Abstract
According to the moral licensing literature, moral self-perceptions induce compensatory behavior: People who feel moral act less prosocially than those who feel immoral. Conversely, work on moral identity indicates that moral self-perceptions motivate behavioral consistency: People who feel moral act more prosocially than those who feel less so. In three studies, the authors reconcile these propositions by demonstrating the moderating role of conceptual abstraction. In Study 1, participants who recalled performing recent (concrete) moral or immoral behavior demonstrated compensatory behavior, whereas participants who considered temporally distant (abstract) moral behavior demonstrated behavioral consistency. Study 2 confirmed that this effect was unique to moral self-perceptions. Study 3 manipulated whether participants recalled moral or immoral actions concretely or abstractly, and replicated the moderation pattern with willingness to donate real money to charity. Together, these findings suggest that concrete moral self-perceptions activate self-regulatory behavior, and abstract moral self-perceptions activate identity concerns.
Take a moment to recall something nice you did for someone else. Perhaps you covered a class for a colleague, supported a friend going through difficult times, or helped an elderly neighbor complete her chores. Now, suppose you are faced with a new opportunity to help: A charity volunteer solicits your donation for a worthy cause. How would you behave?
People may be more likely to act prosocially after recalling previous instances of moral behavior than if past moral missteps are salient. Thinking of past moral actions may highlight one’s self-conception as a moral individual and motivate one to act in line with this view of one’s moral self. In this view, conceiving of oneself as a moral individual constrains one to continue acting in a moral fashion, lest one violate one’s sense of self (Blasi, 1980). Indeed, there is empirical evidence that people induced to view themselves as moral subsequently act more prosocially (e.g., Kraut, 1973; Reed, Aquino, & Levy, 2007).
Alternatively, people may be less likely to act prosocially after recalling past moral behavior. Because acting morally has already validated their moral qualities, people might feel absolved from the need to prove themselves further. In this view, moral self-perceptions act as a self-regulatory mechanism motivating compensatory behavior: Feeling relatively moral reduces the motivation to act prosocially, whereas feeling relatively less moral increases the motivation to act prosocially (Sachdeva, Iliev, & Medin, 2009). This view is also supported by empirical evidence (e.g., Miller & Effron, 2010; Monin & Miller, 2001).
Existing research, then, indicates that considering past moral behavior may either increase or decrease subsequent prosocial behavior. In the present work, we identify a moderator that determines when each response is likely: conceptual abstraction. Conceptualizing moral behavior in a concrete fashion might focus people on the specifics of the action itself, the way in which they have helped and supported another person. This level of construal might then lead them to think they have fulfilled their share of prosocial helping and “license” selfish behavior. Conversely, conceptualizing moral behavior in an abstract sense may activate moral identity concerns, motivating people to uphold their sense of self by acting in identity-consistent ways (Blasi, 1980; Reed et al., 2007). In three studies, we examine how conceptual abstraction might moderate how thinking about past moral behavior impacts future moral behavior.
Moral Identity Motivates Consistent Moral Behavior
Part of people’s sense of self is their moral identity—namely, a self-schema organized around a set of moral trait associations that motivates moral behavior (Aquino & Reed, 2002; Blasi, 1980). As with other aspects of identity, people are motivated to act consistent with their self-view (e.g., dissonance theory; Festinger, 1957), lest they violate their self-concept (Blasi, 1980; Shao, Aquino, & Freeman, 2008). Accordingly, individuals who view themselves as moral tend to act prosocially: Moral identity predicts not only spontaneous moral self-descriptions but also moral behavior (for reviews, see Hardy & Carlo, 2005; Shao et al., 2008). For example, moral identity predicted volunteering and donations to a local food bank (Aquino & Reed, 2002), donations to outgroup charities (Reed & Aquino, 2003), fewer antisocial sport behaviors (Sage, Kavussanu, & Duda, 2006), and willingness to donate time instead of money to prosocial organizations (Reed, Aquino, & Levy, 2007).
The importance of moral identity for the self-concept can fluctuate according to the situational context (Shao et al., 2008). People must balance many, sometimes competing, identities, and only a subset of identities will be active in any given context (Markus & Kunda, 1986). Therefore, moral identity should be particularly influential in determining prosocial behavior when it is activated by contextual cues. Reed, Aquino, and Levy (2007; Study 3) primed moral identity by asking participants to describe themselves using either moral or neutral words. Those in the moral identity prime condition reported greater willingness than participants in the control condition to donate time (vs. money) to a prosocial organization.
As self-perceptions are influenced by perceptions of one’s past behavior (Bem, 1972), recalling past moral behavior ought to activate moral identity, motivating further prosocial behavior. Indeed, the striving for consistent self-perceptions was theorized to underlie the classic foot-in-the-door phenomenon, in which individuals are more likely to engage in substantial helping behavior when they are first asked to perform a minor helpful act (Freedman & Fraser, 1966). Of course, the effect can work the other way: Recalling past moral failings may reduce moral behavior. Accordingly, Kraut (1973) found that charity donors who were labeled charitable subsequently donated more than unlabeled donors, whereas nondonors labeled uncharitable subsequently donated less than unlabeled nondonors. Thus, recalling past moral or immoral behavior should activate the relevant identity and motivate consistency in further prosocial behavior.
