An alternative model of self-forgiveness.

McConnell, John M. ; Dixon, David N. ; Finch, W. Holmes 等

Throughout life, people inevitably transgress by offending others, themselves, or a religious/spiritual figure by failing to uphold others', personal, or spiritual standards. When transgressors have empathy for their victims, take responsibility, and their transgressions are sufficiently severe, they often experience remorse as guilt and self-condemnation through shame (Tangney & Dearing, 2002). In other words, guilt and shame are painful intropunitive feelings focused on behaviors or characterological flaws, respectively (Fisher & Exline, 2006; Lewis, 1971; Tangney & Dearing, 2002). Although transgressors can experience guilt and shame simultaneously, it is important to distinguish the behavioral nature of guilt from the characterological focus of shame. Whereas shame-free guilt is associated with interpersonal and intrapsychic reparative actions, guilt-free shame is related to social withdrawal and avoidance motivation (Tangney & Dearing, 2002). Living through these feelings often diminishes transgressors' self-worth and self-respect (Dillon, 2001). In order to resolve negative self-evaluations, shame-free guilty transgressors may become concerned with self-forgiveness. Guilt-free shameful persons, however, may avoid self-forgiveness altogether unless they are able to reframe their offenses from character flaws into behavioral terms (Hall & Fincham, 2008; Tangney & Dearing, 2002).

To accomplish self-forgiveness, transgressors often apologize to and seek expressions of forgiveness from their victims. Hall and Fincham (2005) suggested that transgressors' various levels of attributions (e.g., internal vs. external cause), transgression severity, empathy, guilt, shame, conciliatory behaviors (e.g., apologies), and perceived interpersonal forgiveness all play roles in the process of self-forgiveness. In a follow up study, however, Hall and Fincham (2008) found only severity of transgressions, guilt, conciliatory behaviors, and perceived forgiveness related to self-forgiveness above and beyond the variance accounted for by the passage of time. Further, attributions, empathy, and shame were unrelated to self-forgiveness when they accounted for time. The results of Hall and Fincham (2008) suggest (1) taking responsibility is a mere precondition to self-forgiveness, (2) empathy plays an insignificant role in self-forgiveness, and (3) shameful individuals may not accomplish self-forgiveness. In the current study, we compared models derived from Hall and Fincham's 2005 and 2008 models (Figures 1 & 2) against our a priori alterative mediation model (Figure 3) to investigate self-forgiveness following interpersonal offenses.

Self-Forgiveness Path Models

We present an overview of Hall and Fincham's (2005) proposed antecedents of self-forgiveness followed by descriptions of models (Figures 1 & 2) derived from Hall and Fincham (2005/2008) and our a priori alternative mediation model (Figure 3).

Attributions

Various attributions about transgressions, or the extent to which transgressors assume blame, may impact transgressors' guilty and shameful feelings. External, unstable, global, uncontrollable, and prideful attributions are likely to superficially increase self-forgiveness because they minimize taking responsibility for offenses (Fisher & Exline, 2006; Hall & Fincham, 2005; Tangney, Boone, & Dearing, 2005), thereby decreasing levels of guilt and shame. For instance, self-forgiveness was negatively correlated with guilt and shame, but positively correlated with narcissism and victim blaming (Strelan, 2007; Zechmeister & Romero, 2002). These attributions are likely to be indicative of pseudo-self-forgiveness (Enright, 1996; Hall & Fincham, 2005). In other words, transgressors may quickly "forgive" themselves by essentially ignoring their culpabilities.

For example, someone may believe, "I forgive myself for hitting him. He deserved it anyway." Because real self-forgiveness may involve more time and effort (Fisher & Exline, 2006), pseudo-self-forgiveness may appear on the surface to have the same benefits of self-forgiveness, but it may not produce the same lasting transformational rewards as its more extensive counterpart (Hall & Fincham, 2005). As transgressors increasingly take ownership of their faults, they may experience increasing levels of guilt and shame. However, attaining self-forgiveness might be especially difficult because internal, stable, specific, controllable, and humble attributions may inhibit forgiveness of self (Fisher & Exline, 2006; Hall & Fincham, 2005; Tangney et al., 2005). However, Hall and Fincham (2008) found changes in attributions were not associated with self-forgiveness beyond the variance accounted for by time. In this sense, taking responsibility may be a necessary precondition to self-forgiveness, but relate little to the process of self-forgiveness.

Severity of transgressions

Severity of transgressions seems to impact transgressors' experiences of guilty and shameful feelings. Indeed, severe offenses can lead to greater guilt and shame (Fisher & Exline, 2006; Ingersoll-Dayton & Krause, 2005). Transgressors perhaps experience deeper feelings of guilt and shame due to their beliefs that they caused more harm to their victims. Hall and Fincham (2008) found changes in perceptions of transgression severity impacted levels of self-forgiveness beyond the variance accounted for by time. Researchers have infrequently explored severity of transgressions within the self-forgiveness literature. To our knowledge, researchers have yet to develop an empirically validated scale to assess transgression severity.

Empathy

Empathy, a multidimensional construct (Davis, 1980, 1983), is the cognitive ability of transgressors to accurately perspective-take and to recognize their victims' affective experiences combined with the ability to personally experience their victims' affective and cognitive experiences (Coke, Batson, & McDavis, 1978; Feshbach, 1975). Empathic feelings may be an important tool for transgressors in first recognizing that they require self-forgiveness (Enright, 1996). Transgressors, knowing and feeling the true positions in which they put their victims, may have strong feelings of guilt and shame associated with their transgressions. Yet transgressors' levels of empathy appear more strongly related to guilt than shame (Tangney & Dearing, 2002).

These higher levels of painful intropunitive feelings may cause transgressors to not feel or believe others have forgiven them, even if they have explicitly stated, "I forgive you." Some transgressors with high levels of empathy may continue to believe self-forgiveness is not acceptable because it signals disrespect (Hall & Fincham, 2005). Empathic transgressors may feel uncomfortable elevating themselves by letting go of guilt and shame if victims are still experiencing any deprivation. Given that empathy has a long empirical history of facilitating other-forgiveness (e.g., McCullough, et al., 1998; McCullough, Worthington, & Rachal, 1997), some may have difficulty believing empathic feelings do not play key roles in self-forgiveness. To date, researchers have provided inconsistent evidence for the directionality and strength of the relationship between empathy and self forgiveness (cf. Barbetta, 2002; Hall & Fincham, 2008; Macaskill, Maltby, & Day, 2002; Rangganadhan & Todorov, 2010; Zechmeister & Romero, 2002).

[FIGURE 1 OMITTED]

Guilt and shame

Individuals who commit transgressions may experience guilt and shame through tension, sorrow, or regret, although researchers have observed important differences between guilt and shame (Tangney, 1995). Transgressors are most likely to experience guilt in relation to specific behaviors, whereas they are more likely to experience shame in relation to some apparent character flaws (Lewis, 1971). When individuals believe they are inherently flawed, they may ignore the outside world because they remain focused on their own personal distress (Tangney & Dearing, 2002). Shame may be unrelated or weakly related with self-forgiveness because it is naturally more "self-centered" than guilt (Hall & Fincham, 2005, 2008; Lewis, 1971; Tangney, 1995). The self-focus and painful nature of shame is more likely to lead transgressors into destructive behaviors, self-criticisms, avoidance, denials, and lashing out at those who pose threats to their characters (Tangney, 1995; Tangney et al., 1996; Tangney, Miller, Flicker, & Barlow, 1996). Shameful transgressors may avoid conciliatory behaviors because they believe changing themselves is improbable. Several independent trait and state studies have exhibited this empathic concern of guilt and egocentric nature of shame (Tangney & Dearing, 2002). Therefore, transgressors who experience guilt may initiate healthy forms of reparative interactions, such as admissions, conciliatory behaviors, and reconciliations. On the other hand, transgressors who experience shame may be more likely to initiate maladaptive behaviors, such as denials and separations. This may shed an interesting light on why Hall and Fincham (2008) found decreases in guilt, but not shame, correlated with higher levels of self-forgiveness beyond the variance accounted for by time. Transgressors, who on the surface appear to be blameless and deny their need for self-forgiveness, may actually be feeling great internal pain through shame (Tangney & Dearing, 2002).

