monkey see monkey do the influence of work groups on the antisocial behavior of employees (Robinson & O'leary Kelly, 1998).pdf

* Academy of Management fournal1998, Vol. 41, No. 6, 658-672.

MONKEY SEE, MONKEY DO: THE INFLUENCE OF WORKGROUPS ON THE ANTISOCIAL BEHAVIOR OF EMPLOYEES

SANDRA L. ROBINSONUniversity of British Columbia

ANNE M. O'LEARY-KELLYUniversity of Arkansas

This cross-level field study, involving 187 employees from 35 groups in 20 organiza-tions, examined how individuals' antisocial behaviors at work are shaped by theantisocial behavior of their coworkers. We found a positive relationship between thelevel of antisocial hehavior exhibited by an individual and that exhibited hy his or hercoworkers. We also found that a numher of factors moderated this relationship.Finally, we found that dissatisfaction with coworkers was higher when individualsengaged in less antisocial behavior than their coworkers.

The prevention of antisocial actions in organiza-tions is increasingly important to American man-agers and organizational scholars. Recent estimatessuggest there is good reason for both managers andresearchers to take a closer look at these actions.Some research reports that as many as 42 percent ofwomen have been victims of sexual harassment atwork (Gruber, 1990), that as many as 75 percent ofemployees have stolen from their employers(McGurn, 1988), and that 33 to 75 percent of allemployees have engaged in behaviors ranging frominsubordination to sabotage (Harper, 1990). Theseactions, of course, represent some of the most seri-ous forms of antisocial behavior. Yet we must alsonote the apparent prevalence of less serious, yetstill harmful, actions, such as lying (DePaulo &DePaulo, 1989), spreading rumors (Skarlicki &Folger, 1997), withholding effort (Kidwell & Ben-nett, 1993), and absenteeism (Johns, 1997), thatmay violate work norms and therefore may be an-tisocial (Robinson & Bennett, 1995).

One of the complexities of initial research in thisarea has been the use of diverse labels to describethese actions. For example, Robinson and Bennett(1995) used the term "deviant behavior," O'Leary-Kelly, Griffin, and Glew (1996) described "aggres-sive work behavior," and Vardi and Wiener (1996)discussed "organizational misbehavior." In this ar-ticle, we use the broad term "antisocial behavior"to describe negative behaviors in organizations. We

Tbe data collection for tbis study was supported by anAmerican Psycbological Association Science DirectorateDissertation Researcb Award. We would like to tbankKatberine Karl and Dean Stilwell for tbeir belpful com-ments on earlier versions.

chose this expansive term because, like the well-established and related term "prosocial behavior"(Brief & Motowidlo, 1986; George, 1990), "antiso-cial behavior" captures a wide range of actions. Inaddition, the term captures the harmful nature ofthese acts, the fact that they have the potential tocause harm to individuals and/or the property of anorganization (Giacalone & Greenberg, 1996). Thisdimension of potential harmfulness is the criticalfocus of most definitions of related constructs (for areview, see Robinson and Greenberg [1998]).

To date, forms of antisocial behavior in organiza-tions have been examined from various theoreticalperspectives. For example, using social learningtheory (Bandura, 1977), O'Leary-Kelly and col-leagues (1996) identified a number of individualand environmental antecedents. Martinko and Zel-lars (1996) expanded this framework, incorporatingattribution theory into the social learning explana-tion. Other examples include Greenberg's (1990,1993) work on theft and Skarlicki and Folger's (1997)research on retaliation, both of which used principlesof justice theory to explain antisocial employee ac-tions. These approaches, well groimded in estab-lished theory, have resulted in important advances inthe understanding of why and when employees en-gage in antisocial behavior.

Although this previous research makes signifi-cant contributions, it also is limited because anti-social behaviors have predominately been exam-ined as individual-level phenomena. This focus isreasonable in that decisions to exhibit any behav-ior, whether antisocial or prosocial, are made byindividuals. However, additional insights might beacquired if these behaviors were examined withinthe social context of work groups. The purpose of

658

1998 Robinson and O'Leary-Kelly 659

this research was to address antisocial behavior asa group-related activity. Specifically, we examinedthe extent to which individuals' antisocial actionsare shaped by the group context within which theywork. It should be emphasized that we see thisfocus as a supplement, not an alternative, to indi-vidual-level explanations.