Moral Licensing Motivates Compensatory Moral Behavior
Remembering past moral acts might have consequences besides making moral identity salient. Recalling previous good deeds might also be taken as reassurance that one is, indeed, a moral individual, and is now free to indulge one’s selfish impulses. In other words, previous moral behavior can endow people with a psychological license, a “perception that they are permitted to take an action or express a thought without fear of discrediting themselves” (Miller & Effron, 2010, p. 116). 1 Therefore, recalling past moral behavior might also reduce subsequent prosocial behavior.
Indeed, there is now a great deal of evidence that performing moral behavior makes people more willing to act in morally questionable ways (for reviews, see Miller & Effron, 2010; Zhong, Liljenquist, & Cain, 2009). For example, male participants given an opportunity to reject sexist statements were more likely to discriminate against a female in a subsequent hiring task than males in a control condition (Monin & Miller, 2001), and White participants given an opportunity to endorse Barack Obama were more likely to discriminate against the Black community (Effron, Cameron, & Monin, 2009). Conversely, people who feel morally deficient are motivated to restore moral worth by engaging in moral cleansing behaviors (Zhong, & Liljenquist, 2006), such as performing prosocial behavior (Baumeister, Stillwell, & Heatherton, 1994).
Moral licensing is not induced only by actually acting in moral ways; simply recalling past moral behavior can reduce subsequent prosocial behavior. Jordan, Mullen, and Murnighan (2011) found that participants assigned to write about a time they helped someone subsequently expressed less motivation to engage in further prosocial behavior than participants who wrote about a nonmoral accomplishment. Importantly, this effect did not occur when participants wrote about other people’s behavior, suggesting that people were not responding to the moral content per se but rather to the way in which this moral behavior reflected on themselves. Similarly, Sachdeva et al. (2009) found that participants who were asked to write stories about themselves using moral adjectives were less likely to donate to charity than those in the neutral condition, who were less likely to donate than participants who were asked to describe themselves using immoral adjectives. Again, this pattern disappeared when participants described other people.
In sum, there is ample evidence showing that, sometimes, perceiving oneself as moral leads to a reduction in moral behavior, whereas perceiving oneself as less moral motivates increases in moral behavior. This pattern has been interpreted in terms of moral self-regulation, balancing moral self-consistent behavior against the costs inherent in prosocial behavior (Sachdeva et al., 2009; Zhong et al., 2009). Thus, to the degree that recalling past moral behavior licenses people to act selfish, they ought to demonstrate reductions in prosocial motivation. This pattern of behavior may be characterized as compensatory motivated moral behavior.
Resolving the Discrepancy: The Moderating Role of Abstraction
According to moral identity theorists, moral self-perceptions increase prosocial behavior; according to moral licensing theorists, moral self-perceptions decrease prosocial behavior. How might these conflicting findings be resolved? We hypothesize that there is an important moderator that influences the consequences of salient moral acts: conceptualization of previous moral behavior on an abstract or concrete level.
According to the construal level theory (CLT; Trope & Liberman, 2003, 2010) and the action-identification theory (Vallacher & Wegner, 1987), actions and events can be construed in either an abstract or concrete fashion. Abstract construals are schematic, decontextualized, detail-poor representations that capture superordinate, central features. Conversely, concrete construals are contextualized, detail-rich representations that capture subordinate, incidental features (Trope & Liberman, 2010). One way to operationalize the level of abstraction is temporal distance (Trope & Liberman, 2010). Temporally close events tend to be construed concretely and engender focus on the characteristics of behavior—how one performs a specific action. In contrast, temporally distant events tend to be construed abstractly and engender a focus on the purpose of behavior—why one performs a certain action (Liberman, Sagristano, & Trope, 2002; Nussbaum, Trope, & Liberman, 2003). Another way of inducing abstraction is a focus on central self-characteristics (Wakslak & Trope, 2009). Undergoing a self-affirmation procedure during which central values were listed led to a more abstract mind-set compared with participants who did not undergo a self-affirmation procedure. Thus, focusing on one’s own central identity can serve as an abstract construal prime and influence subsequent processing of information (Wakslak & Trope; Study 3) or task performance (Wakslak & Trope; Study 4).
However, the link between identity and construal works both ways: Conceptualization of events or actions in a relatively abstract manner can influence the degree to which core identity values are salient. When considering distant rather than close events, people predicted acting more in line with their personal values (Eyal, Sagristano, Trope, Liberman, & Chaiken, 2009; Giacomantonio, De Dreu, Shalvi, Sligte, & Leder, 2010), judged actors who violated moral principles more negatively (Eyal, Liberman, & Trope, 2008), predicted acting more prosocially and feeling guiltier for selfish behavior (Agerström & Björklund, 2009), and actually made more prosocial decisions (Rogers & Bazerman, 2008).
Moreover, construal level affects how people interpret past behavior: Temporally distant actions tend to be construed as evidence of commitment to pursuing a goal, whereas temporally recent events are interpreted as progress toward the goal (Fishbach, Dhar, & Zhang, 2006). In turn, when people perceive past behavior as evincing commitment to a goal, they increase their efforts toward that goal (i.e., demonstrate consistency). Fishbach and Dhar (2005) argued that “when people view the pursuit of a focal goal as a defining characteristic of their self-concept, pursuing it cannot possibly justify withdrawing its pursuit” (p. 371). Conversely, when people interpret past behavior as progress toward a goal, they feel licensed to relax goal strivings and switch to pursue other goals instead (i.e., demonstrate compensatory behavior). Although Fishbach and colleagues examined nonmoral goal behavior (e.g., studying), Miller and Effron (2010) speculated that similar mechanisms may moderate whether recalling past moral behaviors elicits consistent or compensatory moral behavior. The current work is the first to examine that possibility.
