Moral pendulum 1 On the pendulum of moral action - CiteSeerX

7 downloads 280 Views 199KB Size Report
... a life of crime (Bold, March 23, 1999); after a life of devotion, an Omaha nun stole ... identity assurance from moral behavior allows people to relax their moral ...
Moral pendulum

On the pendulum of moral action: The effects of own and others’ past moral actions on future moral behavior

Jennifer M. Jordan University of Groningen Elizabeth Mullen Stanford University J. Keith Murnighan Northwestern University

Please do not cite or distribute without the author(s) permission.

1

Moral pendulum

2

Abstract Moral decisions often create an internal tension between temptation and doing the right thing. The current research stimulated individuals’ moral self-images by asking them to recall instances of their own moral or immoral behavior, prior to providing them with an opportunity to report the extent to which they engage in moral behavior, their intention to engage in moral behavior, or to actually engage in immoral behavior. Results indicated that recalling immoral behavior led individuals to: report greater participation in moral activities, report greater intent to engage in prosocial activities, and engage in less cheating than recalling moral behavior. Rather than compensatory behavior, however, recalling others’ moral and immoral behaviors led to similar behavior. Using self-completion theory as a framework, this research combines evidence of moral cleansing (remunerative moral strivings), moral licensing (relaxed moral strivings), and moral social standards (following others’ moral behavior) to illuminate the dynamics of moral behavior.

Keywords: self-completion theory, morality, ethics, moral identity, social norms, cheating

Moral pendulum

3

“Search others for their virtues, thyself for thy vices.” -Benjamin Franklin Moral temptations are an everyday event. Downloading music illegally, exceeding the speed limit, and taking home supplies from the office are just three examples of the numerous mundane temptations that individuals regularly confront. Although people value a moral selfimage (Monin & Jordan, 2008) and want to see themselves as moral actors (Mazar, Amir, & Ariely, 2008; Nisan, 1992), they also succumb, at least occasionally, to temptation. The media provides many examples of moral oscillation: a prominent governor, lauded by Fortune for espousing “moral clarity” (Gimein, September 16, 2002), resigned in disgrace for acts of moral turpitude; a former drug addict and ex-convict now mentors parolees and helps deter at-risk youth from a life of crime (Bold, March 23, 1999); after a life of devotion, an Omaha nun stole thousands of dollars from her archdiocese (Associated Press, April 11, 2006); and twenty violent criminals completed their prison terms, earned college degrees, and vowed to uphold morallyupright lives (Rubenstein, June 6, 2004). These instances of unstable moral action highlight the dynamic nature of moral behavior. The current research investigates whether individuals’ attempts to balance temptation with the drive to maintain a favorable moral self-image results in predictable oscillations in moral behavior. We present three studies that illuminate how prior behavior influences individuals’ moral self-perceptions, intentions, and subsequent moral behavior. In doing so, we hope to lay the groundwork for a more nuanced theory of dynamic moral behavior. People highly value maintaining a moral self-image, both privately and publicly (Mazar et al., 2008; Monin & Jordan, 2008). For instance, people’s mean self-ratings on a measure of internal moral identity (i.e., how central moral traits are to an individual’s self-concept) were 4.6

Moral pendulum

4

on a 5-point scale (Aquino & Reed, 2002) and 6.28 on a 7-point scale (Reed, Aquino, & Levy, 2007).1 Individuals also use moral behavior as an indicator of others’ ‘true natures’ (Reeder, Kumar, Hesson-McInnis, & Trafimow, 2002; Singh & Teoh, 2002). Self-completion theory (SCT; Gollwitzer & Kirchhof, 1998; Gollwitzer & Wicklund, 1982) provides a framework to explain the dynamics of moral behavior. SCT proposes that individuals’ personal (or group) identities are goals and that individuals acquire identity-relevant symbols to attain these goals (Barry & Tyler, 2009; Ledgerwood, Liviatan, & Carnevale, 2007). It also asserts that individuals feel a sense of incompleteness when they receive unfavorable feedback or do not meet a critical performance threshold related to their cherished identity (i.e., lack an identity relevant symbol). Tension from this sense of incompleteness results in compensatory action aimed at acquiring alternative symbols of the cherished identity. In contrast, when individuals experience a completed identity (i.e., they acquire a symbol of their identity), they then relax their identity-relevant strivings (Gollwitzer, Wicklund, & Hilton, 1982). In this paper, we extend SCT into the moral domain by suggesting that immoral and moral behavior lead to tension-inducing and tension-reducing forces, respectively: the tension that results from immoral behavior motivates subsequent moral action, and the tension-reducing identity assurance from moral behavior allows people to relax their moral strivings. Thus, we predict that identity-threatening immoral behavior pushes people to increase their moral strivings to rebuild their moral selves and identity-boosting moral behavior completes individuals’ moral selves, allowing them to relax their moral strivings. We suggest that these compensatory effects are a function of individuals’ private, personal desires to see themselves as moral. Although SCT predicts that both the effects of an initial threat and subsequent goal striving are strongest when they are publically-acknowledged (Brunstein & Gollwitzer, 1996; Gollwitzer, 1986), we suggest

Moral pendulum

5

that compensation in the moral domain can emerge in private as well as in public. Although public-acknowledgement may heighten the impact of individuals’ behaviors on their moral selfimages, we suggest that the activation of the moral domain reduces the necessity for public witnesses. That is, because of the importance individuals place on possessing a moral identity (Aquino & Reed, 2002; Monin & Jordan, 2008; Tetlock et al., 2000), as well as the quotidian temptations to act immorally, we theorize that the behavioral effects resulting from perceptions of a complete or an incomplete moral self will occur even without an audience. Research on moral cleansing and moral licensing supports our predictions. Moral Cleansing and the Need for Self-completion Consistent with SCT, research on moral cleansing suggests that acting immorally, which poses a threat to one’s moral self-image, pushes people to engage in actions that figuratively cleanse themselves of their past transgression and reassert their moral selves. Tetlock, Kristel, Elson, Green, and Lerner (2000), for instance, found that individuals who witnessed actions that threatened their sacred values subsequently behaved in ways that figuratively cleansed themselves. The mere contemplation of a “taboo tradeoff” (e.g., the choice between spending money to provide a life-saving organ transplant versus upgrading hospital equipment) increased individuals’ intentions to volunteer (e.g., donate their time to the organ donation community) relative to the contemplation of a tradeoff that did not violate their moral values. Similarly, individuals who wrote self-relevant stories using a list of negatively-valenced words (e.g., greedy, disloyal) subsequently indicated that they would donate more money to charity and engage in more environmentally-friendly actions than did those who used positively-valenced words (e.g., caring, generous) (Sachdeva, Illev, & Medin, 2009). More literally, Zhong & Liljenquist (2006) found that recalling their own immoral (versus moral) behavior led individuals

