A dyadic assessment of forgiveness in intimate relationships

17
A dyadic assessment of forgiveness in intimate relationships MYRON D. FRIESEN, GARTH J. O. FLETCHER, AND NICKOLA C. OVERALL University of Canterbury, New Zealand Abstract This study investigated forgiveness by examining couplesÕ recollections and perceptions of specific incidents of transgressions in their relationships. The results replicated previous research but also produced some novel findings. Results showed that more positive attributions and relationship quality independently predicted higher internal forgiveness, whereas expressed forgiveness was related only to relationship quality. Overall, the sample was negatively biased in their perceptions of their partner’s forgiveness, but those participants who tended toward a positive bias were happier with their relationships as were their partners. Results are discussed in terms of prior research and theories of forgiveness and related social judgments in intimate relationships. To err is human; to forgive, divine. —Alexander Pope (1711). An essay on Criticism: part 2 Until recently, the clergy and theologians were the primary authorities on forgiveness. Indeed, as recently as 1987, forgiveness and its related synonyms were not even indexed in Psychological Abstracts (Fincham, 2000). However, significant advances have been forged in the scientific study of forgiveness over the last decade or so (see Enright, 2001; Fincham; McCullough & Witvliet, 2002). As often noted, forgiveness is an inherently interpersonal process (Fincham; North, 1987, 1998). Yet, in spite of its interpersonal nature, researchers have primarily investi- gated forgiveness from an individual perspec- tive, largely ignoring dyadic influences and frequently employing hypothetical transgres- sions rather than actual incidents of offense. In this study, we investigated forgiveness in a sample of intimate couples (dating, cohabitating, and married) who reported on specific occasions of offense or transgression in their relationships. Few studies have assessed forgiveness with both members of the relational dyad, and no published work (to our knowledge) has assessed forgiveness with the dyad referencing the same relational real-life transgressions. Using this design, we were able to investigate both intrapsychic associations and partner effects. For example, if Michael wholeheartedly forgives his part- ner, does this positively influence the relation- ship satisfaction and that of both Michael and his partner? Moreover, by requiring both partners to judge their own level of for- giveness (self-reported forgiveness) as well as their perceptions of their partner’s forgiveness (perceived-partner forgiveness) concerning the same transgressions, we were also able to investigate the nature and role of bias in for- giveness judgments. The present study was carried out as part of the Master’s research of the first author and supervised by Garth Fletcher. The authors would like to thank the relationship research group in the Psychology Department at the University of Canterbury for their assistance in recruiting and running participants, and their comments and feedback. Correspondence should be addressed to Myron D. Friesen, University of Canterbury, Psychology Depart- ment, Christchurch, New Zealand, e-mail: mdf26@student. canterbury.ac.nz. Personal Relationships , 12 (2005), 61–77. Printed in the United States of America. Copyright Ó 2005 IARR. 1350-4126=02 61

Transcript of A dyadic assessment of forgiveness in intimate relationships

A dyadic assessment of forgiveness in

intimate relationships

MYRON D. FRIESEN, GARTH J. O. FLETCHER, AND

NICKOLA C. OVERALL

University of Canterbury, New Zealand

AbstractThis study investigated forgiveness by examining couples� recollections and perceptions of specific incidents of

transgressions in their relationships. The results replicated previous research but also produced some novel findings.

Results showed that more positive attributions and relationship quality independently predicted higher internal

forgiveness, whereas expressed forgiveness was related only to relationship quality. Overall, the sample was

negatively biased in their perceptions of their partner’s forgiveness, but those participants who tended toward a

positive bias were happier with their relationships as were their partners. Results are discussed in terms of prior

research and theories of forgiveness and related social judgments in intimate relationships.

To err is human; to forgive, divine.

—Alexander Pope (1711).

An essay on Criticism: part 2

Until recently, the clergy and theologians

were the primary authorities on forgiveness.

Indeed, as recently as 1987, forgiveness and

its related synonyms were not even indexed

in Psychological Abstracts (Fincham, 2000).

However, significant advances have been

forged in the scientific study of forgiveness

over the last decade or so (see Enright, 2001;

Fincham; McCullough & Witvliet, 2002). As

often noted, forgiveness is an inherently

interpersonal process (Fincham; North, 1987,

1998). Yet, in spite of its interpersonal

nature, researchers have primarily investi-

gated forgiveness from an individual perspec-

tive, largely ignoring dyadic influences and

frequently employing hypothetical transgres-

sions rather than actual incidents of offense.

In this study, we investigated forgiveness

in a sample of intimate couples (dating,

cohabitating, and married) who reported on

specific occasions of offense or transgression

in their relationships. Few studies have

assessed forgiveness with both members of

the relational dyad, and no published work

(to our knowledge) has assessed forgiveness

with the dyad referencing the same relational

real-life transgressions. Using this design, we

were able to investigate both intrapsychic

associations and partner effects. For example,

if Michael wholeheartedly forgives his part-

ner, does this positively influence the relation-

ship satisfaction and that of both Michael

and his partner? Moreover, by requiring both

partners to judge their own level of for-

giveness (self-reported forgiveness) as well as

their perceptions of their partner’s forgiveness

(perceived-partner forgiveness) concerning

the same transgressions, we were also able to

investigate the nature and role of bias in for-

giveness judgments.

The present study was carried out as part of the Master’sresearch of the first author and supervised by GarthFletcher. The authors would like to thank the relationshipresearch group in the Psychology Department at theUniversity of Canterbury for their assistance in recruitingand running participants, and their comments andfeedback.

Correspondence should be addressed to Myron D.Friesen, University of Canterbury, Psychology Depart-ment, Christchurch, New Zealand, e-mail: [email protected].

Personal Relationships, 12 (2005), 61–77. Printed in the United States of America.Copyright � 2005 IARR. 1350-4126=02

61

We addressed three broad issues, each rep-

resenting an extension of previous research

or a new line of inquiry: (a) the roles of both

internal and expressed forgiveness, (b) the

links among relationship evaluations, blame

attributions, and forgiveness, and (c) the

nature and correlates of bias in forgiveness

judgments. We next briefly discuss each of

these issues and formulate our predictions.

Internal and expressed forgiveness

Baumeister, Exline, and Sommer (1998) have

noted that forgiveness can be a ‘‘process that

goes on entirely inside the mind of the vic-

tim, or . a transaction that occurs between

two people, even without much in the way of

inner processing’’ (p. 86). Thus, what Bau-

meister et al. (1998) term silent forgiveness

is characterized by the presence of internal

forgiveness but without any behavioral expres-

sion. Exline and Baumeister (2000) present

several ‘‘barriers’’ that could inhibit a victim’s

expression of forgiveness even if the internal

process has previously occurred. For example,

a victim may choose not to express their for-

giveness in order to keep the perpetrator mor-

ally indebted and/or to maintain certain

benefits or rights that are acquired with victim

status. The opposite of silent forgiveness is

termed hollow forgiveness (Baumeister et al.)

and is characterized by the interpersonal ex-

pression of forgiveness without any internal

transformation. For example, a victim could

express forgiveness because it is the ‘‘right’’

thing to do or to maintain the public appear-

ance of a harmonious relationship.

While silent and hollow forgiveness may

be used in some interpersonal contexts, it is

more likely that most forgiving experiences

are not so extreme, and an individual’s inter-

nal transformation and interpersonal expres-

sion of forgiveness would fall somewhere in

between these two states. It is plausible that

the internal and expressed dimensions of for-

giveness are correlated and governed by the

same forces, but there is also the possibility

that separate factors uniquely influence each

dimension and independently determine the

degree that forgiveness is inwardly experi-

enced and portrayed.