THE INFLUENCE OF GROUPS ON INDIVIDUALANTISOCIAL ACTIONS

George (1990; George & James, 1993) used agroup level of analysis to examine prosocial behav-iors in organizations. Her research showed that thelevels of positive and negative affectivity within awork group influenced the affective tone of thegroup and the group's general level of prosocialbehaviors. Given the importance of work groups inpredicting prosocial actions, it seems appropriateto consider whether and how work groups affectantisocial actions.

A variety of theoretical perspectives support thenotion that individuals' work groups will influencethe likelihood of their behaving in antisocial ways.In this research, we invoked three: (1) the attrac-tion-selection-attrition perspective, (2) social infor-mation processing theory, and (3) social learningtheory.

The Attraction-Selection-Attrition Perspective

A basic assumption underlying the attraction-selection-attrition framework is that individualscarefully analyze their work environments and ad-just their individual actions accordingly (Schnei-der, 1975). Individuals with antisocial tendenciesare more likely to be attracted to, and selected into,the group environments that fit well with thosetendencies. In addition, most individuals willlikely adapt some of their behaviors, cognitions,and attitudes to better fit with the social environ-ment in which they work. Those that adapt well aremore likely to remain with the organization,whereas those who do not sufficiently adapt aremore likely to leave. Thus, employees within workgroups should tend to be relatively homogeneousin terms of their attitudes and behavior regardingantisocial behavior because they are generally sim-ilar individuals who are experiencing comparableconditions and are trying to adapt to their commonenvironment (Schneider, 1987; Schneider & Reich-ers, 1983). In other words, we would expect a pos-itive relationship between a given individual'slevel of antisocial behavior and the level of antiso-cial behavior of his or her coworkers.

Social Information Processing Theory

Social information processing theory would alsosupport the predictions that group-level antisocialbehavior will influence the antisocial behavior ofindividual members and that, over time, individualmembers will come to have more similar levels ofantisocial behavior. According to the social infor-mation processing approach, individuals use infor-mation from their immediate social environmentsto interpret events, develop appropriate attitudes,and understand expectations concerning their be-havior and its consequences (Salancik & Pfeffer,1978). The social context greatly determines howindividuals behave by influencing how they thinkand feel about aspects of their work environment(Salancik & Pfeffer, 1978). Applying this perspec-tive to antisocial behavior suggests that individualgroup members, working in a shared social envi-ronment, will receive similar social cues that con-vince them that certain types and levels of antiso-cial behavior are acceptable adaptations to theirshared working conditions.

Social Learning Theory

O'Leary-Kelly and colleagues (1996) used Bandu-ra's (1977) social learning perspective to examinefactors that encourage antisocial behavior. Onesuch factor was the presence of role models withina work context. They argued that if individualswork in environments that include others whoserve as models for antisocial behavior, these indi-viduals are more likely themselves to behave inantisocial ways. When individuals operate withingroup settings, they are typically able to observeother group members, which creates the opportu-nity for these members to serve as models. In addi-tion, Bandura's research on disengagement ofmoral control suggests that diffusion of responsibil-ity, a common outcome in group contexts, can leadindividuals to disconnect the self-regulatory sys-tems that typically govern moral conduct (Bandura,1990, 1991; Bandura, Barbaranelli, Caprara, & Pas-torelli, 1996).

Integrating social learning theory with the attrac-tion-selection-attrition perspective and the socialinformation processing approach, we considered itlikely that members of groups who are analyzingtheir social environments for information regardingthe appropriateness of particular beliefs, attitudes,and behaviors would often use other group mem-bers as role models. If other group members serveas role models, groups may have a significant in-fluence on individual antisocial behavior throughthis role-modeling process. Drawing on these three

660 Academy of Management Journal

theoretical foundations, we hypothesized the fol-lowing:

Hypothesis 1. There will be a positive relation-ship between the level of antisocial behaviorwithin a group and the level of antisocial be-havior of individual group members.

