Content uploaded by Nathaniel G Wade
Author content
All content in this area was uploaded by Nathaniel G Wade on Jan 06, 2014
Content may be subject to copyright.
Journal of Consulting and Clinical
Psychology
Efficacy of Psychotherapeutic Interventions to Promote
Forgiveness: A Meta-Analysis
Nathaniel G. Wade, William T. Hoyt, Julia E. M. Kidwell, and Everett L. Worthington, Jr.
Online First Publication, December 23, 2013. doi: 10.1037/a0035268
CITATION
Wade, N. G., Hoyt, W. T., Kidwell, J. E. M., & Worthington, E. L., Jr. (2013, December 23).
Efficacy of Psychotherapeutic Interventions to Promote Forgiveness: A Meta-Analysis. Journal
of Consulting and Clinical Psychology. Advance online publication. doi: 10.1037/a0035268
Efficacy of Psychotherapeutic Interventions to Promote Forgiveness:
A Meta-Analysis
Nathaniel G. Wade
Iowa State University
William T. Hoyt
University of Wisconsin–Madison
Julia E. M. Kidwell
Iowa State University
Everett L. Worthington, Jr.
Virginia Commonwealth University
Objective: This meta-analysis addressed the efficacy of psychotherapeutic interventions to help people
forgive others and to examine moderators of treatment effects. Method: Eligible studies reported
quantitative data on forgiveness of a specific hurt following treatment by a professional with an
intervention designed explicitly to promote forgiveness. Random effects meta-analyses were conducted
using k⫽53 posttreatment effect sizes (N⫽2,323) and k⫽41 follow-up effect sizes (N⫽1,716) from
a total of 54 published and unpublished research reports. Results: Participants receiving explicit
forgiveness treatments reported significantly greater forgiveness than participants not receiving treatment
(⌬
⫹
⫽0.56 [0.43, 0.68]) and participants, receiving alternative treatments (⌬
⫹
⫽0.45 [0.21, 0.69]).
Also, forgiveness treatments resulted in greater changes in depression, anxiety, and hope than no-
treatment conditions. Moderators of treatment efficacy included treatment dosage, offense severity,
treatment model, and treatment modality. Multimoderator analyses indicated that treatment dosage (i.e.,
longer interventions) and modality (individual ⬎group) uniquely predicted change in forgiveness
compared with no-treatment controls. Compared with alternative treatment conditions, both modality
(individual ⬎group) and offense severity were marginally predictive (ps⬍.10) of treatment effects.
Conclusions: It appears that using theoretically grounded forgiveness interventions is a sound choice for
helping clients to deal with past offenses and helping them achieve resolution in the form of forgiveness.
Differences between treatment approaches disappeared when controlling for other significant modera-
tors; the advantage for individual interventions was most clearly demonstrated for Enright-model
interventions, as there have been no studies of individual interventions using the Worthington model.
Keywords: forgiveness, interventions, efficacy, treatment, anger
The psychological study of forgiveness has grown dramatically
in the past two decades (Fehr, Gelfand, & Nag, 2010; Worthing-
ton, 2005), especially in the exploration of interventions designed
explicitly to promote forgiveness. Initial evidence supports the
efficacy of these forgiveness interventions, showing that they help
participants increase their degree of forgiveness for an offense or
injury, increase hope and psychological well-being, and decrease
depression, anxiety, and anger (Baskin & Enright, 2004; Wade,
Worthington, & Meyer, 2005).
Definition of Forgiveness
What is “forgiveness”? According to the emerging consensus
among intervention researchers, forgiveness can include both (a) the
reduction in vengeful and angry thoughts, feelings, and motives that
may be accompanied by (b) an increase in some form of positive
thoughts, feelings, and motives toward the offending person (Wade &
Worthington, 2003). Thus, forgiveness is understood as primarily an
intrapersonal experience that does not include reconciliation with the
offending person even though reconciliation might accompany it.
Most researchers agree that forgiveness is not forgetting, condoning,
or excusing the wrongdoing, nor is it simply the opposite or absence
of bitterness and vengefulness (i.e., unforgiveness, Wade & Wor-
thington, 2003; see essential agreement among 20 research teams in
Worthington, 2005).
Given this definition, seeking to promote forgiveness in psycho-
therapy is more than simply reducing anger, bitterness, and vengeful
rumination. With many clients, the simple reduction or elimination of
negative thoughts and feelings would be considered a psychothera-
peutic success. However, some psychotherapists have wondered,
What more can be done for my clients who have experienced signif-
icant hurts? (e.g., DiBlasio & Benda, 1991). In response to this
question, researchers and clinicians have proposed that helping clients
Nathaniel G. Wade, Department of Psychology, Iowa State University;
William T. Hoyt, Department of Educational Psychology, University of
Wisconsin–Madison; Julia E. M. Kidwell, Department of Psychology,
Iowa State University; Everett L. Worthington, Jr., Department of Psy-
chology, Virginia Commonwealth University.
Correspondence concerning this article should be addressed to Nathaniel
G. Wade, who is now at Department of Psychology, University of Iowa,
w112 Lagomarcino Hall, Ames, Iowa, 50011. E-mail: nwade@iastate.edu
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Consulting and Clinical Psychology © 2013 American Psychological Association
2013, Vol. 82, No. 1, 000 0022-006X/13/$12.00 DOI: 10.1037/a0035268
1
to forgive could be a useful focus of psychotherapy (Enright, 2001;
Worthington, 2001). This is in line with a positive psychotherapy
perspective that seeks to attend to and develop strengths rather than
just minimize problems (Gelso & Woodhouse, 2003; Seligman,
Rashid, & Parks, 2006). Thus, the promotion of forgiveness as a
psychotherapeutic technique is more than simply reducing negative
thoughts and feelings but also includes helping clients move toward
more positive, even optimal functioning.
Explicit Forgiveness Interventions
As a result, several theoretical forgiveness models have been
developed to promote forgiveness (e.g., Enright, 2001; Luskin,
2007; Worthington, 2001). Research groups headed by Enright and
Worthington have led the way in investigating the efficacy of these
interventions. Enright’s treatment model contains 20 steps (Enright
& Fitzgibbons, 2000), which are summarized in four phases:
Uncovering (negative feelings about the offense), Decision (to
pursue forgiveness for a specific instance), Work (toward under-
standing the offending person), and Discovery (of unanticipated
positive outcomes and empathy for the offending person). Each of
these phases includes several smaller steps within them. For ex-
ample, within the Work phase, clients work toward understanding
the offender, developing compassion, accepting/absorbing the
pain, and considering giving a gift of forgiveness to the offender
(Enright, 2001). The efficacy of the Enright model has been shown
with groups as diverse as adult incest survivors (Freedman &
Enright, 1996), parents who have adopted special needs children
(Baskin, Rhody, Schoolmeesters, & Ellingson, 2011), and inpa-
tients struggling with alcohol and drug addiction (Lin, Mack,
Enright, Krahn, & Baskin, 2004).
The other primary research group has conducted research orga-
nized around Worthington’s (2001) REACH Forgiveness model.
Each letter in the acronym REACH represents a major component
in the forgiveness process. In the first step of this model, partici-
pants recall (R) the hurt they experienced and the emotions asso-
ciated with it. Next, participants work to empathize (E) with their
offender, take another’s perspective, and consider factors that may
have contributed to their offender’s actions. This is done without
condoning the other’s actions or invalidating the often-strong
feelings the offended person has as a response. Third, participants
explore the idea that forgiveness can be seen as an altruistic (A)
gift to the offender. Participants learn that forgiveness can be
freely given or legitimately withheld and recall times when others
forgave them. Fourth, participants make a commitment (C) to
forgive. This includes committing to the forgiveness that one has
already achieved as well as committing to work toward more
forgiveness, knowing that it is a process that often takes time to
fully mature. Last, participants seek to hold (H) onto or maintain
their forgiveness through times of uncertainty or a return of anger
and bitterness (e.g., if they get hurt again in a similar way).
Previous meta-analyses have indicated that interventions of this
nature can effectively promote forgiveness (Baskin & Enright,
2004; Wade et al., 2005). In one of the first meta-analyses of the
efficacy of forgiveness interventions, Baskin and Enright found
that in nine studies of individual and group therapy (N⫽330
participants), explicit forgiveness interventions increased forgive-
ness, hope, and self-esteem, and reduced anxiety and depression.
Baskin and Enright claimed that interventions that were process
based (in which forgiveness is understood as a process that unfolds
over time through a series of developmental steps) were more
effective than interventions that were decision based (in which
forgiveness is understood as a conscious choice made by the
person who was injured). However, treatment categories were
confounded with the amount of time spent intervening. That is, the
individual counseling and process-based group forgiveness models
had considerably longer treatment durations than the decision-
based interventions. Thus, that meta-analysis left the question
unanswered whether treatment model made a difference over and
above time spent intervening. In earlier meta-analytic studies,
Worthington, Sandage, and Berry (2000; 13 studies) and Wor-
thington, Kurusa, et al. (2000; 25 studies) found that the duration
of treatment and effect size were correlated about .75.
In another meta-analysis, Wade and colleagues (2005) examined
the efficacy of forgiveness interventions, focusing on group treat-
ments in 39 studies. This result was also confirmed in a separate
meta-analysis of group treatments that examined 13 published
studies up to 2006 (Rainey, Readdick, & Thyer, 2012). In addition,
Wade et al. controlled for treatment duration. They reported that
full forgiveness interventions (treatments that incorporate all com-
ponents of an intervention model) were, in fact, more effective
than partial interventions (dismantled treatments that used only
certain components of a model), even when they controlled for
treatment duration. Wade et al. also found that time spent on
certain elements, for example developing empathy, was positively
related to the efficacy of the treatments. However, the analyses of
Wade et al. and Rainey et al. were limited to interventions pro-
vided in a group format. Therefore, neither meta-analysis assessed
whether individual counseling interventions differed after control-
ling for treatment duration nor did they assess potential differences
between counseling formats (e.g., individual, group, or couples).
Finally, Wade et al. and Rainey et al. did not assess outcomes other
than forgiveness. Therefore, the effects of group forgiveness treat-
ment on outcomes such as depression, anxiety, and hope, while
controlling for intervention duration, are still unknown.
Potential Moderators of Forgiveness
Intervention Efficacy
Although forgiveness interventions appear effective for promot-
ing forgiveness and perhaps even mental health, questions about
moderators that affect the forgiveness process remain unaddressed.
Specifically, what factors are likely to facilitate a participant’s
response to treatment? Are some treatment approaches better than
others if the effects of treatment duration are controlled? Does
treatment modality make a difference? How do these interventions
affect different outcomes?
