This is an open access article distributed under the terms of the Creative Commons
Attribution License (“To forgive is the highest, most beautiful form of love. In return, you will receive untold peace and happiness.” Robert Muller (1998, p. 6)
Public discourse surrounding forgiveness provides considerable anecdotal evidence that Western society views forgiveness through a very positive lens. Popular aphorisms cast forgiveness as divine, self-help books extoll its praises, and experts expound on its benefits. We agree that there is much to like about forgiveness; nevertheless, we explore the possibility that people’s views on forgiveness are more complex than such discourse captures. We argue that attitudes toward forgiveness are ambivalent, and we explore whether such ambivalence predicts variation in people’s propensity to forgive and their tendency to view forgiveness as desirable/virtuous. Such ambivalence might have important but as yet unacknowledged implications for understanding whether and when people forgive.
Because humans are profoundly social beings, factors that shape whether and when people forgive are of special significance in understanding relationships. Relationships are critical for our health and wellbeing, fulfill important needs for belonging and felt security, and contribute to our happiness and life satisfaction (Baumeister & Leary, 1995; Kiecolt-Glaser & Newton, 2001). Relationships are also, however, a common context in which the need for forgiveness arises. The interdependence that characterizes our closest relationships, in particular, makes conflicts of interest almost unavoidable and places relationship partners at considerable risk of causing hurt or injury (Fincham, 2000; Murray & Holmes, 2011). If attitudes toward forgiveness are ambivalent, that fact is certain to have important implications for relationship maintenance.
An attitude is ambivalent when inconsistencies between and/or within its different components result in the co-existence of positive (favorable) and negative (unfavorable) evaluations of the attitude object (Thompson et al., 1995). Researchers have not yet demonstrated empirically that people’s attitudes toward forgiveness are ambivalent, but anecdotal evidence and research findings are consistent with this possibility. Lay views commonly portray forgiveness as a virtue and condemn those who withhold forgiveness or seek vengeance (e.g., McCullough, 2008) and research suggests that forgiveness contributes to the health of body, mind, and relationships (e.g., Bono et al., 2008; Karremans & Van Lange, 2004; Lawler-Row et al., 2008). At the same time, laypeople believe there are limits to what deserves to be forgiven (Macaskill, 2005a), believe forgiving can cause emotional and other problems (Kanz, 2000), and identify costs associated with forgiving (Strelan et al., 2016). Additionally, research highlights contexts in which forgiving may be maladaptive (e.g., Gordon et al., 2004; McNulty, 2010) or compromise a victim’s ability to feel safe and respected (Luchies et al., 2010).
The present paper presumes that whether people’s attitudes toward forgiveness are ambivalent may be of both empirical and practical consequence. Correctly forecasting an individual’s response to wrongdoing may, for example, be considerably more difficult when their attitudes toward forgiveness are ambivalent. Highly ambivalent attitudes are subject to greater variability over time and changing circumstances than attitudes low in ambivalence because they are both weaker (Conner & Sparks, 2002; Jonas et al., 2000) and less temporally stable (Krosnick & Petty, 1995). In addition, when attitudes are ambivalent, contextual cues can prime either the favorable or unfavorable aspects of these attitudes in relative isolation from each other, resulting in corresponding changes in orientation toward the attitude object depending on the particular aspects primed (Bell & Esses, 1997). Evidence that attitudes toward forgiveness are ambivalent may thus have implications for our ability to explain and predict whether and when people will forgive wrongdoings.
If ambivalent attitudes are weaker, less stable, and more context dependent than those that are univalent, ambivalence in attitudes toward forgiveness may also diminish individuals’ actual propensity to forgive. As indicated previously, research suggests that forgiveness may confer important benefits for victims’ physical and psychological wellbeing and the health and quality of their relationships with their offenders. Should ambivalence in attitudes toward forgiveness attenuate individuals’ readiness or inclination to forgive, it may have very real consequences for their private and personal lives, particularly their health, wellbeing, and capacity to maintain relationships with those who wrong them.
Does Attitudinal Ambivalence Toward Forgiveness Predict Variation in Important Forgiveness-Related Variables?
Our literature search revealed just two studies examining ambivalence and forgiveness. Kachadourian et al. (2005) assessed whether ambivalence in attitudes toward marital partners predicted forgiveness of a recalled transgression. Kline Rhoades et al. (2007) operationalized ambivalence as participants’ uncertainty about whether they would/could forgive those responsible for September 11th and compared the self-reported psychological wellbeing of participants classified as “ambivalent” with those classified as anti- or pro-forgiveness.
The results of both studies support the proposition that ambivalence may matter for forgiveness. Ambivalence in attitudes toward an offender was negatively associated with forgiveness for both husbands and wives in Kachadourian et al. (2005), for instance, even controlling for severity of the recalled offenses and frequency of rumination. In Kline Rhoades et al. (2007), individuals who were ambivalent about whether they would or could forgive those responsible for 9/11 reported greater psychological distress three to six weeks after the attacks than those who opposed or endorsed forgiving the terrorists. Together, these findings highlight the value in exploring points of intersection between ambivalence and forgiveness. They provide little basis, however, for generating hypotheses about whether ambivalence toward forgiveness might be associated with individuals’ readiness to forgive.