[FIGURE 2 OMITTED]

Conciliatory behaviours

Some researchers have found conciliatory behaviors positively related with self-forgiveness (Fisher & Exline, 2006; Hall & Fincham, 2008). In one study, however, apology and self-forgiveness were both positively and negatively correlated (Exline, DeShea, & Holeman, 2007). Moreover, evidence suggests conciliatory behaviors aimed at transgressors' religious/spiritual figures, rather than their victims, may act to inhibit their self-forgiveness (Hall & Fincham, 2008). Overall, research suggests that conciliatory behaviors may play a role in self-forgiveness, but future research still needs to explore more fully how tertiary factors may impact the effects of conciliation.

Perceived forgiveness

Several studies have demonstrated that transgressors' perceived forgiveness from victims or religious/spiritual figures is positively associated with self-forgiveness (Hall & Fincham, 2008; Martin, 2008; McConnell & Dixon 2012, in press; Witvliet, Ludwig, & Bauer, 2002). Nevertheless, Layer, Roberts, Wild, and Walters (2004) found some transgressors felt that religious/spiritual figures had forgiven them, yet still struggled with the process of self-forgiveness. Conciliatory behaviors are thought to increase perceived forgiveness from victims (Hall & Fincham, 2005), perhaps due to their ability to actually increase empathy and forgiveness from victims (e.g., McCullough et al., 1997; 1998). Transgressors may move past associated feelings by having senses of being restored through forgiveness (Bauer et al., 1992). For transgressors, it may be a small jump from believing their victims have forgiven them to believing it is now acceptable to self-forgive.

Model 1: Hall and Fincham's (2005) self-forgiveness model (see Figure 1)

To be concerned with self-forgiveness, transgressors must commit actions against victims; for clarity, these actions will remain interpersonal, rather than intrapersonal or spiritual. The current study did not measure transgressions, as they are exogenous variables indicating that an interpersonal transgression has occurred. When transgressions occur, attributions and severity of transgressions produce the experiences of transgressors' guilty, shameful, or self-forgiving feelings. First, attributions that take responsibility increase transgressors' experiences of guilt and shame (Hall & Fincham, 2005; Tangney, Wagner, & Gramzow, 1992). Further, attributions that deny blame may be indicative of "pseudo-self-forgiveness," while attributions that reflect taking responsibility will amplify the inability to self-forgive. Likewise, less severe transgressions will increase self-forgiveness, whereas more severe transgressions will increase guilt and shame (Hall & Fincham, 2005). Subsequently, once guilt and shame are experienced, they must be resolved for self-forgiveness to be accomplished. Hall and Fincham (2005) did not clarify any mechanism that resolves shame, as it is a more persistent emotion and is often associated with avoidance responses (Tangney, 1995). Therefore, lower levels of shame will be indicative of self-forgiveness, while higher levels of shame will intensify intrapersonal unforgiveness (Hall & Fincham, 2005). In the same way, lower levels of guilt will increase self-forgiveness, however higher levels of guilt will increase transgressors' use of conciliatory behaviors (Hall & Fincham, 2005). Subsequently, conciliatory behaviors will positively impact perceived forgiveness and self-forgiveness, while perceived forgiveness also increases the likelihood that transgressors will forgive themselves (Hall & Fincham, 2005). According to this model, persons who feel guilty will have greater empathic feelings towards their victims and will experience lower levels of self-forgiveness.

[FIGURE 3 OMITTED]

Model 2: Hall and Fincham's (2008) self-forgiveness model

As previously discussed, Hall and Fincham (2008) reported that only severity of transgressions, guilt, conciliatory behaviors, and perceived forgiveness predicted forgiveness after accounting for time. Therefore, attributions, empathy, and shame may be unrelated or weakly related to transgressors experiences of self-forgiveness. In order to test the model commensurate with Hall and Fincham (2008), we removed attributions, empathy, and shame. For this alternative model (Figure 2), all other relations remain unchanged.

Model 3: Alternative model

Mediation occurs when a third variable (e.g., perceived forgiveness) best explains the relation between a predictor (e.g., conciliatory behavior) and an outcome (e.g., self-forgiveness) as outlined in Frazier, Tix, & Barron (2004). Mediation models inherently infer an experimental-causal chain. Researchers, however, often pragmatically use mediation models when using measurement-of-mediation survey designs (Spencer, Zanna, & Fong, 2005).

Some believe apologies and forgiveness from victims are prerequisites for self-forgiveness (e.g., Griswold, 2007; Hughes, 1994). Further, apologies from transgressors typically influence victims, through their empathic responses, to forgive their transgressors (McCullough et al., 1997; 1998). As victims make their forgiveness known to their transgressors, it is likely to increase transgressors' feelings of being forgiven, and thus, increases self-forgiveness. Therefore, conciliatory behavior may mediate the relation between perceived forgiveness and self-forgiveness, which may clarify the inconsistent role of apologies with self-forgiveness (Exline et al., 2007; Fisher & Exline, 2006; Hall & Fincham, 2008). Further, since severity of transgressions appears more distally related to self-forgiveness in Hall and Fincham (2005 and 2008), the other antecedents in the model may mediate the relation between transgression severity and self-forgiveness. To test this line of reasoning, we created an alternative, more parsimonious model (Figure 3) a priori by removing the paths between severity of transgressions/conciliatory behaviors and self-forgiveness. If model fit--as indicated by parsimony-adjusted indices--increases from Model 2 to Model 3, it would indicate that our alternative model is more parsimonious. Finally, if relations between severity of transgressions/conciliatory behaviors and self-forgiveness were negligible in Model 2, it would indicate full, rather than partial, mediation (Frazier et al., 2004).

The Present Research

Investigating how people forgive themselves is important because people often believe self-forgiveness is more difficult than forgiving others (e.g., Worthington, 2006), yet accomplishing self-forgiveness is more strongly associated with overall well-being (Avery, 2008). Hall and Fincham (2005) made a significant advancement in the field of self-forgiveness when they proposed their theoretical model, yet researchers have attempted little confirmation of the model. Recent evidence (Hall & Fincham, 2008) suggests that time better accounted for the variance explained in self-forgiveness than attributions, empathy, shame. Severity of transgressions, guilt, conciliatory behaviors, and perceived forgiveness maintained strong relations to self-forgiveness after Hall and Fincham (2008) accounted for time.

Contrarily, Rangganadhan and Todorov (2010) found that their path model, which highlighted the importance of shame and personal-distress empathy in dispositional self-forgiveness, fit better than a portion of Hall and Fincham's (2005) model. However, Rangganadhan and Todorov (2010) did not test Hall and Fincham's entire model because they only included dispositional aspects. A second limitation of Rangganadhan and Todorov (2010) was their use of path diagram modeling, rather than a full structural model that uses multiple measurements to create latent variables. The current study extended the self-forgiveness literature by both exploring the whole model in relation to specific interpersonal offenses and through the use of structural models. The present study shed light on the confusing nature of attributions, empathy, guilt, shame, and conciliatory behaviors in previous literature. Via structural equation modeling using maximum likelihood estimation (AMOS 16.0; Arbuckle, 2007), the current study sought to test Hall and Fincham's (2005 and 2008) models against an alternative self-forgiveness model. We hypothesized that an alternative model would be the best fitting, most parsimonious model.