Treatment Duration
One of the most well-established moderators of treatment effi-
cacy in the general psychotherapy outcome literature is treatment
duration (Howard, Kopta, Krause, & Orlinsky, 1986). Dose–
response curves have shown that improvement in client concerns
increases considerably with more treatment, until about 28 coun-
seling sessions, at which point it reaches an asymptote. Thus, in a
short-term model (i.e., less than 28 sessions), duration should be
expected to play an important role with forgiveness interventions
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2WADE, HOYT, KIDWELL, AND WORTHINGTON
as well. In fact, past summaries of the forgiveness outcome liter-
ature have indicated that this is the case (Worthington, Kurusa, et
al., 2000; Worthington, Sandage, & Berry, 2000). Therefore, it is
crucial to attend to the duration of the separate treatments to
control for that factor when analyzing other moderators and to
replicate the dose–response result with a more comprehensive list
of forgiveness intervention studies.
Theoretical Intervention Model
The specific forgiveness model used to intervene is another
important potential moderator of response to forgiveness treat-
ment. Two forgiveness intervention models have been used in the
majority of forgiveness intervention research: Enright’s (2001) and
Worthington’s (2006). In addition, other researchers have devel-
oped and tested intervention models that are distinct from either
the Enright or Worthington approach. For example, Luskin, Gin-
zburg, and Thoresen (2005) developed a cognitive behavioral
model; Greenberg, Warwar, and Malcolm (2008, 2010) used an
emotionally focused therapy approach; and DiBlasio and Benda
(2008) used an explicitly decision-based model. Although the
alternatives have not yet been investigated often enough to be
examined separately, together they do provide an additional treat-
ment model category against which Enright’s and Worthington’s
models can be compared. An examination of the different forgive-
ness intervention models can provide information about the poten-
tial differences in treatment efficacy.
Individual, Couple, or Group Treatment Modality
A third potential moderator is treatment modality. Although the
majority of forgiveness interventions have been conducted in
group formats (Wade et al., 2005), there are a growing number of
studies on individual and couple treatments. Previous reviews have
attempted to answer this question as well. In their meta-analysis,
Baskin and Enright (2004) compared individual process-based
interventions to process- and decision-based group interventions.
In that analysis, individual interventions were more effective than
either group or couple interventions. However, this analysis did
not consider treatment duration, which was confounded with treat-
ment modality (process-oriented individual treatments averaged 36
hr; process-based group interventions averaged 7 hr; decision-
based interventions averaged 4 hr). So, for forgiveness interven-
tions, differences among modalities are possible but have not been
effectively tested without substantial confounding of variables.
Offense Severity
Finally, another possible moderator of the forgiveness process is
offense severity (McCullough & Hoyt, 2002). A variety of researchers
have postulated that a relationship between these variables exists.
Ohbuchi, Kameda, and Agarie (1989) examined the effects of offense
severity, along with the importance of an apology from the offender,
in encouraging forgiveness. They found that offenses that are more
severe, even when an individual receives an apology from the of-
fender, are more difficult to overcome and may require a more
extensive and persuasive apology for anger and aggression to be
mitigated. Fincham, Jackson, and Beach (2005) suggested that severe
offenses are more difficult to overcome and may require a more
complex understanding of forgiveness. Thus, it appears that the more
severe the transgression, the more difficult it is to forgive. For exam-
ple, Krumrei, Mahoney, and Pargament (2011) found that when
divorce was considered a sacred loss or desecration of something
sacred (i.e., marriage), it was particularly difficult to forgive. Al-
though Exline, Worthington, Hill, and McCullough (2003) suggested
that people maintained an intuitive cognitive accounting of the net
injustice of offenses and their aftermath; there might be something
qualitatively different about very severe transgressions, not just quan-
titatively different. This has not been investigated per se.
When applied to intervention research, the role of offense severity
becomes even more nuanced. Often by design, participants recall a
specific offense that was hurtful to them. The reported events vary in
terms of external or objective severity, from murdered family mem-
bers (Luskin & Bland, 2000) to feeling neglected by one’s parents
(Al-Mabuk, Enright, & Cardis, 1995). However, these events were
recalled as something hurtful, and therefore, they often have little
variability in self-reported severity. This is particularly evident in
studies that recruited participants with a range of reported hurts (e.g.,
McCullough, Worthington, & Rachal, 1997; Wade, Worthington, &
Haake, 2009). Typically, in studies that assess the perceived severity
of the offense, almost all of the participants report that the offense was
severe. Because of the design of such research, this result is under-
standable, but it does not provide information about the effect of
offense severity on response to forgiveness treatment. In addition, no
intervention studies of which we are aware address offense severity or
attempt to investigate the role of offense severity that has been
objectively rated by coders. One might suspect that participants with
severe hurts would respond less favorably to forgiveness interventions
or might need interventions of longer durations than those with less
severe hurts (Worthington, Sandage, & Berry, 2000). However, sev-
eral studies have shown that people experiencing significant offenses,
including incest (Freedman & Enright, 1996) and murder of a family
member (Luskin & Bland, 2000), respond favorably to treatment.
Still, none of these studies have directly examined the influence of
severity on response to treatment. One other consideration is impor-
tant. In severe hurts, there is less possibility of a floor effect. Namely,
in a mild hurt, effect sizes expected after intervention are limited by
the amount of change that is achievable. However, with severe hurts,
the amount of forgiveness to be achievable is considerable.
Purpose
The purpose of the present investigation was to expand the
forgiveness intervention literature by (a) systematically comparing
forgiveness and mental health outcomes between treatment and
control groups from pre- to posttreatment and from pretreatment to
follow-up, (b) examining the potential moderators of treatment
efficacy that have been identified but not yet included in meta-
analytic research, and (c) including research studies conducted
since previous meta-analyses (about 7 years of research since
Wade et al., 2005). We analyzed changes in forgiveness, anxiety,
depression, and hope across time. In addition, we explored the
moderating effects of treatment duration, psychotherapeutic
model, treatment modality, and offense severity on forgiveness.
Because considerably fewer studies included depression, anxiety
and hope as outcomes, we did not analyze moderators of these
outcomes.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
FORGIVENESS META-ANALYSIS
Method
This meta-analytic review was based on 54 published and un-
published reports of studies of forgiveness interventions. We
searched the following areas to locate studies for inclusion: (a)
computerized search of the PsycINFO (1872–2011) database using
keywords forgiveness,intervention,psychotherapy,and treatment;
(b) manual search of references listed in all located studies; and (c)
contacting known forgiveness researchers for unpublished studies.
Studies were included if they (a) examined effects of a psycho-
therapeutic intervention specifically designed to promote forgive-
ness, (b) offered the intervention in-person by a trained facilitator,
(c) used a quantitative measure of forgiveness for a specific
offense as an outcome, (d) were written in English, and (e) were
completed prior to 2012. Studies were excluded if they were
self-help rather than therapist-led (e.g., client-directed from an
online resource or book); focused on developing forgiveness gen-
erally, but not for a specific offense; or did not measure forgive-
ness as an outcome. In Figure 1, we display the numbers of found,
eligible, and ineligible studies.
A case can be made for excluding studies lacking a no-treatment
control group from meta-analyses of interventions. When treat-
ment effects in such studies are computed by comparing pre- and
posttreatment means, they have been shown to yield inflated
estimates of effect size (Lipsey & Wilson, 1993). Carlson &
Schmidt (1999) showed that this bias was attributable to the failure
to account for spontaneous improvement among untreated (con-
trol) participants, and Becker (1998) proposed that this source of
bias could be obviated by computing an effect size (⌬) comparing
Identification
Screening
Eligibility
Included
Records identified
through database
searching (k = 226)
Records identified through other
sources (k = 29) e.g., conference
presentations, email solicitation
Records identified after
duplicates removed and
screened (k = 197)
Records ineligible
(k = 132)
• Not an intervention
study
• No state measure of
forgiveness
• No facilitator or
therapist
• Publication not in
English
Full-text articles
ineligible
• No state measure of
forgiveness (k = 3)
• No facilitator or
therapist (k = 4)
• Not a psychotherapy
intervention (k = 2)
• Post-test only design
(k = 1)
• Publication not in
English (k =1)
Reports included in the meta-
analyses (k = 54)
Full text documents reviewed
for eligibility (k = 65)
Figure 1. Flow diagram of studies included in the meta-analyses.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4WADE, HOYT, KIDWELL, AND WORTHINGTON
improvement among treated participants to that among untreated
participants, which makes possible imputation of control group
data for studies without a control group, as described later. As
described in the Results section, we tested to see whether there
were systematic differences between studies with and without a
control group after imputation and found none. We therefore
included both controlled and uncontrolled studies in our analyses.
Assessing Methodological Quality
We assessed the methodological quality of the studies included
in the meta-analysis with three ratings: whether the study was
peer-reviewed, whether participants in the study were randomly
assigned to treatment, and the retention rate of participants in the
study. These three were chosen as estimates of the quality of each
study’s methodology because these are common indicators across
different measures of quality (e.g., Downs & Black, 1998; Higgins
& Altman, 2008). Peer review and random assignment were coded
as dichotomous variables (yes or no) based on the report within the
document to which we had access. To calculate the retention rate
(i.e., the percentage of participants who remained in the study from
pretest and treatment to posttest), we divided the number of par-
ticipants who completed the treatments and pre–post question-
naires and were included in the analyses by the number of partic-
ipants who were assigned to treatments. Because each of these
characteristics can be considered an independent measure of study
quality, we conducted a single-moderator analysis to determine
whether each of these characteristics significantly moderated the
reported effect size. None of these analyses revealed a significant
difference in effect size based on these measures. (The pvalues for
these tests in the forgiveness dataset with no-treatment compari-
sons—the largest and therefore most statistically powerful set of
tests—were .94, .19, and .53 for published, random assignment,
and retention rate, respectively.) These results indicated that treat-
ment effect size was not related to the variables we used to assess
methodological quality.
Coding of Moderators
Four raters, two males and two females, coded the severity of
offenses described in each individual study. Offenses were coded
on the basis of the Holmes–Rahe Social Readjustment Scale (Hol-
mes & Rahe, 1967). The Holmes–Rahe provides estimates of the
magnitude of stress for 43 particular life events, such as change in
financial situation and death of a close friend. Many of these
events directly reflect the offenses participants were experiencing
(e.g., divorce). So, in many cases, the offenses that the treatment in
a particular study was designed to address already had a numerical
value on the Holmes–Rahe Scale. For those that did not (e.g.,
infidelity), a numerical value was determined prior to coding based
on events on the Holme–Rahe that were judged by consensus of
the raters and authors as similar in nature (e.g., sex difficulties).
The greater the severity of the offense, the higher the offense was
rated on a scale from 1 to 100. The four ratings were then averaged
to produce a single index of offense severity for each study.
Interrater reliability for this index (Shrout & Fleiss, 1979; intra-
class correlation [ICC] 34) was .94. Treatment duration, psycho-
therapeutic model, and treatment modality were taken directly
from the research report (i.e., article, chapter, dissertation, or
manuscript) by the first author.
Computation of Effect Sizes
Outcome studies can use a variety of designs. Two common
designs include posttest only with control group and pretest–
posttest with control group. Each of these designs allows for a
comparison of outcomes between the two groups. If participants
are randomly assigned to groups (i.e., randomized controlled trial,
RCT), this provides a sound basis for attributing group differences
in outcomes to the effects of the intervention. The effect size
computed from the well-designed, pretest–posttest RCT has been
called the “standard of accuracy” (Carlson & Schmidt, 1999, p.