Importantly, differences in how dimen-

sions of intrapsychic and transactional for-

giveness are conceptualized account for

much of the variation among proposed for-

giveness models (Enright, 2001; McCullough

et al., 1998; Worthington, 1998). The present

study involves a preliminary attempt to assess

how both the internal and the interpersonal

aspects of forgiveness are related and associ-

ated with other important variables (e.g.,

attributions, relationship quality, and incident

severity).

To appraise the internal motivational

aspects of forgiveness, we employed a modi-

fied version of McCullough’s Transgression-

Related Interpersonal Motivations (TRIM)

Inventory (McCullough et al., 1998), and for

expressed (communicated) forgiveness, we

used an additional single-item assessment.

These two measures of forgiveness consti-

tuted the main dependent variables in this

research. A great deal of prior empirical work

has conceptualized and assessed forgiveness

as an intrapsychic variable. In the absence of

prior empirical work to guide us, we pro-

posed two straightforward predictions. First,

reports of greater internal forgiveness should

be correlated with reports of more expressed

forgiveness. Second, we expected that the

pattern of correlates with internal forgiveness

would be similar to those obtained with ex-

pressed forgiveness.

The role of attributions and

relationship satisfaction

Relational factors, such as higher commit-

ment, closeness, and satisfaction, have re-

peatedly been found to be associated with

greater forgiveness (e.g., Finkel, Rusbult,

Kumashiro, & Hannon, 2002; Karremans,

Van Lange, Ouwerkerk, & Kluwer, 2003;

McCullough et al., 1998). Recent research

has found that in committed relationships,

a lack of forgiveness can lead to increased

psychological tension and decreased psycho-

logical well-being (Karremans et al., 2003).

Similarly, increased blame and responsibil-

ity attributions also predict lower levels of

forgiveness (Boon & Sulsky, 1997; Bradfield

& Aquino, 1999; Fincham, 2000; Weiner,

62 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

Graham, Peter, & Zmuidinas, 1991). How-

ever, the present study, following the lead of

Fincham and colleagues (Fincham; Fincham,

Paleari, & Regalia, 2002) and Finkel et al.

(2002), explored the possibility that the asso-

ciations among relationship satisfaction,

blame attributions, and forgiveness might

involve a mediational relationship.

In his study with British couples, Fincham

(2000) demonstrated that forgiveness medi-

ated the link between partner-blame and

reported behavior. In further work, Fincham

et al. (2002) found that responsibility attribu-

tions mediated the link between marital qual-

ity and forgiveness. However, these studies

either used global measures of relationship

quality and forgiveness or measured forgive-

ness related to fictional accounts of transgres-

sion. When broad measures of relationship

quality and forgiveness tendencies are

employed, it is possible that the results are

influenced by participants holding general

positive (or negative) attitudes toward the

relationship rather than on the basis of for-

giveness processes per se. Additionally, when

forgiveness is assessed based on reading fic-

tional transgression accounts, it is possible

that participants might assign greater weight

to cues about the offense (McCullough &

Hoyt, 2002) and fail to consider the various

relationship factors (i.e., level of trust, com-

mitment, investment) that would influence

forgiveness in real-life transgressions. The

only research thus far that has used real-life

transgressions (Finkel et al., 2002) found that

both cognitive interpretations and emotional

reactions partially mediated the link between

relationship quality and forgiveness.

Using the model in Figure 1 as a guide for

our predictions,1 we hypothesized that the

results would replicate prior findings and

attributions of blame would mediate (at least

partly) the link between relationship quality

and forgiveness. In other words, better rela-

tionship quality and decreased blame would

be associated with increased forgiveness.

However, when blame is entered into the

relationship quality and forgiveness equation,

the direct path from relationship quality to

forgiveness will significantly decrease (par-

tial mediation) or become nonsignificant (full

mediation). In addition, we measured and

controlled for the severity of the incident

(a potential artifact) when analyzing the links

between satisfaction, blame, and forgiveness.

Bias in forgiveness perception

The assessment of bias was made possible in

this study by measuring forgiveness from the

perspectives of both partners. Each partici-

pant first reported how much he or she for-

gave his or her partner for the offense in

question and then reported the extent to

which he or she perceived his or her partner’s

forgiveness for the same incident. Thus,

Mary’s bias can be assessed by comparing

her judgment of the extent to which Michael

forgave her with Michael’s report of how

much he actually forgave her.

The relevant literature and theory suggests

two different hypotheses concerning bias and

its links with perceived relationship quality.

On the one hand, it has been well documented

that people are generally positively biased in

their perceptions of their intimate partners and

that this bias is strongly associated with more

positive relationship satisfaction (for reviews

see Fletcher, 2002; Murray, 2001). Murray

and her colleagues argue (with considerable

evidential support) that those in more success-

ful and happier relationships tend to adopt

Figure 1. Proposed mediational model link-

ing forgiveness with relationship quality.

Double-headed arrows represent correlations.

1. While Figure 1 illustrates the paths tested by the pres-ent research, it should be noted that this figure doesnot portray a complete representation of all the pathsassociated between relationship quality and forgive-ness. Most notably absent is a feedback loop from for-giveness to relationship quality.

Dyadic forgiveness 63

positive illusions and see their partners and

relationships through rose-colored glasses.

However, according to error management

theory (Haselton & Buss, 2000), when the

options of a positive versus a negative bias

are considered, a negative bias can sometimes

constitute the safest and most often adopted

alternative, depending on the associated costs

and benefits. For example, Haselton and Buss

demonstrated that women (but not men) un-

derperceived the commitment levels of poten-

tial mates. This finding was predicted on the

basis that women would suffer greater costs

than men if they were to err on the over-

optimistic side.

Error management theory appears to apply

nicely to the case of perceiving forgiveness

(for both sexes). If Michael optimistically

assumes that Mary has forgiven him, when in

fact she has not, such behavior might serve to

reinforce Mary’s revenge and avoidance moti-

vations, and strengthen her perception of

Michael as insensitive, uncaring, and self-

absorbed. On the other hand, a negative for-

giveness bias (the tendency to perceive the

partner as less forgiving than is the case)

might prompt Michael to stress his remorse

and contrition, and motivate conciliatory be-

haviors. Thus, for most people, a slight neg-

ative bias might be more constructive for their

relationships as compared with a positive

bias. We suspected therefore that overall we

would find evidence of a negative forgiveness

bias (controlling for the severity of the

transgression).

Error management theory, however, does

not necessarily contradict the work of Murray

and others. It may well be wise for people to

adopt a default negative forgiveness bias, but

exaggerating this tendency for specific trans-

gressions hardly seems likely to encourage

(or to be associated with) higher levels of

relationship satisfaction for self or partner. If

Michael refuses to believe that Mary has for-

given him, in the face of clear evidence to

the contrary, this might serve to infuriate

Mary and convince her that Michael really

does not love her. Thus, we expected to find

that, at the individual-difference level, more

positive relationship satisfaction would gen-

erally be associated with less negative for-

giveness bias, consistent with well-replicated

findings in the relationship domain (see

Murray, 2001).

Overview of the present study

The present research recruited couples in inti-

mate relationships to participate in a question-

naire-based study on the processes couples

use to cope with negative events in their rela-

tionships. Both partners responded to paper

and pencil questionnaires that assessed their

current thoughts and feelings about the same

two past specific incidents of offense or trans-

gression in their relationship. The path dia-

gram in Figure 1 illustrates the hypothesized

links between relationship satisfaction, attri-

butions, and forgiveness and the between-

partner paths that this research investigated. It

is important to note that Figure 1 displays

a model that assumes a causal progression

measured from left to right. Although the data

were cross-sectional, this is a reasonable

assumption, given that relationship quality is

a more global variable, in comparison to the

levels of attribution and forgiveness that con-

cerned specific behavioral events.

