row presents findings regarding the association of
commitment with a single criterion (e.g., with positive behavioral tenden-
cies). The

column presents standardized coefficients for associations
with commitment, and the F column presents F statistics for the commit-
ment effect.
* p ⬍ .05. ** p ⬍ .01.
968
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
and cognition. In this analysis, the association with cognition was
significant,

(86) ⫽ .17, p ⬍ .03, and the association with com-
mitment declined somewhat: excluding cognition,

(87) ⫽ .48,
p ⬍ .01; including cognition,

(86) ⫽ .30, p ⬍ .01. A test of the
significance of mediation revealed that cognition significantly (yet
partially) mediated the association with commitment (z ⫽ 2.06,
p ⬍ .04). Similarly, when we regressed behavioral tendencies onto
commitment and emotion, the association with emotion was sig-
nificant,

(86) ⫽ .51, p ⬍ .01, and the association with commit-
ment declined somewhat: including emotion,

(86) ⫽ .24, p ⬍ .01.
A test of the significance of mediation revealed significant (yet
partial) mediation by emotion (z ⫽ 3.97, p ⬍ .01). Thus, in Study 3
the association of commitment with interpersonal forgiveness was
significantly (yet partially) mediated not only by cognitive inter-
pretations (as in Study 2) but also by emotional reactions (unlike
Study 2).
9
Why does commitment promote forgiveness? What is it about
the commitment construct that accounts for the association of
commitment with interpersonal forgiveness? To address this ques-
tion, we examined the associations of Time 1 and 2 measures of
the three components of commitment—intent to persist, long-term
orientation, and psychological attachment—with interaction-based
reports of behavioral tendencies. Preliminary analyses performed
separately for Time 1 and Time 2 components revealed parallel
results, and Time 1 and Time 2 measures of each component were
positively correlated, rs(147) ranged from .81 to .83, all ps ⬍ .01.
Therefore, we performed key analyses using averaged measures of
Time 1 and Time 2 intent to persist, long-term orientation, and
psychological attachment.
First, we examined the simple association of each commitment
component with behavioral tendencies. We represented behavioral
tendencies as a lower level criterion and represented each com-
mitment component, in turn, as an upper level predictor. These
analyses revealed significant or marginal associations with behav-
ioral tendencies for all three components—intent to persist,

(88)
⫽ .48, p ⬍ .01, long-term orientation,

(88) ⫽ .24, p ⬍ .07, and
psychological attachment,

(88) ⫽ .46, p ⬍ .01. Next, we per-
formed hierarchical analyses in which we pitted all possible pairs
of commitment components against one another as predictors of
behavioral tendencies. Coefficients for intent to persist consis-
tently were significant,

s(87 or 88) ⫽ .76 and .60, both ps ⬍ .01,
whereas coefficients consistently were nonsignificant for long-
term orientation,

s(87 or 88) ⫽⫺.14 and .19, both ns, and
psychological attachment,

s(87 or 88) ⫽⫺.37 and .23, both ns.
Of course, these analyses should be interpreted as exploratory, in
that they rest on a relatively new instrument for assessing the three
components of commitment. However, these findings suggest that
the commitment–forgiveness association may rest more on simple
intent to persist than on components of commitment that capture
broadened temporal or interpersonal interests.
Ruling out alternative explanations. Is it possible that the
commitment–forgiveness association is attributable to confounds
such as strength of betrayal? To explore this possibility, we first
performed hierarchical analyses to examine the association of
interaction-based measures of commitment and behavioral tenden-
cies with each of several potential confounds; for duration of
relationship, we examined both simple scores and log-transformed
scores. Commitment was negatively associated with severity of
incident and strength of betrayal,

s(87) ⫽⫺.45 and ⫺.35, both
ps ⬍ .01, was positively associated with duration of relationship
(simple and transformed),

s(88) ⫽ .20 and .20, both ps ⬍ .04, and
impression management,

(88) ⫽ .19, p ⬍ .08, and was essentially
unrelated to long-distance involvement and self-deception,

s(87) ⫽ .12 and ⫺.05, both ns. The measure of behavioral
tendencies was negatively associated with severity of incident and
strength of betrayal,

s(87) ⫽⫺.32 and ⫺.46, both ps ⬍ .01, was
positively associated with impression management,