Fortunately, findings from the broader ambivalence literature offer a foundation for such predictions. First, research shows that ambivalence is associated with less extreme and less certain attitudes (Conner & Sparks, 2002). Those with highly ambivalent attitudes toward forgiveness should thus be less polarized and less certain in their evaluations of forgiveness than those with more uniformly positive attitudes. If so, increasing attitudinal ambivalence should be associated with reduced readiness to forgive, more open attitudes toward vengeance, and less idealized views of forgiving because individuals high in ambivalence toward forgiveness should be more lukewarm about forgiveness and therefore more hesitant to make judgments or take actions that are charitable toward the offender.
Second, ambivalence is most likely when an attitude object produces both strong negative and strong positive evaluations (Conner & Sparks, 2002). The co-existence of strong positive and strong negative evaluations can produce psychological tension when they are simultaneously accessible (Newby-Clark et al., 2002). On its own or in combination with the heightened uncertainty characteristic of ambivalent attitudes, such tension might dampen people’s enthusiasm toward forgiving and/or leave them reluctant to indulge in judgments or courses of action that grant the offender the benefit of the doubt. Compared to those whose attitudes toward forgiveness are more uniformly positive, then, those with ambivalent attitudes may be slower to forgive, more open to vengeance, and less inclined to idealize forgiving as virtuous and admirable.
Overview and Hypotheses
As part of a larger study, participants completed measures of ambivalence toward forgiveness and scales assessing their tendencies to be forgiving versus vengeful and to view forgiveness as moral and desirable. We posed two hypotheses:
H1: Attitudes toward forgiveness will exhibit evidence of ambivalence.
H2: The more ambivalent their attitudes toward forgiveness, the less participants will report being inclined to forgive, the more they will report being inclined to seek vengeance, and the less they will endorse beliefs that forgiving is a morally desirable response to wrongdoing.
Affective ambivalence exists when people experience both positive and negative emotions toward the attitude object, thus feeling “torn” (e.g., wanting to forgive but fearing that forgiving might result in further hurt). Cognitive ambivalence occurs when people hold mixed or conflicting beliefs about the attitude object (e.g., believing that forgiving is a virtue while believing that an offense is unforgivable). In addition to experiencing intracomponental ambivalence, or inconsistency within a single component of the attitude, people also experience intercomponental ambivalence, or inconsistency between attitude components, when their hearts and their heads conflict (e.g., not wanting to forgive but believing that forgiving is necessary to move forward). The literature further discriminates between subjective and objective ambivalence (e.g., Conner & Sparks, 2002). Subjective ambivalence refers to the phenomenological experience of ambivalence and is typically measured directly by asking people to rate the degree to which their thoughts and feelings toward the attitude object conflict. Objective ambivalence taps into inconsistency in evaluations as assessed indirectly through measures that gauge people’s thoughts and feelings concerning the attitude object and allows for the possibility that inconsistency may exist even when individuals are unaware of it (e.g., Jonas et al., 2000). We measured each of these kinds of ambivalence.
In sum, this study makes two important contributions to the literature on forgiveness. First, it permits empirical exploration of the possibility that, despite considerable pro-forgiveness discourse in our society, people’s attitudes toward forgiveness may be more nuanced than this discourse often supposes. Few studies have examined lay conceptualizations of forgiveness (for exceptions see Carr & Wang, 2012; Friesen & Fletcher, 2007; Kanz, 2000; Kearns & Fincham, 2004). None have directly examined ambivalence in forgiveness attitudes. The present study thus extends the literature on lay views of forgiveness by examining the complexity of these views. Second, through investigating whether increased ambivalence in attitudes toward forgiveness is associated with being less disposed to forgive, more inclined to retaliate, and less prone to perceive forgiving as admirable and worthy, this study provides a foundation for future research directed at examining whether such ambivalence may have theoretical and practical implications (e.g., does ambivalence predict/explain whether and when people will forgive? Does it influence personal and relational wellbeing in turn?).
Method
Participants
Undergraduates from a university in western Canada (N = 159; 86 females) received partial course credit for completing an online survey about attitudes toward forgiveness. On average participants were 20.7 years old (SD = 3.60, range 17.1 to 24.3). The majority self-identified as European/White (approximately 60%), Chinese (22%), or South Asian (11%); 7% reported other ethnicities.
Materials and Procedure
We assessed ambivalence, vengefulness, and dispositional forgiveness with multiple scales to determine whether our findings converged across measures capturing different aspects of these constructs. Unless otherwise noted, items employed a 1 completely disagree to 7 completely agree scale. We describe the measures below in the order in which they appeared. Tables 1 and 2 present reliabilities and descriptive statistics.