Moral pendulum

6

to use more cleansing-related words in a word fragment completion task and to rate cleansing products (e.g., soap) as more desirable than neutral products (e.g., pens). Taken together, these studies suggest that immoral actions push individuals to figuratively and literally cleanse themselves of moral violations. Thus, in concert with SCT, we suggest that threats to individuals’ moral self-images will lead them to engage in moral behavior to reassure themselves (and others) that they are moral individuals. Moral Licensing and the Need for Self-completion Research also suggests that moral action can earn individuals moral credentials, figuratively licensing them to relax their moral strivings and engage in morally-questionable behavior. Monin and Miller (2001), for instance, showed that individuals who signaled their gender or race egalitarianism were subsequently more likely to recommend a man for a traditionally male job or a Caucasian for a job in an environment that seemed inhospitable to racial minorities. The authors argued that the initial moral behavior helped allay individuals’ concerns that their subsequent behavior would be interpreted as morally-questionable, thereby licensing them to engage in dubious behavior. These compensatory actions were not due to selfpresentational motives – they surfaced even when participants believed they were being observed by two distinct audiences. Similarly, Sachdeva et al. (2009) found that individuals who wrote positively-valenced stories about themselves indicated that they would donate less money to charity and engage in fewer environmentally-friendly actions than individuals who wrote negatively-valenced stories. Khan and Dhar (2006) also found that people who had imagined performing a prosocial task (e.g., donating to charity or volunteering) were subsequently more likely than control participants to choose a luxurious rather than a utilitarian purchase. This licensing effect seemed

Moral pendulum

7

to operate beneath individuals’ conscious awareness: when asked why they chose the more hedonic item, only one participant in six noted that her choice was related to her initial prosocial action. Thus, research on moral licensing and SCT suggests that engaging in moral deeds can lead people to relax their guard against temptation, making them more susceptible to engaging in immoral behavior. The Current Research Despite this rich set of findings, several important questions about the dynamics of individuals’ moral behavior remain. The first concerns whether compensatory moral behavior is driven by the elicitation of moral or immoral thoughts about the self, in particular, or by the elicitation of positively- or negatively-valenced thoughts about the self, more generally (Forgas, 1998; Tamir & Robinson, 2007). For example, although Sachdeva et al. (2009) demonstrated that writing a personal story subsequently led to moral actions, the stories included positive and negative adjectives (e.g., friendly, indifference) that may have stimulated positive or negative images about the self, rather than changes to individuals’ moral self-images. A second unanswered question concerns the potency of prior and subsequent moral and immoral acts. Although research has documented that moral licensing can lead individuals to: engage in behavior that supports gender stereotypes (Monin & Miller, 2001); express less cooperative intentions for the good of the environment (Sachdeva et al., 2009); and consider purchasing luxurious items for themselves (Khan & Dhar, 2006), it is unclear whether moral licensing can encourage actual immoral behavior. Similarly, although unethical action leads people to wash their hands and choose soap over a pen (Zhong & Liljenquist, 2006), cleaning oneself is not the same as engaging in compensatory moral behavior. Previous research has shown that moral and immoral behaviors are not simply opposing poles on a single scale;

Moral pendulum

8

instead, people think about immoral and moral behavior as independent dimensions (Schwartz & Bilsky, 1990; Tannenbaum & Bartels, 2010). In other words, although donating blood may be perceived as a moral act, not donating blood is not perceived as immoral. Thus, we attempt to provide stronger evidence of the compensatory effects of past moral and immoral behavior by investigating whether initial moral behavior affects actual immoral behavior (i.e., cheating). Third, previous research has not addressed whether moral compensation is inversely proportional to the magnitude of the initial moral action. If individuals want to fortify their moral self-images, more egregious immoral behavior should require more intense remunerative action. We explored this question, as well as whether behaviors that create greater moral credentials might also lead to greater subsequent relaxation. Finally, previous research has often observed the effects of prior moral action in one domain on subsequent moral action within the same domain (e.g., recommending a man for a traditionally male job role after responding to questions about women and domestic life, Monin & Miller, 2001). This raises the question of whether the dynamic nature of moral action is domain-specific or possibly domain-general, suggesting that people combine rather than differentially categorize their moral behaviors as they formulate their moral self-images. Thus, the current research also extends previous work by investigating whether moral recollections in one domain influence subsequent moral behaviors in another, completely unrelated domain. In sum, the current research builds on previous research by demonstrating that (1) effects of initial moral or immoral behaviors can be distinguished from effects due to positively- or negatively-valenced non-moral activities, (2) moral licensing can lead individuals to engage in immoral behavior, (3) compensatory processes are inversely proportional to the magnitude of their initial stimulating event, and (4) the effects of prior moral behavior on subsequent moral

Moral pendulum

9

behavior are domain-general. Our findings, as well as research on SCT (Braun & Wicklund, 1989; Gollwitzer & Wicklund, 1982) and moral identity (Aquino & Reed, 2002; Reed et al., 2007), contribute to an integrated framework for understanding the impact of past moral behavior on future moral intentions and action. Overview We conducted three studies to illustrate the effects of past moral behavior on individuals’ self-reported moral behaviors, prosocial intentions, and immoral behavior. Study 1 examined the effects of recalling one’s own past moral or immoral behavior on subsequent, explicit attempts at symbolizing the moral self-image. It provided support for the effects of both moral cleansing and licensing: relative to control participants, individuals who recalled their past immoral behavior reported that they had a stronger symbolic moral identity (i.e., they participated in more activities that symbolized the centrality of their moral selves to others), whereas individuals who recalled their past moral behavior reported a weaker symbolic moral identity. It also demonstrated that these compensatory effects were inversely proportional to the moral magnitude of the previous behavior. Study 2 focused on prosocial intentions. It demonstrated that, in comparison to a control condition, recalling one’s own past moral behavior decreased subsequent prosocial intentions whereas recalling one’s own past immoral behavior increased subsequent prosocial intentions. It also demonstrated that these compensatory effects were unique to prior moral or immoral behavior as they did not emerge for individuals who recalled positive or negative non-moral behaviors. Lastly, like Study 1, the compensatory effects were inversely proportional to the moral magnitude of the previous behavior. .

Moral pendulum

10

Study 3 investigated the effects of recalling either one’s own or others’ moral behavior on subsequent, immoral behavior (i.e., cheating), and persistence in completing a challenging task. It demonstrated that the moral magnitude of recalled behaviors affected the magnitude of subsequent immoral behavior and persistence. In addition to replicating the compensatory effects of recalling one’s own actions, it found that recalling others’ moral behaviors led to consistent rather than compensatory moral action. Study 1: Moral Compensation and Moral Identity Study 1 examined the effects of individuals’ recollections of their own prior moral or immoral behavior on their moral identity. Individuals’ moral identities, which indicate how much they have “a self-conception organized around a set of moral traits” (Aquino & Reed, 2002, p. 1424), have two dimensions, internal and symbolic, both of which motivate moral choices and actions (Reed et al., 2007). To measure moral identity, participants consider nine traits that are exemplars of a moral individual and indicate the importance of possessing these traits (internal) and of demonstrating them to others (symbolic). Being a moral person, which is central to most individuals’ identities (Aquino & Reed, 2002; Monin & Jordan, 2008), leads to high scores on internal moral identity. We predict that, regardless of whether they recall their past moral or immoral behavior, the importance of possessing moral traits will remain strong. However, consistent with SCT, we predict that individuals who recall a time when they acted immorally will compensate by boosting their symbolic moral identity and that individuals who remember a time when they acted morally will reduce their symbolic moral identity. Thus, we predicted that internal moral identity scores would remain stable and high across experimental conditions but symbolic moral identity scores