We hypothesize that abstract conceptualization of past moral behaviors highlights moral values associated with the behavior and induces people to act consistent with the salient past behavior. That is, recalling prosocial behavior in an abstract fashion should induce people to act prosocially, whereas abstractly recalling previous selfish behavior should induce people to act selfishly. In contrast, we expect that concrete conceptualization of past moral behavior might remind people that they have already fulfilled moral obligations and may allow them to relax subsequent efforts. In other words, recalling past good deeds in a concrete fashion might license more selfish, compensatory behavior, whereas recalling past selfish behavior in a concrete fashion might motivate people to compensate through more prosocial behaviors.
There is some evidence consistent with our hypothesis. Eyal et al. (2009) demonstrated that participants’ behavioral intentions corresponded more with their values when they imagined events in the distant rather than close future. Psychological distance increased consistency motivation, similar to our expectation that psychological distance to one’s own past behavior will induce consistency motivation. In addition, Nelson and Norton (2005) primed participants with an abstract category of moral actors (superheroes) or a concrete exemplar of a moral actor (Superman). Participants primed with the abstract category demonstrated consistency effects: They were more likely than controls to volunteer to tutor their peers and more likely show up at a volunteer meeting. As they interpreted their findings in terms of assimilation or contrast of the self to a (heroic) social category, their reasoning is consistent with ours. We build on these studies by examining the role of recalling one’s own (im)moral behavior on behavior intentions. Note that we do not expect effects of other’s moral behavior in our studies because we examine others who are not self-relevant rather than others who are members of an elite moral category that invites assimilation or contrast of the self.
Overview of Current Work
In three studies, we examine consequences of salient moral behavior on subsequent prosocial behavioral intentions and prosocial behavior. We hypothesize that level of abstraction of the past moral behavior determines whether prosocial intentions are subsequently increased or decreased. When people recall their behavior in a manner that induces a concrete mind-set (e.g., recalling recent behavior), they should demonstrate compensatory effects: Feeling moral should reduce prosocial intentions, whereas feeling immoral should increase prosocial intentions. In contrast, when people recall their behavior in a manner consistent with an abstract mind-set (e.g., distant behavior), they should demonstrate consistency effects: Feeling moral should increase prosocial intentions, whereas feeling immoral should decrease prosocial intentions. We manipulated level of abstraction or concreteness by varying temporal distance to the recalled behavior (Studies 1 and 2) and by varying whether participants recalled concrete actions or abstract values (Study 3). We assessed behavioral intentions, such as willingness to volunteer (Studies 1 and 2), as well as actual prosocial behavior—donation to charities (Study 3). We expected more prosocial intentions if participants recalled a distant moral behavior than distant immoral behavior (a consistency effect) and more prosocial intentions if participants recalled a recent immoral behavior than a recent moral behavior (a compensatory effect).
Study 1
In Study 1, we randomly assigned participants to describe either a moral or immoral behavior that they performed either during the previous week or over 1 year ago. We assessed their willingness to engage in further prosocial behavior in the future. We predicted a significant interaction between type of behavior recalled and the temporal distance of the recalled action. We expected that participants who considered a recent event would be more motivated to act prosocially if they recalled an immoral behavior than if they recalled a moral behavior. In contrast, participants who considered a distant event should be less motivated to act prosocially if they recalled immoral behavior than if they recalled moral behavior. Thus, we predicted an interaction in which the moral-distant conditions and immoral-recent condition would lead to greater prosocial intentions than the immoral-distant and moral-recent conditions. We tested this specific hypothesis with an a priori contrast.
Method
Participants
We recruited 101 American participants (69 female and 32 male), with a mean age of 43.91 (SD = 14.19). 2 Participants were recruited from the website Mechanical Turk (Amazon, 2011) and paid for their participation. Data obtained via Mechanical Turk demonstrate psychometric properties similar to laboratory samples (Buhrmester, Kwang, & Gosling, 2011).
Materials and procedure
Participants completed all materials online. Participants were randomly assigned to write about either a moral or immoral behavior that they performed in either the recent or distant past. Following this, they completed two measures of moral behavioral intentions: a measure of willingness to help and a measure of willingness to volunteer. They then completed the positive and negative affect schedule (PANAS), two manipulation checks, and demographic measures, before being thanked and debriefed.
Moral behavior recall manipulation
Participants completed a moral behavior recall task (adapted from Sachdeva et al., 2009). In the moral behavior recall condition, participants were instructed to recall a time when they acted in such a way that they felt righteous or honorable. We clarified, “Perhaps you were loyal to a friend, were generous when you could have been selfish, were kind to someone for no particular reason, or caring toward someone who needed you.” Note that these terms were identified as prototypically moral (Aquino & Reed, 2002). In the immoral behavior recall condition, participants were instructed to recall a time when they acted in such a way that they felt guilty or ashamed, such as when “You were disloyal to a friend, were greedy when you should have shared, were mean to someone for no particular reason, or uncaring toward someone who needed you.” When recalling the aforementioned event, participants were asked to describe an event that occurred either within the past week (recent condition) or over 1 year ago (distant condition).
The moral events that participants wrote about were typically mundane interpersonal prosocial actions, such as standing up for a girl who is getting teased or volunteering at a homeless shelter. The immoral events were also typically mundane interpersonal grievances, such as acting unsympathetic to a friend in need or avoiding playing with one’s children (despite their pleading). 3 On average, participants wrote 439 words (SD = 205.91). The length of the event description did not differ by distance condition, F(1, 97) = 0.06, p = .810, or event valence, F(1, 97) = 0.12, p = .732.