Table 1
Descriptive Statistics and Reliabilities: Tendencies to Forgive, Seek Revenge, and View Forgiving as Desirable
| Statistic | Variable
|
|||||
|---|---|---|---|---|---|---|
| Mullet Forgiveness Scale
|
TTF | ATF | Vengeance | |||
| Blockage | Circumstance | Forgiving | ||||
| M | 2.36 | 4.90 | 3.75 | 3.45 | 5.01 | 3.00 |
| SD | 1.07 | 0.82 | 1.20 | 1.22 | 0.95 | 0.91 |
| Range | 1–5.80 | 1.43–6.57 | 1–6.50 | 1–6.67 | 2.40–6.80 | 1.35–5.83 |
| Alpha | .86 | .65 | .85 | .73 | .63 | .85 |
Note. TTF = Tendency to Forgive; ATF = Attitudes Toward Forgiveness.
Table 2
Descriptive Statistics and Reliabilities: Ambivalence Measures
| Statistic | Ambivalence Measure
|
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Subjective Ambivalence
|
Objective Thompson
|
Objective Maio
|
|||||||||
| Intra-component
|
Inter-component | ||||||||||
| Cacioppo | P&P | Jamieson | Average | Cognitive | Affective | Average | Cognitive | Affective | Average | ||
| M | 3.87 | 4.02 | 3.63 | 3.83 | 0.98 | 1.85 | 1.41 | 24.82 | 28.37 | 26.58 | 24.13 |
| SD | 1.20 | 1.48 | 1.65 | 1.32 | 2.31 | 2.24 | 2.00 | 8.80 | 8.40 | 6.58 | 5.58 |
| Range | 1.33–6.44 | 1–7 | 1–7 | 1.11–6.74 | -2–7 | -2–7 | -2–6 | 3–46 | 8–48 | 9–44.50 | 9–40.50 |
| Theoretical Range | 1–7 | 1–7 | 1–7 | 1–7 | -2–7 | -2–7 | -2–7 | 0–60 | 0–60 | 0–60 | 0–60 |
| Alpha | .89 | .87 | .82 | .93 | — | — | .79 | — | — | — | — |
Note. P&P = Petty & Priester measure; Average = mean of the three subjective ambivalence measures.
Ambivalence
Two measures assessed objective ambivalence. First, participants completed a task adapted from Maio et al. (2000). To measure cognitive ambivalence, participants listed up to ten words or phrases that described their thoughts/beliefs about forgiveness. They then rated the valence of each word or phrase using a -3 very negative to +3 very positive scale. Participants repeated the procedure for affective ambivalence, listing words or phrases that described their feelings about forgiveness.
We generated scores for both intra- and intercomponent ambivalence using formulas in Maio et al. (2000). To calculate cognitive ambivalence, for example, we calculated the sum of the ratings for the thoughts rated negative in valence (N) and the sum of the ratings for the thoughts rated positive in valence (P) and then applied the following formula:
Intracomponent Ambivalence = P + |N| - 2 * |P + N| + 30.
Following their lead, we also calculated average intracomponent ambivalence (i.e., the mean of affective and cognitive ambivalence). We then generated intercomponent ambivalence scores by calculating net evaluations for both the cognitive and affective components (i.e., the sums of the ratings for the thoughts (B) and feelings (F) participants listed) and applying the following formula:
Intercomponent Ambivalence = (|B| + |F| - 2 * |B + F| + 60)/2
Theoretically, scores on the Maio measures range from 0 to 60. Higher scores indicate greater ambivalence.
Next, participants completed a measure adapted from Thompson et al. (1995) that addressed feelings toward (i.e., affective component), thoughts about (i.e., cognitive component), and overall evaluations of forgiveness. For the three items assessing positive aspects of their attitudes, participants answered considering only their positive feelings/thoughts/evaluation while ignoring their negative feelings/thoughts/evaluation (e.g., “Considering only your feelings of satisfaction toward forgiving and ignoring your feelings of dissatisfaction, how satisfied do you feel about forgiving?”). They then answered the same questions considering only their negative feelings/thoughts/evaluation, ignoring their positive feelings/thoughts/evaluation. Participants answered each question using a 1 not at all to 7 very scale. We excluded the overall evaluation judgments from analysis because there was no equivalent index from the Maio measure.
We used the formula in Thompson et al. (1995) to generate intracomponent ambivalence scores (Thompson et al. did not provide a formula for intercomponent ambivalence):
Ambivalence = (P + N)/2 − |P − N|
P refers to the sum of the values for the positive ratings and N refers to the sum of the values for the negative ratings. Scores fall on a theoretical scale from -2 to +7. Higher scores indicate greater ambivalence. Zero indicates low ambivalence, not no ambivalence. To facilitate comparison with the Maio measure, we also computed an average Thompson objective ambivalence index.
Finally, participants completed three subjective ambivalence measures adapted to a forgiveness context. First, using a 1 very slightly or not at all to 7 extremely scale, participants completed nine items from Cacioppo et al. (1997) in which they rated how much they experienced various feelings (e.g., muddled, conflicted) in response to the question “When you think about your attitude toward forgiveness, to what extent do you feel…?” Next, participants completed three items (e.g., “When I think about my attitude toward forgiveness, I experience: 1 completely one-sided reactions to 7 completely mixed reactions”) from Priester and Petty (1996). Finally, they used a 1 extremely uncharacteristic of my attitude toward forgiveness to 7 extremely characteristic of my attitude toward forgiveness scale to complete three items (e.g., “I find myself feeling torn about whether my attitude towards forgiveness is positive or negative”) developed by Jamieson (1993; cited in Thompson et al., 1995). As the three subjective ambivalence measures were highly correlated (rs .71 to .80), we calculated an average subjective ambivalence index.