To this point, we have conceptualized social cli-mates as relatively fixed phenomena, as conditionsthat either exist or do not. A more realistic por-trayal, and one more consistent with the attraction-selection-attrition framework (Schneider, 1975,1987; Schneider & Reichers, 1983), would repre-sent such environments according to their degree ofstrength. A group's climate reflects the aggregateperceptions of group members regarding a particu-lar aspect of the work setting, perceptions that in-fluence the types of behaviors that are exhibitedwithin the group. When there is strong similarity inmembers' perceptions and behaviors, the socialcontext is most potent and thus most capable ofhaving a profound influence on member behavior.Therefore, we expected that the degree of similarityin group members' levels of antisocial behaviorwould moderate the extent to which a group's gen-eral level of antisocial behavior would influence anindividual group member's level.

Salancik and Pfeffer's (1978) discussion of socialinformation processing theory is consistent withthe above argument. Salancik and Pfeffer positedthat the effect of a particular social environment onindividual attitudes and behavior depends on thedegree to which there are shared beliefs within thesocial environment. Social learning theory alsowould be consistent with this moderating effect. Asargued above, the antisocial behavior of individualgroup members may be influenced by the role mod-els they encounter within a group. To the extentthat potential role models exhibit similar levels ofantisocial behavior, there is a stronger probabilitythat the individual member will choose a rolemodel that reflects the group's norms. For example,if most members of a work group behave in antiso-cial ways, the likelihood that a new group membermight choose a role model who exhibits antisocialbehavior is increased, and the chance that the new-comer will develop antisocial actions is alsogreater. Drawing on these arguments, we predictthe following moderated effect:

Hypothesis 2. The degree of similarity or ofvariance in antisocial behavior within a groupwill moderate the relationship between groupantisocial behavior and individual members'antisocial behavior in such a way that the

December

greater the similarity (the lower the variance),the stronger the relationship.

The impact of a group on individuals is alsolikely to be enhanced as the members' time in thegroup increases. Numerous theoretical frameworkssupport the existence of this moderating effect. Forexample, social impact theory (Latane, 1981) sug-gests that the extent of social influence that anyindividual has over others is a function of, amongother factors, the proximity in time and space be-tween the relevant parties, which has been labeled"immediacy." Attraction-selection-attrition theoryalso supports this argument. Compared to a new-comer, an individual who has been a member of agroup for some time is likely to have acquired moreaccurate perceptions of the group's attributes. Thisindividual is, therefore, in a better position to de-termine the degree of fit between personal andgroup attributes. Attraction-selection-attrition prin-ciples suggest that this individual will either adjustpersonal behavior to fit the work environment orleave the organization (Schneider, 1975; Schneider& Reichers, 1983). The longer a member remains,then, the more likely it is that this person haschosen to remain with the group and to behave inaccordance with the group's climate.

Social information processing theory and sociallearning theory are also consistent with this mod-erating effect. With the former, the longer an indi-vidual retains membership in the group, the greateris the group's ability to provide social informationthat shapes the individual's beliefs, attitudes, andbehaviors (Salancik & Pfeffer, 1978). According tosocial learning theory, individuals determine theutility of modeled actions by watching the model'sinteractions with the environment (Bandura, 1986).Central to the notion of modeling, then, is the as-sumption that individuals have the opportunity toobserve the model. Certainly, the longer an individ-ual's tenure in a work group, the greater his or heropportunity to observe role models and thus, thestronger the impact of antisocial behavior role mod-els.

Hypothesis 3. An individual member's tenurein a group will moderate the effect of thegroup's antisocial behavior on the individual'santisocial behavior in such a way that the re-lationship is stronger for members with longertenure in the group.

In both the attraction-selection-attrition and so-cial information processing theories, the social con-text is represented as the fundamental determinantof behavior (Salancik & Pfeffer, 1978; Schneider,1975). Although the social context is not as explicit


in social learning theory, this perspective is alsoconsistent with the notion that interactions be-tween people determine individual behavior. Thusfar, we have not addressed the question of howother factors, such as objective characteristics of awork group (for example, its structure and technol-ogy), might influence individual behavior. Accord-ing to Schneider (1987), these factors will influenceindividual behavior indirectly, through their abil-ity to enhance or limit interactions between people.