Method

Participants

Participants (male [n = 148], female [n = 257], unspecified [n = 1]) were a convenience sample of 406 undergraduate students at a Midwestern university who completed the questionnaires for partial fulfillment of a psychology course requirement. Individuals were recruited via a posting on a university psychology department website and were allowed to participate if they were at least 18 years old. An initial 530 participants signed up, however 124 people were not included in the study due to incomplete data or random responding. Participants' ages ranged from 18-41 (M = 20.34, SD = 2.80) and they represented various ethnic, religious, and marital statuses. We discuss implications of these demographics in the limitations section.

Procedure

Our institution's ethical review board approved our study prior to us initiating data collection. After viewing a posted recruitment letter online, participants sent an email to the primary investigator indicating their interest in completing the research experiment and were provided hyperlinks to online modules. The online modules instructed participants to recall an event in the past two years in which they had offended someone by something they said or did. The event only needed to be one in which other persons felt that participants harmed them, regardless if participants believed they were innocent or at fault. Participants filled in counterbalanced questionnaires in approximately 25 to 35 minutes. We collected data from July 2008 through December 2008.

Materials

The Causal Dimension Scale (CDS; Russell, 1982).

The CDS consists of three separate subscales--Causality, Stability, and Controllability--which all have previously demonstrated desirable levels of internal consistency reliabilities ([alpha] = .87, .84, and .73, respectively) as well as discriminant validities when compared to each other (Russell, 1982). Participants were instructed to fill in the CDS in respect to their "impressions or opinions of the cause or causes of the event." All nine-items were rated with a nine-point bipolar-type scale (e.g., Is the cause(s) something: Permanent vs. Temporary ; Inside of You vs. Outside of You). Higher scores reflected beliefs of more internal causation, permanence, and controllability.

Multidimensional Forgiveness Inventory (MFI; Tangney, Boone, Dearing, & Reinsmith, 2002). The MFI consists of sixteen scenarios (e.g., "Imagine that you have sex with you best friend's boyfriend/girlfriend/spouse") on which participants rate the degrees to which they believe they would forgive, ask for forgiveness, self-forgive, take time to forgive others, take time to self-forgive, blame others, blame self, be hurt, and be angry. Cronbach's alpha for the MFI ranged from .73 to .85 in a previous sample (Tangney et al., 2002). Further, the MFI has demonstrated desirable levels of convergent and discriminant validities (Tangney et al., 2002). We used the Propensity to Blame Others (BO), Propensity to Blame Self (BS), Propensity to Ask Forgiveness (AF), and Propensity to Forgive Self (PFS) subscales. Responses to each scenario were, "How likely would you be to: 'Try to blame someone or something else for the event (BO); 'Think about the situation over and over blaming yourself for the damage done' (BS); 'Ask them to forgive you' (AF); and 'Forgive yourself' (PFS)." All subscales were quantified on a five-point Likert-type scale (1: Not at All; 5: Very Likely), with higher scores reflecting more blaming others, blaming self, asking for forgiveness, and self-forgiveness.

Test of Self-Conscious Affect--Version 3 (TOSCA-3; Tangney, Dearing, Wagner, & Gramzow, 2000). The TOSCA-3 consists of sixteen scenarios (e.g., "You make plans to meet a friend for lunch. At 5 o'clock, you realize you stood him up") that participants rate the degrees to which they believe they would feel various thoughts and emotions (e.g., externalization, guilt, shame). We used the Externalization (PE; e.g., "My boss distracted me just before lunch"), Guilt (e.g., "You'd think you should make it up to him as soon as possible"), and Shame (e.g., "I'm inconsiderate") subscales, which have previously demonstrated desirable levels of internal consistency reliabilities ([alpha] = .80, .83, .88, respectively; Tangney & Dearing, 2002) and discriminant validity in non-clinical samples when compared to objective measures (Rusch, et al., 2007). All subscales were rated on a five-point Likert-type scale (1: Not at All; 5: Very Likely), with higher scores reflecting more blaming others, guilt, and shame.

Transgression Semantic Differentiation (TSD). We developed the TSD to quantify participants' perceived severity of their transgressions. Its 24-items are intended to quantify participants' beliefs about (1) how they view their transgressions (good vs. bad) and (2) the victims' views of their transgressions (good vs. bad) by using a thesaurus and having two experts in the field of forgiveness review items for face validity. The scale consisted of two subscales--Transgressor Perception (TSD-TP) and Victim Perception (TSD-VP)--each with 12 items rated on a seven-point bi-polar scale (e.g., 1: Mild vs. 7: Harsh; 1: Harmless vs. 7: Harmful; 1: Inoffensive vs. 7: Offensive). Instructions for both subscales were for participants to select the number that most applies to their beliefs or beliefs about how their victims view the event. Higher scores reflected more severe offenses. Internal consistency for the TSD-TP and TSD-VP are reported in Table 1.

Communication Emotional Response Scale (CERS; Batson, Bolen, Cross, & Neuringer-Benefiel 1986). The CERS consists of six adjectives (sympathetic, compassionate, tender, softhearted, moved, and warm) embedded in a list of 16 distracter adjectives. Participants responded to these items while considering how they felt towards their victims. The CERS has demonstrated a desirable level of internal consistency reliability ([alpha] = .82) as well as convergent validity with other empathy scales in a previous study (Batson et al., 1986). The 22 adjectives were quantified on a seven-point Likert-type scale ranging from 1 (Not at All) to 7 (Extremely), with greater scores indicating higher degrees of empathy.

Interpersonal Reactivity Index (IRI; Davis, 1980). The IRI consists of four separate subscales--Perspective Taking (PT), Fantasy (FS), Empathic Concern (EC), and Personal Distress (PD)--that participants respond to as they typically feel. Because the orthogonal nature of the IRI would have caused issues with creating a latent variable (Davis, 1980; 1983; Kline, 2005), we used only the two highly correlated affective (EC) and cognitive (PT) subscales. EC (e.g., "I often have tender, concerned feelings for people less fortunate than me") and PT (e.g., "I try to look at everybody's side of a disagreement before I make a decision") have demonstrated desirable levels of internal consistency reliabilities for males ([alpha] =.72 & .75) and females ([alpha] = .70 & .78) in a previous study (Davis, 1980). Additionally, EC's and PT's test-retest reliabilities for males (.72 & .61) and females (.70 & .62) have been satisfactory (Davis, 1980). The IRI has previously demonstrated desirable levels of convergent and discriminant validities (Davis, 1983). All 28-items were quantified on a five-point Likert-type scale ranging from 1 (Does Not Describe Me Well) to 5 (Describes Me Very Well), with higher scores reflecting greater empathy.

Personal Feelings Questionnaire-2 (PFQ-2; Harder & Zalma, 1990). The PFQ-2 consists of six guilt (e.g., Mild Guilt), ten shame (e.g., Self-Consciousness), and six distracter adjectives (e.g., Sadness). Participants responded to these items while considering how they felt towards their victims. The PFQ-2 has demonstrated desirable levels of internal consistency reliabilities for the guilt ([alpha] = .72) and shame ([alpha] = .78) scales as well as desirable test-retest reliabilities (.85 & .91) in a previous study (Harder & Zalma, 1990). Further, the PFQ-2 has previously demonstrated desirable levels of convergent and discriminant validities (Harder, Rockart, & Cutler, 1993; Harder & Zalma, 1990). The PFQ-2 is rated on a five-point Likert-type scale, ranging from 0 (You Never Experience the Feeling) to 4 (You Experience the Feeling Continuously or Almost Continuously), with higher scores reflecting greater empathy.