(88) ⫽ .37,
p ⬍ .01, and was essentially unrelated to duration of relationship
(simple and transformed), long-distance involvement, and self-
deception,

s(88) ranged from ⫺.03 to .18, all ns.
To ensure that the earlier reported association with commitment
was evident even when we controlled for each potential confound,
we performed auxiliary analyses, including interaction-based re-
ports of commitment along with each potential confound, in turn,
as predictors. The earlier reported analyses revealed a significant
association of commitment with behavioral tendencies; auxiliary
analyses revealed that when we controlled for each potential
confound, the association with commitment remained significant,
Fs(1, 86 to 87) ranged from 5.79 to 21.77, all ps ⬍ .02. Thus, the
commitment–forgiveness association is not entirely attributable to
any of several variables with which commitment and interpersonal
forgiveness are associated.
10
9
At Time 1 and Time 2 sessions, we also administered the commitment
instrument used in Studies 1 and 2 (
␣
s ⫽ .95 and .93, respectively). We
were mainly interested in associations with interaction-based reports of
commitment, in that reports from interaction records represent what we
assume to be the proximal cause of forgiveness. In addition assuming that
Time 1 and Time 2 instruments assess relatively stable levels of commit-
ment, we replicated key analyses to examine associations with these
measures. Preliminary analyses performed separately for Time 1 and
Time 2 commitment measures revealed parallel results, and Time 1 and 2
commitment were positively correlated, r(147) ⫽ .83, p ⬍ .01, so we
performed key analyses using an averaged measure of Time 1 and 2
commitment. A hierarchical analysis representing commitment as an upper
level variable revealed that commitment was positively associated with
behavioral tendencies,

(88) ⫽ .47, p ⬍ .01. In a mediation analysis
including cognition as a covariate, the association of behavioral tendencies
with cognition was significant,

(87) ⫽ .20, p ⬍ .01, and the association
with commitment declined somewhat,

(87) ⫽ .27, p ⬍ .01. A test of the
significance of mediation revealed significant (yet partial) mediation by
cognition (zs ⫽ 2.11, p ⬍ .03). Beyond the association with commitment,
emotional reactions, too, accounted for unique variance in behavioral
tendencies,

(87) ⫽ .54, p ⬍ .01. We could not test the significance of
mediation by emotion because our measure of emotional reactions was not
significantly associated with commitment,

