Mechanisms of Moral Disengagement in the Exercise of …eushe2/Bandura/Bandura1996JPSP.pdfof moral justification, detrimental conduct is made personally and socially acceptable by

Journal of Personality and Social Psychology1996, Vol. 71, No. 2, 364-374

Copyright 1996 by the American Psychological Association, Inc.0022-3514/96/S3.O0

Mechanisms of Moral Disengagement in the Exercise of Moral Agency

Albert BanduraStanford University

Claudio Barbaranelli, Gian Vittorio Caprara,and Concetta Pastorelli

University of Rome, "La Sapienza"

This research examined the role of mechanisms of moral disengagement in the exercise of moralagency. Regulatory self-sanctions can be selectively disengaged from detrimental conduct by con-verting harmful acts to moral ones through linkage to worthy purposes, obscuring personal causalagency by diffusion and displacement of responsibility, misrepresenting or disregarding the injuriouseffects inflicted on others, and vilifying the recipients of maltreatment by blaming and dehumanizingthem. The study examined the structure and impact of moral disengagement on detrimental conductand the psychological processes through which it exerts its effects. Path analyses reveal that moraldisengagement fosters detrimental conduct by reducing prosocialness and anticipatory self-censureand by promoting cognitive and affective reactions conducive to aggression. The structure of thepaths of influence is very similar for interpersonal aggression and delinquent conduct. Although thevarious mechanisms of moral disengagement operate in concert, moral reconstruals of harmfulconduct by linking it to worthy purposes and vilification of victims seem to contribute most heavilyto engagement in detrimental activities.

Psychological theories of moral agency focus heavily onmoral thought to the neglect of moral conduct. The limited at-tention to moral conduct reflects both the rationalistic bias ofmany theories of morality (Kohlberg, 1984) and the conve-nience of investigatory method. It is much easier to examinehow people reason about hypothetical moral dilemmas than tostudy how they behave in difficult life predicaments. Peoplesuffer from the wrongs done to them, regardless of how perpe-trators might justify their inhumane actions. The regulation ofconduct involves much more than moral reasoning. A theory ofmorality must specify the mechanisms by which people cometo live in accordance with moral standards. In social cognitivetheory (Bandura, 1991), moral reasoning is translated into ac-tions through self-regulatory mechanisms through which moralagency is exercised.

In the course of socialization, moral standards are con-structed from information conveyed by direct tuition, evalua-tive social reactions to one's conduct, and exposure to the self-evaluative standards modeled by others. Once formed, suchstandards serve as guides and deterrents for action. People reg-ulate their actions by the consequences they apply to them-

Albert Bandura, Department of Psychology, Stanford University;Claudio Barbaranelli, Gian Vittorio Caprara, and Concetta Pastorelli,University of Rome, "La Sapienza," Rome, Italy.

The research reported in this article was supported by grants from theSpencer Foundation to Albert Bandura and from the Johannn JacobsFoundation to Gian Vittorio Caprara. We thank Delbert Elliott for hisassistance in the early phases of the development of the scale of moraldisengagement.

Correspondence concerning this article should be addressed to AlbertBandura, Department of Psychology, Stanford University, Stanford,California 94305-2130; or to Gian Vittorio Caprara, Dipartimento diPsicologia, Universita Degli Studi di Roma, "La Sapienza," Via deiMarsi, 78,00185 Roma, Italy.

selves. They do things that give them satisfaction and a sense ofself-worth. They refrain from behaving in ways that violate theirmoral standards, because such behavior will bring self-censure.In the face of situational inducements to behave in inhumaneways, people can choose to behave otherwise, by exerting coun-teracting self-influence. Anticipatory self-sanctions thus keepconduct in line with internal standards. It is through the ongo-ing exercise of self influence that moral conduct is motivatedand regulated.

Social cognitive theory grounds moral agency in a self-regu-latory system that operates through three major subfunctions.These include self-monitoring, judgmental, and self-reactivesubfunctions. Self-monitoring of one's conduct is the first steptoward exercising control over it. Action gives rise to self-reac-tions through a judgmental function in which conduct is evalu-ated against internal standards and situational circumstances.Moral judgment sets the occasion for self-reactive influence.People get themselves to behave in accordance with their moralstandards through anticipatory positive and negative self-reac-tions for different courses of action.

Development of self-regulatory functions does not create aninvariant control system within a person, as implied by theoriesof internalization that incorporate entities such as consciences,superegos, or moral principles as perpetual internal overseers ofconduct. Self-reactive influences do not operate unless they areactivated, and there are many psychosocial processes by whichself-sanctions can be disengaged from inhumane conduct(Bandura, 1990, 1991). Selective activation and disengagementof internal control permits different types of conduct with thesame moral standards. Figure 1 summarizes schematically thefour major points in the self-regulatory system at which internalmoral control can be disengaged from detrimental conduct.Self-sanctions can be disengaged by reconstruing the conduct,obscuring personal causal agency, misrepresenting or disregard-ing the injurious consequences of one's actions, and vilifying therecipients of maltreatment by blaming and devaluating them.

364

MORAL DISENGAGEMENT 365

MORAL JUSTIFICATIONPALLIATIVE COMPARISONEUPHEMISTIC LABELING

MINIMIZING, IGNORING,OR MISCONSTRUING THE

CONSEQUENCES

Figure 1. Mechanism through which moral self-sanctions are selectively activated and disengaged fromdetrimental behavior at different points in the self-regulatory process. From Social Foundations of Thoughtand Action: A Social Cognitive Theory (p. 376) by A. Bandura, 1986. Copyright 1986 by Prentice-Hall, Inc.Reprinted by permission of Prentice-Hall, Inc., Upper Saddle River, NJ.

A key set of disengagement practices operates on the con-strual of injurious behavior itself. People do not ordinarily en-gage in reprehensible conduct until they have justified to them-selves the Tightness of their actions. What is culpable can bemade righteous through cognitive reconstrual. In this processof moral justification, detrimental conduct is made personallyand socially acceptable by portraying it in the service of valuedsocial or moral purposes (Kelman & Hamilton, 1989; Kramer,1990; Sanford & Comstock, 1971). People then act on a socialor moral imperative. In the transactions of everyday life, a lot ofaggressive behavior gets justified in the name of protectinghonor and reputation (Cohen & Nisbett, 1994).

Language shapes people's thought patterns on which theybase many of their actions. Activities can take on markedlydifferent appearances depending on what they are called. Eu-phemistic language thus provides a convenient tool for maskingreprehensible activities or even conferring a respectable statusupon them (Bolinger, 1982; Lutz, 1987). Through sanitizedand convoluted verbiage, destructive conduct is made benignand those who engage in it are relieved of a sense of personalagency. Laboratory studies have revealed the disinhibitorypower of euphemistic language (Diener, Dineen, Endresen,Beaman, & Fraser, 1975). People behave much more aggres-sively when assaulting a person is given a sanitized label thanwhen it is called aggression.