Structural equation modeling (SEM) was

employed to deal with the problem of nonin-

dependence across partner data and to carry

out path analyses. This type of analysis pos-

sesses several critical advantages, which were

all applicable in this research. First, SEM

deals with the problem of shared variance

between partners. Second, it allows analysis

of both within-partner processes and across-

partner processes. Third, SEM can test for

significant gender differences across paths

and allows paths to be pooled across gender

if it proves appropriate.

In summary, this study assesses forgive-

ness with both members of relational dyads

referencing the same real-life relational trans-

gressions. Our predictions for this study can

be readily summarized. First, we predicted

that less negative blame attributions and more

positive perceptions of relationship quality

would be related to more forgiveness. How-

ever, we expected that attributions would

mediate the links between relationship evalu-

ations and forgiveness. Second, we expected

64 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

that participants would produce an overall

negative forgiveness bias (controlling for the

severity of the transgression). Third, we pre-

dicted that relationship satisfaction would be

positively related to greater positive forgive-

ness bias, consistent with well-replicated find-

ings in the relationship domain.

Method

Sample

A total of 39 couples (N ¼ 78) participated

in this research, recruited from students at the

University of Canterbury, New Zealand.

Eighteen couples were in dating non-cohabit-

ing relationships, 14 couples were living

together, and 7 couples were married. The

mean length of relationships was 37.8 months

(SD ¼ 62.3 months), and the mean age was

23.7 years (SD ¼ 7.45 years). No other

demographic information was requested.

Procedure

Each couple met with the primary researcher

in a social psychology laboratory. Partici-

pants were thoroughly briefed on the proce-

dure, assured that no deception was in any

way employed in the course of the study, and

reminded of the confidentiality and anonym-

ity of their responses prior to signing a con-

sent form.

Participants were first separated from their

partner and seated at tables in different loca-

tions in the laboratory. Their initial task was

to recall up to four incidents of transgression

or offense in the relationship, occurring

within the last year. Participants were asked

to list two incidents in which the attitudes,

actions, or words of their partner had been

offensive to them and two in which they

believed their attitudes, actions, or words had

offended their partner. Participants were then

asked to write a one-sentence description of

these incidents, so that their partner could

recognize which incident was being recalled,

but at the same time avoided disclosing any

substantive details of the transgression. Part-

ners also rated how distressing each incident

was on a 7-point scale (end points ¼ not at

all to extremely). The descriptions from the

men were then displayed to the women and

vice versa, by the experimenter. After re-

viewing the incident descriptions with each

partner, the researcher selected two incidents

that were recognized and recalled by both

partners, attempting to counterbalance roles

of victim and perpetrator across gender

between the two incidents when this was pos-

sible and favoring those incidents that were

more recent and more serious. This procedure

was completed successfully by all couples.

There were a wide variety of offensive

incidents provided; these included inconsid-

erate and unsupportive behavior (e.g., failing

to contact, ignoring, choosing family over

partner, preoccupation with work or study,

etc.), verbal insults (e.g., name calling, disre-

spect, gossip, slandering friends or family

members, personal criticisms), issues regard-

ing alcohol and drug use, infidelity or the

threat of infidelity (e.g., flirtatious behavior,

contact with ex-partner, etc.), issues regard-

ing relational commitment and personal inde-

pendence, disagreements and arguments,

broken promises, financial indiscretions, re-

fusals to communicate (e.g., ‘‘I clam-up and

refuse to talk about an issue’’), sexual issues

(e.g., ‘‘he/she pressures me to have sex’’, ‘‘he

evaluated our date on the basis of getting

physical’’), and accidents or mistakes in judg-

ment (e.g., ‘‘I played a joke on her that

backfired’’).

After the two incidents had been selected,

the researcher wrote down a short title for each

episode at the top of the next set of scales, and

participants were instructed to answer all the

questions (described later) for each section

specifically regarding the identified incidents.

After the incident-related scales, participants

completed a measure assessing global rela-

tionship quality and the demographic queries.

To enhance participants� perceptions of ano-

nymity, participants sealed all of their infor-

mation in a large envelope and placed it in a

locked mailbox. Finally, the couple was brought

back together, debriefed, and then paid ($5.00

NZ for each person). The tasks took 30 to

45 min to complete.

One reason for attempting to counterbal-

ance the roles of victim and perpetrator was

Dyadic forgiveness 65

to avoid gender being confounded with vic-

tim versus perpetrator status. The self- and

partner-blame attribution means were very

similar across men and women and across

incidents (see Table 1), and the results of two

2 (men vs. women) � 2 (Incident A vs. Inci-

dent B) repeated measures analyses of vari-

ance (ANOVAs) revealed no significant main

effects or interactions. Thus, our aim here

was achieved.

A second reason for attempting to counter-

balance was the possibility of including

perpetrator versus victim as a variable in sub-

sequent analyses. However, the evidence

indicated that it was not possible to make

clean and consistent distinctions between vic-

tims and perpetrators for specific episodes.

First, some couples could only identify inci-

dents in which one person seemed to be the

main perpetrator. Second, many incidents

were described in dyadic terms (e.g., argu-

ment after dinner last Friday) or revealed

disagreement between the partners about per-

petrator roles. Third, correlations across sex

for the blame attributions within incidents

revealed a weak pattern of negative correla-

tions (with only one being significant) (see

Table 1). This pattern suggests that attri-

butions for the same situations by couples

were often independent, rather than exhibit-

ing agreement concerning perpetrator versus

victim status.

Finally, we addressed this issue further by

analyzing a series of multiple regressions with

forgiveness entered as the dependent variable

and with two main effects of partner-blame

and self-blame (e.g., female forgiveness re-

gressed on male partner-blame and female

self-blame) and the interaction term entered

as independent variables. Significant results

Table 1. Means, standard deviations, and correlations across partners for the major

variables

Men (n ¼ 39) Women (n ¼ 39)

Correlations

across partners

Major variables Incident A Incident B Incident A Incident B Incident A Incident B

Negativity/severity 4.9 (1.6) 4.9 (1.2) 5.0 (1.4) 4.8 (1.2) .70** .38*

Partner-blame 3.1 (1.5) 2.9 (1.6) 3.5 (1.5) 3.2 (1.6) �.18 �.23

Self-blame 3.4 (1.3) 3.6 (1.4) 3.3 (1.5) 3.3 (1.5) �.34* .01

Self-report

forgiveness

(TRIM)

6.6 (0.49) 6.6 (0.55) 6.6 (0.81) 6.6 (0.65) .34* �.01

Self-report

expressed

forgiveness

5.2 (1.9) 5.3 (1.9) 5.3 (1.6) 4.8 (2.2) .04 .14

Partner-perceived

forgiveness

(TRIM)

6.5 (0.67) 6.4 (0.69) 6.5 (1.1) 6.6 (0.75) .00 .32

Partner-perceived

expressed

forgiveness

4.8 (1.9) 4.2 (2.0) 4.9 (1.9) 5.1 (2.0) .01 .06

Relationship

quality

6.1 (0.54) 6.0 (0.65) .45**

Note. All means and standard deviations are based on a 7-point scale. The forgiveness measures were reverse coded so

that higher scores represent higher levels of forgiveness. Standard deviations are in parentheses. TRIM ¼ Transgression-

Related Interpersonal Motivations.

*p , .05. **p , .01.