(88) ⫽ .08, ns.
10
Participants described betrayal incidents that varied in severity of
incident (potential for serious harm) and strength of betrayal (extent to
which the incident broke the rules). Does commitment exert differential
effects depending on severity of incident or strength of betrayal? To
address this question, we performed a hierarchical analysis on interaction-
based reports of behavioral tendencies, including as lower level predictors
interaction-based reports of commitment, severity of incident, and the
interaction of these variables; scores for commitment and severity were
centered. The interaction of Commitment ⫻ Severity was nonsignificant,
F(1, 85) ⫽ 2.15, ns. We performed a parallel analysis substituting strength
of betrayal for severity of incident; the interaction of Commitment ⫻
Strength was nonsignificant, F(1, 85) ⫽ 0.08, ns. Thus, the commitment–
969
COMMITMENT AND FORGIVENESS
General Discussion
Why Do Victims Forgive?
The present work seeks to illuminate our understanding of why
victims forgive norm violations by examining the role of commit-
ment in motivating forgiveness. Three studies using diverse meth-
ods and diverse measurement techniques provide converging evi-
dence relevant to this hypothesis. All three studies reveal positive
associations of commitment with interpersonal forgiveness. The
results of Study 1 are particularly noteworthy, in that they provide
the first extant evidence that commitment level can be experimen-
tally primed, enabling us to examine the causal effects of commit-
ment. In all three studies, we took care to demonstrate that the
commitment–forgiveness association is not attributable to poten-
tial confounds such as the severity or recency of betrayal, proper-
ties of relationships such as duration or long-distance involvement,
or socially desirable response tendencies. Also, in Studies 2 and 3
we demonstrate that the strength of the commitment–forgiveness
association does not differ significantly as a function of severity of
incident or strength of betrayal (see Footnotes 7 and 10).
In Study 1 we used the exit, voice, loyalty, and neglect typology
to determine whether our commitment prime exerted differential
effects on constructive versus destructive and active versus passive
forms of forgiveness (Rusbult, 1993). The results reveal that strong
commitment reliably inhibited inclinations to react to betrayal with
destructive exit or neglect (or, alternatively, weak commitment
failed to yield such inhibition). However, the effect of commitment
was not significant for constructive voice and loyalty. These re-
sults are particularly interesting in light of prior evidence that in
ongoing relationships, the harmful effects of negative acts are
substantially greater than the beneficial effects of positive acts—
that is, it is less important that close partners enact constructive
behaviors than that they not enact destructive behaviors (e.g.,
Drigotas, Whitney, & Rusbult, 1995; Rusbult, Johnson, & Mor-
row, 1986). To the extent that these results are replicated in future
work, such findings suggest that commitment exerts its mo-
tivational effects precisely where such effects are most critical,
functioning as a deterrant to potentially devastating destructive
impulses.
Precisely why does commitment promote forgiveness? In
Study 3 we administered a new instrument to measure the three
components of commitment—intent to persist, long-term orienta-
tion, and psychological attachment. These data allow us to explore
what it is about the commitment construct that accounts for its
association with interpersonal forgiveness. Earlier, we suggested
that commitment-relevant interests may be relatively direct and
primitive (as embodied in simple intent to persist) or may be
relatively more extended, resting on broadened temporal interests
(as embodied in long-term orientation) or on broadened interper-
sonal interests (as embodied in psychological attachment). In
Study 3 we performed analyses to pit these components against
one another. The results of these analyses suggest that in under-
standing the commitment–forgiveness association, it is not so
much that commitment involves the broadened temporal interests
embodied in long-term orientation, yielding reciprocal cooperation
or long-term aggregation of beneficent acts. Nor is it that commit-
ment involves the broadened interpersonal interests embodied in
psychological attachment, yielding self–other merger or commu-
nal orientation. Instead, the commitment–forgiveness association
appears to rest on simple intent to persist. The more individuals
intend to persist and remain dependent on their relationship, the
more they are willing to forego vengeance to hold on to what they
have.
Granted, it could be argued that intent to persist may not be so
simple and may, in itself, imply temporally or interpersonally
extended concerns. However, if such concerns were central to why
commitment motivates forgiveness, then measures of long-term
orientation and psychological attachment should have captured
unique variance in forgiveness. Instead, we found that for individ-
uals to forgive their partner it is not necessary that they adopt the
long view from either a temporal or an interpersonal point of view;
it is only necessary that they intend to remain in their relationship.
Of course, it is possible that intent to persist dominated the
commitment–forgiveness association because this component
tapped temporally or interpersonally extended concerns that are
not tapped by the other two components of commitment (e.g.,
social obligations, moral commitment) or because this component
was more reliably measured (alphas were somewhat higher for this
component). Moreover, the instrument we used is relatively new,
and we were able to perform these analyses in only one of three
studies. Therefore, we should form tentative conclusions about the
centrality of intent to persist in explaining the commitment–
forgiveness association, pending replication in future research and
in work concerning maintenance acts other than forgiveness.
How Do Victims Forgive?
How does commitment promote forgiveness? The present work
examines the role of mental events in bringing about prorelation-
ship transformation of motivation. Interdependence theory sug-
gests that the transformation process sometimes is automatic and
habit driven and sometimes is mediated by internal events. There-
fore, we reasoned that mental events would partially mediate the
association of commitment with behavioral tendencies. We tested
this prediction in Studies 2 and 3.
Both studies revealed that the commitment–forgiveness associ-
ation was significantly mediated by positive cognitive interpreta-
tions—by the extent to which individuals discounted internal
causes, identifying extenuating circumstances and forming rela-
tively more positive, external explanations for partner betrayal
forgiveness association did not differ significantly as a function of either
severity of incident or strength of betrayal. (Indeed, parallel analyses
examining possible interactions with commitment for the remaining five
potential confounds revealed no significant interactions. Thus, our findings
do not appear to be moderated by variations in commitment level.)
Which factor plays a greater role in influencing interpersonal forgive-
ness—the degree to which an incident has the potential to harm a relation-
ship (severity of incident), or the degree to which an incident represents a
norm violation (strength of betrayal)? We regressed interaction-based
reports of behavioral tendencies simultaneously onto the severity and
strength variables, both of which were represented as lower level variables.
Behavioral tendencies were significantly influenced by strength of be-
trayal,