Attitudes Toward Forgiveness Scale
The 6-item Attitudes Toward Forgiveness Scale (ATF; Brown, 2003) gauges the extent to which individuals view forgiving as desirable (e.g., “I believe that forgiveness is a moral virtue”). Higher values on the index created by averaging items reflect a tendency to value forgiveness.
Tendency to Forgive Scale
The 4-item Tendency to Forgive Scale (TTF; Brown, 2003) assesses how readily individuals forgive (e.g., “I tend to get over it quickly when someone hurts my feelings”). Higher values on the index created by averaging items reflect a tendency to be charitable or quick to forgive. Contrasted with the ATF, the TTF measures people’s general tendency to be forgiving, rather than their beliefs about the inherent value of forgiving.
Mullet Forgivingness Scale
The Mullet Forgivingness Scale (MFS; Mullet et al., 1998) gauges several aspects of forgivingness. The 5-item Forgiveness Blockage subscale (Blockage) assesses lack of the capacity to forgive (e.g., “The way I consider the world has brought me to never forgive”); the 7-item Personal and Social Circumstances subscale (Circumstances) measures the extent to which circumstances influence one’s capacity to forgive (e.g., “I feel it is easier to forgive when my family or my friends have invited me to do so”); and the 6-item Revenge versus Forgiveness subscale (Forgiving) assesses a tendency to forgive rather than seek revenge (e.g., “I think it is better to forgive than to seek revenge”). After combining items for each subscale, higher scores indicate a reduced capacity to forgive, increased susceptibility to the influence of conditions that promote forgiving, and a dispositional inclination toward forgiving rather than seeking revenge, respectively.
Vengeance Scale
The 20-item Vengeance Scale (Stuckless & Goranson, 1992) assesses pro-vengeance attitudes (e.g., “revenge is morally wrong,” reverse-scored) and the proclivity to be vengeful (e.g., “It is important for me to get back at people who have hurt me”). Higher values on the index created by averaging items reflect greater dispositional vengefulness.
Descriptive statistics and reliabilities for the forgiveness/vengeance indices are reported in Table 1.
Results
Are Attitudes Toward Forgiveness Ambivalent?
H1 predicted that individuals would exhibit ambivalence in their attitudes toward forgiveness. Three lines of evidence support this hypothesis.
First, the means for the ambivalence measures consistently fell toward the middle or lower middle of their respective theoretical ranges rather than the low end. As Table 2 shows, this is true whether we consider subjective (direct) or objective (indirect) indices of ambivalence, affective or cognitive ambivalence (i.e., intracomponent ambivalence), or intercomponent ambivalence.
Second, levels of objective ambivalence observed here are comparable to or exceed levels of objective ambivalence reported for attitudes toward social issues such as euthanasia, capital punishment, and abortion, as well as attitudes toward nuclear power and parents—all attitude objects for which we might expect individuals to exhibit ambivalence. For example, rescaling the mean score for our Thompson cognitive ambivalence index (the lower of our two Thompson ambivalence indices) from our -2 to 7 metric to Thompson and Zanna’s -.5 to 4 metric produces a value of 0.99. Thompson and Zanna reported mean ambivalence scores ranging from .23 (for drinking and driving) to 1.39 (for both AIDS and euthanasia), with a median of 0.84 (Thompson et al., 1995, report separate ambivalence means for each of the three ratings participants made. As indicated in the method, we collapsed across the three ratings to create a single composite index).
Our mean score on the Thompson affective ambivalence index is also comparable in magnitude to the mean level of ambivalence in attitudes toward nuclear power reported in Visschers and Wallquist (2013; Ms = 1.87 and 1.67, respectively, assessed prior to and two weeks after the Fukushima accident using the -2 to 7 metric we used; note that Visschers & Wallquist, 2013, used the same formula to calculate ambivalence scores that Thompson et al. did, thus placing their values and our values on a common metric. However, they measured ambivalence using a 12-item scale assessing attitudes toward nuclear power).
Similarly, levels of objective ambivalence on the Maio measure in our study are comparable to, if not higher than, levels of ambivalence in attitudes toward parents reported in Maio et al. (2000; Ms = 22.83 and 24.42 for fathers and mothers, respectively, collapsing across intra- and intercomponent indices; no study in our literature review reported mean scores for subjective ambivalence).
Third, one-sample t-tests comparing mean scores on the overall/average ambivalence measures (selected for comparison to reduce the number of statistical tests performed) for the subjective, Thompson, and Maio indices against their lowest possible theoretical values (see Table 2) revealed significant differences in every case. Admittedly, these t-tests provide a crude test of the magnitude of ambivalence in participants’ responses. Nevertheless, the results are congruent with the pattern expected if participants’ scores reflected ambivalence: for overall subjective ambivalence t(155) = 26.67, p < .001, 95% CI [2.62, 3.04]; average Thompson t(156) = 23.91, p < .001, 95% CI [3.32, 3.91]; average (intracomponent) Maio t(158) = 50.95, p < .001, 95%CI [25.55, 27.61]; intercomponent Maio t(153) = 53.66, p < .001, 95% CI [23.24, 25.02]).