Behavior can also assume very different qualities dependingon what it is contrasted with. By exploiting advantageous com-parison with more reprehensible activities, injurious conductcan be rendered benign or made to appear to be of little conse-quence. The more flagrant the contrasted activities, the morelikely it is that one's own injurious conduct will appear triflingor even benevolent (Bandura, 1991). Cognitive transformationof harmful conduct into good conduct through moral justifi-cations and palliative characterizations by euphemistic labelingand behavioral contrasts is the most effective psychologicalmechanism for disengagement of self-sanctions. This is becauseinvesting injurious means with high social or moral purpose not

only eliminates self-deterrents but also engages self-approval inthe service of harmful exploits. What was once morally censur-able becomes a source of positive self-valuation.

Self-sanctions are activated most strongly when personalagency for detrimental effects is acknowledged. The second setof dissociative practices operates by obscuring or distorting theagentive relationship between actions and the effects they cause.Under displacement of responsibility, people view their actionsas springing from the social pressures or dictates of others ratherthan as something for which they are personally responsible(Andrus, 1969). Because they are not the actual agents of theiractions, they are spared self-censuring reactions. Hence, theyare willing to behave in ways they normally repudiate if a legiti-mate authority accepts responsibility for the effects of their ac-tions (Diener, 1977; Milgram, 1974).

The exercise of moral control is also weakened when personalagency is obscured by diffusion of responsibility for detrimentalconduct. This is achieved in several ways. Responsibility can bediffused by division of labor for a venture with different mem-bers performing subdivided aspects that seem harmless inthemselves but harmful in its totality (Kelman, 1973). Groupdecision making is another common practice, one that enablesotherwise considerate people to behave inhumanely. When ev-eryone is responsible, no one really feels responsible. Group ac-tion is still another expedient for weakening moral control. Anyharm done by a group can always be attributed largely to thebehavior of others. People behave more cruelly under group re-sponsibility than when they hold themselves personally ac-countable for their actions (Bandura, Underwood, & Fromson,1975; Diener, 1977;Zimbardo, 1969, 1995).

Additional ways of weakening self-deterring reactions operateby disregarding or distorting the consequences of action. Whenpeople pursue activities harmful to others for personal gain, orbecause of social inducements, they avoid facing the harm theycause, or they minimize it. They readily recall prior informa-tion given them about the potential benefits of the behavior butare less able to remember its harmful effects (Brock & Buss,

366 BANDURA, BARBARANELLI, CAPRARA, AND PASTORELLI

1962, 1964). In addition to selective inattention and cognitivedistortion of effects, the misrepresentation may involve activeefforts to discredit evidence of the harm they cause. As long asthe detrimental results of one's conduct are ignored, mini-mized, distorted, or disbelieved, there is little reason for self-censure to be activated. In his studies of commanded aggres-sion, Milgram (1974) obtained diminishing obedience as vic-tims' pain became more evident and personalized.

The final set of disengagement practices operates on the re-cipients of detrimental acts. The strength of moral self-sanc-tions depends partly on how perpetrators view the people theymistreat. To perceive another as human activates empatheticand vicarious emotional reactions through perceived similarity(Bandura, 1992; McHugo, Smith, & Lanzetta, 1982). The joysand suffering of those with whom one identifies are more vicar-iously arousing than are those of strangers, out-group members,or those who have been divested of human qualities. It is, there-fore, difficult to mistreat humanized persons without riskingpersonal distress and self-censure. Subhumans are regarded asnot only lacking sensitivities but also as being influenceableonly by harsh means. In research on the dynamics of victimiza-tion, Perry, Williard, and Perry (1990) reported that aggressivechildren exhibit little sympathetic concern over hurting deval-ued peers. Habitual aggressors care less about inflicting suffer-ing on victimized classmates than on those who are not cast inthe devalued-victim role.

Self-censure for injurious conduct can be disengaged orblunted by dehumanization that divests people of human qual-ities or attributes bestial qualities to them. Once dehumanized,they are no longer viewed as persons with feelings, hopes, andconcerns but as subhuman objects (Haritos-Fatouros, 1988;Keen, 1986; Kelman, 1973). In experimental studies in whichotherwise considerate people are given punitive power, theytreat dehumanized individuals much more harshly than hu-manized ones (Bandura et al., 1975). Dehumanization fostersdifferent patterns of thought. People enlist moral justificationsfor punitive conduct directed toward individuals who have beendeprived of humanness, but they disavow punitive actions andcondemn them on moral grounds toward individuals depictedin humanized terms.

Blaming one's adversaries or circumstances is still anotherexpedient that can serve self-exonerative purposes. In moraldisengagement by attribution of blame, people view themselvesas faultless victims driven to injurious conduct by forcible prov-ocation. Punitive conduct thus becomes a justifiable defensivereaction to instigations. Victims get blamed for bringing suffer-ing on themselves (Ferguson & Rule, 1983). Self-exoneration isalso achievable by viewing one's harmful conduct as forced bycompelling circumstances rather than as a personal decision.Even very young children are quite skilled in using mitigatingfactors to excuse harmdoing (Darley, Klosson, & Zanna,1978). Children who are ready aggressors are quick to ascribehostile intent to others, which provides justification for preemp-tive retaliatory acts (Crick & Dodge, 1994). By fixing the blameon others or on circumstances, not only are one's own injuriousactions excusable but one can even feel self-righteous in theprocess.

The disinhibitory effects of the various forms of moral disen-gagement have been extensively documented in the perpetration

of large-scale inhumanities (Andrus, 1969; Keen, 1986; Kel-man & Hamilton, 1989; Rapoport & Alexander, 1982; Reich,1990). Laboratory studies have further verified that conditionsconducive to disengagement of moral self-sanctions heightenpunitive behavior (Bandura et al., 1975; Diener, 1977; Dieneret al., 1975; Milgram, 1974; Tilker, 1970; Zimbardo, 1969).However, because of the lack of measures of moral disengage-ment, the mediation of the effect of the manipulated conditionson punitive behavior through self-exonerative processes hasbeen presumed rather than assessed.

The present research addressed several key issues concerningthe exercise of moral agency. Research in this area has beenseriously hampered by the lack of measures of moral disengage-ment. The instrument devised and tested in this program ofresearch is grounded in a sociocognitive theory of moral agencythat specifies the different loci in the self-regulatory systemwhere moral self-sanctions can be effectively disengaged. Boththe naturalistic and laboratory investigations usually examineonly a single or a subset of disengagement mechanisms. Thepresent study investigated how the full set of moral disengage-ment mechanisms operate in concert on socially injurious andantisocial conduct under naturally occurring conditions.