853) for intervention research, in that it controls for differences at
pretest while comparing outcomes at posttest.
Another important consideration is the nature of the comparison
group. Researchers commonly compare the intervention group to a
wait-list or no-treatment control group. Another common strategy
is to compare a forgiveness intervention to some alternative type of
treatment (e.g., “placebo” condition; treatment as usual; partial
forgiveness intervention). The effect size computed on this com-
parison estimates the additional gain in efficacy from employing a
forgiveness-focused intervention in comparison to what would be
expected under an alternative treatment approach. Because the two
types of comparison groups address different research questions,
we analyzed effect sizes of each type in separate meta-analyses.
Analyses
First, we computed within-groups (pre–post) standardized mean
differences and sampling variances for both treatment and control
groups. As recommended by Borenstein, Hedges, Higgins, and
Rothstein (2009), these were then corrected to produce unbiased
effect sizes (g
T
and g
C
) and sampling variances (v
gT
and v
gC
) for
treatment and control groups, respectively. Becker’s (1988) ⌬is a
comparison of change in the treatment group to change in the
control group (taking baseline scores into account):
⌬⫽gT⫺gC(1)
⌬⫽
gT
⫹
gC(2)
Becker’s ⌬was the basic unit of effect size for this meta-analysis,
and studies were weighted by 1/v
⌬
in aggregation and significance
testing (Becker, 1988; Hedges & Olkin, 1985).
1
Effect size com-
putation and aggregation of dependent effect sizes was conducted
using the R package “MAd” (Del Re & Hoyt, 2010), and omnibus
analyses, heterogeneity tests, and moderator analyses were con-
ducted using the R package “metafor” (Viechtbauer, 2010).
1
In these calculations, the pre–post correlation is needed to compute the
variance of the within-group (i.e., treatment or control) effect size. This
correlation is rarely reported in research studies and therefore has to be
estimated. This is essentially a test–retest correlation coefficient, although
it may be somewhat smaller than a typical reliability coefficient because
the effects of intervention likely decrease the stability of the scores. We
conservatively assumed r⫽.6 for all outcomes. The sampling variance is
a function of (1 ⫺r), so larger values of rlead to smaller sampling
variances and standard errors. The value chosen for rdoes not change the
effect size but does affect the sampling variance, which is the basis for
weighting that effect size in computing the aggregate effect size. When the
same value of ris used for all studies, the relative values of their sampling
variances do not change radically for small changes in r.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
FORGIVENESS META-ANALYSIS
Effect sizes were treated as random effects in all analyses, based
on our assumption that there were systematic differences among
studies related to intervention efficacy, and our desire to generalize
conclusions beyond the specific studies examined here (Hedges &
Vevea, 1996).
Results
Preliminary Analyses
Studies lacking no-treatment controls. The 54 research re-
ports included data on 62 interventions designed to promote forgive-
ness. However, 20 of these intervention groups were in studies that
did not include a no-treatment comparison condition. Most of these
studies compared a forgiveness intervention to a nonspecific treatment
(k⫽9) or to a placebo group (k⫽4); the remaining studies (k⫽7)
used a single-group, pre–post design.
2
Placebo controls can enhance
outcomes in their own right (Wampold, Minami, Tierney, Baskin, &
Bhati, 2005), and nonspecific interventions are also expected to result
in improvement in outcomes (Wampold, 2001). To produce a com-
mon effect size metric, one must avoid comparing some treatments to
no-treatment controls (T–NT) and others to placebos or other alter-
native treatments (T–AT).
One option for computing comparable effect sizes from studies
lacking a no-treatment condition is to compute the standardized
mean difference comparing postscores to prescores for these treat-
ments. Carlson and Schmidt (1999) noted that these effect sizes are
likely biased relative to those computed by comparing treatment
and no-treatment groups (Lipsey & Wilson, 1993). Instead, Becker
(1988) recommended meta-analyzing control group effect sizes
from studies that include a no-treatment control group. The results
of this preliminary analysis permit imputation of control data for
studies lacking a no-treatment control group.
Of the 42 studies that included data for a no-treatment condition,
four were duplicates—two different interventions that were com-
pared with the same no-treatment control group. When the 38
independent no-treatment control groups were meta-analyzed, they
yielded an aggregate effect size of g
⫹
⫽0.05, 95% confidence
interval (CI) [0 –.06, 0.17], reflecting the expected increase in
forgivingness in the absence of treatment. Thus, expected sponta-
neous improvement over this relatively short interval (Mdn ⫽6.33
weeks; range ⫽1–57) was small and not statistically significant
(95% CI includes 0). Although the control group effect sizes were
heterogeneous—Q(37) ⫽75.17, p⫽.0002 (I
2
⫽53%)—we
found no evidence of moderation by any of the study characteris-
tics tested as moderators in the main analyses. We therefore used
the aggregate effect size of g
C
⫽0.05 (v
gc
⫽.0033) as the control
group effect size for studies lacking a no-treatment control condi-
tion, so that T–NT effect sizes could be computed for all treatment
groups. Because alternative-treatment conditions varied by study,
we did not impute values for studies without alternative-treatment
groups. Consequently, the T–AT analyses include only effect sizes
derived from studies that included such conditions.
Examination of outliers. We followed recommended proce-
dures (Hedges & Olkin, 1985; Viechtbauer, 2010) for outlier analysis
to identify studies with effect sizes so deviant that it appears unlikely
they were drawn from the same population as the remaining studies in
the sample. We relied on the “externally standardized” residuals
(Viechtbauer, 2010), based on the final models (i.e., models including
significant moderator variables) reported. For the T–NT analysis, we
identified two studies with |z|⬎2, where zis the standardized
residuals divided by its standard error: Alvaro (2001) and Sells,
Giordano, and King (2002). The first study had an extreme positive
residual of 3.53 (⌬⫽3.80) and the second had a negative residual of
–3.69 (⌬⫽2 –.92). Visual inspection confirmed that each of these
studies was well into the tails of the distribution of effect sizes. We
therefore excluded these two studies from all reported analyses. A
third study did not contribute to the final analysis because of missing
moderator information but was a distant outlier in the omnibus anal-
ysis (Gambaro, 2002; standardized residual ⫽6.19; ⌬⫽6.81). We
therefore excluded this study from all analyses as well. A second
outlier analysis in the new data set (with the three studies excluded)
revealed no further outliers. Similar outlier analyses for T–AT com-
parisons identified a single outlier (Gambaro, 2002; standardized
residual ⫽6.11; ⌬⫽6.56), which was also excluded from reported
analyses.
3
The outlier analysis of effect sizes derived from baseline-
to-follow-up comparisons revealed no outliers in this data set.
Main Analyses: Forgiveness as Outcome
Omnibus analysis. In Figures 2 and 3 we provide effect sizes
(with 95% CIs) and moderator information for the studies included
in the T–NT and T–AT meta-analyses, respectively. In Table 1, we
display the omnibus effect sizes and homogeneity tests for both
T–NT and T–AT comparisons for forgiveness as an outcome. In
comparison with untreated participants, those receiving forgive-
ness interventions reported substantially greater increases in for-
giveness (⌬
⫹
⫽0.56). Put another way, the average participant in
the intervention group showed greater improvement over the
course of treatment than 71% of those in the no-intervention group.
T–AT comparisons yielded an aggregate effect size of ⌬
⫹
⫽0.45,
almost as large as that for the T–NT comparisons. In addition,
heterogeneity tests were statistically significant for both types of
comparisons, with a high proportion of effect size variance (I
2
⫽
72% and 77%, respectively) attributed to systematic sources be-
yond the variance expected due to sampling error. Thus, we
conducted planned moderator analyses for both sets of effect sizes.
Single moderator analyses. We first tested each potential mod-
erator variable individually, for both T–NT and T–AT comparisons.
Results are shown in Table 2 (continuous moderators) and Table 3
2
Because participants in single group, pre and post (SGPP) test design
studies are not randomized to condition, such studies are not as rigorous at
controlling for threats to internal validity (e.g., client self selection into
treatment as a confound). However, inclusion of SGPP studies, after
establishing their effect sizes in relation to imputed control group data to
eliminate potential bias (Becker, 1988), is in keeping with the goal of
meta-analysis to summarize all empirical literature relevant to the research
question of interest. Based on a reviewer’s concern that there may still be
bias in these effect sizes, we conducted post hoc analyses in which we
included a dummy variable (contrasting studies with and without imputed
control group data) in the final model and found that, when significant
moderators are accounted for, there are no systematic differences between
these two sets of studies (ps⬎.20).
3
The general pattern of findings does not change with the exclusion of
outliers. As expected, the homogeneity statistics (Q;I
2
) are substantially
reduced when outliers are excluded. Also, because two of three outliers
were positive (i.e., extreme, positive effect sizes), the omnibus effect size
is somewhat reduced when they are excluded. For the full data set, ⌬
⫹
⫽
0.62, 95% CI [0.45,0.80]; Q(61) ⫽350, I
2
⫽89%; after exclusion of
outliers, ⌬
⫹
⫽0.56, 95% CI [0.43,0.68], Q(52) ⫽188.91, I
2
⫽72%.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6WADE, HOYT, KIDWELL, AND WORTHINGTON
(categorical moderators). For continuous moderators, in Table 2, we
report intercepts (B
0
) as well as slopes (B
1
). It is the slope that
quantifies the degree of association between the moderator variable
and outcomes; the intercept reflects the estimated improvement when
the value of the moderator is 0. For example, dosage significantly
predicts treatment– control effect sizes, with B
0
⫽0.102 and B
1
⫽
0.047. This means that the predicted effect size (⌬) is a function of the
treatment duration: Predicted ⌬⫽B
0
⫹B
1
ⴱ
(treatment hours).
Thus, the predicted ⌬for an intervention of 1hr’s duration is
0.124 ⫹0.046 ⫽0.17 (a relatively weak effect size), whereas
that for a 10-hr intervention is 0.124 ⫹0.46 ⫽0.584 (a
moderate effect size and close to the omnibus effect size re-
ported in Table 1). The interventions in these studies varied
widely in their duration (Min ⫽1 hr; Max ⫽57 hr; M⫽10.3,
SD ⫽8.8), and this variability helped to account for the
variation in effect sizes between studies. For the T–AT com-
parisons, dosage was also a significant predictor of treatment
efficacy, with a slope very similar to that observed for the
T–NT effect sizes (B
1
⫽0.043).
Offense severity was also a significant predictor of study effect
size (B
1
⫽0.012) in the T–NT comparisons. In addition, severity
of offense was a significant moderator of T–AT effect sizes, with
a much steeper slope (B
1
⫽0.041) than that for T–NT effect sizes.
As the severity of the offense (and presumably the difficulty of
forgiving) increases, the advantage of forgiveness treatments over
generic treatments increases.