Behavioral intentions
To measure willingness to help others, participants completed a five-item willingness to volunteer scale (Devoe & Pfeffer, 2007). Participants responded to items such as “Volunteering is a worthwhile use of my time even if I do not get paid” on scales from 1 = completely disagree to 7 = completely agree. Items were combined into one overall scale (α = .92).
Participants also read four brief vignettes in which they were asked to imagine other people requiring a minor degree of everyday assistance. The scenarios included paying a few extra cents for someone else’s restaurant bill, delivering a parcel, allowing someone to move ahead in line, and lending a special garden tool. Participants indicated whether they would help on scales from 1 = very unlikely to 7 = very likely. There was low but acceptable reliability across scenarios, considering the diversity of helping domains (α = .56), and responses were aggregated into an index of willingness to help.
Manipulation checks
Participants rated event positivity (“The event I wrote about made me feel good about myself”) and perceived temporal distance (“The event I wrote about happened a long time ago”) on scales ranging from 1 = completely agree to 7 = completely disagree.
PANAS
Participants completed Watson, Clark, and Tellegen’s (1988) 20-item PANAS, to be used as a covariate. We assessed affect to ensure that differences on the behavioral intention measures were independent from the result of mood differences between groups. For this measure, participants indicated the extent to which they currently experienced 10 positive emotions (e.g., interested, excited) and 10 negative emotions (i.e., depressed, upset) on scales ranging from 1 = not at all to 5 = very strongly. Items were averaged into a negative emotion subscale (α = .88) and a positive emotion subscale (α = .88).
Results
Manipulation checks
A 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) analysis of variance (ANOVA) showed the expected main effect of event valence on event positivity ratings, F(1, 96) = 287.96, p < .001, η2 = .75. Participants instructed to write about a moral event reported their event to be significantly more positive (M = 4.25, SD = 0.84) than participants instructed to write about an immoral event (M = 1.70, SD = 0.69). Event distance did not moderate this main effect of event valence, F(1, 96) = 0.70, p = .406, η2 = .01.
A 2 (event valence) × 2 (event distance) ANOVA showed the expected main effect of event distance on perceived distance, F(1, 98) = 83.72, p < .001, η2 = .47. Participants who wrote about a distant event perceived that event to be significantly more distant (M = 3.02, SD = 1.09) than participants who wrote about a recent event (M = 1.40, SD = 0.54). Event valence did not moderate this effect of event distance, F(1, 98) = 1.43, p = .236, η2 = .02. Taken together, these items indicated that our manipulations were successful.
Behavioral intentions
We conducted a 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) ANOVA on participants’ willingness to volunteer. There were no significant main effects of either distance, F(1, 96) = 0.13, p = .716, η2 < .01, or valence, F(1, 96) = 0.07, p = .798, η2 < .01, but, as predicted, there was a significant interaction, F(1, 96) = 5.37, p = .023, η2 = .05. As Figure 1 illustrates, when participants wrote about an event in the recent past, those who recalled an immoral event were more willing to volunteer (M = 5.97, SD = 1.30) than participants who recalled a moral event (M = 5.42, SD = 1.37). Conversely, when participants wrote about a distant event, those who recalled an immoral event were less willing to volunteer (M = 5.26, SD = 1.49) than those who recalled a moral one (M = 5.94, SD = 1.02). As expected, the one-way a priori contrast of our specific hypothesis was significant, t(99) = 2.39, p = .001. This indicates that the two conditions we predicted would lead to greater prosocial intentions (moral-distant and immoral-recent) did, indeed, lead to more prosocial intentions than the other two conditions (moral-recent and immoral-distant).
Participants’ prosocial intentions across valence and distance conditions (Study 1).
We replicated this pattern in a 2 (event valence) × 2 (event distance) ANOVA on participants’ willingness to help the people in the vignettes. Results revealed no significant main effect of either distance, F(1, 96) = 0.36, p = .550, η2 < .01, or valence, F(1, 96) = 1.26, p = .265, η2 = .01. However, as predicted, there was, again, a significant interaction, F(1, 96) = 5.97, p = .016, η2 = .06. When participants wrote about an event in the recent past, those who recalled an immoral event were more willing to help (M = 6.00, SD = 0.73) than participants who recalled a moral event (M = 5.81, SD = 0.70). Conversely, when participants wrote about a distant event, those who recalled an immoral event were less willing to help (M = 5.56, SD = 0.70) than those who recalled a moral one (M = 6.01, SD = 0.78). The one-way a priori contrast was again significant, t(99) = 2.52, p = .007, confirming our specific prediction of greater prosocial intentions in the immoral-recent and moral-distant conditions, compared with the moral-recent and immoral-distant conditions.
Including participants’ scores on the positive PANAS subscale and the negative PANAS subscale as covariates did not affect this pattern of findings—the two-way interaction remained a significant predictor of willingness to volunteer, F(1, 95) = 4.25, p = .042, η2 = .03, and willingness to help, F(1, 95) = 5.36, p = .023, η2 = .04. Indeed, scores on the PANAS did not differ by valence or distance condition or their interaction term. Therefore, our results cannot be attributed to differences in mood between conditions.