Moral disenagement can affect detrimental behavior both di-rectly and by its impact on other theoretically relevant determi-nants. Therefore, this research also tested a conceptual modelof the paths of influence through which moral disengagementproduces its behavioral effects. The directional paths are speci-fied both by theory (Bandura, 1991) and by empirical tests ofparticular links in the model (Caprara & Pastorelli, 1993). Inthe proposed causal structure of the model, which is presentedin Figure 2, moral disengagement influences detrimental behav-ior both directly and through its effects on the following medi-ating factors. People have little reason to be troubled by guilt orto feel any need to make amends for inhumane conduct if theyreconstrue it as serving worthy purposes or if they disown per-sonal agency for it. We therefore predicted that high moral dis-engagement would be accompanied by low guilt, thus weaken-ing anticipatory self-restraints against engagement in detrimen-tal behavior. We further predicted that self-exoneration forharmful conduct and self-protective dehumanization of othersand treating them as blameworthy would spawn a low prosocialorientation. Low prosocialness would, in turn, contribute todetrimental conduct in two ways. Having little sympathy forothers would both remove the restraining influence of empa-thetic considerateness to the mistreatment of others and wouldactivate little anticipatory guilt over such behavior. Effectivemoral disengagement creates a sense of social rectitude and self-righteousness that breeds ruminative hostility and retaliatorythoughts for perceived grievances. People often ruminate hos-tilely but do not act on their feelings. However, freed from therestraint of moral self-sanctions, they are more likely to act outtheir resentments. In this mediated link, moral disengagementfosters aggression proneness indexed by irascibility and hostilerumination which, in turn, heighten the likelihood of aggressiveand transgressive behavior. Thus, in the sociocognitive concep-tual model, moral disengagement affects aggressive and trans-gressive conduct both directly and through its influence on an-ticipatory guilt reactions, prosocial orientation, and cognitiveand affective reactions that are conducive to aggression.

MORAL DISENGAGEMENT 367

MoralDisengagement

AggressionProneness

Guilt andRestitution

Figure 2. Proposed causal structure of the paths of influence through which moral disengagement affectsdetrimental conduct.

Method

Participants

The participants in this study were 124 children in the last year ofelementary school and 675 junior high school students in Grades 6-8.They ranged in age from 10 to 15 years with a mean age of 11.8 years.There were 438 males and 361 females.

The students were drawn from four public schools: two elementaryand two junior high schools in a residential community located nearRome, Italy. This community adheres to a stringent consent procedurefor the conduct of research in the schools. A research proposal mustgain approval from a school council composed of parent and teacherrepresentatives and student representatives, as well, at the junior highschool level. In addition, parents must give consent, and children arefree to decline to take part if they choose. The parents not only con-sented to the study, but the mothers participated in the project them-selves. All of the children enrolled in these grades participated in thestudy unless they happened to be absent from school when the measureswere administered. The study was described to the parents and childrenas a project conducted through the University of Rome to gain betterunderstanding of how children develop.

This community represents a socioeconomic microcosm of the largersociety, containing families of skilled workers, farmers, professionals,and local merchants and their service staffs. Socioeconomic status of thefamily was assessed by father's occupation. Fourteen percent were inprofessional or managerial ranks, 25% were merchants or operators ofother businesses, 31 % were skilled workers, 29% were unskilled workers,and 1% were retired. The socioeconomic heterogeneity of the sampleadds to the generalizability of the findings.

Children were administered the sets of scales measuring the variablesof theoretical interest in their classrooms by two female experimenters.The various measures were administered over a period of several days.To add to the significance of obtained relationships, data for the vari-ables of interest were collected by different methods from parents,teachers, and peers, as well as from the children themselves. The scaleswere administered individually to the teachers and parents.

Moral Disengagement

Extensive prior psychometric analyses were conducted in the devel-opment of the measure of moral disengagement. A large pool of itemstapping the different disengagement mechanisms was constructed on

the basis of the guiding conceptual scheme. They were then pretestedon 251 elementary school children, 249 junior high students, and 315high school students. Items were rewritten to remove ambiguities andeliminated if they were not internally consistent within each of themechanisms. This pilot project revealed that proneness to moral disen-gagement was positively related to aggressive behavior and negativelyrelated to prosocial behavior across the three age groups regardless ofwhether the children's behavior was measured by self-ratings, teacherratings, or sociometric peer ratings.

The final form of the multifaceted scale used in the present studyassessed proneness to moral disengagement of different forms of detri-mental conduct in diverse contexts and interpersonal relationships(Bandura, 1995). Each of the eight mechanisms of moral disengage-ment was represented by a subset of 4 items. The full set of 32 items ispresented in the Appendix. The items tapped children's readiness toresort to moral justification, euphemistic labeling, advantageous com-parison, displacement and diffusion of responsibility, distortion of con-sequences, dehumanization, and attribution of blame for differentforms of transgressive conduct. The transgressive activities involvedphysically injurious and destructive conduct, verbal abuse, deceptions,and thefts. The social contexts encompassed educational, familial, com-munity, and peer relations. For each of the items, children rated on a 3-point Likert-type scale their degree of acceptance of moral exonerationsfor such conduct on an agree-disagree continuum. A principal-compo-nents factor analysis with varimax orthogonal rotation revealed a singlefactor structure. It accounted for 16.2% of the variance. Because nosubfactors emerged, we summed the responses to the set of items toprovide the composite measure of moral disengagement. The alpha re-liability coefficient for this measure is .82.

Aggressive and Prosocial Behavior

We obtained data regarding children's prosocial, aggressive, andtransgressive behavior from different sources, using diverse methods ofassessment. The sources included the children themselves, their par-ents, teachers, and peers. The methods of measurement included per-sonality questionnaires and peer sociometric ratings. To avoid possibleresponse biases, several control items were included in each of thequestionnaires.

The children completed two scales developed by Caprara and his col-leagues to measure prosocial behavior and interpersonal aggression(Caprara, Barbaranelli, Pastorelli, & Perugini, 1994). A 3-point re-

368 BANDURA, BARBARANELLI, CAPRARA, AND PASTORELLI

sponse format was used throughout. The measure of physical and verbalaggression, containing 15 items, assessed the frequency with which chil-dren fought with others, hurt them, and verbally disparaged them. Pro-social behavior was assessed by 10 items in terms of helpfulness, shar-ing, kindness, and cooperativeness.

Teachers rated the children in their classes for physical and verbalaggression and prosocial behavior using the scales administered to thechildren but shortened to six items each and cast in a third-person for-mat. The mothers also rated the frequency with which their childrenexhibited prosocial and aggressive forms of behavior, using the same setof scales as administered to the children. We computed internal consis-tency reliabilities using Cronbaeh's alpha. For the 16 sets of scores, ex-cept for an alpha of .61 for peer ratings of prosocial behavior, the reli-ability coefficients for the four sources of data (self, parents, teachers,peers) across the two educational levels for aggressiveness and proso-cialness were virtually all in the .80s and .90s. The concurrent validityof these measures has been corroborated in studies relating children'sratings of their behavior to level of prosocialness and aggressiveness asrated by parents and teachers and by peers' sociometric nominations(Caprara&Pastorelli, 1993).

Sociometric peer nominations served as another source of assessmentof prosocial and aggressive behavior. Children were presented with abooklet containing the names of children in their class along with 10items, 3 of which measured aggressive and prosocial behavior and 4of which measured peer popularity and rejection. Specifically, in theaggression domain peers circled the names of three classmates who fighta lot, insult other children, and often hurt them. For prosocial behavior,the peers circled the names of three classmates who help others, sharethings, and try to make sad people happier.

We measured peer popularity by having peers select the three class-mates with whom they would like to play and to study. To distinguishbetween children who were disliked by their peers and those who weresimply ignored or regarded indifferently, peers selected three classmateswith whom they would neither want to play nor study as a measure ofpeer rejection. The two aspects of peer popularity were positively corre-lated (r = .68, p < .001), as were the two aspects of peer rejection (r =.83, p < .001). Popularity and rejection were negatively related (r =- .42, p < .001). The multidimensional assessment of both positive andnegative status regarding both social and academic activities provided agood basis for gauging the impact of moral disengagement on quality ofpeer relations.