Figure 2. Forgiveness interventions: Treatment– control comparisons. Tx ⫽treatment model; W ⫽Worthing-
ton, E ⫽Enright, and O ⫽other treatment model. Mode ⫽treatment modality; g ⫽group, c ⫽couples, and
i⫽individual therapy modality. NTgrp ⫽no-treatment control group; Y ⫽yes and N ⫽no. Sev ⫽offense
severity. Hrs ⫽hours; CI ⫽confidence interval; RE ⫽random effects.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
FORGIVENESS META-ANALYSIS
In Table 3, we show the pattern of findings for significant categor-
ical moderators. For both T–NT and T–AT comparisons, intervention
model was a significant moderator of outcome. Interventions based on
the Enright model were significantly more effective than those based
on the Worthington model, and other interventions were intermediate.
For both T–NT and T–AT comparisons, treatment mode was also a
significant moderator. In both cases, individual interventions were
more efficacious than either couple or group interventions.
Multiple moderator analyses. The results of single-moderator
analyses may be misleading due to confounding among moderator
variables. For example, in forgiveness interventions, it is common to
offer longer treatments for more severe offenses (thus confounding
treatment duration with offense severity) and also to offer individu-
alized rather than group treatment for severe offenses (thus confound-
ing treatment modality with offense severity). It is therefore recom-
mended to use meta-regression to examine unique effects of a
moderator variable, controlling for the effects of other study charac-
teristics that may be correlated (Viechtbauer, 2007). In Table 4, we
show the results of meta-regressions of effect size onto those moder-
ators that emerged as significant in the single-moderator analyses, for
both T–NT and T–AT comparison.
4
Categorical moderators were
4
Prior to conducting the meta-regressions, we assessed for collinearity
among the variables in the analyses. VIFs (variance inflation factors) for the
final models displayed on Table 4 were all smaller than 5.0. By convention,
many methodologists use the “rule of 10” to interpret the VIF statistic—that is,
VIF ⬎10 creates doubts about the results of the analysis and triggers steps to
reduce multicollinearity before finalizing the model (O’Brien, 2007).
Figure 3. Forgiveness interventions: Treatment–alternative treatment comparisons. Tx ⫽treatment model; W ⫽
Worthington, E ⫽Enright, and O ⫽other treatment model. Mode ⫽treatment modality; g ⫽group, c ⫽couples, and i ⫽
individual therapy modality. Sev ⫽offense severity. Hrs ⫽hours; CI ⫽confidence interval; RE ⫽random effects.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8WADE, HOYT, KIDWELL, AND WORTHINGTON
dummy coded, with Enright-model interventions as the reference
group for the treatment model variable and individual interventions as
the reference group for the treatment modality variable.
For T–NT comparisons, only dosage (i.e., treatment duration in
hours; B⫽0.033) and modality (specifically, the contrast between
group and individual treatment modes; B⫽⫺0.57) emerged as
unique moderators of study effects. When dosage and modality were
controlled, treatment model was not a significant predictor of study
effect size. This means that the advantage for Enright-model inter-
ventions observed in the single-moderator analyses was an artifact of
the confounding of those study characteristics with treatment dosage
and treatment modality. Enright-model interventions tended to be
substantially longer in duration than Worthington-model interventions
(Ms⫽15.4 hr and 5.6 hr, SDs⫽12.3 and 3.2, respectively). In
addition, nearly one third of studies using the Enright model involved
individual interventions, compared with none of the studies using the
Worthington model. When these differences were controlled, the
intervention models did not differ in efficacy.
Figure 4 shows the practical import of the significant moderator
findings for T–NT comparisons (Johnson & Huedo-Medina,
2011). The three panels of this figure depict separate predicted
dose– effect trend lines for the three treatment modalities (group,
individual, and couples), along with 95% confidence bands for
these predicted effect sizes. In all three panels, the lower confi-
dence band is above the zero point on the yaxis, indicating that
even for very low dosages (i.e., a single hour of intervention) the
predicted effect size differs significantly from zero. Although all
effect sizes in the data set contribute to the estimation of each
meta-regression line, we highlight the points in each panel repre-
senting effect sizes for the relevant treatment modality, with the
data points representing other intervention modalities shown in
light gray. So we can see that, in the first panel, group interven-
tions lasting 1–5 hr are predicted to have relatively weak (although
still statistically significant) effects on forgiveness; interventions
of about 10 hr in duration should have a moderate effect (i.e., ⌬⫽
0.5), and those lasting 18 –20 hr a large effect (i.e., ⌬⫽0.8). By
comparison, an individual intervention lasting only 5 hr is pre-
dicted to have a large effect (⌬⫽0.8), with correspondingly
higher effect sizes for interventions of longer duration.
For T–AT comparisons, none of the moderators emerged as
significant (p⬍.05) in the multiple moderator analysis. Given the
lower power of these tests (k⫽21), we note two marginally
significant findings (p⬍.10), which should be interpreted cau-
tiously. Reflecting the T–NT studies, a marginally significant
advantage was observed for individual treatments, this time rela-
tive to couple treatments (B⫽–1.27, p⫽.067). In addition,
offense severity was found to be a marginally significant moder-
ator of effect size in these studies (B⫽0.045, p⫽.068), indicating
a trend for forgiveness interventions to show more of an advantage
over non-specific treatments when offense severity is higher.
Publication bias. We created funnel plots based on residuals
from the final models for both T–NT and T–AT studies to examine
the pattern of effect sizes for evidence of publication bias. The
residuals (plotted against their SEs) form an inverted funnel; when
publication bias is present, the base of the funnel (corresponding to
studies with the smallest Ns and therefore the largest SEs) may show
only extreme values (i.e., may show a dearth of residual values close
to zero). This pattern was not observed for either T–NT or T–AT
studies. In both cases, the funnel plot was symmetrical.
5
Follow-up data. Follow-up intervals for the different treat-
ments ranged considerably, from a low of 2 weeks to a high of 36
weeks (M⫽11.1, SD ⫽8.4). For the subset (k⫽18) of T–NT
studies that included a follow-up assessment for both treatment
and control groups, the effect size comparing change from baseline
to follow-up was ⌬⫽0.45, 95% CI [0.27, 0.62]. The baseline-to-
postintervention ⌬⫽0.39, 95% CI [0.22 to 0.55], and the
postintervention- to-follow-up ⌬⫽0.06, 95% CI [– 0.07, 0.18],
suggesting that treatment gains were maintained over the
follow-up interval. Moderator tests for this subset of T–NT com-
parisons (baseline to follow-up interval) yielded a significant ef-
fect only for intervention dosage: B⫽0.08, 95% CI [0.02, 0.13].
In Figure 5, we show separate average trajectories over time for
treatment (k⫽41) and control (k⫽18) groups that provided
follow-up assessments. For this larger subset of treatment groups,
g
T
⫽0.78, 95% CI [0.60, 0.95], from baseline to follow-up, with
no significant change in forgiveness from postintervention to
follow-up, g
T
⫽0.07, 95% CI [0 –.03, 0.16]. Improvement in the
control group was significant from baseline to follow-up, g
C
⫽
0.14, 95% CI [0.05, 0.23], and also from postintervention to
follow-up, g
C
⫽0.08, 95% CI [0.02, 0.14], but not over the shorter
interval between pre- and postassessments, g
C
⫽0.06, 95% CI
[0 –.02, 0.15]. In summary, the follow-up analyses suggest a pat-
tern of strong improvement in the treatment group postintervention
followed by maintenance of gains at the follow-up assessment,
with much slower but still significant increases in forgivingness
for the control group over time.
We also conducted moderator analyses for the effect size between
pretreatment and follow-up for studies including both treatment and
control data at these two time points. Treatment dosage was the only
significant moderator for this reduced (k⫽18) subset of studies,
⌬
⫹
⫽0.064, 95% CI [0.005, 0.123]. (None of the studies of individ-
5
Based on a request from a reviewer, we conducted significance tests for
asymmetry in these funnel plots, based on procedures recommended by
Egger, Smith, Schneider, and Minder (1997), regressing effect sizes con-
verted to standard normal deviates (zscores) onto the inverse of the
standard errors for these effect sizes. Sutton (2009) noted that this proce-
dure is not appropriate when significant moderators are observed and
proposed an alternative method (including these study characteristics as
covariates in the regression model) –although he noted that evaluation of
the performance of this approach “is an ongoing work” (p. 441). When
following Sutton’s proposal, we found no evidence of significant asym-
metry in these plots (Bs⫽⫺0.09 and ⫺0.39, ps⫽.47 and .36 for
no-treatment and alternative-treatment plots, respectively).
Table 1
Omnibus Effect Sizes and Heterogeneity Tests With Forgiveness
as a Dependent Variable
Comparison k⌬
⫹
95% CI QpI
2
No treatment 53 0.56 [0.43, 0.68] 188.91 ⬍.0001 72%
Alternative treatment 22 0.45 [0.21, 0.69] 72.39 ⬍.0001 77%
Note. Studies were modeled as random effects. k⫽number of stud-
ies; ⌬
⫹
⫽effect size (standardized mean difference controlling for prein-
tervention scores; Becker, 1988); CI ⫽confidence interval; Q⫽homo-
geneity test; p⫽probability value for Qstatistic under H
0
(df ⫽k⫺1);
I
2
⫽percentage of variance in effect sizes that is attributable to systematic
variation.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
FORGIVENESS META-ANALYSIS
ual or couples interventions included follow-up assessments, so mo-
dality could not be investigated as a moderator at follow-up.)
Additional Treatment Outcomes
In addition to assessing change in forgiveness, some studies
included measures of psychological symptoms often associated
with relational transgressions. Although those psychological
symptoms were not targeted in the intervention, it is reasonable to
hypothesize that if forgiveness was promoted, perhaps mental
health symptoms would be mitigated. The most common of these
symptoms were depression, anxiety, and hopelessness. Each of
these outcomes was measured in only a subset of studies, so to
understand the relative efficacy of forgiveness interventions for
these other common outcomes, we compared the aggregate effect
size for each outcome to the corresponding aggregate effect size
for forgiveness in the same subset of studies. All outcomes were
scaled so that a positive effect size reflects greater improvement
for the intervention group relative to no-treatment controls.
For the subset of studies measuring depression (k⫽10), the
aggregate effect size for comparing change in depression between
the forgiveness treatment and no treatment conditions was ⌬
⫹
⫽
0.34, 95% CI [0.17, 0.52]. In this same group of studies, the
aggregate effect size for forgiveness was ⌬
⫹
⫽0.60, 95% CI
[0.26, 0.94]. The effect size for depression in these studies was
43% smaller than the effect size for forgiveness. Despite this
numerical difference, a null hypothesis of no difference in out-
comes for these two variables could not be rejected (p⫽.09). For
the subset of studies measuring anxiety (k⫽7), the aggregate
effect size for anxiety was ⌬
⫹
⫽0.63, 95% CI [0.0003, 1.26]. In
this same group of studies, the aggregate effect size for forgiveness
was ⌬
⫹
⫽1.34, 95% CI [0.55, 2.12]. The effect size for anxiety
was 50% lower than that for forgiveness; however this average
difference was not significantly different from zero (p⫽.21). For
the subset of studies measuring hope (k⫽6), the aggregate effect
size for hope was ⌬
⫹
⫽1.00, 95% CI [0.38, 1.62]. For these same
studies, the aggregate effect size for forgiveness was ⌬
⫹
⫽0.94,
95% CI [0.16, 1.73]. Again, the difference was not statistically
significant (p⫽.96).