(85) ⫽⫺.33, p ⬍ .01, but not by severity of incident,

(85) ⫽
⫺.10, ns, suggesting that the key issue in understanding behavioral ten-
dencies is the degree to which a given betrayal incident breaks the rules,
not the simple potential for harm inherent in a given incident.
970
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
(i.e., chance or situational variables rather than disposition, effort,
or intent). Thus, commitment is associated with forgiveness in part
because committed individuals develop more benevolent (or less
malevolent) betrayal-relevant cognitions. For example, committed
individuals may engage in relatively more systematic processing—
they may carefully attend to the circumstances surrounding a
betrayal incident, they may be willing to give their partner the
benefit of the doubt, or they may be willing to accept some
personal responsibility for betrayal incidents. In turn, the relatively
benevolent understandings they develop help them find their way
to forgiveness.
We also examined possible mediation by emotional reac-
tions—by the absence of sadness, hurt, and anger. In Study 3,
positive emotional reactions partially mediated the association of
interaction-based reports of commitment with interpersonal for-
giveness. Unfortunately, measures of emotional reactions did not
meet the requirements for the assessment of mediation in Study 2
or in Study 3 analyses that examined commitment as measured at
Time 1 and 2 research sessions (see Footnote 9). Moreover, in all
relevant analyses, both commitment and emotion accounted for
unique variance in positive behavioral tendencies. The full com-
plement of evidence suggests that whether an individual achieves
reduced betrayal-relevant negative emotion may be irrelevant to an
understanding of how commitment promotes forgiveness, in that
committed individuals may experience a good deal of negative
emotion yet still find their way to forgiveness. Thus, to the extent
that the commitment–forgiveness association rests on internal
events, such events appear to be largely cognitive in character.
These findings are relevant to controversy in the forgiveness
literature surrounding the role of mental events in mediating in-
terpersonal forgiveness. As noted earlier, authors operating in the
philosophical and Christian theological traditions tend to hold as a
prototype saintly forgiveness, whereby victims do not cognitively
nullify betrayal and do not absolve the perpetrator of blame yet
find their way to forgiveness. In the present work, committed
individuals were relatively capable of forgiving their partner de-
spite persistent negative affect but were not so saintly that achiev-
ing some degree of benevolent cognition—or understanding—was
irrelevant to the forgiveness process. Of course, it could be argued
that by defining forgiveness in terms of positive behavioral ten-
dencies, we have failed to capture the true meaning of forgiveness.
For example, authors operating in alternative traditions might
define forgiveness in terms of changes in behavioral tendencies
that are not accompanied by parallel changes in cognitive inter-
pretations. At the same time, it is important to note that in the
present work, responses to “I forgive my partner” tended to be
closely aligned with behavioral tendencies; mediation analyses
performed on responses to this item revealed findings that paral-
leled those observed for our measure of behavioral tendencies.
Thus, it seems safe to conclude that when forgiveness is defined in
a manner that aligns with lay construals of this construct, the
commitment–forgiveness association is significantly mediated by
changes in cognitive interpretations.
In Study 2, we pursued a subsidiary, process-relevant goal,
examining temporal properties of forgiveness. Given that betrayal
incidents create an interpersonal debt, we speculated that the
process by which individuals forgive may not be immediate but
instead may unfold over the course of extended interaction.
Study 2 reveals that, in comparison with individuals’ later reac-
tions, immediately following betrayal they exhibit more negative
emotion, cognition, and behavior. Also, the Time ⫻ Commitment
interaction was significant for emotional reactions, such that com-
mitment was marginally negatively associated with immediate
emotional reactions but was marginally positively associated with
later emotional reactions. In addition, the association of commit-
ment with emotion and behavior was marginally stronger to the
extent that more time had elapsed since a given act of betrayal (see
Footnote 6). Thus, the immediate response to norm violations does
appear to be relatively negative. Moreover, the role of commitment
in yielding prorelationship internal events, motivation, and behav-
ior appears to increase over the course of postbetrayal interaction.
Our understanding of interpersonal forgiveness would benefit
from further examination of the temporal properties of this phe-
nomenon. Such an analysis would extend our knowledge of the
precise character of the transformation process—the motivational
shift that is argued to underlie forgiveness. Kelley (1984) proposed
that for some sorts of interdependence situation, it is suitable to
represent interaction using a temporally extended transition list.