66 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

for the partner versus self-blame interaction

term would indicate that forgiveness varies as

a function of the relative amounts of blame

attributed to the same events by the members

of the couple. However, all eight analyses

revealed nonsignificant results for the interac-

tion terms.

In light of these findings, the two selected

offending situations for each couple were ran-

domly assigned as Incident A or Incident B.

Measures

Participants initially answered three items re-

garding the time frame (answered in months

and weeks), negativity, and seriousness of the

incident on 7-point scales. The items were

‘‘Approximately how long ago did this inci-

dent take place?’’, ‘‘How negative was this

incident?’’, and ‘‘How serious was this inci-

dent?’’ (end points were not at all and ex-

tremely). Incident A occurred on average 3.7

months prior to assessment (SD ¼ 5.5), and

Incident B occurred on average 2.6 months

prior to assessment (SD ¼ 2.7). The negativ-

ity and seriousness items for each incident

were positively correlated for both men (Inci-

dent A, r ¼ .62; Incident B, r ¼ .45, p ,

.01) and women (Incident A, r ¼ .72; Inci-

dent B, r ¼ .66, p , .01). Therefore, these

scores were summed to produce an overall

level of incident severity for each event.

Attributions of blame. Participants re-

sponded to six items assessing their judg-

ments of blame for each of the incidents;

three items assessed the participant’s percep-

tions of their partner’s level of responsibility,

and three items assessed the participant’s

perceptions of their own responsibility on

7-point scales. The partner-blame items were

‘‘My partner is to blame for this incident.’’,

‘‘My partner’s behavior was motivated by

selfish concerns.’’, ‘‘My partner’s negative be-

havior was intentional and planned.’’ (end

points were disagree completely and agree

completely). For the self-blame items, the

words ‘‘my partner’’ were deleted and the item

was rewritten in the first person (e.g., ‘‘I am to

blame for this incident.’’). Similar items have

been widely used in attribution research, and

these were adapted from work by Bradbury

and Fincham (1992) and Fletcher and Fitness

(1990). All internal reliabilities were accept-

able with slightly higher alpha levels in gen-

eral for partner attributions (male Incident A,

a ¼ .74, Incident B, a ¼ .71; female Incident

A, a ¼ .67, Incident B, a ¼ .76) as compared

to self-attributions (male Incident A, a ¼ .52,

Incident B, a ¼ .66; female Incident A, a ¼.70, Incident B, a ¼ .64).2 Thus, scores on

these items were summed, producing overall

scores for blame attributions toward self and

partner for each incident.

Forgiveness. A modified version of the

TRIM Inventory (McCullough et al., 1998)

was employed to measure internal forgive-

ness, whereas a single-item question measured

expressed forgiveness (‘‘I have communi-

cated to my partner that I have forgiven him/

her.’’ End points were not at all and very

much). The TRIM is a 12-item measure of

forgiveness, assessing motivations of avoid-

ance (seven items) and revenge (five items).

Because of item-overlap problems with the

relationship quality measure and the need to

apply the questions to specific episodes of

offense, the TRIM was modified to include

only seven items; three items assessed

revenge and four items assessed avoidance.

The items were ‘‘I want to make my partner

pay.’’, ‘‘I want my partner to get what they

deserve.’’, ‘‘I am going to get even.’’, ‘‘I keep

as much distance between us as possible.’’,

‘‘I find it difficult to act warmly toward him/

her.’’, ‘‘I avoid him/her.’’ and ‘‘I withdraw

from him/her.’’ These seven items (and the

expressed forgiveness item) were also re-

worded to assess the participant’s perceptions

of his or her partner’s forgiveness (i.e., all

personal pronouns at the beginning of a state-

ment were changed to ‘‘My partner,’’ and

pronouns referring to the partner at the end

of the statement were changed to ‘‘me’’).

Instructions to participants read, ‘‘For the

questions in this section, please indicate your

2. Note that Cronbach internal reliability (alpha) levelsare substantially reduced with only three itemsbecause of a correction factor in the equation.

Dyadic forgiveness 67

current thoughts and feelings about your part-

ner specifically regarding the offense in ques-

tion. That is, these questions pertain to how

you think and feel about this incident now.

Please circle the appropriate number on the

scale which indicates how much you agree or

disagree with the following statements.’’ All

the forgiveness measures used 7-point Likert

scales.

For all of the TRIM scales (self-report and

perceived-partner), corrected item total corre-

lations ranged from .25 to .95, while alpha

reliability levels ranged from .69 to .93. Thus,

the items on each TRIM scale were summed

to produce four independent TRIM scores:

self and partner scores for Incident A and

Incident B.

Relationship quality. To assess overall

relationship quality, the Perceived Relation-

ship Quality Components Inventory was uti-

lized (Fletcher, Simpson, & Thomas, 2000b).

This 18-item scale measures six components

on 7-point scales: satisfaction, commitment,

intimacy, trust, passion, and love. Previous

research has found that this scale provides

a comprehensive and global assessment of

relationship quality with good internal reli-

ability, convergent validity, and predictive

validity (Fletcher, Simpson, & Thomas,

2000a; Fletcher et al., 2000b). Internal reli-

abilities for the scales in this research

were good; .85 for the males and .88 for the

females. Scores were summed to produce

one rating for perceived relationship quality.

Results

Descriptive analyses and

zero-order correlations

Means, standard deviations, and correlations

across partners for each scale are presented in

Table 1. Both men and women perceived the

incidents to be relatively serious, with associ-

ated moderate levels of blame. The correla-

tions across partners, and within each

incident, revealed that incident severity and

relationship quality were moderately posi-

tively correlated across partners. Interest-

ingly, expressed forgiveness was consistently

rated as being lower than reported levels of

internal forgiveness (ts ranged from 4.7 to

7.7, all p , .001), although this result should

be treated with caution given the nature of the

scales used.

The correlations among the major varia-

bles are presented in Table 2 for the men and

Table 2. Zero-order correlations among variables for men within each incident

Study variables 1 2 3 4 5 6 7 8

Negativity/severity — .14 .21 �.13 .23 .03 �.07 .04

Partner-blame .32a — �.29a �.51** �.31a �.14 �.20 �.24

Self-blame .09 �.28 — .16 .27 �.09 .00 .11

Self-report forgiveness

(TRIM)

�.27 �.29 .07 — .45** .41** .11 .40*

Self-report expressed

forgiveness

.17 �.22 �.01 .36* — .35* .24 .42**

Perceived-partner

forgiveness (TRIM)

�.21 .06 �.39* .41** �.04 — .45** .52**

Perceived-partner

expressed forgiveness

.14 �.26 �.09 �.02 .38* .13 — .25

Relationship quality �.02 �.23 .07 .50** .42** .37* .24 —

Note. Incident A is presented to the left of the diagonal, and Incident B is presented to the right. Figures in bold face-

type represent significant or marginally significant correlations. n ¼ 39. TRIM ¼ Transgression-Related Interpersonal

Motivations.ap , .10. *p , .05. **p , .01.

68 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

in Table 3 for the women. The results for the

men and women were similar. Across both

incidents, more positive relationship quality

was significantly associated with greater inter-

nal forgiveness, higher levels of expressed

forgiveness, and higher perceptions of the

partner’s internal forgiveness. These positive

associations between the forgiveness varia-

bles and relationship quality replicate prior

research (Fincham, 2000; McCullough, 2000)

and support our hypotheses. Additionally for

both incidents, greater internal forgiveness

was associated with higher expressed for-

giveness and more positive perceptions of

the partner’s internal forgiveness. Replicating

Fincham, for both incidents, the more individ-

uals blamed their partner, the less likely they

were to forgive him or her. Also consistent

with prior research (e.g., Fletcher & Thomas,

2000), partner-blame predicted other varia-

bles more substantially than self-blame. Thus,

partner-blame, and not self-blame, was used

in all subsequent analyses.