In summary, forgiveness interventions, although not targeting
mental health symptoms directly, resulted in reductions in depres-
Table 2
Significant Single-Moderator Analyses—Continuous Moderators
Variable kB
0
B
1
95% CI (B
1
)z(B
1
)p
Treatment–no treatment comparisons
Dosage (hours) 51 0.102 0.047 [0.029, 0.064] 5.33 ⬍.0001
Offense severity 47 0.120 0.012 [0.002, 0.022] 2.25 .024
Treatment–alternative treatment comparisons
Dosage (hours) 21 ⫺0.024 0.043 [0.009, 0.077] 2.45 .014
Offense severity 21 ⫺1.072 0.041 [0.017, 0.065] 3.30 .001
Note. Univariate analyses used a mixed model (studies random, levels of moderator variables fixed). Signif-
icant categorical moderators are tabulated separately. k⫽number of studies; B
0
⫽intercept; B
1
⫽slope; CI ⫽
confidence interval; z(B
1
)⫽zstatistic for B
1
.
Table 3
Significant Single-Moderator Analyses–Categorical Moderators
Variable k⌬
⫹
95% CI Qdf p I
2
Treatment–no treatment comparisons
Intervention model 8.54 2 .014
Enright 20 0.82
a
[0.60, 1.03] 61.10 19 ⬍.0001 69%
Worthington 18 0.35
b
[0.16, 0.54] 20.53 17 .248 17%
Other 14 0.55
ab
[0.33, 0.77] 78.30 13 ⬍.0001 83%
Treatment mode 16.44 2 .0003
Individual 6 1.44
a
[0.99, 1.89] 14.03 5 .049 64%
Couple 6 0.75
b
[0.44, 1.06] 16.60 5 .005 70%
Group 40 0.44
b
[0.31, 0.56] 115.59 39 ⬍.0001 65%
Treatment–alternative treatment comparisons
Intervention model 8.00 2 .019
Enright 11 0.78
a
[0.46, 1.10] 44.60 10 ⬍.0001 78%
Worthington 9 0.17
b
[⫺0.11, 0.46] 11.09 8 .197 28%
Other 2 0.26
ab
[⫺0.37, 0.89] 0.08 1 .773 0%
Treatment mode 11.28 2 .004
Individual 2 1.98
a
[1.03, 2.92] 0.52 1 .47 0%
Couple 3 0.22
b
[⫺0.27, 0.71] 0.27 2 .87 0%
Group 17 0.37
b
[0.15, 0.60] 53.80 21 .001 70%
Note. Univariate analyses used a mixed model (studies random, levels of moderator variables fixed). Means that do not share a subscript differ
significantly (p⬍.05). Significant continuous moderators are tabulated separately. k⫽number of studies; ⌬
⫹
⫽effect size; CI ⫽confidence interval;
Q⫽homogeneity test. Qfor the moderator assesses homogeneity between groups; Qs for the levels assess homogeneity within groups.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 WADE, HOYT, KIDWELL, AND WORTHINGTON
sion and anxiety and increases in hope; all effect sizes for mental
health symptoms did not contain zero. Effect sizes comparing
forgiveness interventions with no treatment ranged from .34 for
depression to 1.00 for hope, indicating that forgiveness interven-
tions may also help clients with psychological outcomes other than
forgiveness. Direct comparisons of the effects for hope and the
mental health symptoms showed that there were no significant
differences between effects for these outcomes and effects for
forgiveness. However, the significance test results must be inter-
preted with caution because of the small number of studies (and
consequent lack of statistical power). Based on the numerical
differences, effects of forgiveness interventions for reducing neg-
ative affect (depression and anxiety) were 40%–50% lower than
those for forgiveness in the same subset of studies. Effects for
increasing hope were similar in magnitude to those for forgiveness.
Discussion
Several notable findings emerged from this meta-analysis.
First, interventions designed to promote forgiveness are more
effective at helping participants achieve forgiveness and hope
and reduce depression and anxiety than either no treatment or
alternative treatments. Additionally, the specific treatment
model used did not make a difference in outcomes. From our
Table 4
Multiple-Moderator Analyses—Significant Single Predictors Only
Variable B95% CI p
Treatment–no treatment comparisons
Intercept 0.92
ⴱ
[0.15, 1.70] .020
Dosage (hours) 0.033
ⴱ
[0.012, 0.055] .002
Offense severity ⫺0.001 [⫺0.013, 0.012] .873
Treatment model
Worthington vs. Enright ⫺0.19 [⫺0.52, 0.14] .270
Other vs. Enright ⫺0.25 [⫺0.58, 0.09] .150
Modality
Group vs. individual ⫺0.57
ⴱ
[⫺1.05, ⫺0.08] .021
Couple vs. individual ⫺0.31 [⫺0.84, 0.22] .251
Treatment–alternative treatment comparisons
Intercept ⫺0.25 [⫺2.77, 2.28] .849
Dosage (hours) ⫺0.01 [⫺0.057, 0.036] .658
Offense severity 0.045
a
[⫺0.003, 0.093] .068
Treatment model
Worthington vs. Enright ⫺0.03 [⫺0.69, 0.63] .923
Other vs. Enright ⫺0.14 [⫺1.32, 1.03] .810
Modality
Group vs. individual ⫺0.79 [⫺2.00, 0.42] .201
Other vs. Enright ⫺1.27
a
[⫺2.63, 0.09] .067
Note. Overall tests of model significance were Q(6) ⫽35.74 and 17.82; ps⫽⬍.0001 and ⫽.007; ks⫽46
and 21 for no-treatment and alternative-treatment comparisons, respectively. CI ⫽confidence interval.
a
Marginally significant (.05 ⬍p⬍.10).
ⴱ
p⬍.05.
Figure 4. Dose– effect relationship for three treatment modalities (treatment vs. no treatment comparisons):
group, individual, and couples interventions. Del ⫽delta.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
FORGIVENESS META-ANALYSIS
results, it appears that using theoretically grounded forgiveness
interventions is a sound choice for helping clients to deal with
past offenses and helping them achieve resolution in the form of
forgiveness.
Second, treatment dosage was an important predictor of forgive-
ness as an outcome when comparing forgiveness interventions to
no treatment, though individual treatment (vis-a-vis group treat-
ment) also makes a difference. As measured in this study, treat-
ment duration was the amount of time that therapists worked
specifically with the participants to help them forgive. This fits
with findings from past meta-analyses (e.g., Worthington,
Sandage, & Berry, 2000), research on typical patterns of forgive-
ness in college students (e.g., McCullough, Fincham, & Tsang,
2003), and similar changes following various interventions to
promote forgiveness (e.g., Wade et al., 2009). With more time,
people are generally able to develop more forgiveness, even those
not receiving a forgiveness intervention (McCullough, Luna,
Berry, Tabak, & Bono, 2010). For those who are receiving for-
giveness treatment, shorter interventions promote less forgiveness
than do longer interventions. However, the specific forgiveness
model does not seem to make a difference when duration of
treatment and modality are controlled. The relationship between
duration and effect size also seems to account for other potential
moderators such as severity of the offense.
Another finding of note was that offense severity was positively
correlated with forgiveness as an outcome for the forgiveness
versus alternative treatment comparisons. One possible reason for
this correlation is that a confound exists between severity and
duration of treatment. Severe transgressions tend to be treated
longer. The results of the multiple moderator analyses do not fully
support this explanation; a relationship between forgiveness out-
come and offense severity was still suggested (p⫽.068) after
including treatment dosage in the prediction model. If a relation-
ship between offense severity and forgiveness outcomes does
exist, another factor that might explain this relationship is that
those who were more severely offended may have had more room
to change in terms of forgiveness. If more severe offenses result in
less forgiveness (which basic research supports they do; e.g.,
Fincham et al., 2005), then those with more severe offenses at the
start of the intervention may have had the opportunity for greater
changes in forgiveness than those who experienced less severe
offenses. In addition, those with more severe offenses may have
responded more positively to the explicit forgiveness interventions
than those with less severe offenses. Perhaps those who have been
more dramatically hurt may need more focused attention on the
hurt and the healing and forgiveness process.
Finally, in the analyses of studies that included follow-up data
collection, the overall delta estimating change in forgiveness in-
dicated that on average clients achieve about .78 standard devia-
tions of change at post-treatment and maintain that change at
follow-up (see Figure 4). Furthermore, these changes appear to
persist over time, suggesting that not only do forgiveness inter-
ventions help clients achieve forgiveness but that forgiveness is
maintained following treatment (e.g., Blocher & Wade, 2010).
Implications
The present findings have several implications. First, many
psychotherapists use forgiveness interventions, presumably be-
cause interpersonal difficulties are prevalent in most counseling.
Thus, focus on forgiveness can be expected to provide not only an
experience of increased forgiveness, but it can also provide psy-
chotherapeutic benefit in treatment of depression and anxiety, and
it can provide a benefit of hope, illustrating that forgiveness
interventions might not only help to remediate problems (e.g.,
depression) but enhance human functioning as well.
Second, it seems not to matter as much which program is
employed (i.e., Enright’s, Worthington’s, or some other) as how
long the psychotherapist and client work on forgiving. That is not
to say necessarily that any treatment is equally efficacious for
developing forgiveness. Genuine forgiveness interventions
showed clearly superior efficacy over alternative treatments. Un-
fortunately, the alternative treatments were not one single alterna-
tive treatment, but included a range of treatments, some of which
were true alternative psychotherapies and some not. Still on aver-
age, explicit forgiveness interventions were more effective at pro-
moting forgiveness than were the alternatives. Individual research
projects indicate that forgiveness interventions might be more
effective than typical psychotherapeutic interventions (e.g., Lin et
al, 2004; Reed & Enright 2006) for dealing with some problems.
Still only a few psychotherapeutic treatments that have been tested
against explicit forgiveness interventions, so it is currently impos-
sible to draw definitive conclusions about the superior efficacy of
explicit forgiveness interventions.
Third, treatment duration seems to be a crucial element in
promoting forgiveness and other mental health benefits associated
with forgiveness interventions. In most general psychotherapy,
attention to forgiveness takes a minor part of psychotherapy—
perhaps as little as 2 or 3 hr (DiBlasio & Benda, 1991). Given the
strong relationship between time spent in explicit forgiveness
intervention and the promotion of forgiveness, general psychother-
apy might be supplemented by adjunctive forgiveness-promoting
treatments such as psychoeducational groups. General psychoedu-
cational groups to promote forgiveness can include people with a
variety of interpersonal transgressions within any group. Thus,
adjunctive psychoeducational groups could be extend the amount
of forgiveness and benefit to improving depression, anxiety, and
hope. This is especially important given that modality of delivery
(i.e., to individuals, couples, or groups) seemed to matter little in
the amount of benefit participants derived.