State Shame and Guilt Scale (SSGS; Marschall, Sanftner, & Tangney, 1994). The SSGS consists of 15-items assessing guilt (e.g., "I feel bad about what I did"), shame (e.g., "I feel small"), and pride (e.g., "I feel proud"), with five-items each. We used the Guilt and Shame subscales, which have demonstrated desirable levels of internal consistency reliabilities for guilt ([alpha] = .82) and shame ([alpha] = .89) in previous research (Tangney & Dearing, 2002). The SSGS was developed from a rich empirical and theoretical background and has demonstrated adequate construct validity (Tangney & Dearing, 2002). Participants were asked to respond to items in relation to their offenses. The SSGS is quantified on a five-point Likert-type scale ranging from 1 (Not Feeling This Way at All) to 5 (Feeling This Way Very Strongly), with higher scores reflecting greater levels of guilt and shame.

Conciliatory Behaviors Scale (CBS; McCullough et al., 1997). We modified the CBS to more fully represent the construct of conciliatory behaviors. The original scale consists of two items ("I tried to make amends or compensations" and "I took steps toward reconciliation: Wrote them, called them, expressed love, showed concern, etc.") that have demonstrated a desirable level of internal consistency reliability (a = .74). Additionally, one item ("I attempted to say I was sorry") from the three-item apology scale (McCullough et al., 1997) was modified and added to the CBS. The underlined components of the item represent the modification. Further, two items ("I have expressed personal responsibility and guilt/shame for the offense," and "I have expressed my intentions to not repeat the offense") were added in order to include the missing components of an effective apology (Olshtain, 1989; Weiner et al., 1991). Participants were asked to respond to items in relation to their offenses. The five-item scale was quantified on a seven-point Likert-type scale ranging from 1 (Strongly Disagree) to 7 (Strongly Disagree), with higher scores reflecting greater conciliation. We created two parcels from these items in order to provide the latent construct of conciliatory behaviors with a sufficient amount of observed variables. We further discuss the controversial methodology of parceling in the results section.

The Heartland Forgiveness Scale (HFS; Thompson et al., 2005). To quantify participants' levels of perceived forgiveness, we modified the Heartland Forgiveness of Others subscale (HFSO). We adjusted all six items of the HFSO to assess transgressors' perceptions of the victims' forgiveness by changing the subject of each sentence to assess perceptions of receiving forgiveness, rather than extending forgiveness (e.g., "The victim will continue to punish me"). We created two parcels from these items in order to provide the latent construct of perceived forgiveness with a sufficient amount of observed variables.

To quantify different items relating to forgiveness of self, we modified the HFS Self-Forgiveness subscale (HFSS). For the focus of this study, we modified all six items of the HFSS to correctly assess self-forgiveness in relation to a specific event, rather than dispositional self-forgiveness. For example, "Although I feel bad at first when I mess up, over time I can give myself some slack," was modified to "Although I felt bad when I messed up, over time I have given myself some slack."

The original HFS subscales have shown Cronbach's alphas ranging from .72 to .87 across three samples as well as test-retest reliabilities for a 3-week interval, .72-.77, and a 9-month interval, .68-.69) (Thompson et al., 2005). The original HFS has demonstrated desirable levels of convergent validity with other forgiveness measures (Thompson et al., 2005). All items of the HFSO and HFSS were quantified with a seven-point scale ranging from 1 (Definitely False) to 7 (Definitely True), with greater scores reflecting greater perceived forgiveness and self-forgiveness.

Transgression-Related Interpersonal Motivations Inventory (TRIM; McCullough et al., 1998). The 12-item TRIM consists of two oblique subscales--Avoidance (e.g., "He/she avoids me") and Revenge (e.g., "He/she will make me pay")--which have both demonstrated desirable levels of internal consistency reliabilities ([alpha] = .86 & a = .90, respectively), test-retest reliability, and convergent and discriminant validity in a previous study (McCullough et al., 1998). To quantify the degree to which the participants' believed their victims forgave them we modified all twelve items of the TRIM to assess correctly the transgressors' perception of the victims' forgiveness. As with the HFSO, we changed the subject of each sentence to assess perceptions of receiving forgiveness, rather than extending forgiveness. Participants were asked to respond to items in relation to their offenses. The TRIM was quantified on a seven-point Likert-type scale ranging from 1 (Definitely False) to 7 (Definitely True), with greater scores reflecting greater perceived forgiveness.

State Self-Forgiveness Scale (SSFS; Wohl, DeShea, & Wahkinney, 2008). The SSFS consists of two oblique subscales--Self-Forgiving Feelings and Actions (SSFA; a = .86) and Self-Forgiving Beliefs (SFB; a = .91)--which have demonstrated desirable levels of internal consistency reliabilities as well as convergent and discriminant validities when compared to objective measures in a previous study (Wohl et al., 2008). Participants were asked to respond to questions preceded by "As I consider what I did that was wrong ..." Example items include, "I punish myself" (SSFA) and "I believe I am acceptable" (SFB). All items were quantified with a four-point scale ranging from 1 (Not at All) to 4 (Completely"), with higher scores representing greater self-forgiveness.

The Marlowe-Crowne Social Desirability Scale (M-C SDS; Crowne & Marlowe, 1960). To observe if any scales related to socially desirability, we utilized the M-C SDS. The M-C SDS contains 33-items (e.g., "I have never intensely disliked anyone") that, if reporting in socially desirable ways, participants are expected to answer True (T) or False (F). Quantifying the M-C SDS involves assigning a value of one (T = T or F = F) or a value of zero (T [not equal to] T or F [not equal to] F) to the participants' responses. Therefore, higher scores on the M-C SDS represent higher levels of social desirability. The M-C SDS has demonstrated desirable levels of internal consistency reliability ([alpha] = .88) and test-retest reliability (.89) (Crowne & Marlowe, 1960).

Attention questions

Due to the large amounts of items necessary for this study, we randomly added two items within the main scales to check for attention (e.g., "After reading this question please select the number 4").

Demographic questionnaire

To assess participants' demographics, we administered a demographic questionnaire that consisted of questions regarding age, gender, race, religious affiliation, marital status, and semesters in college.

Results Data Preparation Excluding cases

We excluded participants based on (1) not recalling an offense, (2) not filling in any questionnaires, (3) incorrectly marking either of the two attention questions, or (4) completing the module in less than 25 minutes. To determine the minimum duration for valid participants, we rank-ordered participants' total response times across the instruments in juxtaposition to their responses on the two attention questions. Ninety-two percent of participants who completed modules in less than 25 minutes also failed to correctly mark the attention questions. Based on these four criteria, we eliminated 124 of 530 participants, leaving N = 406 for analyses.

Missing data imputation

We assumed missing data were Missing Completely At Random (MCAR) because the data appeared randomly dispersed among participants and individual items. In addition, the low percentage of missing data (.02%) allowed for confidence in concluding MCAR. We utilized the statistical package AMOS 16.0 to conduct multiple imputation (i.e., generating and pooling multiple plausible values for missing data) in order to complete the data set (Arbuckle, 2007).