There were also a few differences in the

correlations according to sex. However, as

previously noted, the pattern of similarities

and differences according to sex are difficult

to interpret because of shared variance across

the partners. Thus, we move to the SEM anal-

yses, testing the path analysis model shown in

Figure 1.

SEM analyses with internal forgiveness

as the dependent variable

Design strategy and path analysis. EQS

for Windows version 5.7b (Bentler & Wu,

1995) was used to analyze these results. Sep-

arate SEM analyses were conducted for each

incident as the results varied somewhat

across incidents. In order to test for gender

differences in the paths, the relationship qual-

ity / attribution and attribution / forgive-

ness paths were constrained as equal across

partners. The Lagrange multiplier (LM) test

was then used to determine if significant

amounts of variance were explained if the

constraints were released. For example (see

Figure 2), the path from male relationship

quality to male partner-blame was initially

set as equal to the path from female relation-

ship quality to female partner-blame. The

LM test then revealed that releasing the con-

straint would not add significant explained

variance to the model; that is, the two paths

are not significantly different.

The results revealed that only the

between-partner paths (diagonal paths) from

Table 3. Zero-order correlations among variables for women within each incident

Study variables 1 2 3 4 5 6 7 8

Negativity/severity — .43** �.20 �.02 �.07 .21 �.09 .18

Partner-blame .37* — �.35* �.29a .05 .00 .08 �.15

Self-blame .23 �.33* — .00 �.31a �.24 �.29a �.02

Self-report forgiveness

(TRIM)

�.42** �.25 �.11 — .36* .78** .33* .55**

Self-report expressed

forgiveness

�.07 �.09 �.04 .54** — .33* .71** .34*

Perceived-partner

forgiveness (TRIM)

�.38* �.30a �.11 .85** .51** — .44** .58**

Perceived-partner

expressed forgiveness

.04 �.15 .03 .35* .73** .49** — .39*

Relationship quality �.12 �.04 �.07 .44** .45** .35* .46** —

Note. Incident A is presented to the left of the diagonal, and Incident B is presented to the right. Figures in bold face-

type represent significant or marginally significant correlations. n ¼ 39. TRIM ¼ Transgression-Related Interpersonal

Motivations.ap , .10. *p , .05. **p , .01.

Dyadic forgiveness 69

partner-blame to internal forgiveness were

significantly different across sex. Thus, the

constraints were released for this pair of paths,

and the remaining paths were left as pooled

across partners for both incidents. Note that

the relevant regression coefficients were set as

equal in their unstandardized form, which

means that the standardized path coefficients

(as shown in Figure 2) may vary somewhat

across gender. The perceived negativity rat-

ings of the incident were included mainly to

control for this variable as a possible artifact.

For example, more blame might be related to

less forgiveness because more blame is asso-

ciated with more negative incidents, with the

negativity of the incidents accounting for

the association. However, we have included

the findings for this variable given that the

detailed results were of some interest.

The results for Incident A and Incident B

are displayed in Figure 2, with Incident A

paths presented above the Incident B paths.

Fit indices indicated a better fit for Incident

B than for Incident A (Incident A, v2 (13, 39)

¼ 21.78, p ¼ .06, RMSEA ¼ .14, CFI ¼0.89; Incident B, v2 (13, 39) ¼ 4.20, p ¼ .99,

RMSEA ¼ 0.00, CFI ¼ 1.00).3

Within-individual paths. First, as pre-

dicted, after controlling for incident negativ-

ity, more positive perceptions of relationship

quality were related to less blame (although

this pattern was significant only for Incident

B). Second, more blame was significantly

related to lower levels of forgiveness. Third,

transgressions judged as more serious and

negative resulted in significantly increased

partner-blame (after controlling for relation-

ship quality).

Figure 2. Standardized path coefficients derived from a structural equation modeling test of

the overall model from Incident A (upper figures) and Incident B (lower figures) data. *coeffi-

cients were significant at the p , .05 two-tailed level.

3. Overall, these standard measures of fit are acceptable;however, they need to be treated with caution because(a) there are relatively few paths of theoretical interestset to zero and (b) we included variables (such as inci-dent severity) primarily to control for artifacts, andsome of the related paths set to zero are of little theo-retical interest. Diagnosing model fit is not the primarypurpose of SEM in these analyses, but rather, wewanted to analyze the predicted paths in the model.

70 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

Across-partner paths. Previous research

has shown that partner-blame can have sig-

nificant effects on the behavior of the blamed

partner (Fletcher & Thomas, 2000). As can

be seen, the only significant across-partner

path showed that (in Incident A) men who

blamed their partners less had partners who

forgave them more for the same incident.

The across-partner (diagonal) paths from

relationship quality / partner-blame were

also tested, but all were close to zero and

were nonsignificant (and are thus not dis-

played in Figure 2).

Mediational analyses. As can be seen in

Figure 2, both men and women who reported

more positive relationship quality experi-

enced significantly more internal forgiveness.

However, the paths from relationship quality

to forgiveness remained positive and signifi-

cant when male blame and female blame

were controlled. Moreover, the decrease from

the direct to the indirect paths was small and

nonsignificant. Thus, our hypothesis that

blame attributions would play a mediating role

was clearly not supported. Instead, partner-

blame and relationship quality were indepen-

dently associated with internal forgiveness

(controlling for incident negativity).

SEM analyses with expressed

forgiveness as the dependent variable

The SEM analyses were repeated with ex-

pressed forgiveness entered as the dependent

variable. The findings revealed a weaker pat-

tern of results than with internal forgiveness.

The only significant paths (both within indi-

viduals and across partners) revealed that

both men and women, across both incidents,

who perceived more positive relationship

quality expressed higher levels of forgiveness

(paths ranged from .30 to .41, p , .05).

Bias in forgiveness

To analyze bias we initially conducted a 2

(female partner attribution vs. male partner

attribution) � 2 (partner attributed forgive-

ness vs. self-reported forgiveness) � 2 (inter-

nal forgiveness vs. expressed forgiveness) �

2 (Incident A vs. Incident B) ANOVA (note

all variables are within-subject variables).

The second factor is the critical element as it

represents bias. That is, it compares Partner

A’s rating of how much Partner B forgave

him or her with how much Partner B actually

reported forgiving Partner A.4

The means and standard deviations of the

variables in this analysis are listed in Table 1.

This revealed a significant main effect for

bias, F(1, 38) ¼ 13.51, p , .01. As expected,

the associated means revealed that both the

men and the women held a negative forgive-

ness bias, undercutting their partner’s forgive-

ness (perceived-partner internal and expressed

forgiveness, M ¼ 5.62; self-reported internal

and expressed forgiveness, M ¼ 5.89). There

was also a significant main effect for forgive-

ness type, F(1, 38) ¼ 98.54, p , .001, and

the associated means revealed that internal

forgiveness (M ¼ 6.56) was reported and per-

ceived more than expressed forgiveness (M ¼4.96). However, this (unpredicted and inciden-

tal) particular finding should be treated with

caution and may not be interpretable given the

different nature of the two measures. Finally,

these two main effects were qualified by a

bias � forgiveness type interaction, F(1, 38) ¼5.31, p , .05 (shown in Figure 3).