Affective and Cognitive Aspects of Aggression

Students at the junior high school level were administered two addi-tional scales that measured the affective and cognitive aspects of aggres-sive and transgressive conduct for matters especially relevant for olderchildren. The hostile rumination measure assessed with 15 items thelevel of preoccupation with personal grievances and retaliatory action.The irascibility measure, comprising 14 items, tapped petulance in so-cial transactions and weak restraint over anger even to slight provoca-tions. The alpha reliability coefficients were .86 for hostile ruminationand .84 for irascibility. The predictive validity of these measures hasbeen corroborated experimentally under simulated conditions in whichparticipants can inflict shocks of varying intensity on a provocateur.Individuals who have a low threshold for anger arousal and are prone tohostile rumination behave more punitively than those who are slower toanger and disinclined to dwell on grievances and possible retaliations(Caprara, Coluzzi, Mazzotti, Renzi, & Zelli, 1985; Caprara, Renzi, Al-cini, DTmperio, & Travaglia, 1983; Caprara, Renzi, Amolini, DTmp-erio, & Travaglia, 1984; Caprara et al., 1986).

Self-Sanctions for Transgressive Behavior

The scale measuring guilt and restitution, which included 15 items,dealt with the self-regulation of transgressive conduct by anticipatory

self-sanctions. It assessed the degree of guilt, remorsefulness, and self-criticism anticipated for transgressive conduct and the need to makerestitution if it were carried out. Factor analysis of the items revealed asingle factor. The alpha reliability coefficient for this scale was .79.

Delinquent Behavior

Delinquent behavior was measured by the relevant items from theChild Behavior Checklist developed by Achenbach and Edelbrock(1978), Both the reliability and predictive validity of this measure ofproblem behavior are well established (Achenbach, McConaughy, &Howell, 1987). The Delinquency subscale, comprising 22 items formales and 19 items for females, covers a wide range of transgressivebehaviors, including theft, cheating, lying, destructiveness, truancy, anduse of alcohol and drugs. The Parental scale includes 12 items, whichoverlap with the children's version. Both the mothers and the childrenthemselves recorded whether they engage in such antisocial activitiesand, if they do, whether they do so only occasionally or often. The reli-ability coefficients were .77 for parents, .77 for females, and .85 formales.

Results

Children's proneness to moral disengagement was unrelatedto familial socioeconomic status, and it did not differ as a func-tion of age. However, males exhibited higher moral disengage-ment than did females, F = 22.17, p < .0001. The major sourcesof this difference were the males' greater readiness to providemoral justifications for detrimental conduct, F = 45.81, p <.0001; to mask it in euphemistic language, F = 33.81, p < .0001;to minimize its injurious effects, F = 6.14, p < .025; and todehumanize victims, F = 26.60, p < .0001, and attribute blameto them, F = 9.92, p < .002. The degrees of freedom for theseanalyses are df= 1, 789.

Although the analyses indicate that the various mechanismsof moral disengagement operate in concert in the self-regulatoryprocess, they varied somewhat in degree of enlistment. Con-struing injurious behavior as serving righteous purposes, dis-owning responsibility for harmful effects, and devaluing thosewho are maltreated were the most widely used modes of exon-erative disengagement of self-sanctions. Masquerading censur-able activities in palliative language or rendering them benignby advantageous comparison, both of which require dexterouscognitive skills, were used less often.

Pattern of Relationships

Table 1 presents the relationships between moral disengage-ment and prosocial and detrimental conduct. The correlationsare highly consistent across different sources of data, and thecorrelates did not differ significantly on any of the measuresacross the two educational levels. Compared to individuals whomaintain a high level of moral agency, those who are highlyprone to moral disengagement tend to be more irascible, rumi-nate about perceived grievances, and are neither much troubledby guilt nor feel the need to make amends for harmful conduct.They also engage in a higher level of interpersonal aggressionand delinquent behavior.

Moral disengagement is related to prosocial behavior as wellas to transgressive activities. High moral disengagers are lessprosocially oriented and more likely to be rejected by peers.

MORAL DISENGAGEMENT 369

Table 1Relationship of Moral Disengagement to Prosocial,Aggressive, and Delinquent Behavior

Social and transgressivebehavior

Moral disengagement

Elementary Junior high

Self-ratingsPhysical and verbal aggressionProsocial behaviorIrascibilityHostile ruminationGuilt and restitutionDelinquency

Teacher ratingsPhysical and verbal aggressionProsocial behavior

Peer ratingsPhysical and verbal aggressionProsocial behaviorPopularityRejection

Parent ratingsPhysical and verbal aggressionProsocial behaviorDelinquency

.52***-.40***

————

.22**-.30***

.46***-.16*-.04

.23**

.38***-.17*

.20*

.36***-.28***.36***.38***

-.18***.45***

.06***-.18***

.29***-.17***-.07*.22***

.29***-.11***.27***

Note. Dashes indicate data that were not collected for this sample ofchildren.*/)<.05. **p<.01. ***p<.001.

However, there is no consistent relationship between moral dis-engagement and peer popularity. One can be rejected by proso-cial peers and gravitate toward, and gain acceptance from, dis-social or deviant peers (Cairns, Cairns, Neckerman, Gest, &Gariepy, 1988;Dishion, 1990).

The correlations among the other set of variables were similarin direction and magnitude across the two educational levels.Except for one correlation, which will be noted in the text assignificant beyond the .01 level, all of the relationships amongthe different variables are significant beyond the .001 level. Par-ticipants who are prosocially oriented anticipate guilt reactionsfor detrimental acts (r = .32), exhibit low aggression pronenessas reflected in irascibility (r = —.13) and ruminate hostility (r= —.19), and are disinclined to behave aggressively (r = —.27)or transgressively (r = —.32).

Participants who experience guilt over detrimental conductcannot stop thinking about their troublesome experiences andperceived grievances (r = .09, p < .01) and refrain from aggres-sive behavior (r = —. 15) or engagement in delinquent activities(r = -.25). Those who ruminate hostilely behave more aggres-sively (r = .32) and transgressively (r = .29). Irascibility is sim-ilarly related to aggressive behavior (r = .42) and pursuit of de-linquent activities (r = .30).

Although the various mechanisms of moral disengagementoperate interrelatedly as a single factor, there is some indicationthat they differ in their relative contribution to detrimental be-havior. The various disengagement mechanisms were groupedinto the subsets depicted in Figure 1 depending on whether theyoperated on reconstruing the conduct, obscuring personal re-sponsibility, misrepresenting injurious consequences, and vili-fying the victims. Responses to items were summed within sub-

sets and correlated separately with aggressive and delinquentbehavior. For delinquent behavior, the correlates were r = .42for moral reconstrual of harmful behavior, r = . 19 for obscuringresponsibility, r = .21 for misrepresenting harmful conse-quences, and r = .39 for vilifying the victims by blaming anddehumanizing them. For aggressive behavior, the correlateswere r = .34 for moral reconstrual, r = .16 for obscuring re-sponsibility, r = .15 for misrepresenting consequences, and r= .33 for vilifying the victims. Each of these correlations wassignificant beyond the .0001 level. Thus, across both classes ofdetrimental activities, the reconstrual of injurious behavior asserving worthy purposes and vilifying the victim exerted thegreater disinhibitory impact. It is easy to hurt others when suchconduct is viewed as doing worthy things with unworthy people.