Parceling

When there are not enough observed variables to sufficiently create a latent variable, as is the case within the newly developing field of self-forgiveness, researchers may choose to parcel, or distribute single items into several subset totals (Little, Cunningham, Shahar, & Widaman, 2002). We parceled the CBS and HFSO with a method recommended by Little and colleagues (2002). Using Principal Axis Factoring with promax rotation, we found that both the CBS and the HFSO were sufficiently unidimensional. Then, we randomly assigned individual items into two parcels for CBS and three parcels for HFSO.

Scale properties

See Table 1 for a summary of scale properties. We conducted internal consistency reliabilities on all scales we used in the current study. Nearly all scales elicited adequate results ([alpha] = .70 - .95) and were used for analyses, except CDS, which was excluded because of low internal consistency ([alpha] = .45 to .57). We conducted multicollinearity diagnostics with variance inflation factors, which ranged from 1.02 to 3.98, indicating little redundancy among observed variables measuring similar constructs. We discuss implications of these scale properties in the limitations section.

Intercorrelations

We used raw data for all structural equation modeling analyses. Table 2 presents the intercorrelation matrix of the analyzed variables [Editor's Note: Due to its size, Table 2 is only available online at www.nspb.net.]. For exploratory purposes, we used the MC-SDS to observe if certain scales related to social desirability. Social desirability was most highly related to severity of transgressions (TSD-TP: r = -.12; TSD-VP: r = .14), attributions (PE: r = -.15), shame (PFQ: r = -.11; TOSCA-3: r = .15), and perceived forgiveness (HFSO, Parcel 1: r = .11) at the .05 level. Other variables were unrelated to social desirability (Table 2). Social desirability only accounted for 2.25% or less of the variability with other scales. These correlations suggested that social desirability minimally related to participants' responses.

Determining and Comparing Model Fits

Model Chi-Square ([chi square]) compares the difference in fit between researcher's specified models with just-identified models (i.e., all parameters estimated). It has been demonstrated that high correlations among the observed variables and large sample sizes tend to inflate [chi square], which often leads to erroneously rejecting true models (Kline, 2005). However, Normed Chi-Square ([chi square]/df) is less sensitive to sample size than [chi square]. Some have offered several different cutoff values for [chi square]/df (see Kline, 2005). Therefore, for the purpose of the current study, various cutoff values indicated great (< 2), good (2-3), adequate (3-5), and poor model fit (> 5).

The Comparative Fit Index (CFI) and the Tucker-Lewis Index (TLI) are both incremental fit indices, which compare increases in fit relative to independence models; that is, models assuming zero covariance matrices. TLI also penalizes model complexity and therefore prefers models that are more parsimonious (Kline, 2005). Researchers have long sought a "golden-rule" cutoff value for interpreting incremental fit indices, but have experienced considerable difficulty (Hu & Bentler, 1999; Marsh, Hau, & Wen, 2004). Some suggested conventional cutoff values (i.e., [greater than or equal to] .90) might even be too stringent (Marsh et al., 2004), whereas others proposed more rigorous values (i.e., [greater than or equal to] .95) (Hu & Bentler, 1999). Thus, values [greater than or equal to] .90 signify reasonably good fit, while values [greater than or equal to] .95 point to excellent fit (Kline, 2005). The Root Mean Square Error of Approximation (RMSEA), also a parsimony-adjusted index (i.e. models with more estimated parameters are penalized), is immune to sample size fluctuations because it computes the model-implied covariance matrix's ability to reproduce the population covariance matrix. Values of [less than or equal to] .05 evidenced close, .05-.08 indicated adequate, and > .08 represented poor approximate fit (Kline, 2005). For non-hierarchical models, that is to say, models that are not a subset of one another (cf. Model 1 vs. Model 2 or 3), we used Akaike information criterion (AIC) and Bayesian information criterion (BIC). AIC and BIC are computed by recreating equally sized, hypothetically replicated samples (Kline, 2005). Both of these fit indices are parsimony-adjusted, however BIC adjusts for model complexity more so than AIC (Kline, 2005). Unlike previously mentioned fit indices, AIC and BIC both have added advantages of not assuming multivariate normality, which is not always observed in psychological data (Micceri, 1989). Although regression weights are usually fairly accurate under cases of multivariate non-normality, some indices are inappropriately inflated (e.g., [chi square], [chi square]/df, & RMSEA) or deflated (e.g., CFI, TLI), leading to greater chances of rejecting true models (Kline, 2005). There are no set cutoff values for these comparative fit indices, however, models with lower values provide better fit to the data (Kline, 2005).

Structural Equation Modelling

Model 1: Hall and Fincham (2005). Initial computation of Hall and Fincham's (2005) SEM elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -25.11, SE = 17.70, p =.16). Therefore, we computed an admissible solution by fixing its variance to zero. Unexpectedly, more externalizing attributions were positively, not negatively, related with guilt ([beta] = .19, p = .003) and shame ([beta] = .15, p = .003); empathy ([beta] = .28, p < .001) was positively, not negatively, related to self-forgiveness; and attributions (P = .10, p = .09) along with conciliatory behaviors ([beta] = -.03, p = .63) were unrelated to self-forgiveness. All other predicted relations were maintained or weakly maintained (Figure 1). Nonetheless, these predicted relations should not be entirely trusted because of poor model fit [[chi square] = 1563.73, df = 262, p < .001; [chi square]/df = 5.97; CFI = .730; TLI = .691; RMSEA = .111 (CI = .105-.116); AIC = 1689.73; BIC = 1942.13]; however, significant multivariate non-normality (Table 3) might have increased the chance of rejecting a true model.

Model 2: Hall and Fincham (2008). Initial computation of Hall and Fincham's (2008) SEM elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -124.90, SE = 76.13, p = .10). Therefore, we fixed the variance at zero. Whereas severity of transgressions ([beta] = .06, p = .17) and conciliatory behaviors ([beta] = .03, p = .59) were unrelated with self-forgiveness, all other predicted relations were maintained at the .001 level (Figure 2). This pattern of results indicated that offense severity and conciliatory behaviors were mediated by guilt and perceived forgiveness. Model 2 elicited poor to adequate model fit [[chi square] = 358.14, df = 98, p < .000; [chi square]/df = 3.65; CFI = .918; TLI = .899; RMSEA = .081(CI = .072-.09); AIC = 434.14; BIC = 586.38]. Nevertheless, multivariate non-normality (Table 3) may have decreased model fit. In this case, Hall and Fincham (2008) might truly approximate a great fitting model and thus the results of Model 2 are inconclusive.

Model 3: Alternative model

Initial computation of the alternative model elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -136.95, SE = 86.23, p = .12). Thus, we again fixed the variance at zero. All predicted relations were strongly maintained at the .001 level (Figure 3). Most notably, guilt accounted for 47.61% of the variance in self-forgiveness. The alternative model elicited adequate to good model fit [[chi square] = 360.20, df = 100, p < .000; [chi square]/df = 3.60; CFI = .918; TLI = .901; RMSEA = .080(CI = .071-.89); AIC = 432.20; BIC = 576.43]. Nevertheless, multivariate non normality (Table 3) may have decreased model fit and thus, the alternative model might truly approximate a great fitting model.