The significant interaction demonstrates

that the negative forgiveness bias was more

marked for expressed forgiveness than for

internal forgiveness. Planned comparisons

revealed that the difference between partner-

perceived (M ¼ 5.62) and self-reported (M ¼

4. This ANOVA does not test whether bias might bea function of the sex of the attributor or the target per-son. The first-level variable (female partner attribu-tion vs. male partner attribution) is not interpretableas a main effect. The reason is that each level of thisvariable mixes partner and self-judgments from differ-ent sexes; that is, the mean of one level equals thesum of the male partner attributions of forgivenessand the self-forgiveness attributions of their femalepartners (and vice versa for the other level). Interac-tions with this variable are also not meaningful, withthe critical exception of those involving bias (partnerattributed forgiveness vs. self-reported forgiveness);for example, it is possible that women produce morebias than men. However, none of these particularinteractions were significant (Fs , 1.0). These resultsshow that bias is not a function of the sex of theattributor or the target person.

Dyadic forgiveness 71

5.89) internal forgiveness was not significant

(F ¼ 1.09) but that perceivers rated their

partners as expressing their forgiveness sig-

nificantly less (partner-perceived, M ¼ 4.74)

than the levels of forgiveness their partners

claimed (self-reported, M ¼ 5.17, F(1, 78) ¼4.07, p , .05). No other main effects or in-

teractions were significant.

To assess individual differences in for-

giveness bias, we calculated residual varia-

bles from regression equations in which the

participant’s perceptions of his or her part-

ner’s forgiveness was entered as the depen-

dant variable and the partner’s self-reported

forgiveness for the same incident was entered

as the independent variable. A positive resid-

ual for a participant means that he or she per-

ceived more forgiveness than was reported

by the partner (relative to the sample),

whereas a negative residual is equivalent to

a negative bias (relative to the sample). This

exercise was completed for both sexes sepa-

rately for Incidents A and B and separately

for internal and expressed forgiveness. These

residuals were positively correlated across

incidents both for men (internal forgiveness,

r ¼ .32; expressed forgiveness, r ¼ .42) and

women (internal forgiveness, r ¼ .68; ex-

pressed forgiveness, r ¼ .59). Thus, the

scores were summed across incidents.

Next we used an SEM approach to model

the links between bias and relationship qual-

ity. As can be seen in Figures 4 and 5, this

analysis controls for the shared variance

across partners in levels of bias and calculates

both within-individual and across-partner

effects. In these analyses we also controlled

for perceptions of incident negativity (a

potential artifact), but the inclusion of these

variables did not alter the four paths dis-

played, thus these variables were left out of

Figures 4 and 5. As previously, we also tested

for significant differences across gender.

There were none, so the paths shown were

pooled across gender. The fit indices were

good (internal forgiveness bias, v2 (7, 39) ¼7.20, p ¼ .41, RMSEA ¼ .03, CFI ¼ 1.00;

expressed forgiveness bias, v2 (7, 39) ¼ 9.08,

p ¼ .25, RMSEA ¼ .09, CFI ¼ .93). This

analysis revealed strong positive associations

between bias and relationship quality for men

and women, both within and across partners

Figure 4. Standardized path coefficients

derived from a structural equation modeling

test of the links between internal forgiveness

bias and relationship quality for Incident A

and Incident B data combined. *coefficients

were significant at the p , .05 two-tailed

level.

Figure 3. General forgiveness bias (partner-

perceived vs. self-reports) for internal and ex-

pressed forgiveness. Higher numbers represent

greater forgiveness or perceived forgiveness.

Figure 5. Standardized path coefficients

derived from a structural equation modeling

test of the links between expressed forgive-

ness bias and relationship quality for Incident

A and Incident B data combined. *coeffi-

cients were significant at the p , .05 two-

tailed level.

72 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

and for both internal and expressed forgive-

ness. Thus, even though the sample held an

overall negative bias, men and women who

held a less negative or a more positive for-

giveness bias (relative to the sample) judged

their relationship as more satisfying and had

partners who also perceived their relationship

more positively.5

An alternative explanation. Returning to

Tables 2 and 3, there is a pattern of positive

and significant zero-order correlations be-

tween self-perceptions of forgiveness and

partner-perceived forgiveness. These results

suggest an alternative explanation to the find-

ings just described. Perhaps people who

exhibit more forgiveness bias tend to be hap-

pier because they project more and assume

their partners are similar to self. To test this

hypothesis, we recalculated the same SEM

analyses, but with the addition of two new

variables: male and female self-reported for-

giveness (summed across incidents). These

variables were added to the model shown in

Figures 4 and 5 as new independent variables

(connected via double-headed arrows to the

bias residuals and to relationship quality with

direct paths). The results were essentially

unchanged, with all four of the original paths

from bias to relationship quality maintaining

their positive and significant associations

(paths ranged from .34 to .43, p , .05).

Discussion

This research used a relatively novel design

in which internal and expressed forgiveness

was assessed with both members of the rela-

tional dyad and with reference to the same

specific relational transgressions. However,

the results of this study generally replicated

prior attribution and forgiveness research that

has typically used individuals and hypotheti-

cal scenarios. Men and women in happier

relationships blamed their partner less and for-

gave them more (although the association

between relationship quality and blame was

significant only for Incident B after control-

ling for incident severity). We also found that

more negative and serious transgressions

resulted in more partner-blame and less for-

giveness. Against predictions, however, we

found no evidence that blame attributions

mediated the link between relationship evalua-

tions and forgiveness. More positive perceived

relationship quality and less partner-blame

were independently related to elevated inter-

nal forgiveness, whereas higher levels of ex-

pressed forgiveness were only associated with

more satisfying relationships. Both men and

women tended to perceive their partner as less

forgiving than they actually were (an overall

negative forgiveness bias), although this find-

ing needs to be treated with caution, given

that this effect was only significant for ex-

pressed forgiveness. However, participants in

happier relationships were markedly less prone

to negative bias. Finally, these results held up

when several artifacts were controlled for,

including seriousness of the incident and

assumed similarity.

Theoretical implications

Internal and expressed forgiveness. Ex-

pressed forgiveness in our research was

measured with a single-item question, as

compared to the modified TRIM Inventory

that contained seven items. Thus, it is possi-

ble that the lower reliability of a single-item

might explain the relatively low correlations

between expressed forgiveness and attribu-

tions. However, the same single-item ex-

pressed forgiveness measure produced quite

strong and significant correlations with rela-

tionship satisfaction and with several other

variables, which counts against this methodo-

logical explanation. Forgiveness can be ex-

pressed in a variety of ways, including verbal

and nonverbal behavior, presumably either

consciously or unconsciously. Because of the

way in which the expressed forgiveness item

was worded, it is likely that the primary focus

of recollection in this study was conscious

5. In order to control for possible overlapping variancebetween the internal and expressed forgiveness biasvariables, we performed the SEM analyses again withthe relevant variables added into the equation. Wheninternal forgiveness bias was the predictor, we con-trolled for expressed forgiveness bias and vice versa.All paths in both models exhibited slight increases intheir effect sizes when run in this manner.

Dyadic forgiveness 73

verbal behavior (e.g., saying, ‘‘I forgive

you.’’ or ‘‘Don’t worry about it.’’) rather than

implicit verbal or nonverbal behavior (e.g.,

normal everyday communication with a pleas-

ant tone of voice or allowing interpersonal

interaction to return to normal).