The transition list representation is predicated on the assumption
that, in addition to selecting specific behaviors, consciously or
unconsciously, interacting individuals also select future interac-
tions, creating some sorts of future interaction opportunities and
eliminating others.
The transition list representation highlights the role of both
partners’ actions in promoting forgiveness, including the actions of
victim and perpetrator. For example, in The War of the Roses,
when Oliver destroys the fish entre´e that Barbara has prepared for
her friends, he does more than simply betray her—his act leads the
two down a fork in the road on which forgiveness by Barbara is not
an available response. Had Oliver exhibited remorse and offered
amends, he would have made available a domain of interactions in
which forgiveness continued to be viable. An act of atonement
would have provided Barbara with a more constructive (and po-
tentially forgiving) set of behavioral options, thereby not only (a)
enhancing the positivity of the immediate interaction but also (b)
enhancing the positivity of the future interactions available to the
pair. In ongoing work (Hannon, Finkel, Rusbult, & Kumashiro,
2001), we are examining the effects of perpetrator behavior on the
forgiveness process.
Sex Differences in Reactions to Betrayal
Studies 1 and 2 reveal evidence of sex differences in inclinations
toward forgiveness. In Study 1, men were less forgiving than
women, as evidenced in their lesser inclination to react to betrayal
with constructive voice; this difference is not attributable to so-
cially desirable responding. In Study 2, men exhibited more for-
giving feelings, thoughts, and behavioral tendencies than did wom-
en; this difference is not attributable to time since betrayal,
properties of relationships, or socially desirable responding, but
there was inconclusive evidence that severity of betrayal might be
an issue in understanding this effect. We believe it is somewhat
imprudent to speculate on the meaning of these effects, given that
sex differences were observed in only two of three studies and
given that the direction of such effects differed. We leave it to
future work to determine whether there are reliable differences
between women and men in inclinations toward forgiveness and to
971
COMMITMENT AND FORGIVENESS
determine whether men may appear more forgiving because they
tend to suffer less severe betrayals than do women.
Broader Implications
What does an interdependence-based model buy us that other
orientations may not? First, our interdependence theoretic analysis
highlights the importance of norms in governing interaction in
ongoing relationships. Earlier, we suggested that although norms
initially may emerge as a matter of convenience, over time such
rules acquire the properties of moral obligation. Accordingly,
individuals feel betrayed when their partner violates relationship-
relevant norms. Betrayals involve norm violations that harm the
victim, so such incidents produce an interpersonal debt, yielding
characteristic profiles of negative emotion, cognition, and behav-
ior. Indeed, findings from Study 3 suggest that the harmful con-
sequences of a given incident result more from the fact that the
incident broke the rules than from the negativity of the act per se
(see Footnote 10). Thus, norm violations appear to be relatively
distinct and powerful threats to couple well-being, differing in
important ways from other sorts of noncorrespondent interactions,
such as accommodative dilemmas.
Second, whereas many social scientists might feel inclined to
explain forgiveness largely by reference to properties of individ-
uals (e.g., dispositions such as vengefulness or a forgiving person-
ality), interdependence theory suggests that behavior frequently is
shaped by properties of relationships (cf. Kirby, Rusbult, & Kil-
patrick, 2001). (It sometimes seems that the field of psychology is
subject to the fundamental attribution error that governs lay cog-
nition—the ready inclination to identify internal, individual-based
causes and to downplay situational causes, e.g., properties of
relationships.) Of course, interdependence theory does not focus
exclusively on properties of relationships as the cause of interper-
sonal behavior. This theory merely argues that prosocial motiva-
tion may be embodied not only in properties of individuals but also
in properties of dyads and groups.
Third, many social scientists might feel inclined to explain
forgiveness largely by reference to the process by which forgive-
ness comes about—in terms of the cognitive or affective concom-
itants of forgiveness or in terms of critical events en route to
forgiveness. An interdependence analysis also highlights the mo-
tivational basis of forgiveness: In addition to explaining how
Oliver comes to forgive Barbara, it may be equally important to
ask why he forgives her. Typically, answers to “how?” comple-
ment answers to “why?” Such complementarity becomes evident
when one considers the therapeutic implications of answers to
“how?” and “why?”: An emphasis on how informs interventions
oriented toward skills training relevant to achieving forgiveness
(e.g., perspective taking). An emphasis on why informs comple-