Comparing structural models

Because Model 1 vs. Model 2 or 3 are non-hierarchical comparisons, we could use only AIC and BIC to directly test the equal fit hypotheses. When comparing Model 1 with Model 2/3, both AIC ([DELTA]1255.59/[DELTA]1257.53) and BIC ([DELTA]1355.75/[DELTA]1365.70) had large value decreases, suggesting that Hall and Fincham (2008) and the alternative are more preferred over Hall and Fincham (2005). Model 2 elicited negligible relations for severity of transgressions and conciliatory behaviors when predicting self-forgiveness. In addition, there was a small increase in fit between Model 2 [[chi square] = 358.14, df = 98, p < .000; [chi square]/df = 3.65; CFI = .918; TLI = .899; RMSEA = .081(CI = .072-.09)] and Model 3 [[chi square] = 360.20, df = 100, p < .000; [chi square]/df = 3.60; CFI = .918; TLI = .901; RMSEA = .080(CI= .071-.89)]. Taken together, this supports full that are not a subset of one another (cf. Model 1 vs. Model 2 or 3), we used Akaike information criterion (AIC) and Bayesian information criterion (BIC). AIC and BIC are computed by recreating equally sized, hypothetically replicated samples (Kline, 2005). Both of these fit indices are parsimony-adjusted, however BIC adjusts for model complexity more so than AIC (Kline, 2005). Unlike previously mentioned fit indices, AIC and BIC both have added advantages of not assuming multivariate normality, which is not always observed in psychological data (Micceri, 1989). Although regression weights are usually fairly accurate under cases of multivariate non-normality, some indices are inappropriately inflated (e.g., [chi square], [chi square]/df, & RMSEA) or deflated (e.g., CFI, TLI), leading to greater chances of rejecting true models (Kline, 2005). There are no set cutoff values for these comparative fit indices, however, models with lower values provide better fit to the data (Kline, 2005).

Structural Equation Modeling

Model 1: Hall and Fincham (2005). Initial computation of Hall and Fincham's (2005) SEM elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -25.11, SE = 17.70, p =.16). Therefore, we computed an admissible solution by fixing its variance to zero. Unexpectedly, more externalizing attributions were positively, not negatively, related with guilt ([beta] = .19, p = .003) and shame ([beta] = .15, p = .003); empathy ([beta] = .28, p < .001) was positively, not negatively, related to self-forgiveness; and attributions ([beta] = .10, p = .09) along with conciliatory behaviors ([beta] = -.03, p = .63) were unrelated to self-forgiveness. All other predicted relations were maintained or weakly maintained (Figure 1). Nonetheless, these predicted relations should not be entirely trusted because of poor model fit [[chi square] = 1563.73, df = 262, p < .001; [chi square]/df = 5.97; CFI = .730; TLI = .691; RMSEA = .111 (CI = .105-.116); AIC = 1689.73; BIC = 1942.13]; however, significant multivariate non-normality (Table 3) might have increased the chance of rejecting a true model.

Model 2: Hall and Fincham (2008). Initial computation of Hall and Fincham's (2008) SEM elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -124.90, SE = 76.13, p = .10). Therefore, we fixed the variance at zero. Whereas severity of transgressions ([beta] = .06, p = .17) and conciliatory behaviors ([beta] = .03, p = .59) were unrelated with self-forgiveness, all other predicted relations were maintained at the .001 level (Figure 2). This pattern of results indicated that offense severity and conciliatory behaviors were mediated by guilt and perceived forgiveness. Model 2 elicited poor to adequate model fit [[chi square] = 358.14, df = 98, p < .000; [chi square]/df = 3.65; CFI = .918; TLI = .899; RMSEA = .081(CI = .072-.09); AIC = 434.14; BIC = 586.38]. Nevertheless, multivariate nonnormality (Table 3) may have decreased model fit. In this case, Hall and Fincham (2008) might truly approximate a great fitting model and thus the results of Model 2 are inconclusive.

Model 3: Alternative model. Initial computation of the alternative model elicited a negative error variance associated with TSD-TP's error term. However, the negative value was not significantly different from zero (V = -136.95, SE = 86.23, p = .12). Thus, we again fixed the variance at zero. All predicted relations were strongly maintained at the .001 level (Figure 3). Most notably, guilt accounted for 47.61% of the variance in self-forgiveness. The alternative model elicited adequate to good model fit [[chi square] = 360.20, df = 100, p < .000; [chi square]/df = 3.60; CFI = .918; TLI = .901; RMSEA = .080(CI = .071-.89); AIC = 432.20; BIC = 576.43]. Nevertheless, multivariate non-normality (Table 3) may have decreased model fit and thus, the alternative model might truly approximate a great fitting model.

Comparing structural models. Because Model 1 vs. Model 2 or 3 are non-hierarchical comparisons, we could use only AIC and BIC to directly test the equal fit hypotheses. When comparing Model 1 with Model 2/3, both AIC ([DELTA]1255.59/[DELTA]1257.53) and BIC ([DELTA]1355.75/[DELTA]1365.70) had large value decreases, suggesting that Hall and Fincham (2008) and the alternative are more preferred over Hall and Fincham (2005). Model 2 elicited negligible relations for severity of transgressions and conciliatory behaviors when predicting self-forgiveness. In addition, there was a small increase in fit between Model 2 [[chi square] = 358.14, df = 98, p < .000; [chi square]/df = 3.65; CFI = .918; TLI = .899; RMSEA = .081(CI = .072-.09)] and Model 3 [[chi square] = 360.20, df = 100, p < .000; [chi square]/df = 3.60; CFI = .918; TLI = .901; RMSEA = .080 (CI= .071-.89)]. Taken together, this supports full conciliatory behaviors and perceived forgiveness from their victims. According to the results of this study, conciliatory behaviors do not directly impact self-forgiveness, rather perceived forgiveness fully mediates their relations. However, guilt appeared to impact self-forgiveness most strongly when compared to other variables explored in this study. Therefore, transgressors' conciliatory behaviors and perceived forgiveness may partially mediate the relations between guilt and self-forgiveness. Alternatively, there also may be other mechanisms unaccounted for in the alternative model that dissolve guilt, such as relationship closeness, beliefs surrounding the appropriateness of self-forgiveness, self-respect, or existential variables. Future studies should pay close attention to the possible indirect or direct relations between guilt and self-forgiveness.

Therapeutic Implications

The results of the current study, alongside Hall and Fincham (2008), suggest therapists may aid clients in forgiving themselves by focusing on dissolving guilt. For instance, clients might lessen their guilt by exploring the possibility of unrealistic beliefs about transgression severity (Worthington, 2006). In this sense, transgressors may catastrophize the extent to which their actions impacted their victims. They also may resolve their guilt by utilizing conciliatory behaviors, real or symbolic, in attempts to make amends (Hall & Fincham, 2005; 2008). Through this process, clients may experience a sense of being restored by their victims and consequently their journeys toward self-forgiveness could be accelerated (Enright, 1996). On the other hand, even if their victims forgave them, clients still could hold feelings of unforgiveness. Therapists are in key positions to challenge their clients' maladaptive beliefs while also providing empathic, safe environments in which their clients can learn to let go (Worthington, 2006). Further, some clients may experience shame and appear uninterested in self-forgiveness. Therapists could perhaps aid their clients in shifting from experiencing shame to feeling guilt by changing foci from unchangeable character flaws to unfixed behavioral errors (Hall & Fincham, 2005; 2008; Tangney & Dearing, 2002). Mental health workers also may find Enright's (1996), Flanigan's (1996), and Worthington's (2006) stage models of self-forgiveness helpful in treating clients who desire to forgive themselves.

Conclusion

With the recent expansion of self-forgiveness literature, researchers identified a number of variables believed to play roles in the process of self-forgiveness (cf. Hall & Fincham, 2005, 2008). We found several of these proposed antecedents--severity of transgressions, guilt, conciliatory behaviors, and perceived forgiveness--appeared to represent the best fitting model of self-forgiveness (Figure 3). Moreover, each of the proposed paths elicited highly significant or strong relations. Contrary to a recent study investigating the dispositional aspects of Hall and Fincham's (2005) model (Rangganadhan & Todorov, 2010), we found that a mediating model quite similar to Hall and Fincham (2008) elicited the best fitting and most parsimonious model when we included both dispositional and trait measures in a structural model.