McCullough and colleagues� conceptuali-zation of forgiveness as a transformation from

revenge and avoidance motivations toward

positive, relationship-enhancing motivations

proved to be a useful basis for studying

specific incidents of offense and transgres-

sion in couples� relationships (McCullough,

Worthington, & Rachal, 1997; McCullough

et al., 1998). While the results for internal for-

giveness were generally stronger than for

expressed forgiveness, we believe the distinc-

tion between internal and expressed forgive-

ness remains an important one. It is plausible,

for example, that in most cases of offense and

transgression, the expression of forgiveness

would be produced (in part) as a function of

internal forgiveness. The correlations between

internal and expressed forgiveness in this

study were positive and significant (from .36

to .54) but not so high as to suggest that they

tap into an identical construct. Baumeister

and Exline’s (Baumeister et al., 1998; Exline

& Baumeister, 2000) work suggests that the

victim’s analysis of costs and benefits may

inhibit or facilitate the expression and com-

munication of forgiveness. Indeed, there will

be many variables, situational (e.g., external

pressure) and dispositional (e.g., self-monitor-

ing), which could moderate the links between

internal and behavioral forgiveness processes.

Attributions and forgiveness. As previ-

ously noted, most prior forgiveness research

has measured attributions and forgiveness ten-

dencies in relationships in a general fashion

or assessed hypothetical responses to fictional

transgressions. These practices may limit the

generalizability of the findings or open the

possibility that previously documented posi-

tive associations among attributions, forgive-

ness, and relationship satisfaction are products

of general attitudes toward the relationship.

The present study addressed this issue by hav-

ing both partners in a relationship focus on

two specific negative incidents, which they

had previously experienced. Nevertheless, we

generally replicated prior findings. One nota-

ble exception is that we found no evidence

that attributions mediated the link between

relationship evaluations and forgiveness.

One explanation for our null findings con-

cerning mediation is simply that measuring

the variables at different levels of specificity

might serve to dilute mediation. However,

there are other differences across studies that

might be involved. For example, the Fincham

et al. (2002) study that reported full mediation

assessed hypothetical forgiving situations in

long-term married couples who were in their

midforties and had been married approxi-

mately 20 years. In contrast, we assessed per-

ceptions of specific real-life transgressions

from a sample of relatively young university

students in relationships that were compara-

tively less developed. Thus, the specific rela-

tionship between attributions and forgiveness

may change across different populations or

come and go according to measurement pro-

cedures. However, it is important to note that,

although we failed to find mediation, our

results affirmed the role of both relationship

quality and blame attributions in predicting

internal forgiveness.

Forgiveness bias. A novel result of this

study was that people perceived their partners

as forgiving them less than what their part-

ners actually reported (a negative bias),

although this tendency was only statistically

significant for expressed rather than internal

forgiveness. This finding seems counterintui-

tive (Why would participants show a greater

bias for the form of forgiveness that is more

explicit?) and should be considered as pre-

liminary until it is replicated in further

research. Nevertheless, the general pattern of

results is consistent with error management

theory (Haselton & Buss, 2000).

Haselton and Buss (2000) use the house-

hold smoke alarm to illustrate the functional

value of a built-in negative bias. Because the

consequences of a fire are severe, the smoke

alarm is designed to be sensitive to any form

of heat or smoke. The result is that false

alarms often occur (false negatives) from

minor kitchen mishaps, such as burning the

74 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

morning toast. Thus, a negative forgiveness

bias could act as a relationship alarm oper-

ating via an acute sensitivity to a lack of

forgiveness (with the associated cost of occa-

sionally raising false alarms). This perspec-

tive is similar to what Holmes and Rempel

(1989, p. 211) refer to as a risk-aversive

strategy. In the present study, perpetrators

may have recognized that their wrongdoing

had placed them in a vulnerable position if

they wished to maintain the relationship and

positive conclusions about forgiveness could

have proved risky (also see Murray, Bellavia,

Rose, & Griffin, 2003).

Evidence of an overall negative forgive-

ness bias might seem to suggest that forgive-

ness is a domain, which is free from the

positive biases found in most relationship

research (e.g., see Murray, 2001). However,

our individual-difference analyses indicated

that excessive negative bias (either internal

or expressed) is not conducive to happiness.

Consistent with the work by Murray and her

colleagues, the association between positive

forgiveness bias and relationship quality re-

mained strong, significant, and positive for

both men and women. Perhaps even more

striking, higher levels of positive bias also

inflated the partner’s level of relationship

quality (controlling for the partner’s level of

forgiveness bias). This set of findings is the

clearest example in this research of dyadic

influences in relationships.

It is important to note that our evidence

and claims in this study concern bias, not

accuracy. These two constructs are frequently

confounded in social psychological and rela-

tionship research, but they can operate quite

independently (Fletcher, 2002; Gagne &

Lydon, 2001). For example, men and women

may accurately track the levels of their part-

ner’s forgiveness but be systematically biased

at the same time (positively or negatively).

Michael forgives Mary on three separate

occasions that Mary transgresses; he forgives

her a great deal (Time 1), a moderate amount

(Time 2), and not at all (Time 3). Mary accu-

rately tracks Michael’s intrapsychic forgive-

ness states but does so through rose-colored

glasses—she attributes forgiveness as com-

pletely (Time 1), a lot (Time 2), and slightly

(Time 3). Thus, in this example, Mary is

accurate but positively biased. Extending

the possibilities, a given sample may be un-

biased and wildly inaccurate or unbiased and

accurate.

To provisionally assess the overall accu-

racy of the current sample in their forgiveness

judgments, we correlated the forgiveness judg-

ments of the men with the self-reports of their

female partners (and vice versa). We did this

for both internal and expressed forgiveness

and for both Incident A and Incident B, which

produced eight correlations. All the correla-

tions were positive, and four were significant

at the p , .05 level (ranging from .31 to .44).

None of these results were changed when we

controlled for relationship satisfaction.

These findings suggest that the sample was

quite accurate in assessing the extent to which

their partners forgave them and confirm the

importance of carefully interpreting bias and

associated analyses in relationship settings.

While error management theory (Haselton &

Buss, 2000) does not distinguish between

accuracy and bias, their approach would pre-

sumably predict that an acutely sensitive

negative bias should be associated with a ten-

dency toward accurate tracking at the same

time (as we found in the present study).

Limitations

We measured and controlled for some plausi-

ble third variables in this study, including

assumed similarity and incident severity/neg-

ativity, and these generally did not influence

the results. However, because the study is cor-

relational, the possibility of alternative third

variables must always be considered. In addi-

tion, the predicted model in Figure 1 assumed

a causal path from the global evaluations of

relationship quality (that preexisted the spe-

cific incidents of offense and transgression) to

the specific attributions of offense and for-

giveness responses. Although a reasonable

assumption, specific incidents of offense or

transgression, especially those judged as

highly serious and injurious, could cause the

relationship to be reevaluated and thus re-

verse the causal direction shown in Figure 1

(Fitness, 2001).

Dyadic forgiveness 75

The power of this study was relatively low

because of the sample size. With more partic-

ipants, it would have been possible to further

investigate the relationship of attributions

and forgiveness by dividing the sample ac-

cording to incidents of joint and individual-

dominated responsibility. Such a division

would have allowed the analysis of how the

effects vary when one partner is to blame or

when blame is roughly reciprocal. This is an

important direction for further research.

Another limitation of this study was our

measurement of expressed forgiveness. As

discussed above, this single-item measure

probably did not capture the variety of ways

in which individuals communicate forgive-

ness. The vast majority of published forgive-

ness scales measure internal forgiveness,

although Hargrave and Sells (1997) and

Enright (Enright, 2001; Subkoviak et al.,

1995) include behavioral and communicative

items. However, these two dimensions of for-

giveness (the internal and interpersonal) need

to be examined in greater detail to tease apart

their unique properties and causes.

Kelley (1998) reported that nearly one half

of his participants described using only

implicit strategies to communicate forgive-

ness when recalling a recent transgression.