mentary interventions centering on forgiveness-relevant motiva-
tion. For example, our Study 1 priming results suggest that in-
creasing the salience of commitment-relevant thoughts may
enhance motivation to forgive (e.g., on awaking in the morning,
one may ask, “Why am I committed to thee? Let me count the
ways”).
Fourth, because an interdependence-based analysis emphasizes
the causes and consequences of interaction, this orientation helps
us identify important impediments to reconciliation following be-
trayal. In previous work, we have demonstrated that commitment-
inspired interaction behaviors such as accommodation and will-
ingness to sacrifice strengthen partner trust (Wieselquist et al.,
1999). Specifically, people come to trust their partner as a conse-
quence of observing the partner behave well in situations involving
noncorrespondent preferences or as a consequence of observing
the partner depart from his or her direct self-interest to benefit the
relationship (cf. Holmes & Rempel, 1989). Given that commitment
reliably motivates such prosocial acts, trust can be construed as an
implicit gauge of the strength of a partner’s commitment. Thus,
acts of betrayal may be problematic in part because such acts
imply a lack of commitment, undermine trust, and upset the
balance of power. If this is so, then it becomes even more evident
why betrayal incidents are a seriously destabilizing force. In on-
going work, we are examining the role of trust in the forgiveness
process, both as cause (e.g., commitment and trust may interact in
motivating forgiveness) and as effect (e.g., restoring trust may be
a key issue in understanding reconciliation following betrayal).
Limitations and Directions for Future Research
Before closing, we comment on several limitations of this work.
One limitation centers on our participant populations. We ex-
amined the forgiveness process among young adults who were
largely North American, most of whom were involved in relatively
short-term relationships and few of whom were married to their
partner. It is possible that in other populations, the commitment–
forgiveness association does not rest so thoroughly on the emer-
gence of benevolent betrayal-relevant cognition. For example, for
married partners, among deeply religious individuals, or in cul-
tures that regard marriage as relatively inviolable, forgiveness may
be so thoroughly obligatory that forgiveness following betrayal is
a relatively habitual phenomenon (e.g., partners may rather auto-
matically and effortlessly turn the other cheek). On the other hand,
we suspect that the general principles outlined herein—for exam-
ple, the principle that forgiveness is more probable given strong
interdependence—may be valid across specific populations, situ-
ations, and times. In future work, it will be important to examine
forgiveness in relationships of varying type and duration as well as
among individuals from varying cultures. In particular, it will be
important to examine forgiveness among marital partners, for
whom there may be strong norms regarding betrayal and forgive-
ness and for whom variables other than (or in addition to) com-
mitment may be relevant to understanding the motivation to
forgive.
A second limitation centers on the fact that our studies rely
entirely on self-report measures. Thus, our work is subject to
common critiques of self-report measurement—our findings might
be colored by socially desirable responding, acquiescence, retro-
spective reconstruction, and the like. On the other hand, across the
three studies we obtained not only retrospective but also concur-
rent reports, we assessed both internal events and behavioral
tendencies, and we ensured that our findings were not attributable
to self-deception or impression management. Moreover, in
Study 1, commitment was an experimentally manipulated variable
rather than a self-report measure (i.e., we did not simply examine
correlations among multiple self-report variables), in Studies 2
and 3 our findings were replicated in analyses controlling for other
self-report measures that arguably share related response biases,
and in Study 3 our findings were replicated in analyses using a
972
FINKEL, RUSBULT, KUMASHIRO, AND HANNON
within-subject analysis strategy (i.e., a strategy that controls for
individual differences in response bias). In future work, research-
ers might explore the forgiveness process through the use of
physiological indices, unobtrusive measures, or alternative behav-
ioral indices—for example, by coding videotapes of betrayal-
relevant conversations and by examining unambiguous indices of
commitment such as persistence in a relationship. (However, for
inherently internal experiences e.g., commitment, it could be ar-
gued that self-report indices are preferable to behavioral indices.)
Conclusions
We advance a conceptual analysis of forgiveness using the
principles of interdependence theory. On the basis of existing
work, it seems clear that betrayal incidents are problematic, yield-
ing a signature constellation of negative affect, cognition, and
behavior. One obstacle to proceeding on a positive path in the
aftermath of betrayal is that the victim’s immediate inclinations
generally are antithetical to forgiveness, favoring self-oriented