Figure 5. Growth in forgiveness for treatment (T) and no-treatment
control (C) groups, baseline to follow-up.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 WADE, HOYT, KIDWELL, AND WORTHINGTON
Fourth, the advantage of individual interventions over the more
common group modalities was apparent for T–NT comparisons,
and individual interventions were marginally superior to couples,
but not group, interventions for T–AT comparisons. Relatively few
studies have examined the effectiveness of individual counseling
for forgiveness, and of the six we located for this meta-analysis,
five (83%) used the Enright model (the sixth was a test of emotion-
focused therapy to promote forgiveness; Greenberg et al., 2008).
Thus, while these results are important and suggest that further
study of individual forgiveness interventions is warranted, one
should be cautious about generalizing beyond the types of treat-
ments examined here. Perhaps the surest conclusion is that indi-
vidual implementations of the Enright model have been substan-
tially more effective than group implementations (although the
group implementations themselves lead to significant improve-
ment relative to what would be expected in the absence of treat-
ment).
Fifth, questions arise about the active ingredients within for-
giveness treatments. Wade and colleagues (2005) tried to describe
the effects of seven elements that were roughly modeled on Wor-
thington’s REACH steps (plus explicitly defining forgiveness and
incorporating other interventions like relaxation or anger-
management methods). Given the strong dose–response relation-
ship we have found, it is reasonable to inquire whether any
particular technique contributes to outcomes more than any other.
Yet common sense suggests that there would be some aspects of
any treatment that are not therapeutically active. Future research
must determine what might be omitted from treatment protocols to
provide a more time-efficient intervention.
Sixth, much is still unknown about the nature of the trans-
gressions that are most appropriate as targets for a forgiveness
intervention. Is there an optimal time to intervene after a
transgression is experienced? Is there an optimal amount of
severity of initial harm? Given that time seems to decrease
unforgiving emotions and motivations following a power law
(McCullough et al., 2010), is there an optimal amount of
residual unforgiveness that would be most responsive (or least
responsive) to an intervention? Sixth, no effort was made to
evaluate which treatments, if any, would be particularly effec-
tive for which types of transgression. For example, the emo-
tionally focused treatment (EFT) by Greenberg, Warwar, and
Malcolm (2008, 2010) might be particularly effective for cou-
ples in couple therapy, given the status of EFT as empirically
supported (see Baucom, Shoham, Mueser, & Daituo, 1998), and
Worthington’s REACH model might be particularly effective
for couple enrichment, given the status of his hope-focused
couple approach as an empirically supported couple enrichment
intervention (see Jakubowski, Milne, Brunner, & Miller, 2004).
However, this is speculation because this was not tested in the
present study. We note that Enright’s model has successfully
been used as psychotherapy for several severe problems. Even
though statistically significant, our regression analyses show
that duration of treatment might be the key variable in treatment
success, we must note that Worthington’s model has not been
tested in long interventions with severe offenses. Thus, research
is needed to test whether it would actually be as successful as
Enright’s or other models with such severe problems and long
durations.
Limitations
Although we located an adequate set of studies that measured
forgiveness as an outcome, the outcomes of depression, anxiety,
and hope were reported less often. Therefore, our analyses assess-
ing the effects of the interventions on these variables were limited.
Although we were able to assess change over time, there were not
enough studies to conduct the same moderator analyses used with
forgiveness for these outcomes. The results showing change in
depression, anxiety, and hope are based on a smaller, more select
set of studies and therefore should be viewed with caution. Like-
wise, the follow-up assessments for the control groups, especially
the no treatment conditions, were limited. Although a respectable
percentage of studies included follow-up assessments of the treat-
ment conditions, fewer reported follow-up data for control partic-
ipants who often entered the treatment phase following the post-
assessment. Therefore, this limits our confidence in our results
about how people who are not in treatment or who are waiting for
treatment change in terms of forgiveness over a longer time.
Another limitation of this review is that the only measures used
to assess forgiveness in these studies were self-report measures.
Although self-report measures are crucial for assessing internal
and subjective experiences such as forgiveness, these measures
may include biases from socially desirable responding or halo
effects. Related to this limitation is the limitation of the offense
severity rating system. Although we had strong reliability across
raters, the degree of severity was grouped by study, making it a
much broader measure than would be the case if individual client
offenses were rated. Unfortunately, we did not have that data for
most studies and could not provide that level of detail. Third, our
measure of methodological quality was limited by the scope of the
questions we assessed. Although our measures were not related to
effect size, this does not mean that methodological quality was
unrelated to outcomes. If assessed with different measures, effect
size might be related to quality. In addition, we used methodolog-
ical quality as a predictor of effect size whereas it could be used as
a cut-off for only including studies in the meta-analysis with a
certain level of quality. Because we wanted to cast a wide net on
this literature and because quality was not related to effect size in
our analyses, we included all studies that met our initial criteria.
A fourth limitation is that the moderator analyses are only
correlational in nature. These were not included as part of an
experimental design that would provide evidence of causation.
Instead, the moderators we examined were only correlated with
outcome and therefore do not indicate causation. For example,
treatment duration is certainly related to outcome but we, cannot
say from these data alone that the time spent intervening caused
larger effect sizes. Finally, the failure of some studies to include a
control or comparison group (i.e., use of the SGPP design) is a
serious weakness, although we believe this design limitation can
be largely overcome using the imputation procedures recom-
mended by Becker (1988). This places severe restrictions on the
ability to draw causal conclusions about improvement for these
participants based on the findings of a single study, as the SGPP
design fails to rule out several important threats to internal validity.
However, the meta-analysis of control group improvement pre-
sented here provides a basis for comparison for these studies that
helps to overcome this limitation and can eliminate the bias found
by Carlson & Schmidt (1999) in simple pre-post effect sizes. The
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
FORGIVENESS META-ANALYSIS
logic of this method is similar to that of benchmarking (using data
from large-scale clinical trials as a basis for comparison; e.g.,
Minami, Serlin, Wampold, Kircher, & Brown, 2008), which is
another technique for enhancing the validity of conclusions drawn
from client data in the absence of statistical controls.
Future Research
Based on this review and meta-analysis, the results of research
on basic outcome studies of individual and group treatments are
apparent and robust; forgiveness interventions are helpful for
many people and many different kinds of hurts. However, beyond
the most basic questions, few questions have been answered.
Specifically, researchers might conduct intervention studies that
examine the relationship of counseling processes to forgiveness
outcomes, comparing and contrasting what has already been es-
tablished as important processes in the more general psychother-
apy literature. In addition, future research could further specify the
active ingredients of these interventions, focusing on common
versus specific mechanisms of change (Wampold, 2001). Under-
standing who most benefits from these interventions is also a
question that has not been thoroughly examined and would be very
helpful for mental health professionals. Specifically, work with
minorities (racial/ethnic, religious, sexual orientation) could be
especially valuable to understand potential interactions between
social justice, advocacy, and forgiveness intervention efforts. For
example, researchers might examine interventions that help
lesbian-gay-bisexual-transgender clients forgive experiences of
discrimination in a way that promotes their individual mental
health but does not limit their motivation to work for social change
and to seek justice for themselves and others. Finally, forgiveness
interventions provided in couple formats have not received near
the research attention that individual and group modalities have.
Therefore, more research into couple therapy that explicitly pro-
motes forgiveness (e.g., Makinen & Johnson, 2006) and more
general couple therapy would help to develop our understanding
more in this important area.
In addition, future research utilizing different methods to mea-
sure forgiveness would help to advance the field. Although for-
giveness is in many ways a subjective and internal experience,
careful operationalization of forgiveness definitions and creative
methodology to assess forgiveness in ways other than client self-
report would be beneficial. Some intervention studies in the past
have used an observer report (e.g., romantic partner, close friend;
Rye et al., 2005). Measurements of behaviors associated with
forgiveness might also be useful to validate self-reported forgive-
ness and add additional dimensions to the existing outcome stud-
ies.
Because forgiveness intervention appears successful in relieving
depression and anxiety and promoting hope, researchers should
consider including other psychological variables in future studies
of forgiveness interventions. These might include posttraumatic
stress, hostility, self-control, relationship satisfaction, well-being,
spirituality, and job performance. Even physiological responses
such as heart rate (and other peripheral physiological measures),
heart rate variability (as a sympathetic nervous system variable), or
brain activity might be fruitful avenues to pursue following for-
giveness interventions. Because forgiveness seems to have a
marked effect on people, it is likely that there are other factors
associated with forgiveness and receiving a forgiveness interven-
tion. In addition, research on other psychological interventions
might include the development of forgiveness as a potential me-
diator or moderator of mental health. For example, researchers
examining the efficacy of treatments for posttraumatic stress dis-
order (PTSD) could examine the development of forgiveness for
an offender as a mediator of improved mental health (i.e., PTSD
treatment improves forgiveness which in term reduces anxiety,
intrusive thoughts, and the like). Future research could explore
these and other ancillary benefits of forgiveness in more detail and
specificity, which would further delineate possible benefits of
receiving forgiveness interventions.
Conclusion
Overall, the status of the research to date suggests that forgive-
ness is a viable and evidence-based treatment for dealing with
transgressions. These interventions are more effective than alter-
native treatments and no treatment in promoting forgiveness of the
offender and hope for the future and reducing depression and
anxiety. Results from this meta-analysis indicate that forgiveness
treatments are robust and effects are maintained following the
termination of the treatment. Although not enough research has
been conducted to answer various specific questions about the
efficacy of forgiveness interventions, it appears that the duration of
the treatment is directly linked to amount of forgiveness achieved,
whereas the specific treatment packages (e.g., Enright, Worthing-
ton) and modalities (e.g., individual, group) do not differentially
predict outcome.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
ⴱ
Al-Mabuk, R. H., Enright, R. D., & Cardis, P. A. (1995). Forgiveness
education with parentally love-deprived late adolescents. Journal of
Moral Education, 24, 427– 444. doi:10.1080/0305724950240405
ⴱ
Alvaro, J. A. (2001). An interpersonal forgiveness and reconciliation
intervention: The effect on marital intimacy. Dissertation Abstracts
International: Section B. Sciences and Engineering, 62(3-B), 1608.
Baskin, T. W., & Enright, R. D. (2004). Intervention studies on forgive-
ness: A meta-analysis. Journal of Counseling & Development, 82, 79 –
90. doi:10.1002/j.1556-6678.2004.tb00288.x
ⴱ
Baskin, T. W., Rhody, M., Schoolmeesters, S., & Ellingson, C. (2011).
Supporting special-needs adoptive couples: Assessing an intervention to
enhance forgiveness, increase marital satisfaction, and prevent depres-
sion. The Counseling Psychologist, 39, 933–955. doi:10.1177/
0011000010397554
Baucom, D. H., Shoham, V., Mueser, K. T., & Daiuto, A. D. (1998).