Paths of Influence

The posited causal structure was tested with data from thejunior high students because they provide the full set of theoret-ically relevant variables. The factors in the hypothesized modelincluded moral disengagement, prosocialness, anticipatoryguilt and restitutive reactions for harmful conduct, and a latentfactor labeled Aggression Proneness. This factor was composedof the measured variables of hostile rumination and irascibility,which heighten propensity to detrimental action. The high load-ings of .71 and .72, respectively, for the model concerning delin-quency, and .67 and .76 for the model concerning aggression,show it to be a well-defined latent construct. The outcome vari-ables in the structural model were physical and verbal aggres-sion and engagement in delinquent activities. Delinquent acts,involving as they do more serious offenses than expression ofverbal or physical aggression, require a greater exercise of moraldisengagement. Hence, we conducted separate analyses on im-pact of moral disengagement on these two classes of behavior.Socioeconomic level was not included as a control variable be-cause it was related neither to the predictive factors nor to theoutcome variables. We tested the conceptual model on the co-variance matrices using the EQS program (Bentler, 1989).

The results of the structural equation modeling for delin-quent behavior are presented in Figure 3. The goodness of fit ofthe model to the data was corroborated by all of the fit indicesconsidered. The tests yielded a nonsignificant x2(3, N = 659)of 3.29, p = .35, a Normed Fit Index (NFI) of .996, a Non-Normed Fit Index (NNFI) of .998, and a Comparative Fit In-dex (CFI) of 1.00. rtests revealed each of the constituent pathsto be an essential part of the model.

In accord with the posited model, moral disengagement in-fluenced delinquent behavior both directly and by reducingprosocialness and anticipatory guilt over transgressions and byfostering aggression proneness. Prosocialness and anticipatoryguilt reactions influenced delinquency through their restrainingeffect on such conduct and by their impact on aggression prone-ness. Prosocialness increased feelings of guilt and counteractedpropensity to aggression, whereas guilt roused ruminationsabout perceived grievances and irascibility. Heightened aggres-sion proneness, in turn, increased the level of delinquent behav-ior. The full set of sociocognitive factors accounted for 31 % ofthe variance in delinquent behavior. The direct (.20) and medi-

370 BANDURA, BARBARANELLI, CAPRARA, AND PASTORELLI

.22

ProsocialBehavior

DelinquentBehavior

MoralDisengagement

AggressionProneness

Rumination IrascibilityGuilt and

Restitution -.22

Figure 3. Contribution of moral disengagement to the multivariate determination of delinquent behavior.All paths of influence are significant at/? < .05 or less.

ated (.23) effects of moral disengagement combined for a hightotal impact (.45) on delinquent behavior.

All but one of the posited paths of influence on aggressivebehavior were significant. No direct link was found betweenmoral disengagement and aggressive behavior. The various in-dexes of goodness of fit showed a good fit of the model to thedata—an NFI of .982, an NNFI of .928, and a CFI of .986—but a significant x2( 3,N= 666)= 13.37,/? < .01 was obtained.However, the dependence of this statistic on sample size makesit a less sensitive test with a large sample. Figure 4 presents theresults of the structural equation modeling for aggressive behav-ior with the one nonsignificant path removed.

The influence of moral disengagement on aggressive behaviorwas mediated through prosocialness, guilt, and aggressionproneness. High moral disengagement reduced prosocialnessand guilt reactions and promoted cognitive and affective reac-tions that are conducive to aggression. The three mediating fac-tors operated in the same way as they did in delinquent behavior,with prosocialness and anticipatory moral self-sanctions curb-

ing aggression and aggression proneness heightening it. The setof factors in the model accounted for 34% of the variance inaggressive behavior.

Several alternative causal models also were tested. One suchmodel assumes that weak prosocialness, low guilt, moral disen-gagement, and detrimental behavior are all simply coeffects ofaggression proneness. This causal structure provides a verypoor fit to the empirical data. The tests for goodness of fit fordelinquent behavior yielded a highly significant x2 (9, iV = 659)= 215.54, p < .001, and an NFI of .71, an NNFI of .52, and aCFI of .71. The results of the corresponding tests for aggressivebehavior are x2(9, N = 666) = 130.35, p < .001, NFI = .74,NNFI = .58, and CFI = .75.

Another alternative model posits that aggression pronenessaffects detrimental behavior both directly and through the me-diation of prosocialness, guilt reactions, and moral disengage-ment. This causal structure also provides a poor fit to the em-pirical data. The fitness tests for delinquent behavior yielded ahighly significant x2(6, A' = 659) = 152.49, p < .001, and an

-.08ProsocialBehavior

MoralDisengagement

AggressiveBehavior

AggressionProneness

'f/ .67/ \.76

Rumination IrascibilityGuilt and

Restitution -.21

Figure 4. Contribution of moral disengagement to the multivariate determination of aggressive behavior.All paths of influence are significant a tp < .05 or less.

MORAL DISENGAGEMENT 371

NFI of .79, an NNFI of .49, and a CFI of .80. The results of thetests for aggressive behavior are x2(6, N = 666) = 163.99, p <.001, NFI = .78, NNFI = .45, and CFI = .78.

Discussion

The findings of the present study lend considerable supportto the influential role played by mechanisms of moral disen-gagement both in detrimental and prosocial conduct. In accordwith prediction, high moral disengagers are more readily an-gered and behave more injuriously than those who apply moralself-sanctions to detrimental conduct. Ready moral disengagersare also more prone to engage in thought patterns that are con-ducive to aggression. They ruminate about perceived grievancesand dwell on punitive retaliations. By contrast, when self-re-proof remains engaged to detrimental conduct, adolescents arequicker to forgive and forget. Cognitive self-arousal throughvengeful rumination perpetuates a high level of anger long aftersocial slights or other provocations have ceased (Bandura,1973). Anger arousal primes one for vindictive action(Berkowitz, 1990; Zillman, 1983). Both the findings of thepresent naturalistic study and laboratory tests of heightened pu-nitiveness over a prolonged time span (Caprara et al., 1985;Zelli, 1984) corroborate the link between hostile ruminationand interpersonal aggression.

Whereas moral disengagement weakens self-restraints overinjurious conduct, adherence to moral self-sanctions fostersprosocial relations. It is difficult to hurt others who are human-ized and not blamed entirely for their life predicaments. Adher-ence to self-sanctions against injurious conduct is strengthenednot only by a sense of empathy but also by assuming personalresponsibility for one's actions and not minimizing their injuri-ous effects. The obtained relationship between low moral disen-gagement and prosocial behavior is consistent with evidencefrom controlled experiments in which personal responsibilityand humanization are systematically varied (Bandura et al.,1975). People refuse to behave cruelly, even under high instiga-tion to do so, if they act under personalized responsibility andrecipients are humanized. Prosocialness increases anticipatoryself-reproof for injurious conduct and attempts at restitutionshould self-restraint temporarily fail.