References

Arbuckle, J. L. (2007). Amos 16.0 User's Guide. Chicago, IL: Amos Development Corporation.

Avery, C. M. (2008). The relationship between self-forgiveness and health: Mediating variables and implications for well-being (Doctoral dissertation). Dissertation Abstracts International: Section B. Sciences and Engineering, 69, 1939.

Barbetta, F. (2002). An exploration of the relationship between empathy, forgiving others and self-forgiveness. (Doctoral dissertation). Dissertation Abstracts International: Section B. Sciences and Engineering, 64(1), 407B.

Batson, C. D., Bolen, M. H., Cross, J. A., & Neuringer-Benefiel, H. E. (1986). Where is the altruism in the altruistic personality? Journal of Personality and Social Psychology, 50, 212-220. doi: 10.1037/0022-3514.50.1.212

Bauer, L., Duffy, J., Fountain, E., Halling, S., Holzer, M., & Jones, E.,Rowe, J. E. (1992). Exploring self-forgiveness. Journal of Religion and Health, 31, 149-160.

Coke, J. S., Batson, C. D., & McDavis, K. (1978). Empathic mediation of helping: A two-stage model. Journal of Personality and Social Psychology, 36, 752-766. doi: 10.1037/0022-3514.36.7.752

Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting and Clinical Psychology, 24, 349-354.

Davis, M. (1980). A multidimensional approach to individual differences in empathy. Catalog of Selected Documents in Psychology, 10, 85-103.

Davis, M. (1983). Measuring individual differences in empathy: Evidence for a multidimensional approach. Journal of Personality and Social Psychology, 44, 113-126. doi: 10.1037/0022-3514.44.1.113

Dillon, R. S. (2001). Self-forgiveness and self-respect. Ethics, 112, 53-83.

Enright, R. D. (1996). Counseling within the forgiveness triad: On forgiving, receiving forgiveness, and self-forgiveness. Counseling and Values, 40, 107-127.

Exline, J. J., DeShea, L., & Holeman, V. T. (2007). Is apology worth the risk? Predictors, outcomes, and ways to avoid regret. Journal of Social and Clinical Psychology, 26, 479-504. doi: 10.1521/jscp.2007.26.4.479

Feshbach, N. D. (1975). Empathy in children: Some theoretical and empirical considerations. The Counseling Psychologist, 5(2), 25-30. doi: 10.1177/001100007500500207

Fisher, M. L., & Exline, J. J. (2006). Self-forgiveness versus excusing: The roles of remorse, effort, and acceptance of responsibility. Journal of Self and Identity, 5, 127-146. doi: 10.1080/15298860600586123

Flanigan, B. (1996). Forgiving yourself: A step-by-step guide to making peace with your mistakes and getting on with your life. New York: Macmillan.

Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects in counseling psychology research. Journal of Counseling Psychology, 51, 115-134. doi: 10.1037/0022-0167.51.1.115

Griswold, C. L. (2007). Forgiveness: A philosophical exploration. New York: Cambridge University Press.

Hall, J. H., & Fincham F. D. (2005). Self-forgiveness: The stepchild of forgiveness research. Journal of Social and Clinical Psychology, 24, 621-637. doi: 10.1521/jscp.2005.24.5.621

Hall, J. H., & Fincham F. D. (2008). The temporal course of self-forgiveness. Journal of Social and Clinical Psychology, 27, 174-202. doi: 10.1521/jscp.2008.27.2.174

Harder, D. W., Rockart, L., & Cutler, L. (1993). Additional validity evidence for the Harder Personal Feelings Questionnaire-2 (PFQ2): A measure of shame and guilt proneness. Journal of Clinical Psychology, 49, 345-348.

Harder, D. W., & Zalma, A. (1990). Two promising shame and guilt scales: A construct validity comparison. Journal of Personality Assessment, 55(3 & 4), 729-745. doi: 10.1207/s15327752jpa5503&4_30

Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. doi: 10.1080/10705519909540118

Hughes, P. M. (1994). On forgiving oneself: A reply to Snow. The Journal of Value Inquiry, 28, 557-560.

Ingersoll-Dayton, B., & Krause, N. (2005). Self-forgiveness: A component of mental health in later life. Research on Aging, 27, 267-289. doi: 10.1177/0164027504274122

Kline, R. B. (2005). Principles and Practice of Structural Equation Modeling (2nd ed.). New York: The Guilford Press.

Layer, S. D., Roberts, C., Wild, K., & Walters, J. (2004). Postabortion grief: Evaluating the possible efficacy of a spiritual group intervention. Research on Social Work Practice, 14, 344-350. doi: 10.1177/1049731504265829

Lewis, H. B. (1971). Shame and guilt in neurosis. New York: International University Press.

Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F. (2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling, 9, 151-173.

Macaskill, A., Maltby, J., & Day, L. (2002). Forgiveness of self and others and emotional empathy. The Journal of Social Psychology, 142, 633-665. doi: 10.1080/00224540209603925

Marschall, D., Sanfter, J., & Tangney, J. P. (1994). The State Shame and Guilt Scale. George Mason University, Fairfax, VA.

Marsh, H. W., Hau, K., & Wen, Z. (2004). In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler's (1999) findings. Structural Equation Modeling, 11, 320-341. doi: 10.1207/s15328007sem1103_2

Martin, A. M. (2008). Exploring forgiveness: The relationship between feeling forgiven by God and self-forgiveness for an interpersonal offense. (Doctoral dissertation). Retrieved from University Microforms International (3321224).

McConnell, J. M., & Dixon, D. N. (2012). Perceived forgiveness from God and self-forgiveness. Journal of Psychology and Christianity.

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

McCullough, M. E., Rachal, K. C., Sandage, S. J., Worthington, E. L. Jr., Brown, S. W., & Hight, T. L. (1998). Interpersonal forgiving in close relationships: II. Theoretical elaboration and measurement. Journal of Personality and Social Psychology, 75, 1586-1603. doi: 10.1037/0022-3514.75.6.1586

Micceri, T. (1989). The unicorn, the normal curve, and other improbable creatures. Psychological Bulletin, 105, 156-166. doi: 10.1037/0033-2909.105.1.156

Olshtain, E. (1989). Apologies across languages. In S. Blum-Kulka, J. House, & G. Kasper (Eds.), Cross-cultural pragmatics: Requests and apologies (pp. 157). Norwood, NJ: Ablex.

Rangganadhan, A. R., & Todorov, N. (2010). Personality and self-forgiveness: The roles of shame, guilt, empathy, and conciliatory behavior. Journal of Social and Clinical Psychology, 29, 1-22. doi: 10.1521/jscp.2010.29.1.1

Rusch, N., Corrigan, P. W., Bohus, M., Jacob, G. A., Brueck, R., & Lieb, K. (2007). Measuring shame and guilt by self-report questionnaires: A validation study. Psychiatry Research, 150, 313-325. doi: 10.1016/j.psychres.2006.04.018

Russell, D. (1982). The causal dimension scale: A measure of how individuals perceive causes. Journal of Personality and Social Psychology, 42, 1137-1145. doi: 10.1037/0022-3514.42.6.1137

Spencer, S. J., Zanna, M. P., & Fong, G. T. (2005). Establishing a causal chain: Why experiments are often more effective than meditational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89, 845-851. doi: 10.1037/0022-3514.89.6.845

Strelan, P. (2007). Who forgives others, themselves, and situations? The roles of narcissism, guilt, self-esteem, and agreeableness. Personality and Individual Differences, 42, 259-269. doi: 10.1016/j.paid.2006.06.017

Tangney, J. P. (1995). Recent advances in the empirical study of shame and guilt. American Behavioral Scientist, 38, 1132-1145. doi: 10.1177/0002764295038008008

Tangney, J. P., Barlow, D. H., Wagner, P. E., Marschall, D. E., Borenstein, J. K., Santfer, J., ... Gramzow, R. (1996). Assessing individual differences in constructive vs. destructive responses to anger across the lifespan. Journal of Personality and Social Psychology, 70, 780-796. doi: 10.1037/0022-3514.70.4.780

Tangney, J. P., Boone, A. L., & Dearing, R. L. (2005). Forgiving the self: Conceptual issues and empirical findings. In E. L., Worthington, Jr. (Ed.), Handbook of forgiveness (pp.143-158). New York: Routledge.