Empirically supported couple and family interventions for marital dis-
tress and adult mental health problems. Journal of Consulting and
Clinical Psychology, 66, 53– 88. doi:10.1037/0022-006X.66.1.53
ⴱ
Beck, S. (2006). Efficacy of a forgiveness group intervention for aggres-
sive victims. Dissertation Abstracts International: Section B. Sciences
and Engineering, 66(8-B), 4512.
Becker, B. J. (1988). Synthesizing standardized mean-change measures.
British Journal of Mathematical and Statistical Psychology, 41, 257–
278. doi:10.1111/j.2044-8317.1988.tb00901.x
ⴱ
Blocher, W. G., & Wade, N. G. (2010). Sustained effectiveness of two
brief group interventions: Comparing an explicit forgiveness-promoting
treatment with a process-oriented treatment. Journal of Mental Health
Counseling, 32, 58 –74.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
14 WADE, HOYT, KIDWELL, AND WORTHINGTON
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009).
Introduction to meta-analysis. New York, NY: Wiley. doi:10.1002/
9780470743386
Carlson, K. D., & Schmidt, F. L. (1999). Impact of experimental design on
effect size: Findings from the research literature on training. Journal of
Applied Psychology, 84, 851– 862. doi:10.1037/0021-9010.84.6.851
ⴱ
Coyle, C. T., & Enright, R. D. (1997). Forgiveness intervention with
postabortion men. Journal of Consulting and Clinical Psychology, 65,
1042–1046. doi:10.1037/0022-006X.65.6.1042
Del Re, A. C., & Hoyt, W. T. (2010). MAd: Meta-analysis With Mean
Differences (R package, Version 8) [Computer software]. Retrieved
from http://CRAN.R-project.org/package⫽MAd
DiBlasio, F. A., & Benda, B. B. (1991). Practitioners, religion, and the use
of forgiveness in the clinical setting. Journal of Psychology and Chris-
tianity, 10, 166 –172.
ⴱ
DiBlasio, F. A., & Benda, B. B. (2008). Forgiveness intervention with
married couples: Two empirical analyses. Journal of Psychology and
Christianity, 27, 150 –158.
Downs, S. H., & Black, N. (1998). The feasibility of creating a checklist for
the assessment of the methodological quality both of randomised and
non-randomised studies of health care interventions. Journal of Epide-
miological Community Health, 52, 377–384. doi:10.1136/jech.52.6.377
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in
meta-analysis detected by a simple, graphical test. BMJ: British Medical
Journal, 315, 629 – 634. doi:10.1136/bmj.315.7109.629
Enright, R. D. (2001). Forgiveness is a choice: A step-by-step process for
resolving anger and restoring hope. Washington, DC: American Psy-
chological Association.
Enright, R. D., & Fitzgibbons, R. P. (2000). Helping clients forgive: An
empirical guide for resolving anger and restoring hope. Washington,
DC: American Psychological Association. doi:10.1037/10381-000
Exline, J. J., Worthington, E. L., Jr., Hill, P. C., & McCullough, M. E.
(2003). Forgiveness and justice: A research agenda for social and per-
sonality psychology. Personality and Social Psychology Review, 7,
337–348. doi:10.1207/S15327957PSPR0704_06
Fehr, R., Gelfand, M. J., & Nag, M. (2010). The road to forgiveness: A
meta-analytic synthesis of its situational and dispositional correlates.
Psychological Bulletin, 136, 894 –914. doi:10.1037/a0019993
Fincham, F. D., Jackson, H., & Beach, S. H. R. (2005). Transgression
severity and forgiveness: Different moderators for objective and subjec-
tive severity. Journal of Social and Clinical Psychology, 24, 860 – 875.
doi:10.1521/jscp.2005.24.6.860
ⴱ
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. doi:10.1037/0022-006X.64.5.983
ⴱ
Freedman, S., & Knupp, A. (2003). The impact of forgiveness on ado-
lescent adjustment to parental divorce. Journal of Divorce & Remar-
riage, 39, 135–165. doi:10.1300/J087v39n01_08
ⴱ
Gambaro, M. E. (2002). School-based forgiveness education in the man-
agement of trait anger in early adolescents. Dissertation Abstracts In-
ternational: Section B. Sciences and Engineering, 63(11-B), 5549.
ⴱ
Gassin, E. A. (1995). Social cognition and forgiveness in adolescent
romance: An intervention study. Dissertation Abstracts International:
Section A. Humanities and Social Sciences, 56(4-A), 1290.
Gelso, C. J., & Woodhouse, S. (2003). Toward a positive psychotherapy:
Focus on human strength. In B. W. Walsh (Ed.), Counseling psychology
and optimal human functioning: Contemporary topics in vocational
psychology (pp. 171–197). Mahwah, NJ: Erlbaum.
ⴱ
Goldman, D. B., & Wade, N. G. (2012). Comparison of group interven-
tions to promote forgiveness: A randomized controlled trial. Psycho-
therapy Research, 22, 604 – 620. doi:10.1080/10503307.2012.692954
ⴱ
Greenberg, L. S., Warwar, S. H., & Malcolm, W. M. (2008). Differential
effects of emotion-focused therapy and psychoeducation in facilitating
forgiveness and letting go of emotional injuries. Journal of Counseling
Psychology, 55, 185–196. doi:10.1037/0022-0167.55.2.185
ⴱ
Greenberg, L. S., Warwar, S. H., & Malcolm, W. M. (2010). Emotion-
focused couples therapy and the facilitation of forgiveness. Journal of
Marital and Family Therapy, 36, 28 – 42. doi:10.1111/j.1752-0606.2009
.00185.x
ⴱ
Harris, A. H., Luskin, F., Norman, S. B., Standard, S., Bruning, J., Evans,
S., & Thoresen, C. E. (2006). Effects of a group forgiveness interven-
tions on forgiveness, perceived stress, and trait-anger. Journal of Clin-
ical Psychology, 62, 715–733. doi:10.1002/jclp.20264
ⴱ
Hart, K. E., & Shapiro, D. A. (2002, August). Secular and spiritual
forgiveness interventions for recovering alcoholics harboring grudges.
Paper presented at the annual convention of the American Psychological
Association, Chicago, IL.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
Orlando, FL: Academic Press.
Hedges, L. V., & Vevea, J. L. (1996). Estimating effect size under publi-
cation bias: Small sample properties and robustness of a random effects
selection model. Journal of Educational and Behavioral Statistics, 21,
299 –332. doi:10.2307/1165338
ⴱ
Hepp-Dax S. H. (1996). Forgiveness as an educational goal with fifth-
grade inner-city children (Unpublished doctoral dissertation). Fordham
University, New York, NY.
Higgins, J. P. T., & Altman, D. G. (2008). Assessing risk of bias in
included studies. In J. P. T. Higgins & S. Green (Eds.), Cochrane
handbook for systematic reviews of interventions (Version 5.0.1). Re-
trieved from www.cochrane-handbook.org
Holmes, T. H., & Rahe, R. H. (1967). The Social Readjustment Rating
Scale. Journal of Psychosomatic Research, 11, 213–218. doi:10.1016/
0022-3999(67)90010-4
Howard, K. I., Kopta, S. M., Krause, M. S., & Orlinsky, D. E. (1986). The
dose– effect relationship in psychotherapy. American Psychologist, 41,
159 –164. doi:10.1037/0003-066X.41.2.159
ⴱ
Hui, E. K. P., & Chau, T. S. (2009). The impact of a forgiveness
intervention with Hong Kong Chinese children hurt in interpersonal
relationships. British Journal of Guidance & Counselling, 37, 141–156.
doi:10.1080/03069880902728572
ⴱ
Humphrey, C. W. (1999). A stress management intervention with forgive-
ness as the goal. Dissertation Abstracts International: Section B. Sci-
ences and Engineering, 60(4-B), 1855.
ⴱ
Ingersoll-Dayton, B., Campbell, R., & Ha, J. (2008). Enhancing forgive-
ness: A group intervention for the elderly. Journal of Gerontological
Social Work, 52, 2–16. doi:10.1080/01634370802561901
ⴱ
Jackson, R. E. (1999). Reducing shame through forgiveness and empathy:
A group therapy approach to promoting prosocial behavior. Dissertation
Abstracts International: Section B. Sciences and Engineering, 60(4-B),
1856.
Jakubowski, S. F., Milne, E. P., Brunner, H., & Miller, R. B. (2004). A
review of empirically supported marital enrichment programs. Family
Relations, 53, 528 –536. doi:10.1111/j.0197-6664.2004.00062.x
Johnson, B. T., & Huedo-Medina, T. B. (2011). Depicting estimates using
the intercept in meta-regression models: The moving constant technique.
Research Synthesis Methods, 2, 204 –220. doi:10.1002/jrsm.49
ⴱ
Klatt, J. (2008). Testing a forgiveness intervention to treat aggression
among adolescents in a Type 1 correctional facility: A pilot study.
Dissertation Abstracts International: Section A. Humanities and Social
Sciences, 69(5-A), 1670.
Krumrei, E. J., Mahoney, A., & Pargament, K. I. (2011). Spiritual stress
and coping model of divorce: A longitudinal study. Journal of Family
Psychology, 25, 973–985. doi:10.1037/a0025879
ⴱ
Lampton, C., Oliver, G. J., Worthington, E. L., & Berry, J. W. (2005).
Helping Christian college students become more forgiving: An interven-
tion study to promote forgiveness as part of a program to shape Christian
character. Journal of Psychology and Theology, 33, 278 –290.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
15
FORGIVENESS META-ANALYSIS
ⴱ
Lin, W. N. (1998). Forgiveness as an intervention for late adolescents with
insecure attachment in Taiwan. Dissertation Abstracts International:
Section B. Sciences and Engineering, 59(5-B), 2456.
ⴱ
Lin, W. F. (2010). Treatment of substance abuse disorders by the psy-
chological forgiveness. Bulletin of Educational Psychology, 41, 859 –
883.
ⴱ
Lin, W. F., Mack, D., Enright, R. D., Krahn, D., & Baskin, T. W. (2004).
Effects of forgiveness therapy on anger, mood, and vulnerability to
substance use among inpatient substance-dependent clients. Journal of
Consulting and Clinical Psychology, 72, 1114 –1121. doi:10.1037/0022-
006X.72.6.1114
Lipsey, M. W., & Wilson, D. B. (1993). The efficacy of psychological,
educational, and behavioral treatment: Confirmation from meta-analysis.
American Psychologist, 48, 1181–1209. doi:10.1037/0003-066X.48.12
.1181
ⴱ
Louden-Gerber, G. M. (2009). A group forgiveness intervention for adult
male homeless individuals: Effects on forgiveness, rumination, and
social connectedness. Dissertation Abstracts International: Section A.
Humanities and Social Sciences, 69(12-A), 4640.
Luskin, F. (2007). Forgive for love: The missing ingredient for a healthy
and lasting relationship. San Francisco, CA: HarperOne/HarperCollins.