Note too that, with a single exception (the nonsignificant correlation between Thompson affective ambivalence and Maio cognitive ambivalence), the ambivalence measures were consistently and moderately intercorrelated (see Table 3). Our data thus provide converging evidence of ambivalence in attitudes toward forgiveness across different markers of and types of ambivalence.
Table 3
Correlations and Descriptive Statistics
| Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Blockage | — | |||||||||||||
| 2. Circumstances | -.06 | — | ||||||||||||
| 3. Forgiving | -.49* | -.06 | — | |||||||||||
| 4. TTF | -.40* | -.02 | .45* | — | ||||||||||
| 5. ATF | -.52* | .04 | .52* | .38* | — | |||||||||
| 6. Vengeance | .59* | .03 | -.49* | -.43* | -.60* | — | ||||||||
| 7. Average Subjective | .45* | .05 | -.42* | -.46* | -.39* | .38* | — | |||||||
| 8. Cognitive Thompson | .34* | -.06 | -.31* | -.27* | -.43* | .38* | .40* | — | ||||||
| 9. Affective Thompson | .20* | -.01 | -.21* | -.17* | -.24* | .16* | .38* | .56* | — | |||||
| 10. Average Thompson | .32* | -.05 | -.30* | -.24* | -.37* | .32* | .47* | .80* | .88* | — | ||||
| 11. Cognitive Maio | .14 | -.12 | -.12 | -.15 | -.22* | .17* | .27* | .26* | .13 | .22* | — | |||
| 12. Affective Maio | .24* | -.09 | -.16* | -.17* | -.15 | .23* | .36* | .26* | .26* | .27* | .18* | — | ||
| 13. Average Maio | .24* | -.14 | -.18* | -.21* | -.24* | .26* | .40* | .32* | .24* | .32* | .78* | .76* | — | |
| 14. Intercomponent Maio | .28* | -.12 | -.22* | -.21* | -.18* | .24* | .36* | .29* | .26* | .28* | .54* | .42* | .62* | — |
Note. TTF = Tendency to Forgive; ATF = Attitudes Toward Forgiveness.
*p < .05.
Two additional findings warrant mention. First, participants displayed greater affective than cognitive ambivalence on both the Thompson and Maio objective ambivalence measures (Thompson: t(156) = 5.12, p < .001, d =.46; Maio: t(153) = 4.16, p < .001, d =.32). This suggests that individuals experience more intense conflict among their feelings toward forgiveness than their beliefs about it. Second, males and females did not differ on any ambivalence measure (all p’s > .29).
Does Attitudinal Ambivalence Regarding Forgiveness Predict Readiness to Forgive?
H2 predicted that ambivalence in attitudes toward forgiveness would be associated with people’s general tendencies to be forgiving versus vengeful and to view forgiveness as desirable and virtuous. We predicted that people would be less dispositionally forgiving, more inclined to seek revenge than grant forgiveness, and less disposed to view forgiveness as a virtue the more ambivalent their forgiveness attitudes.
Our findings generally supported this prediction (see Table 3). Each ambivalence measure except Maio cognitive ambivalence was significantly negatively correlated with both the TTF and the Forgiving subscale of the MFS, both of which gauge dispositional forgiveness. Similarly, except for a nonsignificant correlation between Maio cognitive ambivalence and MFS Blockage, the ambivalence measures were significantly positively correlated with the Vengeance Scale, which assesses dispositional vengefulness, and the MFS Blockage subscale, which assesses diminished capacity to forgive. Finally, each ambivalence index, except Maio affective ambivalence, was significantly negatively correlated with the ATF, indicating that participants viewed forgiveness as less desirable and virtuous as their ambivalence toward forgiveness increased. Interestingly, the MFS Circumstances subscale did not correlate with any ambivalence measure (nor with any key variable in this study). This may not be surprising, however, as it gauges the propensity to forgive under particular circumstances—such as when encouraged by others to do so—rather than a general propensity to forgive.
Ancillary Analyses
To gain a deeper understanding of participants’ attitudes toward forgiveness, we investigated potential differences between individuals scoring very high and very low on ambivalence (cf., Bell & Esses, 1997). More particularly, we compared those whose attitudes were highly ambivalent with those with anti-forgiveness and pro-forgiveness attitudes.
We assigned participants to high versus low ambivalence groups based on their scores on the average subjective ambivalence index (average subjective) and the two average intracomponent measures of objective ambivalence. We first performed a tertile split on each ambivalence measure, assigning participants in the bottom third of the distribution to a low ambivalence group and participants in the top third to a high ambivalence group (using both the Thompson & Maio average intracomponent indices allowed us to assess convergence across measures of objective ambivalence. Including average subjective ambivalence allowed us to assess convergence across subjective and objective ambivalence. To reduce the risk of Type I error, and because intercomponent ambivalence was available only for the Maio measure, we excluded the Maio intercomponent index of ambivalence from these analyses).