Discussion
This study demonstrated that temporal distance moderated participants’ response to salient past actions. When participants considered recent actions, they demonstrated compensatory effects: Participants who recalled immoral actions were more willing to help than participants who recalled moral actions. Conversely, when participants recalled distant actions, they demonstrated consistency effects: Those who recalled immoral actions were less willing to help than participants who recalled moral actions. This pattern is consistent with the hypothesis that recalling past moral actions in a concrete mind-set (induced by temporal closeness) motivates moral self-regulation, striving to balance moral actions over time, whereas recalling past moral actions in an abstract mind-set (induced by temporal distance) highlights moral identity and induces participants to act in line with their self-perceptions.
Importantly, both hypothesized processes—balancing one’s own moral actions over time and activating moral identity—involve the self. That is, recalling one’s own actions should be meaningfully different from simply being primed with moral or immoral behavior. For example, thinking of other people’s behavior, as opposed to one’s own past behavior, should not induce compensatory or consistency effects. Others’ actions do not affect either one’s own self-regulatory account of moral actions or one’s own moral identity; thus, they should not systematically impact willingness to engage in prosocial behavior. We examine this hypothesis in the next study.
Study 2
Moral identity is inherently tied to self-perceptions; it is defined as an association between the self and moral traits (Aquino & Reed, 2002). Therefore, it stands to reason that one’s own moral behavior ought to more powerfully impact moral identity than someone else’s behavior. In the next study, we aim to establish that the interaction obtained in Study 1 was due to changes in moral self-perceptions, rather than simply greater activation of moral values in the distant condition. We varied whether participants recalled moral behavior performed by themselves or moral behavior performed by someone else. We expected that compensatory and consistency patterns of subsequent prosocial motivation would occur only for self-relevant past behavior, not for moral behavior by others. Again, we tested our specific hypothesis with an a priori contrast of the immoral-distant and moral-recent condition versus the moral-distant and immoral-recent conditions for self and other, respectively.
It is important to note that we do not examine the effect of moral behavior conducted by the self versus a significant other. There is reason to believe that good deeds by one’s relationship partner (O’Conner, Effron, Mullen, & Monin, 2010) or ingroup (Kouchaki, 2011) can sometimes provide vicarious moral credentials, driven by cognitive overlap between the self and one’s partner or group. In contrast, in the present study, we simply instructed participants to think of moral behavior completed by a general friend—not a relationship partner.
Method
Participants
We recruited 164 American participants for their participation from the website Mechanical Turk. In all, 13 participants were excluded for failing to write about moral behavior, or for writing about moral behavior performed by others that personally victimized or benefitted participants themselves, leaving a final sample of 151 (82 female and 69 male). Mean age was 31.22 years (SD = 12.17). Precautions were taken to exclude participants who completed Study 1.
Materials and procedure
The materials and procedure of Study 2 were identical to those of Study 1, except for two changes. First, we eliminated willingness to help as a dependent measure due to its low reliability in Study 1. Second, we added an additional manipulation: Now, participants were asked to recall behaviors performed either by themselves (self condition) or a friend (other condition). In the self condition, instructions were identical to Study 1. In the other condition, participants were asked to recall a time when a friend performed a specific past behavior. Participants were instructed to avoid describing any behavior that benefitted or victimized them personally to avoid activating concepts such as gratitude or retaliation. In sum, this study used a 2 (actor: self vs. other) × 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) between-subjects design.
Results
Manipulation checks
A 2 (actor: self vs. other) × 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) ANOVA on perceptions of event positivity revealed the expected main effect of valence, F(1, 143) = 111.38, p < .001, η2 = .44. Participants instructed to write about a moral event reported that their event was significantly more positive (M = 3.60, SD = 1.09) than participants instructed to write about an immoral event (M = 1.91, SD = 1.31). There was also a significant interaction between actor and event valence, F(1, 143) = 27.00, p < .001, η2 = .16, indicating that participants viewed moral behaviors they performed (M = 4.10, SD = 0.96) as more positive than moral behaviors others performed (M = 3.14, SD = 1.01), whereas immoral behaviors they committed (M = 1.62, SD = 0.79) were considered more negative than those perpetuated by others (M = 2.26, SD = 1.06). To control for this potential confound, we conducted additional analyses with event positivity as covariate (see the “Behavioral Intentions” section). No other effects were significant.
A 2 (actor) × 2 (event valence) × 2 (event distance) ANOVA on perceptions of event distance revealed the expected distance main effect, F(1, 143) = 61.08, p < .001, η2 = .30. Participants who were asked to write about an event occurring over a year ago (M = 3.06, SD = 1.14) reported that the event was significantly more distant than participants asked to write about an event occurring within the past week (M = 1.78, SD = 0.98). No other effects reached significance. Thus, we concluded that our manipulations were successful.
Behavioral intentions
A 2 (actor: self vs. other) × 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) ANOVA on participant’s willingness to volunteer (α = .84) revealed a significant main effect of the actor condition, F(1, 143) = 4.33, p = .039, η2 = .03. Participants who wrote about another persons’ behavior (M = 5.64, SD = 1.24) were more willing to volunteer than those who wrote about their own behavior (M = 5.23, SD = 1.29). As predicted, results also revealed a significant three-way interaction, F(1, 143) = 5.48, p = .021, η2 = .04 (see Figure 2). No other effects were significant (distance main effect: F = 1.66, p = .200; valence main effect: F = 0.10, p = .758; Valence × Distance interaction: F = 0.16, p = .690; Valence × Actor interaction: F = 1.05, p = .308; Distance × Actor interaction: F < 0.001, p = .970).
Participants’ prosocial intentions across actor, valence, and distance conditions (Study 2).