Moral pendulum

11

would be (inversely) affected by individuals’ recall of their own previous moral or immoral behavior. Method Participants and design. Twenty-three MBA students (Mage = 28.83, SD = 2.71; 48% women) and 43 executive MBA students (Mage = 39.71, SD = 6.14; 24% women) participated as part of a classroom exercise. They were randomly-assigned to the moral or the immoral recall conditions in a between-participants design.2 An additional control condition was collected later from thirty-three adults (Mage = 35.19, SD = 7.73; 42% women) using an online subject pool; they participated in exchange for a $5 gift certificate. Seven participants, who did not complete the recall prompt (1 in the ethical, 4 in the unethical, and 2 in the control condition), were eliminated from the analyses; thus, N = 92 for the analyses. Procedures. Participants completed the materials online and anonymously. The experimental conditions asked them to recall “a time when [they] did something (ethical/unethical)3 in a work or professional setting.” The control condition dropped the word “ethical” or “unethical” from the instructions. Following a filler task (i.e., a work-related decision-making task), they completed the 10-item moral identity scale (Aquino & Reed, 2002). After reading nine moral identity adjectives (helpful, honest, caring, hardworking, compassionate, kind, fair, friendly, generous), participants indicated on 7-point scales how much they agreed with ten statements that reflected the importance of possessing (internal) and demonstrating to others that they possessed (symbolic) these traits. Sample items from the scale were, “It would make me feel good to be a person who has these characteristics,” (internal); and “The types of things I do in my spare time (e.g., hobbies) clearly identify me as having these characteristics” (symbolic). Responses to the internal and symbolic items were averaged

Moral pendulum

12

separately to create measures of internal (M = 6.40, SD = 0.70, α = .78) and symbolic moral identity (M = 4.81, SD = 1.05, α = .80). The subscales were positively correlated, r(92) = .326, p = .002. Story coding. We provided two coders, blind to conditions and hypotheses, with the following definition of moral behavior: “act[ing] in the service of human welfare” (Aquino & Reed, 2002, p. 1423). They then evaluated the morality of the recalled behaviors (7-point bipolar scale: -3 = very immoral; +3 = very moral). Initial inter-rater reliability was high (ICC = .76); coders discussed discrepancies to reach consensus. Results Manipulation check. The recall manipulation had the intended effect: a one-way ANOVA revealed a significant effect of condition on the coders’ morality ratings, F(2, 89) = 103.96, p < .001, see Table 1. Relative to the control condition, stories in the moral condition were rated as more moral, F(1, 89) = 24.90, p < .001, and stories in the immoral condition were rated as less moral, F(1, 89) = 86.49, p < .001. Moral Identity. A 3 (recall: moral, immoral, control) x 2 (moral identity: internal, symbolic) mixed-design ANOVA (with identity as the repeated factor) led to two significant main effects and the predicted interaction. Internal subscale scores were higher than symbolic subscale scores, F(1, 89) = 213.64, p < .001, ηp 2 = .71, and there was a significant main effect of recall condition, F(2, 89) = 3.11, p = .049, ηp 2 = .07. The interaction, F(2, 89) = 3.47, p = .036, ηp 2 = .07 (see Table 1) indicated that, as predicted, recall did not affect individuals’ internal moral identities, F(2, 89) = 2.27, p = .11; these scores were consistently high.4 In contrast, there was a significant effect of recall condition on symbolic moral identity, F(2, 89) = 3.67, p = .03. A significant linear trend indicated that

Moral pendulum

13

participants in the immoral recall condition had the highest symbolic moral identity and participants in the moral recall condition had the lowest; control participants fell in between these two groups, t(89) = 2.70, p = .008. In addition, the correlation between coders’ ratings of the morality of participants’ behavior and their prosocial intentions was negative and significant, r(92) = -.21, p = .049, indicating that increasingly moral behavior was associated with a lower reported symbolic moral identity. We also examined the correlation between morality ratings and prosocial behavior for participants exclusively in the moral and immoral conditions, which revealed a similar result, r(61) = -.28, p = .027. Discussion As predicted, activating memories of their past immoral behavior, compared to memories of their past moral behavior, led individuals to report that they were more moral actors in their everyday lives. The data was consistent with theories of moral cleansing and moral licensing and, as predicted by SCT, these effects were a function of individuals’ motivation to protect their cherished moral self-images. The consistently high mean of individuals’ internal moral identity scores across conditions reiterates the observation that individuals highly value their moral identities. Finally, changes to individuals’ symbolic moral identities were related to the intensity of individuals’ previous moral behavior, with greater moral behavior leading to greater reductions in individuals’ reported symbolic moral identities. Although Study 1 provided preliminary evidence of the compensatory effects of one’s own moral behavior, it also had limitations. First, the symbolic moral identity items assessed individuals’ interpretations of their current behavior rather than their future moral intentions. Thus, individuals’ actions might have reflected attempts at impression management rather than

Moral pendulum

14

attempts to reconstitute their threatened moral self-images (Schneider, 1981). This motivation might have also been stimulated by asking participants to recall their own moral behavior and rate their moral identity in the same session. In addition, our recall instructions were fairly intrusive, particularly in the immoral condition. This may have made participants more attuned to our research goals than to their actual tendencies. Study 2: Moral Compensation and Prosocial Intentions To alleviate these concerns, Study 2 used a different set of recall instructions and a different measure of moral compensation. Specifically, participants recalled previous actions involving help or harm to others (Jones, 1991; Reed et al., 2007; Velasquez & Rostankowski, 1985). We then led participants to believe that they were participating in two separate studies. As before, we included a control condition to determine whether recalling (im)moral behavior affected subsequent moral intentions over a non-moral baseline. And, finally, two additional conditions asked people to recall positive and negative non-moral behaviors to examine whether the mere positive or negative valence of recalled behavior might account for our previous results. We predicted that, in comparison to control participants, recalling moral (immoral) behaviors would decrease (increase) prosocial intentions and that recalling positive or negative non-moral behaviors would have no significant impact on individuals’ prosocial intentions. Thus, we predicted that compensatory effects would only surface in the moral domain. Method Participants and design. We recruited participants (N = 155; 62% women; Mage = 31.61, SD = 10.25) from an online subject pool; they received $5 for participating and were randomlyassigned to the moral, immoral, positive non-moral, negative non-moral, or control condition.

Moral pendulum

15

Seven participants were excluded for not completing the essay task (1 in the negative non-moral, 3 in the immoral, and 3 in the moral condition); thus, N = 148 for the analyses. Procedures. Participants logged-in to the study website and learned that they would be completing two unrelated studies. “Study 1” asked participants to write a story about their typical Tuesday (control) or about a time when they: helped other people (moral); used others to get something they wanted (immoral); achieved an important goal (positive non-moral); or failed to achieve an important goal (negative non-moral). The non-moral conditions captured situations that led to positive or negative self-assessments (Brunstein, 1993; Diener, 1984). In “Study 2,” participants completed a filler-task (i.e., 10 trivia questions) before responding to a series of randomly-ordered items about their likelihood of donating to charity, donating blood, volunteering, vacationing, attending a party, and seeing a movie in the next month (7-point scales: -3 = very unlikely; +3 = very likely). A principal components analysis (varimax rotation) revealed that the first three items loaded onto one, prosocial intentions component (eigenvalue = 2.09, loadings ranged from .68 to .77, Cronbach’s α = .60),5 and the latter three loaded onto another, leisure activities component (eigenvalue = 1.21, loadings ranged from .66 to .77, Cronbach’s α = .59). To enhance the cover story of two separate studies each was preceded by a separate consent form. Story coding. As in Study 1, two coders, blind to conditions and hypotheses, evaluated the morality of the recalled behaviors (7-point bipolar scale: -3 = very immoral; +3 = very moral). Initial inter-rater reliability was high (ICC = .86); coders discussed discrepancies to reach consensus. Results