Next, we used a median split on the ATF to create pro- and anti-forgiveness groups for each ambivalence index, assigning low ambivalence participants to two groups: those perceiving forgiveness as relatively less desirable and virtuous (bottom 50% on the ATF) versus relatively more desirable and virtuous (top 50%). We then ran one-way ANOVAs comparing the three newly created groups (separately for the average subjective, average objective Thompson, and average objective Maio indices) on their scores on the MFS Blockage and Forgiving subscales, the TTF, and the Vengeance Scale. We excluded the MFS Circumstances subscale from analysis because it was unrelated to ambivalence.
Because these analyses were exploratory, we had no specific hypotheses about the results we might observe. Nevertheless, we generally expected the pro-forgiveness group to exhibit the most forgiving responses (e.g., be relatively more inclined than the other two groups to forgive than seek vengeance and score highest on dispositional forgivingness), the anti-forgiveness group to exhibit the least forgiving responses (e.g., be least inclined to forgive rather than seek vengeance and score lowest in dispositional forgiveness), and the high ambivalence group to fall between the two low ambivalence groups. Essentially, we expected individuals high in ambivalence to be less polarized in their responses than those low in ambivalence, because highly ambivalent individuals hold conflicting and contradictory views of/feelings about forgiveness that ought to moderate their inclinations to be vengeful and forgiving (cf. Conner & Sparks, 2002).
As Table 4 shows, the main effect of group was significant for each DV and across all ambivalence indices. Contrary to expectations, however, there was a generally consistent trend in which, rather than falling between the low ambivalence groups, high ambivalence participants either exhibited the least forgiving and most vengeful tendencies or did not differ from their anti-forgiveness counterparts. As expected, post-hoc Tukey HSD tests (see Table 5) revealed that high ambivalence participants reported significantly less forgiving/more vengeful tendencies than pro-forgiveness participants on each DV, regardless of the ambivalence index used to create groups. Counter to expectations, differences between the two low ambivalence groups and between the anti-forgiveness and high ambivalence groups attained significance less consistently. In sum, our results suggest that individuals with highly ambivalent forgiveness attitudes may match and sometimes even exceed their anti-forgiveness counterparts in their tendency to be vengeful and disinclined to forgive.
Table 4
One-way ANOVAs Comparing High Ambivalence, Anti-, and Pro-Forgiveness Groups
| Variable | Average Subjective
|
(Average) Objective Thompson
|
(Average) Objective Maio
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| df | F | p | df | F | p | df | F | p | ||||
| Blockage | 2,103 | 16.25 | .24 | < .001 | 2,101 | 12.80 | .20 | < .001 | 2,104 | 5.94 | .10 | .004 |
| Forgiving | 2,103 | 13.80 | .21 | < .001 | 2,101 | 6.95 | .12 | < .001 | 2,104 | 3.40 | .06 | .04 |
| TTF | 2,104 | 19.45 | .27 | < .001 | 2,102 | 5.37 | .10 | .006 | 2,105 | 4.43 | .08 | .01 |
| Vengeance | 2,103 | 16.87 | .25 | < .001 | 2,101 | 10.42 | .17 | < .001 | 2,104 | 4.98 | .09 | .009 |
Note. TTF = Tendency to Forgive scale.
Table 5
Tukey HSD Comparisons Among High Ambivalence, Anti-, and Pro-Forgiveness Groups
| Variable/Group | Average Subjective
|
(Average) Objective Thompson
|
(Average) Objective Maio
|
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
| MD | SE | p | MD | SE | p | MD | SE | p | ||
| Blockage | ||||||||||
| Anti | Pro | .56 | .28 | .12 | .47 | .30 | .27 | .32 | .32 | .59 |
| High Ambiv | -.69 | .26 | .02 | -.63 | .28 | .07 | -.46 | .30 | .28 | |
| Pro | High Ambiv | -1.25 | .23 | .00 | -1.10 | .22 | .00 | -.78 | .22 | .00 |
| Forgiving | ||||||||||
| Anti | Pro | -.71 | .32 | .07 | -.24 | .35 | .77 | -.17 | .35 | .87 |
| High Ambiv | .63 | .29 | .08 | .67 | .33 | .10 | .45 | .33 | .36 | |
| Pro | High Ambiv | 1.34 | .26 | .00 | .91 | .25 | .01 | .63 | .25 | .03 |
| TTF | ||||||||||
| Anti | Pro | -.50 | .31 | .23 | -.32 | .36 | .65 | -.01 | .34 | 1.00 |
| High Ambiv | .97 | .28 | .01 | .52 | .33 | .27 | .65 | .32 | .11 | |
| Pro | High Ambiv | 1.48 | .24 | .00 | .84 | .26 | .01 | .65 | .24 | .02 |
| Vengeance | ||||||||||
| Anti | Pro | .81 | .25 | .01 | .42 | .27 | .25 | .49 | .28 | .18 |
| High Ambiv | -.34 | .23 | .30 | -.45 | .25 | .16 | -.12 | .27 | .88 | |
| Pro | High Ambiv | -1.15 | .19 | .00 | -.88 | .19 | .00 | -.62 | .20 | .01 |
Note. TTF = Tendency to Forgive. MD = Mean Difference.