ⴱ
Luskin, F. M., & Bland, B. (2000). Stanford-Northern Ireland HOPE–1
Project. Unpublished manuscript, Department of Education, Stanford
University, Palo Alto, CA.
ⴱ
Luskin, F., & Bland, B. (2001). Stanford-Northern Ireland HOPE-2
Project. Unpublished manuscript, Department of Education, Stanford
University, Palo Alto, CA.
ⴱ
Luskin, F. M., Ginzburg, K., & Thoresen, C. E. (2005). The efficacy of
forgiveness intervention in college age adults: Randomized controlled
study. Humboldt Journal of Social Relations, 29, 163–184.
ⴱ
Makinen, J. A., & Johnson, S. M. (2006). Resolving attachment injuries in
couples using emotionally focused therapy: Steps toward forgiveness
and reconciliation. Journal of Consulting and Clinical Psychology, 74,
1055–1064. doi:10.1037/0022-006X.74.6.1055
McCullough, M. E., Fincham, F. D., & Tsang, J. (2003). Forgiveness,
forbearance, and time: The temporal unfolding of transgression-related
interpersonal motivations. Journal of Personality and Social Psychol-
ogy, 84, 540 –557. doi:10.1037/0022-3514.84.3.540
McCullough, M. E., & Hoyt, W. T. (2002). Transgression-related motiva-
tional dispositions: Personality substrates of forgiveness and their links
to the Big Five. Personality and Social Psychology Bulletin, 28, 1556 –
1573. doi:10.1177/014616702237583
McCullough, M. E., Luna, L. R., Berry, J. W., Tabak, B. A., & Bono, G.
(2010). On the form and function of forgiving: Modeling the time-
forgiveness relationship and testing the valuable relationships hypothe-
sis. Emotion, 10, 358 –376. doi:10.1037/a0019349
ⴱ
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. doi:
10.1002/j.2161-007X.1995.tb00387.x
ⴱ
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. doi:10.1037/0022-3514.73.2.321
Minami, T., Serlin, R. C., Wampold, B. E., Kircher, J. C., & Brown, G. S.
(2008). Using clinical trials to benchmark effects produced in clinical
practice. Quality and Quantity, 42, 513–525.
O’Brien, R. M. (2007). A caution regarding rules of thumb for variance
inflation factors. Quality and Quantity, 41, 673– 690.
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.
ⴱ
Osterndorf, C., Enright, R., Holter, A., & Klatt, J. (2011). Treating adult
children of alcoholics through forgiveness therapy. Alcoholism Treat-
ment Quarterly, 29, 274 –292. doi:10.1080/07347324.2011.586285
ⴱ
Park, J. (2003). Validating the effectiveness of a forgiveness intervention
program for adolescent female aggressive victims in Korea. Dissertation
Abstracts International: Section A. Humanities and Social Sciences,
64(5), 1528.
Rainey, C. A., Readdick, C. A., & Thyer, B. A. (2012). Forgiveness-based
group therapy: A meta-analysis of outcome studies published from
1993–2006. Best Practices in Mental Health: An International Journal,
8, 29 –51.
ⴱ
Reed, G. L., & Enright, R. D. (2006). The effects of forgiveness therapy
on depression, anxiety, and posttraumatic stress for women after spousal
emotional abuse. Journal of Consulting and Clinical Psychology, 74,
920 –929. doi:10.1037/0022-006X.74.5.920
ⴱ
Ripley, J. S., & Worthington, E. L., Jr. (2002). Hope-focused and
forgiveness-based group interventions to promote marital enrichment.
Journal of Counseling & Development, 80, 452– 463. doi:10.1002/j
.1556-6678.2002.tb00212.x
ⴱ
Rye, M. S., Fleri, A. M., Moore, C. D., Worthington, E. L., Wade, N. G.,
Sandage, S. J., & Cook, K. M. (2012). Evaluation of an intervention
designed to help divorced parents forgive their ex-spouse. Journal of
Divorce & Remarriage, 53, 231–245. doi:10.1080/10502556.2012
.663275
ⴱ
Rye, M. S., & Pargament, K. I. (2002). Forgiveness and romantic rela-
tionships in college: Can it heal the wounded heart? Journal of Clinical
Psychology, 58, 419 – 441. doi:10.1002/jclp.1153
ⴱ
Rye, M. S., Pargament, K. I., Pan, W., Yingling, D. W., Shogren, K. A.,
& Ito, M. (2005). Can group interventions facilitate forgiveness of an
ex-spouse? A randomized clinical trial. Journal of Consulting and Clin-
ical Psychology, 73, 880 – 892. doi:10.1037/0022-006X.73.5.880
ⴱ
Sandage, S. J., & Worthington, E. L., Jr. (2010). Comparison of two group
interventions to promote forgiveness: Empathy as a mediator of change.
Journal of Mental Health Counseling, 32, 35–57.
Seligman, M. E. P., Rashid, T., & Parks, A. C. (2006). Positive psycho-
therapy. American Psychologist, 61, 774 –788. doi:10.1037/0003-066X
.61.8.774
ⴱ
Sells, J. N., Giordano, F. G., & King, L. (2002). A pilot study in marital
group therapy: Process and outcome. The Family Journal, 10, 156 –166.
doi:10.1177/1066480702102005
ⴱ
Shechtman, Z., Wade, N. G., & Khoury, A. (2009). Effectiveness of a
forgiveness program for Arab Israeli adolescents in Israel: An empirical
trial. Peace and Conflict: Journal of Peace Psychology, 15, 415– 438.
doi:10.1080/10781910903221194
Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in
assessing rater reliability. Psychological Bulletin, 86, 420 – 428. doi:
10.1037/0033-2909.86.2.420
ⴱ
Stratton, S. P., Dean, J. B., Nonneman, A. J., Bode, R. A., & Worthington,
E. L., Jr. (2008). Forgiveness interventions as spiritual development
strategies: Comparing forgiveness workshop training, expressive writing
about forgiveness, and retested controls. Journal of Psychology and
Christianity, 27, 347–357.
Sutton, A. J. (2009). Publication bias. In H. Cooper, L. Hedges, & J. C.
Valentine (Eds.), The handbook of research synthesis and meta-analysis
(2nd ed., pp. 435– 452). New York, NY: Russell Sage Foundation.
ⴱ
Toussaint, L., Peddle, N., Cheadle, A., Sellu, A., & Luskin, F. (2010).
Striving for peace through forgiveness in Sierra Leone: Effectiveness of
a psychoeducational forgiveness intervention. In A. Kalayjian & E.
Dominique (Eds.), Mass trauma and emotional healing around the
world: Rituals and practices for resilience and meaning-making (Vol. 2,
pp. 251–267). Santa Barbara, CA: Praeger/ABC-CLIO.
ⴱ
Toussaint, L., Zoelzer, M., Worthington, E. L., Jr., & Luskin, F. (2010).
[Learning to forgive at Luther College: A randomized, controlled trial of
REACH and Forgive for Good]. Unpublished raw data.
ⴱ
Van Loon, P. C. (1998). A cognitive development intervention for clergy:
Forgiveness education. Dissertation Abstracts International: Section A.
Humanities and Social Sciences, 58(12-A), 4604.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
16 WADE, HOYT, KIDWELL, AND WORTHINGTON
Viechtbauer, W. (2007). Accounting for heterogeneity via random-effects
models and moderator analyses in meta-analysis. Zeitschrift für Psy-
chologie/Journal of Psychology, 215, 104 –121. doi:10.1027/0044-3409
.215.2.104
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor
package. Journal of Statistical Software, 36(1), 1– 48.
ⴱ
Wade, N. G., & Meyer, J. E. (2009). Comparison of brief group inter-
ventions to promote forgiveness: A pilot outcome study. International
Journal of Group Psychotherapy, 59, 199 –220. doi:10.1521/ijgp.2009
.59.2.199
Wade, N. G., & Worthington, E. L., Jr. (2003). Overcoming unforgiveness:
Is forgiveness the only way to deal with unforgiveness? Journal of
Counseling & Development, 81, 343–353. doi:10.1002/j.1556-6678
.2003.tb00261.x
ⴱ
Wade, N. G., Worthington, E. L., Jr., & Haake, S. (2009). Promoting
forgiveness: Comparison of explicit forgiveness interventions with an
alternative treatment. Journal of Counseling & Development, 87, 143–
151. doi:10.1002/j.1556-6678.2009.tb00561.x
Wade, N. G., Worthington, E. L., Jr., & Meyer, J. E. (2005). But do they
work? A meta-analysis of group interventions to promote forgiveness. In
E. L. Worthington, Jr. (Ed.), Handbook of forgiveness (pp. 423– 440).
New York, NY: Brunner/Routledge.
ⴱ
Waltman, M. A., Russell, D. C., Coyle, C. T., Enright, R. D., Holter,
A. C., & Swoboda, C. M. (2009). The effects of a forgiveness interven-
tion on patients with coronary artery disease. Psychology & Health, 24,
11–27. doi:10.1080/08870440801975127
Wampold, B. E. (2001). The great psychotherapy debate: Models, meth-
ods, and findings. Mahwah, NJ: Erlbaum.
Wampold, B. E., Minami, T., Tierney, S. C., Baskin, T. W., & Bhati, K. S.
(2005). The placebo is powerful: Estimating placebo effects in medicine
and psychotherapy from randomized clinical trials. Journal of Clinical
Psychology, 61, 835– 854. doi:10.1002/jclp.20129
Worthington, E. L., Jr. (2001). Five steps to forgiveness: The art and
science of forgiving. New York, NY: Crown.
Worthington, E. L., Jr. (2006). Forgiveness and reconciliation: Theory and
application. New York, NY: Brunner-Routledge.
Worthington, E. L., Jr. (Ed.). (2005). Handbook of forgiveness. New York,
NY: Brunner/Routledge.
ⴱ
Worthington, E. L., Jr., Berry, J. W., Miller, A. J., Sharp, C. B., Canter,
D. E., Hook, J. N.,...Ripley, J. S. (2011). Forgiveness-reconciliation
and communication-conflict-resolution interventions versus retested
controls in early married couples. Unpublished manuscript, Department
of Psychology, Virginia Commonwealth University, Richmond, VA.
ⴱ
Worthington, E. L., Jr., Kurusu, T. A., Collins, W., Berry, J. W., Ripley,
J. S., & Baier, S. N. (2000). Forgiving usually takes time: A lesson
learned by studying interventions to promote forgiveness. Journal of
Psychology and Theology, 28, 3–20.
Worthington, E. L., Jr., Sandage, S. J., & Berry, J. W. (2000). Group
interventions to promote forgiveness: What researchers and clinicians
ought to know. In M. E. McCullough, K. I. Pargament, & C. E. Thoresen
(Eds.), Forgiveness: Theory, research, and practice (pp. 228 –253). New
York, NY: Guilford Press.
Received August 27, 2012
Revision received October 23, 2013
Accepted October 28, 2013 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
17
FORGIVENESS META-ANALYSIS
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Consulting and Clinical Psychology
This content is subject to copyright. Terms and conditions apply.