Moral pendulum

16

Manipulation check. The recall manipulation had the intended effect: a one-way ANOVA revealed a significant effect of condition on the coders’ morality ratings, F(4, 142) = 27.87, p < .001, ηp2 = .44, see Table 2.6 Relative to the control condition, stories in the moral condition were rated as more moral, t(38.79) = 4.87, p < .001, d = 1.37, and stories in the immoral condition were rated as less moral, t(25.95) = 5.14, p < .001, d = 1.72. Stories in the positive and negative non-moral conditions did not differ from the control condition, t(63) = 1.26, p = .21, and, t(45.15) = 1.58, p = .12, respectively. Prosocial intentions. As predicted, recalling moral or immoral behavior led to compensatory prosocial intentions but remembering a positive or negative non-moral behavior did not. A one-way ANOVA on participants’ prosocial intentions revealed a significant effect of condition, F(4, 143) = 3.46, p = .01, ηp2 = .09, see Table 2. Relative to controls, participants in the moral condition had marginally weaker prosocial intentions, F(1, 143) = 3.36, p = .069, d = .44, and participants in the immoral condition had significantly stronger prosocial intentions, F(1, 143) = 4.12, p = .044, d = .56. Also as predicted, participants’ prosocial intentions in the positive and negative non-moral conditions did not differ from controls’, F(1, 143) = 1.27, p = .26, d = .26, and, F(1, 143) = 0.24, p = .63, d = .11, respectively. In addition, replicating the results of Study 1, the correlation between coders’ ratings of the morality of participants’ behavior and their prosocial intentions was negative and significant, r(147) = -.21, p = .01, indicating that increasingly moral behavior was associated with decreasingly prosocial intentions. We also examined the correlation between morality ratings and prosocial behavior for participants exclusively in the moral and immoral conditions, which revealed a similar result, r(50) = -.34, p = .02.

Moral pendulum

17

Leisure activities. A one-way ANOVA revealed that participants’ intentions to engage in leisure activities did not significantly differ as a function of condition, F(4, 142) = 0.83, p = .51, see Table 2. Discussion These results provided additional support for the moral compensation hypotheses using a less intrusive set of recall instructions and a different dependent measure: relative to controls, recalling moral behavior led to (marginally) weaker prosocial intentions whereas recalling immoral behavior led to stronger prosocial intentions. The findings also suggest that changes in prosocial intent were related to the intensity of individuals’ previous moral behavior, and that recalling success or failure at reaching a goal had little impact on prosocial intentions relative to control participants. We should note, however, that the prosocial intentions resulting from the positive non-moral and moral conditions did not significantly differ from one another (although the immoral and non-moral conditions did significantly differ from one another). Finally, by presenting the recall and intention tasks as parts of independent studies, we reduced the possibility that moral compensation was due to motivations to strategically self-present (Schneider, 1971). Studies 1 and 2 utilized participants’ self-reports of moral behavior rather than their actual behavioral. In addition, previous research has considered the presence or absence of moral (or “questionably moral,” viz. Monin & Miller’s [2001] hiring task) intent and behavior, but has rarely examined individuals’ explicitly immoral behavior. Thus, in Study 3 we examine actual immoral behavior (i.e., cheating). Study 3: Moral Compensation, Moral Standards, and Immoral Behavior

Moral pendulum

18

In addition to investigating the effects of one’s own moral behavior, Study 3 investigates the effects of another’s behavior on one’s own behavior to provide a more comprehensive analysis of the dynamics of moral behavior. Although others’ behaviors and achievements could plausibly influence an individual’s own sense of completeness, this indirect effect of others’ behaviors on one’s own identity completeness and subsequent identity strivings has gone relatively unexplored in the SCT literature. This hypothesized transfer from observation of others’ behavior to reflections on one’s own self-image should be particularly likely when the “other” shares an identity with the individual (Cialdini, Reno, & Kallgren, 1990; Larimer, Turner, Mallett, & Geisner, 2004). Thus, we examine the influence of similar others’ (im)moral behavior on one’s own subsequent immoral behavior. Social comparison processes provide important cues for evaluating one’s identity (Festinger, 1954); moral identity should be no exception. We propose that others’ actions can set standards that individuals must reach to maintain their cherished moral self-images. Thus, individuals who fall short these moral standards signaled by others’ behavior, may perceive a threat to their moral self-images, leading them to subsequently increase their own moral behavior. Similarly, observation of others’ immoral action is likely to have the reverse effect, reinforcing the completeness of individuals’ moral selves and providing them with license to relax their moral strivings. . In summary, we predict that the impact of others’ behaviors on individuals’ perceptions of their moral self-images will influence their moral behaviors, but in the opposing direction as an individual’s own actions would: actions that create feelings of moral completion (i.e., one’s own moral behavior or others’ immoral behavior) will lead to moral licensing, while behaviors that threaten one’s moral self-image (i.e., one’s own immoral behavior and others’ moral

Moral pendulum

19

behavior) should lead to moral cleansing. As with the previous studies, we again explored whether the moral magnitude of an individuals’ subsequent behavior would be inversely related to the moral magnitude of their initial behavior, and whether it would be directly related to the moral magnitude of other’s relevant behavior. Thus, recalling potent events should lead to potent effects - in a direction determined by whether the recalled behaviors were one’s own or others. Method Participants and design. The 168 undergraduate participants (Mage = 20.08, SD = 3.01; 69% women), recruited from a volunteer subject pool and paid $15 for participating, were randomly-assigned to conditions in a 2 (target: self, other) x 2 (recall: moral, immoral) betweenparticipants design. Procedures. We told participants that the study was about transferring paper-and-pencil tasks to the computer and that they would be completing both a qualitative (writing) and a quantitative (mathematical) task. To avoid demand characteristics, we used Study 2’s recall instructions. The “other” recall conditions used the same instructions except that participants were asked to write about the past behavior of someone with whom they shared an identity (March, 1994; Terry, Hogg, & White, 1999) rather than their own past behavior. Participants were then asked to solve a series of 15 math problems (i.e., adding 10 twodigit numbers) that required them to submit a correct answer before they could continue to the next problem (adapted from von Hippel, Lakin, & Shakarchi, 2005). The experimenter told participants that if they did not hit the spacebar immediately after the problem appeared onscreen, a programming “bug” would reveal the correct answer. The experimenter also explained that, although no one would know whether the participant had seen the answer, seeing it would sabotage the experiment. The “answer” appeared at the bottom right hand corner of the

Moral pendulum

20

participant’s screen 3.5 seconds after the problem appeared - unless the participant hit the spacebar. Von Hippel and colleagues used the appearance of the answer as a measure of cheating; however, because we were concerned that the answer would appear for slowresponders as well as intentional cheaters, we altered the task so that the answer that appeared was +/-1 from the mathematically-correct answer. For example, a problem string of “-25 + -12 + 18 + 32 + 18 + -19 + -24 + 26 + 14 + -23” would lead to a “6” appearing at the bottom righthand corner of the computer screen if the participant did not hit the space bar within 3.5 seconds (more than sufficient time to hit the spacebar; von Hippel et al., 2005) even though 5 was the correct answer. This change allowed us to identify participants who simply forgot to hit the spacebar from those who intentionally used the provided answer (i.e., cheated). Thus, for the purposes of consistency with previous research, we used von Hippel et al.’s measure of cheating: allowing the answer to appear. However, we also used the participant’s frequency of using the provided answer as a more sensitive measure of cheating. Participants’ scores on both of these measures could range from 0 (indicating they did not cheat on any problem) to 15 (indicating that they cheated on all 15 problems). We also assessed participants’ motivation to persist on the task by measuring the number of problems participants correctly completed before they cheated for the first time (number completed before first cheat). Participants’ scores on this measure could range from 0 (indicating that they cheated immediately) to 15 (indicating that they never cheated). Before leaving the lab, participants completed a questionnaire that included questions about their math skills and how easy they found the task: “How good are you at doing math calculations in your head?” (1 = not good at all; 7 = very good) and “How easy did you find the mental math task to be?” (1 = very difficult; 7 = very easy). We also probed for suspicion using a