Tangney, J. P., Boone, A. L., Dearing, R. L., & Reinsmith, C. (2002, August). Individual differences in the propensity to forgive: Measurement and implications for psychological and social adjustment. Paper presented at the 1999 American Psychological Association Convention, Boston, MA.

Tangney, J. P., Dearing, R. L., Wagner, P.E., & Gramzow, R. (2000). The Test of Self-Conscious Affect--3 (TOSCA-3). George Mason University, Fairfax VA.

Tangney, J. P., & Dearing, R. L. (2002). Shame and guilt. New York: The Guilford Press.

Tangney, J. P., Miller, R. S., Flicker, L., & Barlow, D. H. (1996). Are shame, guilt, and embarrassment distinct emotions? Journal of Personality and Social Psychology, 70, 1256-1269.

Tangney, J. P., Wagner, P. E., & Gramzow, R. (1992). Proneness of shame, proneness to guilt, and psychopathology. Journal of Abnormal Psychology, 103, 469-478. doi: 10.1037/0021 843X.101.3.469

Thompson, L. Y., Snyder, C. R., Hoffman, L., Michael, S. T., Rasmussen, H. N., Billings, L. S., . Roberts, D. E. (2005). Dispositional forgiveness of self, others, and situations. Journal of Personality, 73, 313-360. doi: 10.1111/j.1467-6494.2005.00311.x

Weiner, B., Graham, S., Peter, O., & Zmuidinas, M. (1991). Public confession and forgiveness. Journal of Personality, 59, 281-312. doi: 10.1111/j.1467-6494.1991.tb00777.x

Witvliet, C. V., Ludwig, T. E., & Bauer, D. J. (2002). Please forgive me: Transgressors' emotions and physiology during imagery of seeking forgiveness and victim responses. Journal of Psychology and Christianity, 21, 219-233.

Wohl, M. J. A., DeShea, L., & Wahkinney, R. L. (2008). Looking within: Measuring state self-forgiveness and its relationship to psychological wellbeing. Canadian Journal of Behavioural Science, 40, 1-10. doi: 10.1037/0008-400x.40.1.1.1

Worthington, E. L. (2006). Forgiveness and reconciliation: Theory and application. New York: Routledge.

Zechmeister, J. S., & Romero, C. (2002). Victim and offender accounts of interpersonal conflict: Autobiographical narratives of forgiveness and unforgiveness. Journal of Personality and Social Psychology, 82, 675-686. doi: 10.1037/0022-3514.82.4.675

Table 1--Summary Table of Scale Properties

 [alpha] X SD VIF

 TSD-TP .89 59.53 11.63
 TSD-VP .77 60.74 13.82
 CDS .45- ([dagger]) ([dagger])
 TOSCA-3:PE .73 40.65 8.53
 MFI:BO .82 16.36 6.35
 MFI:BS .76 18.82 5.47
 IRI:EC .70 27.69 4.03
 IRI:PT .78 24.70 4.97
 CERS .87 20.38 8.45
 SSGS: Shame .88 12.94 5.45
 PFQ: Shame .84 27.51 7.84
 TOSCA-3: .76 50.22 9.17
 Shame
 SGGS: Guilt .83 16.61 5.02
 PFQ: Guilt .81 19.31 5.23
TOSCA-3: Guilt .73 65.70 6.84
(Parcel 1) CBS .80 16.93 3.95
(Parcel 2) CBS .81 11.79 2.88
 MFI:AF .76 32.90 5.45
 TRIM .95 61.20 18.83
 (Parcel 1) .80 10.55 3.18
 HFSO
 (Parcel 2) .81 10.49 3.49
 HFSO
 (Parcel 3) .80 10.74 3.16
 HFSO
 MFI:FS .72 22.60 5.25
 SSFA .90 21.30 5.72
 SFB .94 27.96 6.42
 HFSS .74 30.37 6.52
 MC-SDS .78KR 13.20 3.00

 [alpha] Skewness Kurtosis

 TSD-TP 3.91 -.22 .34
 TSD-VP 3.96 -.14 -.14
 CDS ([dagger]) ([dagger]) ([dagger])
 TOSCA-3:PE 1.63 .32 * .38
 MFI:BO 1.60 .74 * .07
 MFI:BS 1.72 .43 * .16
 IRI:EC 1.62 -.43 * .01
 IRI:PT 1.41 -.16 -.14
 CERS 1.37 .20 -.65 *
 SSGS: Shame 3.36 .30 * -.86 *
 PFQ: Shame 2.39 -.25 * -.36
 TOSCA-3: 1.94 -.23 -.30
 Shame
 SGGS: Guilt 3.16 -.33 * -.57 *
 PFQ: Guilt 2.82 -.28 * 0.32
TOSCA-3: Guilt 2.10 -.73 * .87 *
(Parcel 1) CBS 2.84 -1.25 * 1.43 *
(Parcel 2) CBS 2.73 -1.66 * 2.35 *
 MFI:AF 1.25 -.70 * -.07
 TRIM 2.21 -.41 * -.95 *
 (Parcel 1) 2.96 -1.04 * .50 *
 HFSO
 (Parcel 2) 2.31 -.76 * -.41
 HFSO
 (Parcel 3) 3.55 -.84 * -.05
 HFSO
 MFI:FS 1.67 .33 * .05
 SSFA 3.36 -.06 -.71 *
 SFB 3.01 -.62 * -.39
 HFSS 1.93 -.18 -.45
 MC-SDS 1.12 .14 -.50 *

Note. [alpha] = Cronbach's Alpha; KR = Kuder-Richardson-20
Alpha; X = Mean; SD = Standard Deviation; VIF = Multicollinearity.
Items were parceled in all analyses in order to provide enough
indicators for the latent variable.

* Indicates significant non-normality. ([dagger]) The researcher
excluded CDS from all analyses due to poor internal consistency
reliabilities.

Table 3--Summary Table of Goodness-of-Fit Indices

 [chi square] df P [chi square]/df CFI

H&F (2005) 1563.73 262 .000 5.97 .730
H&F (2008) 358.14 98 .000 3.7 .918
Alternative 360.20 100 .000 3.6 .918

 TLI RM-SEA RM-SEA (CI) AIC

H&F (2005) .691 .111 .105-.116 1689.73
H&F (2008) .899 .051 .072-.090 434.14
Alternative .901 .080 .071-.089 432.20

 MIC BIC ABIC Multivariate
 Normality

H&F (2005) -- 1942.13 -- 51.03 *
H&F (2008) -1255.59 586.38 -1355.75 33.62 *
Alternative -1.94 576.43 -9.95 33.62 *

Note. * Indicates significant non-normality