impulses such as grudge, vengeance, and the expectation of atone-
ment. We suggest that interpersonal forgiveness rests on prorela-
tionship transformation of motivation, one cause of which is strong
commitment. Three studies reveal convergent evidence in support
of the claim that commitment promotes prorelationship mental
events, prorelationship motives, and interpersonal forgiveness. The
studies also shed light on the transformation process by which
commitment promotes forgiveness, demonstrating the role of cog-
nitive interpretations in mediating the commitment–forgiveness
association. The present work thus complements existing research
regarding how people come to forgive their partner by highlighting
the role of commitment in explaining why people forgive their
partner.
References
Agnew, C. R., Van Lange, P. A. M., Rusbult, C. E., & Langston, C. A.
(1998). Cognitive interdependence: Commitment and the mental repre-
sentation of close relationships. Journal of Personality and Social Psy-
chology, 74, 939–954.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Thousand Oaks, CA: Sage.
Aron, A., & Aron, E. N. (1997). Self-expansion motivation and including
other in the self. In S. Duck (Ed.), Handbook of personal relationships:
Theory, research, and interventions (2nd ed., pp. 251–270). Chichester,
England: Wiley.
Arriaga, X. B., & Agnew, C. R. (in press). Being committed: Affective,
cognitive, and conative components of relationship commitment. Per-
sonality and Social Psychology Bulletin.
Axelrod, R. (1984). The evolution of cooperation. New York: Basic Books.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychol-
ogy, 51, 1173–1182.
Baumeister, R. F., Exline, J. J., & Sommer, K. L. (1998). The victim role,
grudge theory, and two dimensions of forgiveness. In E. L. Worthington,
Jr. (Ed.), Dimensions of forgiveness (pp. 79–104). Philadelphia:
Templeton Foundation Press.
Baumeister, R. F., Stillwell, A. M., & Heatherton, T. F. (1995). Personal
narratives about guilt: Role in action control and interpersonal relation-
ships. Basic and Applied Social Psychology, 17, 173–198.
Baumeister, R. F., Stillwell, A., & Wotman, S. R. (1990). Victim and
perpetrator accounts of interpersonal conflict: Autobiographical narra-
tives about anger. Journal of Personality and Social Psychology, 59,
994–1005.
Boon, S. D., & Sulsky, L. M. (1997). Attributions of blame and forgiveness
in romantic relationships: A policy-capturing study. Journal of Social
Behavior and Personality, 12, 19–44.
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models:
Applications and data analysis methods. Newbury Park, CA: Sage.
Clark, M. S., & Mills, J. (1979). Interpersonal attraction in exchange and
communal relationships. Journal of Personality and Social Psychol-
ogy, 37, 12–24.
Dorff, E. N. (1992). Individual and communal forgiveness. In D. H. Frank
(Ed.), Autonomy and Judaism (pp. 193–218). Albany: State University
of New York Press.
Drigotas, S. M., Whitney, G. A., & Rusbult, C. E. (1995). On the pecu-
liarities of loyalty: A diary study of responses to dissatisfaction in
everyday life. Personality and Social Psychology Bulletin, 21, 596–609.
Enright, R. D., & Human Development Study Group. (1996). Counseling
within the forgiveness triad: On forgiving, receiving forgiveness, and
self-forgiveness. Counseling and Values, 40, 107–122.
Fagenson, E. A., & Cooper, J. (1987). When push comes to power: A test
of power restoration theory’s explanation for aggressive conflict esca-
lation. Basic and Applied Social Psychology, 8, 273–293.
Freedman, S. R., & Enright, R. D. (1996). Forgiveness as an intervention
goal with incest survivors. Journal of Consulting and Clinical Psychol-
ogy, 64, 983–992.
Gonzales, M. H., Haugen, J. A., & Manning, D. J. (1994). Victims as
“narrative critics”: Factors influencing rejoinders and evaluative re-
sponses to offenders’ accounts. Personality and Social Psychology Bul-
letin, 20, 691–704.
Gonzales, M. H., Manning, D. J., & Haugen, J. A. (1992). Explaining our
sins: Factors influencing offender accounts and anticipated victim re-
sponses. Journal of Personality and Social Psychology, 62, 958–971.
Gordon, K. C., & Baucom, D. H. (1998). Understanding betrayals in
marriage: A synthesized model of forgiveness. Family Process, 37,
425–449.
Hannon, P. A., Finkel, E. J., Rusbult, C. E., Kumashiro, M. A. (2001).
Perpetrator behavior and forgiveness. Unpublished manuscript, Univer-
sity of North Carolina at Chapel Hill.
Hodgins, H. S., Liebeskind, E., & Schwartz, W. (1996). Getting out of hot
water: Facework in social predicaments. Journal of Personality and
Social Psychology, 71, 300–314.
Holmes, J. G., & Rempel, J. K. (1989). Trust in close relationships. In C.
Hendrick (Ed.), Review of personality and social psychology (Vol. 10,
pp. 187–220). London: Sage.
Johnson, D. J., & Rusbult, C. E. (1989). Resisting temptation: Devaluation
of alternative partners as a means of maintaining commitment in close
relationships. Journal of Personality and Social Psychology, 57, 967–
980.
Kelley, H. H. (1983). Love and commitment. In H. H. Kelley, E. Ber-
scheid, A. Christensen, J. H. Harvey, T. L. Huston, G. Levinger, et al.
(Eds.), Close relationships (pp. 265–314). New York: Freeman.
Kelley, H. H. (1984). The theoretical description of interdependence by