Discussion
Across an array of measures, our participants displayed moderate levels of ambivalence in their attitudes toward forgiveness. As their ambivalence increased, they also exhibited a diminished inclination to be forgiving, an enhanced pro-vengeance orientation, and less idealistic views of forgiveness. These results support the conclusion that attitudes toward forgiveness often contain both positive and negative thoughts and feelings. They also provide an empirical basis for believing that such ambivalence might have important implications for understanding whether and when people forgive, by demonstrating that evaluative inconsistency in forgiveness attitudes is associated with variation in self-reported tendencies to be forgiving/vengeful and to perceive forgiveness as virtuous and desirable.
Both societal discourses surrounding forgiveness and scholarly treatments of forgiveness tend to emphasize the benefits of forgiving, painting a rather idealistic portrait that accords little consideration to its potential challenges, pitfalls, and limitations (Strelan et al., 2016). Contrasted with views of forgiveness that dominate both lay and scholarly discussions, then, our results suggest that laypeople’s attitudes toward forgiveness may be more complex and nuanced than public discourse or the scholarly literature typically acknowledge.
A small body of research has explored laypeople’s conceptualizations of forgiveness (e.g., Carr & Wang, 2012; Friesen & Fletcher, 2007; Kanz, 2000; Kearns & Fincham, 2004; Macaskill, 2005a; Mullet et al., 2004). Our results replicate findings indicating that people’s attitudes toward forgiveness contain negative elements (e.g., anxiety about the possibility that the offender may re-offend). They also extend this work in important new directions, however. First, they provide evidence that individuals may experience a degree of tension, or “felt ambivalence” (Newby-Clark et al., 2002), regarding forgiveness. For example, our participants reported feeling moderately muddled, conflicted, torn, confused, and indecisive when thinking about forgiveness. In other words, our participants were aware of the inconsistency in their forgiveness attitudes and experienced that inconsistency in a phenomenological sense. To our knowledge, this is the first evidence of its kind in the forgiveness literature.
Second, our study is the first to show that evaluative inconsistency—measured directly (subjective or felt ambivalence) or indirectly (objective or potential ambivalence)—predicts variation in broader beliefs about forgiveness (i.e., that it is desirable, moral, admirable, virtuous) and dispositional inclinations to be forgiving and to exact revenge. Our research thus moves the literature on lay conceptualizations of forgiveness beyond simple description to show that characteristics of such conceptualizations, in the present case attitudinal ambivalence, are associated with important forgiveness-related variables that may, in their turn, influence whether and when individuals forgive.
The results of our ancillary analyses augment the evidence for concluding that attitudinal ambivalence toward forgiveness may be consequential. Counter to our speculations, those whose attitudes were most ambivalent either scored lower than their anti-forgiveness peers or were statistically indistinguishable from them in their tendencies to be forgiving/vengeful. This suggests that, at high enough levels, ambivalent forgiveness attitudes may operate much like disapproving attitudes. If so, highly ambivalent individuals may be just as disinclined to forgive—and to seek revenge—as those with more uniformly negative attitudes toward forgiving, despite the presence of pro-forgiveness elements in their attitudes (e.g., beliefs that forgiving is healing, brings peace, is beneficial).
Such a possibility aligns well with findings showing that negative events, outcomes, affect, and cognitions are more influential in driving people’s feelings, thoughts, and behavior than their positive counterparts (Baumeister et al., 2001). We should thus perhaps not be surprised if the negative elements in individuals’ attitudes toward forgiveness overshadow the positive elements in their impact. Whatever mechanism(s) might explain such an effect, it may have important implications for both explaining and predicting when people will forgive. Based on previous ambivalence research, we suggested in the introduction that the task of correctly forecasting an individual’s response to a perpetrator’s wrongdoing may be considerably more difficult when their attitudes toward forgiveness are ambivalent than when they are not. The findings from our ancillary analyses suggest otherwise: Above a certain threshold, it may matter little whether an individual’s attitudes toward forgiveness are highly ambivalent or more uniformly disapproving.
Implications
Scholarly treatments of forgiveness have occasionally been criticized for painting forgiveness in idealistic and uncritical terms (e.g., Kearns & Fincham, 2004; Macaskill, 2005b; Murphy, 2005). Our findings lend further support to this criticism, suggesting that laypersons hold more nuanced perspectives on forgiveness than are reflected in much scholarly work in this area. They also call for researchers and theorists to investigate the full complexity of people’s attitudes toward and experiences of forgiving/forgiveness.
Practically, previous work on lay views of forgiveness has highlighted the need for counselors and therapists to define forgiveness for clients, warning them to expect disagreement with and possibly resistance to widely accepted definitions in the literature (e.g., Kearns & Fincham, 2004; Mullet et al., 2004). The present findings suggest that therapists ought also to expect clients to exhibit ambivalence in their forgiveness attitudes. Therapeutic techniques that involve examining the content and valence of clients’ thoughts and feelings about forgiveness with an eye toward exploring their basis, validity, and impact may prove fruitful in enabling people to make decisions about forgiving in ways that allow them to move forward with their lives and in their relationships while fully acknowledging the complexity of their views of forgiveness.