Moral pendulum

21

funneled debriefing technique (Bargh & Chartrand, 2000): we first asked participants what they thought the study was about and then if they found any of the instructions or tasks unusual or confusing. Sixteen participants (4 in the immoral-self, 3 in the moral-self, 7 in the immoral-other, and 2 in the moral-other condition) believed that the computer bug was part of the experimental procedures and were eliminated from the analyses; thus, N = 152 for the analyses.7 Story coding. As in Studies 1 and 2, two coders, blind to conditions and hypotheses, rated the morality of the recalled behavior (7-point bipolar scale: -3 = very immoral; +3 = very moral). Initial inter-rater reliability was high (ICC = .80); coders discussed discrepancies to reach consensus. Results Manipulation check. The recall manipulation had the intended effect: a 2 (target: self, other) x 2 (recall: moral, immoral) ANOVA indicated that the stories in the moral condition were rated as more moral than those in the immoral condition, F(1, 148) = 460.42, p < .001, ηp 2 = .76. We also observed an unpredicted target effect, F(1, 148) = 15.22, p < .001, ηp 2 = .09, indicating that self-stories were judged as more moral than other-stories, and a significant target by recall interaction, F(1, 148) = 6.21, p = .014, ηp 2 = .04, indicating that, although moral-self and moral-other stories did not differ in moral magnitude, F(1, 148) = 0.178, p = .67, d = .24, other-immoral stories were rated as less moral than self-immoral stories, F(1, 148) = 21.53, p < .001, d = .45, see Table 3. Thus, although the magnitude of the difference between moral and immoral stories differed by target, moral stories were judged as more moral than immoral stories for both the self and other conditions. Math skill. On average, participants reported being proficient (i.e., above the scale midpoint) at calculating math problems in their head (M = 4.47, SD = 1.82) and rated the task as

Moral pendulum

22

slightly difficult (M = 3.78, SD = 1.74); their self-ratings were not significantly different across conditions, all ps > .60. Cheating behavior. Of the 152 participants, 71 (47%) cheated at least once. In the selfmoral recall condition, approximately 68% (26) of participants cheated at least once versus 26% (10) in the self-immoral condition, a significant difference, χ2(1, N = 76) = 13.55, p < .001, ηp2 = .42. In contrast, in the other-moral condition, 47% (18) of participants cheated at least once versus 45% (17) in the other-immoral condition, a non-significant difference, χ2(1, N = 76) = .053, p = .82, ηp2 = .03. Results of 2 (target: self, other) by 2 (recall: moral, immoral) ANOVAs on the frequency that individuals allowed the answer to appear, F(1, 148) = 14.32, p < .001, ηp2 = .09, and used the provided answer, F(1, 148) = 14.17, p < .001, ηp2 = .09, revealed the predicted interaction, see Table 4.8 Participants both allowed the answer to appear, t(148) = 3.26, p = .001, d = 1.02, and used the revealed answer, t(148) = 3.43, p = .001, d = 1.00, more after recalling their own moral behavior, compared to recalling their own immoral behavior, but they cheated less, on allowing the correct answer to appear, t(148) = 2.16, p = .03, d = 0.38, and (marginally) less on using the correct answer, t(148) = 1.89, p = .06, d = 0.36, after recalling others’ moral behavior in comparison to recalling others’ immoral behavior.9 Main effects of target and recall were nonsignificant (ps > .23). Persistence. Results of a 2 (target: self, other) by 2 (recall: moral, immoral) ANOVA revealed a significant main effect for recall on the number of problems completed before cheating: participants who recalled immoral behavior completed more problems (M = 11.45, SD = 6.06) than those who recalled a moral behavior (M = 8.70, SD = 6.54) before they cheated for the first time, F(1,148) = 7.71, p = .006, ηp2 = .05. A significant interaction, F(1,148) = 11.62, p

Moral pendulum

23

= .001, ηp2 = .07 (see Table 4), and planned comparisons indicated that participants who recalled their own moral behavior answered fewer problems before their first instance of cheating than participants who recalled their own immoral behavior, t(148) = 4.15, p < .001, d = 2.47, but there was no significant difference in persistence for participants who recalled others’ moral versus immoral behavior, t(148) = 0.65, p = .51.10 Morality ratings and magnitude of cheating and persistence. The coder’s morality ratings were correlated with the magnitude of cheating (i.e., frequency of using the provided answer), r(76) = .34, p = .002, and persistence, r(76) = -.29, p = .010, in the self conditions, and but not significantly related to the magnitude of cheating, r(76) = -.10, p = .35, or persistence, r(76) = .02, p = .85, in the other conditions, providing partial support for our moral magnitude predictions. Discussion Participants in the moral-self condition cheated more than participants in the immoralself condition and participants in the moral-other condition cheated marginally-less than participants in the immoral-other condition. Moreover, individuals in the moral-self condition persisted less on the problems before their first incidence of cheating, in comparison to individuals in the immoral-self condition. However, there was not a significant difference in persistence between those in the moral- and immoral-other conditions. In addition, the moral magnitude of individuals’ own past behavior was directly related to the magnitude of their cheating and persistence. These findings are consistent with SCT (Gollwitzer, 1986; Gollwitzer & Kirchoff, 1998): individuals were influenced more by their own prior moral behavior than by the moral behavior of others (even others with whom they shared an identity). A comparison of

Moral pendulum

24

the effect sizes for own versus others’ behavior also reveals that individuals’ own recalled behavior had larger effects on their subsequent cheating than did others’ recalled behavior. Thus, Study 3 extends Study 1 and 2’s results, indicating that recalling moral (or immoral) behaviors not only affects an individual’s reported moral behavior and moral intentions but also affects their actual moral behavior. Study 3’s findings also indicate that recalling others’ moral behaviors influences an individual’s subsequent moral behaviors, albeit in a consistent rather than a compensatory fashion. Thus, while recalling one’s own immoral behavior leads to increased moral strivings, recalling others’ immoral behaviors relaxes an individual’s subsequent strivings. The effects of others’ behavior, however, were only observed on individuals’ cheating, but not persistence, behavior. General Discussion The current results demonstrate the impact of recalling one’s own or others’ moral behavior as contrasting determinants of moral identity, intentions, and behavior. They also suggest that individuals’ desires to maintain a moral self-image through compensatory action or through actions that are consistent with similar others may be domain-general, as individuals’ moral intentions or behaviors were in different domains from their prior recollections, and depended, in part, on the magnitude of the initial moral or immoral act. Moreover, the findings also suggest that individuals’ recollections of their own moral behavior were powerful enough to license not just a reduced striving for moral behavior (e.g., reduced intentions to donate to charity or volunteer one’s time) but also increased their engagement in actual immoral behavior (i.e., cheating). The Pendulum of Moral Behavior