Psychological theorizing and research tend to emphasize howeasy it is to bring out the worst in people through dehumaniza-tion and other self-exonerative means. Thus, for example, theaspect of Milgram's (1974) research on obedient aggressionthat is widely cited is the evidence that good people can be co-erced into performing cruel deeds. However, to get people tocarry out punitive acts, the overseer had to be physically pres-ent, repeatedly ordering them to act injuriously and absolvingthem of any responsibility for the effects of their actions as theyvoiced their mounting concerns and objections. Orders to esca-late punitiveness to more intense levels are largely ignored orsubverted when remotely issued by verbal command. As Helmand Morelli (1979) noted, this is hardly an example of blindobedience triggered by an authoritative mandate. Moreover,what is rarely noted is the equally striking evidence that mostpeople steadfastly refuse to behave punitively, even in responseto incessant authoritarian commands, if the situation is person-alized by having them see their victim or requiring them to in-

flict pain directly rather than remotely (Milgram, 1974). Theconcern with suspension of self-restraints over detrimental be-havior is understandable considering the prevalence of people'sinhumanities to one another. However, the power of humaniza-tion to counteract human cruelty is of considerable theoreticaland social significance, but it continues to receive compara-tively little attention. Increased research efforts are needed toclarify how the affirmation of common humanity can bring outthe best in others.

If moral standards are disengaged from transgressive conductit can be carried out free from restraints of anticipatory self-censure. Through cognitive reconstruals and disownment of asense of personal agency, negative self-sanctions are unlikely tobe activated. There is little reason to engage in self-reproof forbehavior that has been rendered acceptable or for which oneprofesses no responsibility. Indeed, the findings confirm that thebetter the moral disengagement the weaker the felt guilt and theless the need to undo any harm caused by detrimental behavior.

So far the discussion has centered on aggressive and destruc-tive conduct. Self-regulatory mechanisms preside over trans-gressive activities as well as interpersonal aggression. Highmoral disengagers were much more inclined to engage in delin-quent pursuits than those who adhered to self-regulatory stan-dards under conditions that lend themselves easily to self-exon-eration. High moral disengagers were also less prosocial, lesstroubled by anticipatory feelings of guilt, and more prone toresort to vengeful ruminations and irascible reactions. Thesefactors were, in turn, related to delinquent behavior. The use ofthese mechanisms is by no means confined to habitual freelanc-ing delinquents. Gabor (1994) documented in considerable de-tail the widespread illegalities and inhumanities committed inall walks of life by ordinary citizens with self-exonerativejustifications.

Analysis of the pattern of influences on delinquent behaviorreveals that the proposed conceptual model provides an excel-lent fit to the empirical data. The model not only fits the datawell, but it also yields a better fit than alternative causal struc-tures. Moral disengagement affects delinquent behavior both di-rectly and indirectly through its influence on prosocial behavior,level of guilt, and aggression proneness. Moral disengagementoperates through essentially the same paths of influence on ag-gressiveness except that it involves no direct link. High moraldisengagers are less prosocial and less guilty over detrimentalconduct—both of which, in turn, lessen restraints over aggres-sive actions. Aggression proneness, of course, bears a close re-semblance to aggressive action, which gives it a larger media-tional role. This most likely accounts for a major share of thedifference in its strength of influence across the two classes ofdetrimental behavior.

A number of features of the present research add to the reli-ability and generalizability of the obtained relationships be-tween moral disengagement and detrimental conduct. It is rep-licated across diverse sources of data, different methods of mea-surement, distinct forms of detrimental conduct, and variantsociodemographic characteristics of the participants. Moreover,the predictiveness of proneness to moral disengagement is notconfounded by socioeconomic factors. The causal analysesshould be interpreted with some caution, however, because ofthe cross-sectional design of the research. However, the findings

372 BANDURA, BARBARANELLI, CAPRARA, AND PASTORELLI

of experimental investigations involving the key variables attestto the direction of causality posited in the guiding causal struc-ture. On the moral disengagement side, controlled variations indisplacement and diffusion of responsibility, dehumanization,and euphemistic labeling lead people to behave more aggres-sively (Bandura, 1991; Diener, 1977; Milgram, 1974; Zimb-ardo, 1969). With regard to the mediational paths, hostile ru-mination and irascibility have been shown to heighten aggres-siveness (Capraraetal., 1985). Hence, there are experimentallyvalidated bases for the posited paths of influence. The presentstudy is part of a larger longitudinal project. The causal influ-ence of moral disengagement will be further tested longitudi-nally in a multiple-panel design.

A recently completed study by Elliott and Rhinehart (1995)of serious assaults and transgressions with American youths at-tests to the generalizability of the theory of moral disengage-ment. In accord with the findings from the present study, malesexhibit higher levels of moral disengagement than do females.The cross-cultural replication indicates that some of the genderdifferences in aggression may reside in differential proclivity todisengage moral self-sanctions from injurious conduct. Moraldisengagement also differs by age, with older youths being moreprone to adopt self-exonerative devices in regard to seriousoffenses. There were no significant differences in moral disen-gagement by race or by socioeconomic level. The latter findingalso concurs with that obtained in the present study. Proclivityto moral disengagement predicted both felony and misdemea-nor assaults and thefts. Ready moral disengagement retainedhigh predictiveness regardless of age, sex, race, religious affil-iation, and social class. This high predictive consistency atteststo the pervading role of the self-regulatory system in detrimen-tal behavior.

The differences between attribution theory and social cogni-tive theory regarding the mechanism of victim blaming war-rants some comment. In attribution theory (Weiner, 1986), as-criptions of responsibility affect behavior through the media-tion of emotional reactions. For example, blaming victims fortheir plight arouses anger toward them, whereas placing theblame on situational causes arouses pity (Zucker & Weiner,1993). This interpretation begs the question of the mechanismgoverning when and how emotion gets translated into action.Sometimes people act on their anger, but oftentimes they go togreat lengths to conceal how they feel and are not about to venttheir anger in action. The differential reactions to anger indicatethat emotion is linked to action through a self-regulatory mech-anism. Moreover, many transgressive situations are not emo-tionally arousing, but they lend themselves readily to self-exon-eration for transgressive acts. Consider, for example, the moraldisengagement item "If people are careless where they leavetheir things it is their own fault if they get stolen." In this situa-tion, transgressors do not have to arouse themselves to a stateof anger to pilfer desired goods. They exempt themselves fromrestraining self-sanction for transgressive conduct bycontending that, through their negligence, victims have onlythemselves to blame. This analysis indicates that emotionalarousal may be facilitory, but it is neither necessary nor suffi-cient for the perpetration of transgressive conduct.

In the social cognitive theory of moral agency (Bandura,1991), moral conduct is motivated and regulated mainly by the

ongoing exercise of self-reactive influence. The major self-regu-latory mechanism, which is mobilized in concert with situa-tional factors, operates through three main subfunctions. Theseinclude self-monitoring of conduct, judgment of the culpabilityof conduct in relation to personal standards and environmentalcircumstances, and affective self-reaction. In this conceptualscheme, ascription of blame to victims functions as a disengagerof moral self-sanctions. Indeed, the findings of the present re-search reveal that the disengagement mechanisms operate inconcert on detrimental behavior both directly and by theireffects on anticipatory guilt, prosocialness, and emotion-arous-ing ideation.