To follow up on the significant three-way interaction, we conducted separate 2 (event valence: moral vs. immoral) × 2 (event distance: recent vs. distant) ANOVAS within the self condition and the other condition, respectively. In the self condition, there was no main effect of valence (F = 0.92, p = .342) or distance (F = 0.91, p = .372). A marginally significant valence by distance interaction, F(1, 73) = 3.88, p = .053, η2 = .05, showed that participants were more willing to volunteer after recalling an immoral (M = 5.51, SD = 1.34) than moral event (M = 4.66, SD = 1.53) in the recent past, but less willing to volunteer after recalling an immoral (M = 5.20, SD = 1.10) than moral event (M = 5.49, SD = 1.01) in the distant past. This interaction replicated effects found in Study 1. In the other actor condition, however, the valence by distance interaction was not significant, F(1, 70) = 1.84, p = .180, η2 = .03 (valence main effect: F = 0.25, p = .619; distance main effect: F = 0.86, p = .358).
Once again, we tested the a priori contrast of our specific hypothesis. This was significant for the self condition, t(75) = 2.03, p = .023, but not the other condition, t(72) = 1.20, p = .234. Therefore, the two conditions we predicted would lead to greater prosocial intentions (immoral-recent and moral-distant) did, indeed, lead to greater prosocial intentions than the other two conditions (moral-distant and immoral-recent)—but only when participants considered actions they themselves performed.
Including participants’ scores on the positive PANAS subscale (α = .88) and the negative PANAS subscale (α = .93) as covariates did not affect this pattern of findings—the three-way interaction remained significant, F(1, 141) = 5.57, p = .020, η2 = .05. Indeed, scores on the PANAS did not differ by distance or valence condition or their interaction term. Thus, these findings cannot be attributed to mood differences between the event distance and event valence conditions. However, actor condition affected participants’ reported negative feelings, indicating that participants who wrote about their own behavior reported greater negative mood (M = 1.88, SD = 0.90) than those who wrote about other’s behavior (M = 1.49, SD = 0.62), F = 9.95, p = .002.
There was a significant interaction between actor condition and event valence on the rated positivity of the event; participants rated their own moral deeds as more extremely good or bad than those performed by others. Thus, we included event positivity as a covariate. The three-way interaction remained significant, F(1, 142) = 5.79, p = .017, η2 = .04. Thus, our results cannot be attributed to differences in the extremity of event valence between own and other’s acts.
Discussion
This study extended Study 1 by examining the actor of the recalled past behavior. In this study, we demonstrated that the compensatory and consistency effects of recalling past behavior found in Study 1 were indeed limited to self-relevant past behaviors. When participants recalled their own past actions, recent moral acts induced participants to show compensatory tendencies, whereas distant moral acts induced consistency tendencies. Writing about behavior performed by other people had no impact on willingness to help as a function of distance or valence. This pattern of findings confirmed our hypothesis that the interaction of distance and valence on behavioral intentions is driven by the activation of self-regulatory and identity concerns, rather than the activation of moral conceptions in general.
An unexpected main effect of the actor condition occurred: Participants were more willing to volunteer overall after writing about another person’s behavior than writing about their own behavior. There are at least two possible explanations for this finding. First, writing about another persons’ moral behavior may have elicited moral elevation, a positive moral emotion elicited by witnessing another person act in a virtuous fashion, which motivates prosocial behavior in turn (Schnall, Roper, & Fessler, 2010). Second, writing about another persons’ immoral behavior may have threatened belief in a just world. People are deeply motivated to believe the world is a fair place, and evidence to the contrary—such as other’s immoral behavior—is threatening (Lerner, 1980). Participants in this condition may have been motivated to restore perceptions of a just world by acting in a prosocial manner, which is a common strategy when the situation affords it (Haynes & Olson, 2006). For current purposes, the important thing is that other’s moral behavior failed to demonstrate the interaction pattern evident when participants recalled their own moral behavior.
Study 3
The first two studies demonstrated that recalling one’s moral actions in the recent past led to compensatory behavior, whereas recalling one’s moral actions in the distant past led to consistency-oriented behavior. We argue that this occurred because actions in the recent past are conceptualized on a more concrete level, leading people to focus on the specifics of the action, whereas actions in the distant past are conceptualized more abstractly, leading people to focus on the implications of the event for their self-concept (Liberman et al., 2002). In the next study, we directly tested this proposition. We induced participants to consider either the concrete steps involved in performing a moral or immoral action (i.e., action condition), or the implications of such an action for their personal identity (i.e., identity condition). We expected compensatory behavior patterns if participants were instructed to consider a concrete behavior and expected consistency behavior patterns if participants were instructed to consider the actions’ implications for identity. Once again, we tested our specific hypothesis with an a priori contrast comparing the action-immoral and identity-moral conditions versus the action-moral and identity-immoral conditions. Study 3 also extended previous studies by using a different dependent variable. Instead of assessing general willingness to volunteer, we assessed people’s actual willingness to give up a portion of potential money to charity.
Method
Participants
We recruited 130 American participants via the website Mechanical Turk. In all, 15 participants either failed to complete the study or failed to write about a morally relevant action and were excluded, leaving a final sample of 115 (68 female and 47 male), with a mean age of 35.89 (SD = 13.66).
Materials and procedure
This study used a 2 (moral content: moral vs. immoral) × 2 (focus: identity vs. action) design. Participants described themselves as either possessing moral or immoral traits or performing moral or immoral behaviors, and then allocated donations to charity. Finally, participants completed the PANAS (α = .90 and .92 for the positive and negative subscale, respectively), as in the previous studies.