Moral pendulum

25

The current results, combined with previous findings (e.g., Monin & Miller, 2001; Sachdeva et al., 2009), provide the basis for the formation of a theory of dynamic moral behavior. The motive to maintain a favorable moral self-image (Bolton, Katok, & Zwick, 1998; Gollwitzer et al., 1982), combined with the difficulty of avoiding the temptation of immoral behavior, suggests that individuals’ moral behavior is dynamic, rather than static, and fluctuates over time as a function of the current completeness of an individual’s moral self. Behaving morally earns moral credentials, boosting individuals’ moral self-images and licensing them to relax their moral strivings. Acting immorally, in contrast, threatens individuals’ moral selfimages, requiring cleansing and remunerative moral action to restore their moral self-images. Combining these effects within the same individual suggests that moral behavior will follow a variable oscillation function in response to recent moral behaviors. In contrast, the moral behaviors of similar others seem to act as a signal of moral standards, which also influences the completeness of a person’s moral self-image: seeing others act morally suggests high moral standards, leading individuals to question the completeness of their moral self-images and encourages an increase in their own moral strivings and behavior. Conversely, seeing others act immorally suggests low moral standards, leading individuals to experience a sense of moral self-completion and to subsequent relaxations in their moral strivings and behavior (cf., Mullen & Nadler, 2008). Research on SCT suggests that both an initial threat and subsequent compensatory behavior will have their greatest effects when they are publically-acknowledged (Brunstein & Gollwitzer, 1996; Gollwitzer, 1986); the current findings suggest that public acknowledgment may not be necessary for either moral cleansing or moral licensing effects to emerge. Although our data does not directly resolve whether an audience is necessary, the results provide some

Moral pendulum

26

suggestive evidence. Across all three studies, participants recalled prior moral or immoral behavior relatively anonymously (e.g., in Study 2 they completed the study from the privacy of their own home and were told that the experimenter would not be able to identify their responses) and often believed that their initial recall responses were elicited for a study unrelated to a second study in which the dependent measure appeared. Although the recalled behavior may have had an audience when it originally occurred, their recall of that behavior (and the subsequent behavior it elicited) was completely private. Thus, it appears that privately reflecting on one’s past behavior was enough to produce changes in individuals’ current moral self-image, which, in turn, produced compensatory effects. We leave a more direct test of these issues to future research. In addition, a close examination of our findings suggests that individuals’ past immoral behavior may be more motivating than their past moral behavior. For example, a comparison of the effect sizes in Study 2 reveals that compensatory effects in response to immoral behavior (d = .56), relative to controls, is larger than the compensatory effects produced by moral behavior (d = .44). The desires to engage in moral cleansing following immoral behavior may be stronger than the licensing effects associated with engaging in moral behavior, possibly because the motivation that results from a violated self-image is likely to be quite strong (Gollwitzer et al., 1982). Thus, these findings suggest that individuals who have volunteered to help the needy may be less motivated to subsequently do wrong than are individuals who have stolen from the needy are motivated to subsequently do right. Although theoretical parsimony would argue for symmetric effects for recalling moral and immoral behavior, a positive view of humanity, based on the notion that most people act morally most of the time, clearly suggests that moral licensing should be less potent behaviorally than moral cleansing. Determining whether moral behavior produces

Moral pendulum

27

boosts to individuals’ moral self-images to the same degree that immoral behavior threatens individuals’ moral self-images is also an important avenue for future research. Limitations and Additional Future Directions This research raised a variety of questions. For example, although Study 3 found that recalling others’ moral actions led to subsequently similar moral behavior, Sachdeva et al. (2009) found no effects when they asked participants to write about “someone they knew” (p. 525). The discrepant findings may be a consequence of participants’ identifying conceptually distinct sets of individuals. For instance, Gino, Ayel, & Ariely (2009) recently demonstrated that similar others are necessary for their immoral behavior to influence one’s own (see also Goldstein et al., 2008). Thus, the behavior of others with whom one shares an identity might have more influence on the completeness of a person’s moral self-image than the behavior of unrelated others. Future research might therefore examine the effects of recalling the behaviors of a variety of different “others.” In addition, because our research focused on the direct effects of moral behavior, we did not explore a host of potential moderators of the past-to-future linkages of moral behavior. Future research might examine whether concepts like the cognitive complexity of an individual’s moral reasoning (Rest, 1986) or their proneness toward experiencing moral emotions like guilt or shame (Tangney, Dearing, Wagner, & Gramzow, 2000) might moderate the effects of past moral behavior on future behavior. Future research might also explore the effects of the timing of past moral behavior on subsequent moral behavior. The current studies provided a relatively short time span between the recall task and subsequent responses. Egregious immoral acts, for instance, might have long-term

Moral pendulum

28

effects on an individual’s behavior, and potent, attention-diverting events that follow moral behavior might blunt its subsequent impact on moral licensing. Conceiving of moral striving as goal pursuit might also provide a new frame of reference for theories of morality. We suggest that, just as people who have a goal of losing weight will occasionally engage in counter-goal behaviors (e.g., consume high-calorie foods) (see Fishbach, Zhang, & Koo, 2009 for a review), people who have a goal of being a moral person will occasionally succumb to moral temptations. Research on goal pursuit suggests that the certainty of individuals’ commitments to a goal influences whether their past actions lead to consistent or compensatory future actions: when they are uncertain of their goal commitment, they construe goal-consistent behavior as evidence of commitment to the goal and this, in turn, increases goalconsistent behavior (Koo & Fishbach, 2008). In contrast, when individuals are certain of their goal commitment, they construe goal-consistent actions as evidence of goal progress, which licenses them to engage in less goal-consistent behavior (see Fishbach, Zhang, & Koo, 2009 for a review). In the current context, we suggest that people are relatively certain of their commitment to being a moral person. Thus, consistent with the current findings, being reminded of past goal progress should decrease, rather than increase motivations to achieve their moral self. Future research could examine how manipulating participants’ perceived commitment to their moral selves affects the consistency between past to future moral behavior. Finally, the current study did not empirically investigate the mechanisms behind the selfand other-influences on moral behavior. Consistent with SCT, we suggest that individuals who reflect on their own or others’ moral behavior may be engaging in processes that, at least temporarily, alter their moral self-images. The experience of remembering moral behavior prompts broad, moral considerations that can be intensely personal. Future research might

Moral pendulum

29

explore the multitude of possible effects that can result from these kinds of self- and othercontemplations, as they are often intertwined; Study 2’s results suggest that many effects may be unique to a person’s moral self-image, as they seem to be less affected by the recall of mere positive or negative thoughts about the self. Conclusions This research suggests that people strive to be moral but that their moral behavior is dynamic rather than stable. These findings bring us back to the quote at the beginning of this paper. To facilitate the goal of acting morally, it may pay to heed Benjamin Franklin’s advice, as well as to extend it: to encourage moral behavior, we might remind individuals of their own immoral behavior, or of others’ moral behavior.

Moral pendulum

30

Footnotes 1

Yet participants reported a weaker drive to signal their moral identities to others (i.e.,

their symbolic moral identity): mean scores in this previous research were 3.10 and 4.43, respectively. 2

A 2 (condition: immoral, moral) x 2 (sample: MBA, executive MBA) ANOVA,

revealed a non-significant main effect of sample on internal moral identity, F(1, 57) = 1.77, p = .19, and a non-significant interaction, F(1, 57) < 1. Similarly, results of a 2 (condition: moral, immoral) x 2 (sample: MBA, executive MBA) ANOVA on symbolic moral identity revealed a non- significant main effect for sample and for the sample by condition interaction, Fs < 1. 3

Like other scholars (e.g., Jones, 1991; Trevino, 1986; Weber, 1996), we use the terms

moral and ethical interchangeably. 4

The lack of a difference between conditions on internal moral identity may have been

due to a ceiling effect. 5

An alpha of .60 meets the minimum level of acceptable reliability (Nunnally, 1967) for

our three-item scale. 6

The coder’s morality ratings did not exhibit homogenous variance across conditions,

Levene’s F(2, 142) = 13.35, p < .01, and sample sizes were unequal. Thus, the ANOVA should be interpreted with caution. Thus, we report unequal-variance t-tests in the follow-up comparisons (Welch, 1937). These t- tests led to identical conclusions as equal-variance (i.e., traditional) t-tests. 7

Suspicion rates did not significantly differ by recall, χ2(1, N = 168) = 2.04, p = .15, or

target condition, χ2(1, N = 168) = .181, p = .67. In addition the target x recall interaction was non-significant, Breslow-Day χ2(1, N = 168) = 0.64, p = .43.