Therefore, there may be more general prob-

lems with relying on self-report measures of

expressed forgiveness. A fruitful avenue for

future research would be to adopt a dyadic

approach, as we have done in the present

work, and examine in greater detail how part-

ners both perceive and express forgiveness.

Comparisons between internal and expressed

forgiveness may also shed light on how the

expression of forgiveness can influence

reconciliation.

Conclusion

Exploring forgiveness from a dyadic perspec-

tive, as applied to specific incidents of offense

and transgression in intimate relationships,

has provided new insights into how couples

recover from hurtful events in their relation-

ships. We have replicated a range of findings

from prior research but have also broken

new ground. McCullough describes forgive-

ness as a ‘‘complex’’ of motivational changes

(McCullough, Bellah, Kilpatrick, & Johnson,

2001). By examining the dyad, this research

has begun to deal with some of the complex-

ity in the way forgiveness is experienced and

perceived in self and in intimate others. As

scientists untangle the complex links between

forgiveness and other relationship processes,

this work should both increase our scientific

understanding of relationships and assist the

clergy, the counselor, and the clinician to help

couples achieve healthier relationships.

References

Baumeister, R. F., Exline, J. J., & Sommer, K. L. (1998).The victim role, grudge theory, and two dimensionsof forgiveness. In E. L. Worthington (Ed.), Dimen-sions of forgiveness: Psychological research andtheological perspectives (pp. 79–104). Philadelphia:Templeton Press.

Bentler, P. M., & Wu, E. J. C. (1995). EQS for Windowsuser’s guide. Encino, CA: Multivariate Software.

Boon, S. D., & Sulsky, L. M. (1997). Attributions ofblame and forgiveness in romantic relationships: Apolicy-capturing study. Journal of Social Behaviorand Personality, 12, 19–44.

Bradbury, T., & Fincham, F. D. (1992). Attributions andbehavior in marital interaction. Journal of Personal-ity and Social Psychology, 63, 613–628.

Bradfield, M., & Aquino, K. (1999). The effects of blameattributions and offender likeableness on forgivenessand revenge in the workplace. Journal of Manage-ment, 25, 607–631.

Enright, R. D. (2001). Forgiveness is a choice. Washing-ton, DC: American Psychological Association.

Exline, J. J., & Baumeister, R. F. (2000). Expressingforgiveness and repentance: Benefits and barriers.In M. E. McCullough & K. I. Pargament (Eds.),Forgiveness: Theory, research, and practice (pp.133–155). New York: Guilford.

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

Fincham, F. D., Paleari, G., & Regalia, C. (2002). For-giveness in marriage: The role of relationship quality,attributions, and empathy. Personal Relationships, 9,27–37.

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

Fitness, J. (2001). Betrayal, rejection, revenge and forgive-ness. In M. R. Leary (Ed.), Interpersonal rejection(pp. 73–103). New York: Oxford University Press.

Fletcher, G. J. O. (2002). The new science of intimaterelationships. Malden, MA: Blackwell.

Fletcher, G. J. O., & Fitness, J. (1990). Occurrent socialcognition in close relationship interaction: The roleof proximal and distal variables. Journal of Personal-ity and Social Psychology, 59, 464–474.

76 M. D. Friesen, G. J. O. Fletcher, and N. C. Overall

Fletcher, G. J. O., Simpson, J. A., & Thomas, G.(2000a). Ideals, perceptions, and evaluations in earlyrelationship development. Journal of Personality andSocial Psychology, 79, 933–940.

Fletcher, G. J. O., Simpson, J. A., & Thomas, G.(2000b). The measurement of Perceived RelationshipQuality Components: A confirmatory factor analyticapproach. Personality and Social Psychology Bulle-tin, 26, 340–354.

Fletcher, G. J. O., & Thomas, G. (2000). Behavior andon-line cognition in marital interaction. PersonalRelationships, 7, 111–130.

Gagne, F. M., & Lydon, J. E. (2001). Mind-set in closerelationship: When bias leads to (in)accurate predic-tions. Journal of Personality and Social Psychology,81(1), 85–96.

Hargrave, T. D., & Sells, J. N. (1997). The developmentof a forgiveness scale. Journal of Marital and FamilyTherapy, 23, 41–63.

Haselton, M. G., & Buss, D. M. (2000). Error manage-ment theory: A new perspective on biases in cross-sex mind reading. Journal of Personality and SocialPsychology, 78, 81–91.

Holmes, J. G., & Rempel, J. K. (1989). Trust in closerelationships. In C. Hendrick (Ed.), Close relation-ships (pp. 187–221). Newbury Park, CA: Sage.

Karremans, J. C., Van Lange, P. A. M., Ouwerkerk,J. W., & Kluwer, E. S. (2003). When forgiving en-hances psychological well-being: The role of inter-personal commitment. Journal of Personality andSocial Psychology, 84, 1011–1026.

Kelley, D. (1998). The communication of forgiveness.Communication Studies, 49, 255–271.

McCullough, M. E. (2000). Forgiveness as humanstrength: Theory, measurement, and links to well-being. Journal of Social and Clinical Psychology,19, 43–55.

McCullough, M. E., Bellah, C. G., Kilpatrick, S. D., &Johnson, J. D. (2001). Vengefulness: Relationshipswith forgiveness, rumination, well-being and the bigfive. Personality and Social Psychology Bulletin, 27,601–610.

McCullough, M. E., & Hoyt, W. T. (2002). Transgres-sion-related motivational dispositions: Personality

substrates of forgiveness and their links to the bigfive. Personality and Social Psychology Bulletin, 28,1556–1573.

McCullough, M. E., Rachal, K. C., Sandage, S. J., Wor-thington, E. L., Jr., Brown, S. W., & Hight, T. L.(1998). Interpersonal forgiving in close relationships:II. Theoretical elaboration and measurement. Journalof Personality and Social Psychology, 75, 1586–1603.

McCullough, M. E., & Witvliet, C. V. (2002). Thepsychology of forgiveness. In C. R. Snyder & S. J.Lopez (Eds.), Handbook of positive psychology (pp.446–458). London: Oxford University Press.

McCullough, M. E., Worthington, E. L., Jr., & Rachal,K. C. (1997). Interpersonal forgiving in close rela-tionships. Journal of Personality and Social Psychol-ogy, 73, 321–336.

Murray, S. L. (2001). Seeking a sense of conviction:Motivated cognition in close relationships. In G. J. O.Fletcher & M. S. Clark (Eds.), Blackwell handbookof social psychology: Interpersonal processes (pp.107–126). Malden, MA: Blackwell.

Murray, S. L., Bellavia, G. M., Rose, P., & Griffin,D. W. (2003). Once hurt, twice hurtful: How per-ceived regard regulates daily marital interactions.Journal of Personality and Social Psychology, 84,126–147.

North, J. (1987). Wrongdoing and forgiveness. Philoso-phy, 62, 499–508.

North, J. (1998). The ‘‘ideal’’ of forgiveness: A philoso-pher’s exploration. In R. D. Enright & J. North(Eds.), Exploring forgiveness (pp. 15–34). Madison:University of Wisconsin Press.

Subkoviak, M. J., Enright, R. D., Wu, C.-R., Gassin, E.A., Freedman, S., Olson, L. M., et al. (1995). Mea-suring interpersonal forgiveness in late adolescenceand middle adulthood. Journal of Adolescence, 18,641–655.

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

Worthington, E. L. (1998). An empathy-humility-com-mitment model of forgiveness applied within familydyads. Journal of Family Therapy, 20, 59–76.

Dyadic forgiveness 77