means of transition lists. Journal of Personality and Social Psychol-
ogy, 47, 956–982.
Kelley, H. H., & Thibaut, J. W. (1978). Interpersonal relations: A theory
of interdependence. New York: Wiley.
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social
psychology. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Hand-
book of social psychology (4th ed., Vol. 1, pp. 233–265). Boston:
McGraw-Hill.
Kirby, J. S., Rusbult, C. E., & Kilpatrick, S. D. (2001). Pro-relationship
motives and behavior across multiple relationships: Is “behaving well”
973
COMMITMENT AND FORGIVENESS
an actor-specific disposition or a relationship-specific phenomenon?
Unpublished manuscript, University of North Carolina at Chapel Hill.
Kremer, J. F., & Stephens, L. (1983). Attributions and arousal as mediators
of mitigation’s effect on retaliation. Journal of Personality and Social
Psychology, 45, 335–343.
Marty, M. E. (1998). The ethos of Christian forgiveness. In E. L. Worth-
ington, Jr. (Ed.), Dimensions of forgiveness (pp. 9–28). Philadelphia:
Templeton Foundation Press.
McCullough, M. E., Rachal, K. C., Sandage, S. J., Worthington, E. L. Jr.,
Brown, S. W., & Hight, T. L. (1998). Interpersonal forgiving in close
relationships: II. Theoretical elaboration and measurement. Journal of
Personality and Social Psychology, 75, 1586–1603.
McCullough, M. E., Sandage, S. J., & Worthington, E. L. Jr. (1997). To
forgive is human. Downers Grove, IL: Inter Varsity.
McCullough, M. E., & Worthington, E. L. Jr. (1995). Promoting forgive-
ness: A comparison of two brief psychoeducational group interventions
with a waiting-list control. Counseling and Values, 40, 55–68.
McCullough, M. E., Worthington, E. L. Jr., & Rachal, K. C. (1997).
Interpersonal forgiving in close relationships. Journal of Personality and
Social Psychology, 73, 321–336.
Nietzsche, F. W. (1887). The genealogy of morals. London: SPCK.
North, J. (1987). Wrongdoing and forgiveness. Philosophy, 62, 499–508.
Ohbuchi, K., Kameda, M., & Agarie, N. (1989). Apology as aggression
control: Its role in mediating appraisal of and response to harm. Journal
of Personality and Social Psychology, 56, 219–227.
Paulhus, D. L. (1984). Two-component models of socially desirable re-
sponding. Journal of Personality and Social Psychology, 46, 598–609.
Reis, H. T., & Wheeler, L. (1991). Studying social interaction with the
Rochester Interaction Record. In M. P. Zanna (Ed.), Advances in exper-
imental social psychology (Vol. 24, pp. 269–318). New York: Academic
Press.
Ridley, M. (1998). The origins of virtue: Human instincts and the evolution
of cooperation. New York: Penguin Books.
Rosenzweig-Smith, J. (1988). Factors associated with successful reunions
of adult adoptees and biological parents. Child Welfare, 67, 411–422.
Rusbult, C. E. (1983). A longitudinal test of the investment model: The
development (and deterioration) of satisfaction and commitment in
heterosexual involvements. Journal of Personality and Social Psychol-
ogy, 45, 101–117.
Rusbult, C. E. (1993). Understanding responses to dissatisfaction in close
relationships: The exit, voice, loyalty, and neglect model. In S. Worchel
& J. A. Simpson (Eds.), Conflict between people and groups: Causes,
processes, and resolutions (pp. 30–59). Chicago: Nelson-Hall.
Rusbult, C. E., Davis, J. L., Finkel, E. J., Hannon, P. A., & Olsen, N.
(2001). Forgiveness of betrayal in close relationships: A dual-process
model of the transformation from self-interested impulses to
relationship-oriented actions. Unpublished manuscript, University of
North Carolina at Chapel Hill.
Rusbult, C. E., Johnson, D. J., & Morrow, G. D. (1986). Impact of couple
patterns of problem solving on distress and nondistress in dating rela-
tionships. Journal of Personality and Social Psychology, 50, 744–753.
Rusbult, C. E., Martz, J. M., & Agnew, C. R. (1998). The investment
model scale: Measuring commitment level, satisfaction level, quality of
alternatives, and investment size. Personal Relationships, 5, 357–391.
Rusbult, C. E., & Van Lange, P. A. M. (1996). Interdependence processes.
In E. T. Higgins & A. Kruglanski (Eds.), Social psychology: Handbook
of basic principles (pp. 564–596). New York: Guilford Press.
Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F., & Lipkus, I.
(1991). Accommodation processes in close relationships: Theory and
preliminary empirical evidence. Journal of Personality and Social Psy-
chology, 60, 53–78.
Rusbult, C. E., & Zembrodt, I. M. (1983). Responses to dissatisfaction in
romantic involvements: A multidimensional scaling analysis. Journal of
Experimental Social Psychology, 19, 274–293.
Spanier, G. B. (1976). Measuring dyadic adjustment: New scales for
assessing the quality of marriage and similar dyads. Journal of Marriage
and the Family, 38, 15–28.
Stanley, S. M., & Markman, H. J. (1992). Assessing commitment in
personal relationships. Journal of Marriage and the Family, 54, 595–
608.
Steinmetz, S. (Ed.). (1993). The Random House Webster’s dictionary. New
York: Ballantine Books.
Stillwell, A. M., & Baumeister, R. F. (1997). The construction of victim
and perpetrator memories: Accuracy and distortion in role-based ac-
counts. Personality and Social Psychology Bulletin, 23, 1157–1172.
Tangney, J. P., Wagner