The possibility that high ambivalence and anti-forgiveness attitudes might result in similar beliefs about the value of forgiveness and dispositional tendencies to be forgiving/vengeful may also have implications for the delivery of forgiveness interventions. If, as our results suggest, highly ambivalent forgiveness attitudes impact the processes by which people determine how to respond to wrongdoing in the same way that attitudes more uniformly opposed to forgiveness do, highly ambivalent individuals might benefit from interventions designed to promote forgiving (i.e., when appropriate) as much as those with negative attitudes. Furthermore, people may not need to be aware of inconsistencies in their attitudes for this to be the case: The results of our ancillary analyses were consistent whether we used subjective or objective measures to categorize participants into ambivalence groups.
Limitations and Directions for Future Research
Our decision to employ multiple measures of ambivalence, dispositional forgiveness, and vengefulness is a strength of this research. It allowed us to assess ambivalence in forgiveness attitudes in a way that acknowledged the several forms ambivalence can take while gauging the extent to which our results were consistent across different measures/forms of ambivalence and in relation to several different but related facets of forgivingness and vengefulness.
In hindsight, however, we recognize that the order in which participants completed the various measures may have affected our results in unintended ways. Participants always completed the open-ended Maio task prior to the Thompson measure, the objective measures of ambivalence before the subjective measures, and the ambivalence measures before the other variables. This order may have inflated consistency between scores on the Maio and Thompson measures of objective ambivalence, subjective ambivalence scores, convergence between objective and subjective ambivalence, and associations between the ambivalence and forgiveness/revenge-related variables.
We are inclined to believe, nevertheless, that the impact order had on participants’ responses was likely slight. First, the levels of objective ambivalence we observed are comparable in magnitude to levels reported elsewhere, and the correlations between the various Maio and Thompson indices ranged from weak and nonsignificant (.13) to moderate (.32). The correlations between the various subjective and objective measures of ambivalence were at best moderate in size (.20 to .49), as well, and generally comparable in magnitude to those reported elsewhere (e.g., range .18 to .36 with three of the five rs ≥ .30 in Newby-Clark et al., 2002; range .36 to .44 in Priester and Petty, 1996). Finally, if completing the ambivalence measures before the ATF, TTF, MFS, and Vengeance Scale resulted in elevated scores on these latter scales, we might have expected to see even higher correlations and greater consistency in findings than we did. As discussed previously, there were several exceptions to the generally consistent pattern of associations between the ambivalence measures and DVs. The Maio cognitive ambivalence index, in particular, did not correlate well with the DVs (only two of six possible rs were significant and even these were small) and the Maio affective ambivalence index was uncorrelated with the ATF. Moreover, the results for the two Maio indices of intracomponent ambivalence varied.
In the future, there would be considerable value in investigating potential moderators of the associations between ambivalence and forgivingness-vengefulness. In line with Kachadourian et al. (2005), for instance, it will be important to explore whether ambivalence in attitudes toward forgiveness varies as a function of the quality of individuals’ relationships with their wrongdoers as reflected in the degree to which their attitudes toward their offenders are ambivalent. People may experience greater ambivalence about forgiving offenders toward whom they experience torn feelings, for example. Individuals in dissatisfied relationships, including relationships punctuated by frequent conflict, and those who harbor anxieties about being abandoned by the partner and/or difficulties trusting that the partner will respond to their needs (i.e., those high in attachment anxiety and/or avoidance) may be especially worthy of investigation.
Finally, there is a clear need for research examining whether ambivalence in attitudes toward forgiveness predicts actual forgiving. Our findings suggest that ambivalence is related to the proclivity to respond to wrongdoing with forgiveness, rather than vengeance, and the tendency to view forgiveness as a desirable and virtuous act, but they cannot speak to whether such ambivalence affects people’s actual decisions to forgive. If withholding forgiveness may deprive both individuals and their relational partners of the benefits of forgiveness, or perhaps protect them from its possible harms, this is an important issue to investigate.
Conclusion
We began this paper by quoting from a poem that exhorts us to forgive, promising that forgiving confers benefits from reducing personal resentment and bringing happiness to others to becoming “the master of fate, the fashioner of life, the doer of miracles” (Muller, 1998, p. 6) Such paeans to forgiveness are not uncommon in public discourse. In highlighting the ambivalent nature of people’s attitudes toward forgiveness, we do not mean to discourage forgiveness or diminish its value or potential benefits. Rather, we hope to encourage further inquiry into lay conceptualizations of forgiveness and their role in shaping and guiding people’s decisions concerning whether and when to forgive. Our results suggest the existence of a disconnect between people’s actual attitudes toward forgiveness and popular discourses on the topic and highlight the need for scholars to adopt a more critical stance toward forgiveness that recognizes its possible costs and limitations—or, at the very least, laypeople’s views on these. If relationship scholars are to contribute to the body of knowledge that may be used to help people deal with the hurts and injuries that occur in their relationships with those in their social networks, such a stance will be necessary to ensure our contributions are truly useful.