Theories of aggression typically characterize influences thatreduce restraints over aggression as disinhibitory. Disinhibitiondescribes a process but does not specify the mechanisms gov-erning the behavioral effects. In social cognitive theory(Bandura, 1986), transgressive conduct is regulated by two ma-jor sources of sanctions: social sanctions and internalized self-sanctions. Both control mechanisms operate anticipatorily. Incontrol arising from social sanctions, people refrain from be-having transgressively because they anticipate that such con-duct will bring them social censure or other forms of punishingoutcomes. In control rooted in self-sanctions, people behaveprosocially because it brings self-satisfaction and self-respect,and they refrain from detrimental behavior because it will giverise to self-censure. What is called disinhibition largely reflectsthe disengagement of controlling self-sanctions from detrimen-tal conduct. Understanding of the nature of disinhibition can,therefore, be advanced by increasing knowledge of self-regula-tive disengagement.

The discussion thus far has centered on the role of disengage-ment mechanisms in the regulation of one's own injurious con-duct. These mechanisms also affect how the inhumanities per-petrated by others are viewed. For example, displacement ofresponsibility not only weakens restraints over one's own detri-mental actions but also diminishes concern over the sufferingof those mistreated by others (Tilker, 1970). Collective moraldisengagement can have widespread societal and political ram-ifications by supporting, justifying, and legitimizing inhumanesocial practices and policies.

Psychological theorizing and research on aggression has fo-cused heavily on impulsive aggression. The massive threats tohuman welfare stem mainly from deliberate acts of principlerather than from unrestrained acts of impulse. It is the morallyjustified and principled resort to destructiveness that is of great-est social concern but is largely ignored in psychological analy-ses of inhumanities. Over the years, much reprehensible anddestructive conduct has been perpetrated by ordinary, other-wise considerate people in the name of religious principles,righteous ideologies, nationalistic imperatives, and ruthless so-cial policies (Bandura, 1986; Rapoport& Alexander, 1982; San-ford & Comstock, 1971). There is much to be gained from un-derstanding how the facility for moral disengagement developsand how institutional justificatory strategies are used to enlistpeople for exploitive and destructive purposes.

ReferencesAchenbach, T. M., & Edelbrock, C. S. (1978). The classification of child

psychopathology: A review and analysis of empirical efforts. Psycho-logical Bulletin, S5. 1275-1301.

MORAL DISENGAGEMENT 373

Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications forcross-informant correlations for situational specificity. PsychologicalBulletin, 101, 213-232.

Andrus, B. C. (1969). The infamous of Nuremberg. London: Fravin.Bandura, A. (1973). Aggression: A social learning analysis. Englewood

Cliffs, NJ: Prentice-Hall.Bandura, A. (1986). Social foundations of thought and action: A social

cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.Bandura, A. (1990). Mechanisms of moral disengagement. In W. Reich

(Ed.), Origins of terrorism: Psychologies, ideologies, theologies, andstates of mind (pp. 161-191). Cambridge, England: Cambridge Uni-versity Press.

Bandura, A. (1991). Social cognitive theory of moral thought and ac-tion. In W. M. Kurtines & J. L. Gewirtz (Eds.), Handbook of moralbehavior and development: Theory, research and applications (Vol. 1,pp. 71-129). Hillsdale, NJ: Erlbaum.

Bandura, A. (1992). Social cognitive theory of social referencing. In S.Feinman (Ed.), Social referencing and the social construction of real-ity in infancy (pp. 175-208). New York: Plenum.

Bandura, A. (1995). Multifaceted scale of mechanisms of moral disen-gagement. (Available from A. Bandura, Stanford University, Stan-ford, California 94305-2130.)

Bandura, A., Underwood, B., & Fromson, M. E. (1975). Disinhibitionof aggression through diffusion of responsibility and dehumanizationof victims. Journal of Research in Personality, 9, 253-269.

Bentler, P. M. (1989). Theory and implementation of EQS: A structuralequations program. Los Angeles, CA: BMDP Statistical Software.

Berkowitz, L. (1990). On the formation and regulation of anger andaggression: A cognitive-neoassociationistic analysis. American Psy-chologist, 45, 494-503.

Bolinger, D. (1982). Language: The loaded weapon. London: Longman.Brock, T. C , & Buss, A. H. (1962). Dissonance, aggression, and evalu-

ation of pain. Journal of Abnormal and Social Psychology, 65, 197—202.

Brock, T. C , & Buss, A. H. (1964). Effects of justification for aggressionand communication with the victim on postaggression dissonance.Journal of Abnormal and Social Psychology, 68, 403-412.

Cairns, R. B., Cairns, B. D., Neckerman, H. J., Gest, S. D., & Gariepy,J. (1988). Social networks and aggressive behavior: Peer support orpeer rejection? Developmental Psychology, 24, 815-823.

Caprara, G. V., Barbaranelli, C , Pastorelli, C , & Perugini, M. (1994).Individual differences in the study of human aggression. AggressiveBehavior, 20, 291-303.

Caprara, G. V., Coluzzi, M., Mazzotti, E., Renzi, P., & Zelli, A. (1985).Effect of insult and dissipation-rumination on delayed aggressionand hostility. Archivio di Psicologia, Neurologia e Psichiatria, 46,130-139.

Caprara, G. V., & Pastorelli, C. (1993). Early emotional instability, pro-social behaviour, and aggression: Some methodological aspects. Eu-ropean Journal of Personality, 7, 19-36.

Caprara, G. V., Renzi, P., Alcini, P., D'Imperio, G., & Travaglia, G.(1983). Instigation to aggress and escalation of aggression examinedfrom a personological perspective: Role of irritability and of emo-tional susceptibility. Aggressive Behavior, 9, 345-353.

Caprara, G. V., Renzi, P., Amolini, P., D'Imperio, G., & Travaglia, G.(1984). The eliciting cue value of aggressive slides reconsidered in apersonological perspective: The weapons effect and irritability. Euro-pean Journal of Social Psychology, 14, iM-'ill.

Caprara, G. V., Renzi, P., D'Augello, D., D'Imperio, G., Rielli, G., &Travaglia, G. (1986). Interpolating physical exercise between instiga-tion to aggress and aggression: The role of irritability and emotionalsusceptibility. Aggressive Behavior, 12, 83-91.

Cohen, D., & Nisbett, R. E. (1994). Self-protection and the culture of

honor: Explaining southern violence. Personality and Social Psychol-ogy Bulletin, 20, 551-567'.

Crick, N. R., & Dodge, K. A. (1994). A review and reformulation ofsocial information-processing mechanisms in children's social ad-justment. Psychological Bulletin, 115, 74-101.

Darley, J. M., Klosson, E. C , & Zanna, M. P. (1978). Intentions andtheir contexts in the moral judgments of children and adults. ChildDevelopment, 49, 66-74.

Diener, E. (1977). Deindividuation: Causes and consequences. SocialBehavior and Personality, 5, 143-156.

Diener, E., Dineen, J., Endresen, K., Beaman, A. L., & Fraser, S. C.(1975). Effects of altered responsibility, cognitive set, and modelingon physical aggression and deindividuation. Journal of Personalityand Social Psychology, 31. 328-337.