Moral pendulum 8

31

To rule out the possibility that these effects were due to participants’ organic math skills,

we conducted an ANCOVA for both the cheating and the persistence analyses using selfreported math skill and the ease with which the individuals perceived the task as covariates. The interaction effects remained significant (i.e., all Fs > 12.24, all ps < .001). 9

The cheating and persistence variables did not have homogenous variance across

conditions, all Levene’s Fs(3, 148) > 4.36, ps < .01. However, the equal n in each condition significantly reduced the chances of increased Type 1 error associated with this violation (Mulligan, Wong, Thompson, 1987), leading us to report traditional follow-up comparisons. As a precaution, however, we also conducted unequal-variance t-tests. Doing so, produced identical conclusions as equal variance t-tests for all comparisons except for the other-moral versus otherimmoral comparisons for our cheating variables, in which the unequal variance t-tests produced a larger p-value for allowing the answer to appear (p = .08) and the frequency of using the provided answer (p = .107), respectively. 10

We also tested whether recalling one’s own behavior differed from recalling others’

behavior separately in the moral and immoral recall conditions. Relative to participants who recalled their own immoral behavior, participants who recalled others’ immoral behavior: allowed the answer to appear more frequently, t(148) = 3.65, p < .001, d = 0.80; used the answer more frequently, t(148) = 3.29, p = .001, d = 0.73, and answered fewer problems before first cheating, t(148) = 2.26, p = .03, d = 0.54. In contrast, relative to participants who recalled others’ moral behavior, participants who recalled their own moral behavior: allowed the answer to appear (marginally) more frequently, t(148) = 1.77, p = .08, d = 0.43, used the answer more frequently, t(148) = 2.04, p = .04, d = 0.48, and answered fewer problems before first cheating, t(148) = 2.54, p = .01, d = 0.56.

Moral pendulum

32

References Aquino, K., & Reed, A. (2002). The self-importance of moral identity. Journal of Personality and Social Psychology, 83, 1423-1440. Associated Press (2008, July 11). Nebraska nun gets prison for stealing from church. Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle: A practical guide to priming and automaticity research. Handbook of Research Methods in Social Psychology (pp. 253-285). New York: Cambridge University Press. Barry, H., & Tyler, T. R. (2009). The other side of injustice: When unfair procedures increase group-serving behavior. Psychological Science, 20, 1026-1032. Bold, K. (1999, March 23). Ex-con’s mission: keeping teens out of prison. Los Angeles Times, E-1. Bolten, G. E., Katok, E., & Zwick, R. (1998). Dictator game giving: Rules of fairness and acts of kindness. International Journal of Game Theory, 27, 269-299. Braun, O. L., & Wicklund, R. A. (1989). Psychological antecedents of conspicuous consumption. Journal of Economic Psychology, 10, 161-187. Brunstein, J. C. (1993). Personal goals and subjective well-being: A longitudinal study. Journal of Personality and Social Psychology, 65,1061 - 1070. Brunstein, J. C., & Gollwitzer, P. M. (1996). The effects of failure on subsequent performance: The impact of self-defining goals. JPSP, 70, 395-407. Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social Psychology, 58, 1015-1026. Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95,542-575.

Moral pendulum

33

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117140. Fishbach, A. Zhang, Y., & Koo, M. (2009). The dynamics of self-regulation. European Review of Social Psychology, 20, 15-344. Forgas, J. P. (1998). Asking nicely: The effects of mood on responding to more or less polite requests. PSPB, 24, 173-195. Gimein, M. (2002, September 16). Eliot Spitzer; The enforcer forget the perp walks: What he wants is change – top to bottom. Fortune, 76. Gino, F., Ayal, S., & Ariely, D. (2009). Contagion and differentiation in unethical behavior: The effect of one bad apple on the barrel. Psychological Science, 20(3), 393-398. Goldstein, N. J., Cialdini, R. B., & Griskevicius, V. (2008). A room with a viewpoint: Using social norms to motivate environmental conservation in hotels. Journal of Consumer Research, 35, 472-482. Gollwitzer, P. M. (1986). Striving for specific identities: The social reality of selfsymbolizing. In R. Baumeister (Ed.), Public self and private self (pp. 143-159). New York: Springer-Verlag. Gollwitzer, P. M., & Kirchhof, O. (1998). The willful pursuit of identity. In J. Heckhausen & C. S. Dweck (Eds.), Life-span perspectives on motivation and control (pp. 389423). New York: Cambridge University Press. Gollwitzer, P. M., Sheeran, P., Michalski, V., & Seifert, A. E. (2009). When intentions go public: Does social reality widen the intention-behavior gap? Psychological Science, 20, 612618.

Moral pendulum

34

Wicklund, R. A., & Gollwitzer, P. M. (1982). Symbolic self-completion. Hillsdale, N.J.: Lawrence Erlbaum. Gollwitzer, P. M., Wicklund, R. A., & Hilton, J. L. (1982). Admission of failure and symbolic self-completion: Extending Lewinian theory. Journal of Personality and Social Psychology, 43, 358-371. Jones, T. M. (1991). Ethical decision making by individuals in organizations: An issuecontingent model. Academy of Management Review, 16, 366-395. Khan, U., & Dhar, R. (2006). Licensing effect in consumer choice. Journal of marketing research, 43, 259-266. Larimer, M. L., Turner, A. P., Mallett, K. A., & Geisner, I. M. (2004). Predicting drinking behavior and alcohol-related problems among fraternity and sorority members: Examining the role of descriptive and injunctive norms. Psychology of Addictive Behaviors, 18, 203-212. Ledgerwood, A., Liviatan, I., & Carnevale, P. J. (2007). Group-identity completion and the symbolic value of property. Psychological Science, 18, 873-878. Locke, E. A., & Latham, G. P. (1990). A Theory of Goal Setting and Task Performance. Englewood Cliffs, NJ: Prentice Hall. Mazar N., Amir O. & Ariely D. (2008). The dishonesty of honest people: A theory of self-concept maintenance. Journal of Marketing Research, 45, 633–644. March, J. (1994). A primer on decision making: How decisions happen. New York: Simon & Schuster. Monin, B., & Jordan, A. H. (2008). Dynamic moral identity: A social psychological perspective. In D. Narvaez & D. Lapsley (Eds.), Moral Self, Identity and Character: Prospects for a New Field of Study, Cambridge University Press.