Manipulations
Participants in the moral identity condition completed a task modified from Aquino and Reed’s (2002) moral identity questionnaire. They were presented with nine moral traits—caring, compassionate, fair, friendly, generous, helpful, hardworking, honest, kind. In the immoral identity condition, we instead used the nine antonyms of these words (also see Sachdeva et al., 2009): uncaring, indifferent, unfair, unfriendly, greedy, selfish, lazy, dishonest, mean. Participants were asked to visualize having these characteristics, and wrote about what having these traits would mean for their personality.
Participants in the action conditions were presented with the same nine moral traits, but converted into specific actions (e.g., “act caring [uncaring] toward a person who is sad, act compassionately [indifferent] toward a person in pain, act friendly [unfriendly] toward a stranger, give generously to [take greedily from] a person in need, act kindly [mean] toward another person”). Participants were asked to imagine performing these behaviors, and then to write a paragraph about how they would perform them.
Charity donation task
Participants entered a draw for US$50. They were told that they could allocate part of this amount to a variety of charities and that the allocated amount would be taken off automatically of the winner’s reward. Participants then indicated how many dollars should be donated on their behalf to UNICEF, The American Red Cross, The World Wildlife Federation, and/or a charity of their own choosing, should they win the draw. Charitable donations were aggregated across all charities to indicate the overall amount of money participants were donating to a good cause. A minority of participants (n = 5) declined to enter into the draw, but excluding them did not affect results.
Results
Charity donations
A 2 (moral content: positive vs. negative) × 2 (focus: identity vs. action) ANOVA on participants’ charity donations revealed a significant main effect of moral content, F(1, 111) = 4.11, p = .045, η2 = .04. Participants in the immoral condition donated more (M = US$21.90, SD = 16.24) than participants in the moral condition (M = US$15.87, SD = 16.32). There was no significant main effect of identity-action, F = 0.36, p = .548, η2 < .01. As predicted, the main effect of valence was qualified by a significant interaction, F = 4.06, p = .046, η2 = .04. As illustrated in Figure 3, participants donated more when they imagined performing immoral actions (M = US$24.03, SD = 16.20) than when they imagined possessing immoral traits (M = US$19.77, SD = 16.26), whereas participants donated less when they imagined performing moral actions (M = US$11.85, SD = 14.05) than when they imagined possessing moral traits (M = US$19.74, SD = 17.64). Once again, the one-way a priori contrast comparing the action-immoral and identity-moral conditions with the action-moral and identity-immoral conditions was significant, t(113) = 1.93, p = .028.
Mean charity donations (in dollars) among participants who described themselves as possessing moral or immoral traits or performing moral or immoral actions (Study 3).
The pattern was attenuated when including participants’ scores on the positive and negative subscales of the PANAS as covariates: The interaction remained marginally significant, F(1, 109) = 3.94, p = .050, η2 = .03. However, scores on the PANAS did not differ by condition. Thus, the results cannot be attributed to mood differences between the conditions.
Discussion
The results of Study 3 conceptually replicated Studies 1 and 2 using a different manipulation and measuring actual donation behavior. When thinking of their own moral actions, participants exhibited compensatory behavior, donating less money to charity after thinking about moral actions versus immoral actions. In contrast, when thinking about their own moral identity, participants acted consistent with the identity that was made salient, donating less money after considering immoral, unhelpful aspects of their identity than after considering prosocial, helpful aspects of their identity. These findings support the hypothesis that conceptual abstraction moderates the effect of salient behavior on subsequent behavior. A concrete mind-set, with a focus on actions, appears to elicit contrast effects (i.e., compensatory behavior), whereas an abstract mind-set, with a focus on implications for identity, elicits assimilation to self-perception (i.e., consistency behavior).
Note that in the identity condition in this study, we asked participants to think of the meaning certain moral or immoral traits would have for their identity (instead of asking them to write a story about themselves using these words (cf. Sachdeva et al., 2009). Thus, although our manipulation was superficially similar to the more concrete story-telling manipulation by Sachdeva and colleagues, our manipulation emphasized participants’ moral identity—and, consequently, inspired consistent behavioral intentions rather than moral licensing effects.
General Discussion
Across three studies, we found that conceptualizing moral acts as concrete behavior motivated compensatory prosocial behavior, whereas conceptualizing moral acts in an abstract manner motivated consistency prosocial behavior. Thinking about recent moral behavior or concrete moral actions reduced prosocial motivation, whereas recent or concrete immoral behavior increased prosocial motivation. Conversely, thinking about distant moral behavior or abstract moral traits increased prosocial motivation, whereas thinking about distant or abstract immoral behavior reduced prosocial motivation. This effect was demonstrated to be distinct from simply thinking about moral behavior per se: Only self-relevant past moral behavior affected subsequent prosocial motivation. Considering other people’s morally relevant behavior did not.
Prosocial Motivation and Morality
Throughout this article, we use the terms prosocial and moral interchangeably. By the term moral, we are referring to actions that benefit others at a cost to oneself, in line with most theorists and lay people (e.g., Agerström & Björklund, 2009; Aquino & Reed, 2002; Krebs, 2008). Note that the current work examines variation in prosocial motivation, as opposed to motivation to act in a genuinely immoral fashion. That is, the current work examines only prescriptive morality—the morality of actively helping others—not proscriptive morality, the morality of avoiding harming others, both of which are not psychologically equivalent (Janoff-Bulman, Sheikh, & Hepp, 2009). Although recalling concrete moral behavior may reduce prosocial motivation, we are not arguing that it will actually increase motivation to actively harm others. Nonetheless, future work might examine whether compensatory efforts at moral behavior sometimes include even greater selfish or immoral actions as compensatory behavior. Consistent with this possibility, Jordan and colleagues (2011) demonstrated similar moral licensing effects on both moral behavior (intention to act prosocially) and immoral behavior (cheating).