Dishion, T. J. (1990). Peer context of child and adolescent troublesomebehavior. In P. Leone (Ed.), Understanding troubled and troublesomeyouth (pp. 128-153). Beverly Hills, CA: Sage.

Elliott, D. S., & Rhinehart, M. (1995). Moral disengagement, delin-quent peers and delinquent behavior. Unpublished manuscript, Insti-tute of Behavioral Sciences, University of Colorado.

Ferguson, T. J., & Rule, B. G. (1983). An attributional perspective onanger and aggression. In R. G. Geen & E. I. Donnerstein (Eds.), Ag-gression: Theoretical and empirical reviews (Vol. 1, pp. 41 -74) . NewYork: Academic Press.

Gabor, T. (1994). Everybody does it: Crime by the public. Toronto: Uni-versity of Toronto Press.

Haritos-Fatouros, M. (1988). The official torturer: A learning modelfor obedience to the authority of violence. Journal of Applied SocialPsychology, 18, 1107-1120.

Helm, C , & Morelli, M. (1979). Stanley Milgram and the obedienceexperiment: Authority, legitimacy, and human action. Political The-ory, 7,321 -346.

Keen, S. (1986). Faces of the enemy. San Francisco: Harper & Row.Kelman, H. C. (1973). Violence without moral restraint: Reflections

on the dehumanization of victims and victimizers. Journal of SocialIssues, 29,25-61.

Kelman, H. C , & Hamilton, V. L. (1989). Crimes of obedience: Towarda social psychology of authority and responsibility. New Haven, CT:Yale University Press.

Kohlberg, L. (1984). The psychology of moral development (Vol. 2) .New York: Harper & Row.

Kramer, M. (1990). The moral logic of Hizballah. In W. Reich (Ed.),Origins of terrorism: Psychologies, ideologies, theologies, slates ofmind (pp. 131-157). Cambridge, England: Cambridge UniversityPress.

Lutz, W. D. (1987). Language, appearance, and reality: Doublespeakin 1984. In P.C. Boardman(Ed.), The legacy of language—atributeto Charlton Laird (pp. 103-119). Reno: University of Nevada Press.

McHugo, G. J., Smith, C. A., & Lanzetta, J. T. (1982). The structureof self-reports of emotional responses to film segments. Motivationand Emotion, 6, 365-385.

Milgram, S. (1974). Obedience to authority: An experimental view. NewYork: Harper & Row.

Perry, D. G., Williard, J. C , & Perry, L. C. (1990). Peers' perceptions ofthe consequences that victimized children provide aggressors. ChildDevelopment.61, 1310-1325.

Rapoport, D. C , & Alexander, Y. (Eds.). (1982). The morality of ter-rorism: Religious and secular justification. Elmsford, NY: PergamonPress.

Reich, W. (Ed.). (1990). Origins of terrorism: Psycholgies, ideologies,theologies, states of mind. Cambridge, England: Cambridge Univer-sity Press.

Sanford, N., & Comstock, C. (1971). Sanctions for evil. San Francisco:Jossey-Bass.

374 BANDURA, BARBARANELLI, CAPRARA, AND PASTORELLI

Tilker, H. A. (1970). Socially responsible behavior as a function of ob-server responsibility and victim feedback. Journal of Personality andSocial Psychology, 14, 95-100.

Weiner, B. (1986). An altributional theory of motivation and emotion.New York: Springer-Verlag.

Zelli, A. (1984). Ruolo delle differenze individuali nella condotta ag-gressiva: Variabili di personalita, termperamentali, psicofisiologiche[ Role of individual differences in aggressive behavior: Personality,temperamental and psychophysiological variables]. Tesi di Laurea.Universita Degli Studi di Roma "La Sapienza," Roma.

Zillman, D. (1983). Arousal and aggression. In R. G. Geen & E. Don-

nerstein (Eds.), Aggression: Theoretical and empirical reviews (Vol.1, pp. 75-101). New York: Academic Press.

Zimbardo, P. G. (1969). The human choice: Individuation, reason, andorder versus deindividuation, impulse, and chaos. In W. J. Arnold &D. Levine (Eds.), Nebraska symposium on motivation (Vol. 17, pp.237-309). Lincoln: University of Nebraska Press.

Zimbardo, P. G. (1995). The psychology of evil: A situationist perspec-tive on recruiting good people to engage in anti-social acts. Researchin Social Psychology, 11, 125-133.

Zucker, G. S., & Weiner, B. (1993). Conservatism and perceptions ofpoverty: An attributional analysis. Journal of Applied Social Psychol-ogy, 23, 925-943.

Appendix

Mechanisms of Moral Disengagement

1. It is alright to fight to protect your friends.2. Slapping and shoving someone is just a way of joking.3. Damaging some property is no big deal when you consider that

others are beating people up.4. A kid in a gang should not be blamed for the trouble the gang

causes.5. If kids are living under bad conditions they cannot be blamed for

behaving aggresively.6. It is okay to tell small lies because they don't really do any harm.7. Some people deserve to be treated like animals.8. If kids fight and misbehave in school it is their teacher's fault.9. It is alright to beat someone who bad mouths your family.

10. To hit obnoxious classmates is just giving them "a lesson."11. Stealing some money is not too serious compared to those who steal

a lot of money.12. A kid who only suggests breaking rules should not be blamed if

other kids go ahead and do it.13. If kids are not disciplined they should not be blamed for

misbehaving.14. Children do not mind being teased because it shows interest in

them.15. It is okay to treat badly somebody who behaved like a "worm."16. If people are careless where they leave their things it is their own

fault if they get stolen.17. It is alright to fight when your group's honour is threatened.

18. Taking someone's bicycle without their permission is just "borrow-ing it."

19. It is okay to insult a classmate because beating him/her is worse.20. If a group decides together to do something harmful it is unfair to

blame any kid in the group for it.21. Kids cannot be blamed for using bad words when all their friends

doit.22. Teasing someone does not really hurt them.23. Someone who is obnoxious does not deserve to be treated like a

human being.24. Kids who get mistreated usually do things that deserve it.25. It is alright to lie to keep your friends out of trouble.26. It is not a bad thing to "get high" once in a while.27. Compared to the illegal things people do, taking some things from

a store without paying for them is not very serious.28. It is unfair to blame a child who had only a small part in the harm

caused by a group.29. Kids cannot be blamed for misbehaving if their friends pressured

them to do it.30. Insults among children do not hurt anyone.31. Some people have to be treated roughly because they lack feelings

that can be hurt.32. Children are not at fault for misbehaving if their parents force them

too much.

Note. The following items correspond to the various mechanisms of moral disengagement. Moral justification: 1,9, 17,25. Euphemistic language:2,10, 18,26. Advantageous comparison: 3,11, 19,27. Displacement ofresponsibility: 5, 13,21,29. Diffusion ofresponsibility: 4, 12, 20,28. Distortingconsequences: 6, 14,22,30. Attribution ofblame: 8, 16, 24, 32. Dehumanization: 7, 15,23,31.

Received August 28, 1995Revision received November 20, 1995

Accepted November 22, 1995 •