Moral pendulum

35

Monin, B., & Miller, D. T. (2001). Moral credentials and the expression of prejudice. Journal of Personality and Social Psychology, 81, 33-43. Mullen, E., & Nadler, J. (2008). Moral spillovers: The effect of moral violations on deviant behavior. Journal of Experimental Social Psychology, 44, 1239-1245. Mulligan, G. W., Wong, D. S., & Thompson, P. A. (1987). Robustness Properties of Nonorthogonal Analysis of Variance. Psychological Bulletin, 101, 464-470. Nisan, M. (1991). The moral balance model: Theory and research extending our understanding of choice and deviation. In W. M. Kurtines & J. L. Gerwitz (Eds.), Handbook of Moral Behavior and Development, Vol. 3. Hillsdale, NJ: Lawrence Erlbaum. Nucci, L. (2002). The development of moral reasoning. In U. Goswami (Ed.), Blackwell handbook of child cognitive development (pp. 303-325). Oxford: Blackwell. Nunnally, J. (1967). Psychometric theory (1st ed.). New York: McGraw-Hill. Reed, A., Aquino, K., & Levy, E. (2007). Moral identity and judgments of charitable behaviors. Journal of Marketing, 71, 178-193. Reeder, G. D., Kumar, S., Hesson-McInnis, M. S., & Trafimow, D. (2002). Inferences about the morality of an aggressor: The role of perceived motive. Journal of Personality and Social Psychology, 83, 789-803. Rest, J. R. (1986). Moral development: Advances in research and theory. New York: Preager. Rubenstein, C. (2004, June 6). A bachelor’s in redemption. The New York Times, SECT14WC, 4. Sachdeva, S., Illev, R., Medin, D. (2009). Sinning saints and saintly sinners: The paradox of moral self-regulation. Psychological Science, 20, 523-528.

Moral pendulum

36

Schneider, D. J. (1981). Tactical self-presentations: Towards a broader conception. In J. T. Tedeshi (Ed.), Impression management theory and social psychological research. New York: Academic Press. Schwartz, S. H. and Bilsky, W. (1990). Toward a theory of the universal content and structure of values: Extensions and cross cultural replications. Journal of Personality and Social Psychology, 58, 878-891. Singh, R., & Teoh, J.B.P. (2000) Impression formation from intellectual and social traits: evidence for behavioural adaptation and cognitive processing. British Journal of Social Psycholology, 39, 537–554. Tamir, M., & Robinson, M. D. (2007). The happy spotlight: Positive mood and selective attention to rewarding information. Personality and Social Psychology Bulletin, 33, 1124-1136. Tangney, J.P., Dearing, R., Wagner, P.E., & Gramzow, R. (2000). The Test of SelfConscious Affect – 3 (TOSCA-3). George Mason University, Fairfax VA. Tannenbaum, D., & Bartels, D. (2010, January). Value-driven choice and trade off reasoning in consumer choice. Poster session presented at the annual meeting of the Society for Personality and Social Psychology, Las Vegas, Nevada. Terry, D. J., Hogg, M. A., & White, K. M. (1999). The theory of planned behaviour: Selfidentity, social identity, and group norms,” British Journal of Social Psychology, 38, 225 - 244. Tetlock, P. E, Kristel, O.V., Elson, S.B., Green, M. C. & Lerner, J. S. (2000). The psychology of the unthinkable: Taboo trade-offs, forbidden base rates, and heretical counterfactuals. Journal of Personality and Social Psychology, 78, 853-870. Trevino, L. K. (1986). Ethical decision-making in organizations: A person-situation interactionist model. Academy of Management Review, 11, 601-617.

Moral pendulum

37

Velasquez, M. G., & Rostankowski, C. (1985). Ethics: Theory and practice. Englewood Cliffs, NJ: Prentice-Hall. Von Hippel, W., Lakin, J. L., Shakarchi, R. L. (2005). Individual differences in motivated social cognition: The case of self-serving information processing. Personality and Social Psychology Bulletin, 31, 1347-1357. Weber, J. (1996). Influences upon managerial moral decision making: Nature of the harm and magnitude of consequences. Human Relations, 49, 1-22. Welch, B. L. (1937). The significance of the difference between two means when the population variances are unequal. Biometrika, 29, 350-362. Zhong, C., & Liljenquist, K. (2006). Washing away your sins: Threatened morality and physical cleansing. Science, 313, 1451-1452.

Moral pendulum Table 1. Descriptive statistics for coders’ ratings of the morality of the essay, and participants’ internal and symbolic moral identity scores by recall condition, Study 1. Content of Recalled Story Moral (n = 34)

Control (n = 31)

Immoral (n = 27)

Morality of Essay

1.57 (0.83)a

0.60 (0.77)b

-1.33 (0.76)c

Internal Moral Identity

6.44 (0.54)a

6.21 (0.95)a

6.59 (0.46)a

Symbolic Moral Identity

4.48 (1.11)c

4.85 (1.06)b,c

5.19 (0.84)b

Notes. Morality of the essay could range from -3 (very immoral) to +3 (very moral). Standard deviations are provided in parentheses. Means with different subscripts within each row differ significantly from each other at p < .05.

38

Moral pendulum

39

Table 2. Descriptive statistics for coders’ ratings of the morality of the essay, and participants’ prosocial behavior and leisure activity intentions as a function of condition, Study 2. Moral

Immoral

Control (n = 38)

Positive Non-moral (n = 27)

Negative Non-moral (n = 32)

(n = 30)

(n = 21)

Morality of Essay

0.91 (0.83)a

-0.71 (0.66)c

0.11 (0.38)b

0.26 (0.56)b

0.01 (0.12)b

Prosocial Behavior

-0.73 (1.06)ad

0.57 (1.22)c

-0.15 (1.55)b

-0.52 (1.29)bd

-0.30 (1.24)b

Leisure Activities

0.17 (1.31)a

0.38 (1.78)a

0.48 (1.43)a

-0.21 (1.73)a

0.23 (1.71)a

Notes. Morality of the essay could range from -3 (very immoral) to +3 (very moral); prosocial behavior and leisure activities could range from -3 (very unlikely) to +3 (very likely). Standard deviations are provided in parentheses. Means with different subscripts within each row differ significantly from each other at p < .05, with the exception of the moral vs. control comparison for prosocial behavior where p = .06.

Moral pendulum

40

Table 3. Descriptive statistics for coders’ ratings of the morality of the essay, measures of cheating behavior and persistence as a function of target by recall condition, Study 3. Content of Recalled Story Moral

Immoral

Self 1.42 (0.78)a

-0.89 (0.87)b

Other 1.25 (0.68)a

-1.66 (0.65)c

Morality of Essay

Frequency of Allowing Answer to Appear (Range: 0-15) Self 3.58 (3.83)a

0.71 (1.21)b

Other 2.00 (3.46)a

3.84 (5.56)c

Self 3.03 (3.52)a

0.45 (1.01)b

Other 1.47 (2.59)c

2.82 (4.60)d

Frequency of Using Provided Answer (Range: 0-15)

Number Completed Before First Cheat (Range: 0-15) Self

6.76 (3.37)a

12.89 (0.96)b

Other 10.63 (6.21)c

10.00 (6.59)c

Notes: Coder’s ratings of the morality of the essay could range from -3 (very immoral) to +3 (very moral). Higher values on the frequency of allowing the answer to appear and using the provided answer indicate greater cheating. Lower values on the number of problems completed before first cheating indicate lesser persistence before cheating. Standard deviations are provided in parentheses. Means with different subscripts within the same row and column significantly differ from each other, p < .05, with the exception of the moral other vs. immoral other comparison on frequency of using the provided answer, where p = .06.

Moral pendulum Figure 1. Likelihood of Engaging in Prosocial Behavior as a Function of Condition, Study 2.

Note: Error bars represent ± 1 standard error of the mean.

41

Moral pendulum

42

Authors Note We thank Katherine DeCelles, Adam Galinsky, Katie Liljenquist, Benoît Monin, ChenBo Zhong and members of the Monin/Mullen lab for their comments on an earlier version of this manuscript. We also thank Richard Shakarchi for his generosity with sharing materials for the cheating task used in Study 3.