Although the overall interaction pattern and a priori contrasts were significant across all three studies, not all simple effects were consistently significant, and, moreover, effect sizes were of medium to small size (Cohen, 1973). We would like to caution that, although recalling previous moral or immoral behavior in an abstract or concrete fashion does appear to impact future moral behavior, it likely does not turn sinners into saints or vice versa. There are many other influences on moral behavior, including dispositional factors (Walker, Frimer, & Dunlop, 2010). That said, considering the wide variation in how important the recalled events were and how moral or immoral participants might have acted in the past, the naturalistic setting of the experiment might invite random error in the data. A replication with more controlled behaviors (e.g., where participants all recall the same past event or where they are assigned to act in a moral or immoral way before being asked to help, volunteer, or donate) might yield larger effect sizes.
Moral Self-Regulation
Our work clarifies the role of self-regulation in moral behavior. In line with evolutionary models of morality (e.g., Krebs, 2008), Sachdeva and colleagues (2009) argued that people feel the need to balance moral self-conceptions against the costs inherent in prosocial behavior (see also Zhong et al., 2009). Accordingly, they found that moral self-perceptions motivated contrasting behaviors in a negative feedback loop around a standard set point: Perceiving oneself as moral motivated a reduction in moral behavior, whereas perceiving oneself as less moral motivated increases in moral behavior. The present studies support this view of moral behavior fluctuating in a negative feedback loop, when behavior was construed concretely.
Unlike the self-regulation model, however, the present studies also suggest that past moral actions can lead to consistent behavior—People who feel moral act moral, in a positive feedback loop. Such a positive feedback loop suggests a mechanism by which people can reach a new set point around which moral self-regulation may fluctuate: construing past moral action abstractly. Indeed, real-life moral exemplars tend to describe their actions in broad, abstract, general terms (Colby & Damon, 1992).
It may be that moral exemplars experience moral identity activation under a wider array of situations. As argued by Blasi (1980) and Reed and Aquino (2003), strong moral identities constrain people to act morally, not just sometimes, but consistently across situations. Thus, whereas moral exemplars (e.g., Gandhi) might experience moral self-activation universally, most people may experience moral identity activation in some situations but not others. Accordingly, Reed and Aquino (2003) found that people higher in moral identity donated more to socially distant charities. It may be that people higher in dispositional moral identity are more likely to construe past moral behavior abstractly, and consequently experience moral identity activation more often than people low in moral identity. Future work should investigate whether individual differences in dispositional moral identity influence the construal of past moral behavior.
CLT
Note that we focused on temporal distance and the action/trait distinction to manipulate abstraction of people’s mind-sets. Other construal priming paradigms (e.g., Freitas, Gollwitzer, & Trope, 2004; Navon, 1977) should also moderate the effect of recalling past moral actions. It may also be that other aspects of psychological distance (e.g., social distance, spatial distance) have a similar effect (Trope & Liberman, 2010). For example, thinking about moral acts taking place far away might elicit abstract identity concerns similar to thinking about acts committed long ago. However, this may not necessarily be the case, as different forms of distance sometimes have different effects (e.g., social distance reduces positivity, as distant groups are viewed less positively than one’s own, whereas temporal distance increases positivity, as people are more optimistic about the distant than near future; Trope & Liberman, 2010). Indeed, our finding in Study 2 that other people’s moral behavior did not systematically influence future prosocial intentions suggests that there may be important differences between different types of psychological distance in terms of their effect on prosocial behavior. Future research is needed to examine these differences.
Consequences and Implications
There are different risks involved with maintaining abstract and concrete mind-sets. Considering past behavior abstractly lends weight to one’s actions by increasing their implications for the self. Although this motivates increases in moral behavior when past actions are moral, when past actions are immoral, they may snowball into an unfavorable perception of one’s moral self, leading to less and less effort to act prosocially. However, maintaining a low-level mind-set might lead to moral licensing and social laziness when one recalls moral behavior. To maximize prosocial efforts, then, past good deeds should be highlighted in abstract terms (e.g., their implications for one’s identity), whereas past failures to help or support others should be framed in concrete (e.g., behavioral) terms. Indeed, these prescriptions might apply not only to the framing of past behavior but also to the framing of anticipated future behavior. Thoughts about intended future moral behavior might induce compensatory or consistency motivations, similar to thoughts about past behavior (Caruso, 2010; Eyal et al., 2009).
In sum, the current work presented three studies demonstrating that feeling moral does not always motivate moral behavior, nor does it always reduce moral behavior. Rather, recalling past moral behavior makes people either relax or renew their moral strivings depending on their conceptualization of the behavior. Think back to the good deed you recalled at the beginning of this article. Whether you are motivated to act prosocially may depend on how abstract your recollection of past moral behavior was. Together, the three current studies underscore the importance of mental mind-set and reconcile two opposing literatures on moral behavior.
Footnotes
Acknowledgements
The authors would like to thank Miranda Giacomin, Anita KÖrner, James M. Olson, Christopher J. Wilbur, and anonymous reviewers for providing valuable feedback on drafts of this article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a doctoral fellowship from the Social Sciences and Humanities Research Council of Canada [Award number: 767-2008-1666] to the first author, and by an internal faculty grant of the University of Cologne to the second author.