One objective organizational characteristic thathas a history of importance within group settings istask interdependence, or the degree to which em-ployees in a work group must coordinate their in-dividual efforts. In groups with high task interde-pendence, individual members are likely to havegreater opportunity to interact with others in thegroup. This enhanced interaction allows membersto more easily acquire the social information thatwill, according to both the attraction-selection-attrition and social information processing frame-works, determine their subsequent behavior. In ad-dition, in line with social learning theory (Bandura,1973, 1977), the higher level of interaction amonggroup members will increase the likelihood thatgroup members will be perceived as relevant com-parison others and therefore chosen as role models.Under conditions of high task interdependence,therefore, the influence of a group's antisocial be-havior on individual antisocial behavior should beintensified.

Hypothesis 4. A group's level of task interde-pendence will moderate the relationship be-tween group antisocial behavior and the anti-social behavior of individual members in sucha way that the higher the task interdepen-dence, the stronger the relationship.

Each of the three perspectives provides a concep-tual explanation of how groups might influence theantisocial behavior of individual members. Onlythe attraction-selection-attrition perspective, how-ever, deals directly with the issue of how individ-uals might become alienated from, and ultimatelyleave, a group. As mentioned previously, the attri-tion component of the attraction-selection-attritionframework suggests that individuals who do not fita work environment will wish to leave their orga-nizations (Schneider, 1975, 1987; Schneider &Reichers, 1983). This lack of fit implies that peopleperceive themselves to be significantly different onrelevant attributes from others in the work environ-ment, feel dissatisfled with the poor flt, and want towithdraw from the setting. One might expect,therefore, that individuals whose behaviors arevery different from those of others in their work

group would experience dissatisfaction with mem-bers of the group and would wish to leave it. Withregard to antisocial behavior, it seems quite likelythat there will be individuals who find their per-sonal attributes or behavioral tendencies at oddswith the attributes or behaviors of a work groupthat engages in antisocial behavior because suchbehaviors are, by definition, in violation of gener-ally held (that is, societal level) social mores.

Hypothesis 5. To the extent that individualgroup members exhibit lower levels of antiso-cial behavior than their group exhibits, theywill report lower satisfaction with group mem-bers.

Hypothesis 6. To the extent that individualgroup members exhibit lower levels of antiso-cial behavior than their group does, they wi,llreport greater intentions to leave.

The theories that provide the conceptual founda-tion for our previous hypotheses are useful for de-termining how the group environment, with its ex-plicit and implicit pressures and its formal andinformal social constraints and sanctions, mightaffect an individual's antisocial behavior at work.However, given the potentially negative repercus-sions of antisocial behavior, organizations (that is,the organizational environments that exist beyondimmediate work groups) typically have a clear in-terest in preventing such actions, even if they areexplicitly and/or implicitly encouraged by a workgroup. For example, a work group may ignore em-ployee theft by its members, but organizationalmanagers outside the group want to prevent theft.This contrast presents an interesting dilemma—ifthe group's climate encourages antisocial behav-iors, but these actions are discouraged in the largerenvironment, will individual group members be-have in antisocial ways?

Generally, we expected a group's ability to influ-ence a member's antisocial behavior to depend onthe member's perception that such behavior can beengaged in without negative consequences; there-fore, conditions that enhance the belief that pun-ishment will follow antisocial behavior shouldlessen the group's influence. Two techniques thatmanagers often use to deter antisocial behavior arethe threat of punishment and close supervision ofemployees. We expected the group's influence tobe limited by these managerial actions and pre-dicted the following:

Hypothesis 7. The likelihood of punishment bymanagement will moderate the relationshipbetween group antisocial behavior and indi-vidual antisocial behavior in such a way that

662 Academy of Management fournal December

the greater the likelihood of punishment, theweaker the relationship.

Hypothesis 8. Closeness of supervision willmoderate the relationship between group anti-social behavior and individual antisocial be-havior in such a way that the closer the super-vision, the weaker the relationship.

METHODS

Overview

Sample. Data were collected from 187 full-timeemployees. Average tenure within the firm was5.93 years, and average tenure in the current jobwas 3.87 years. Occupations and the percentages ofrespondents in them were as follows: productionworkers, 24 percent; business consultants, 19 per-cent; general managers, 18 percent; administrative/clerical personnel, 10 percent; real estate agents, 8percent; accountants, 7 percent; human resourceprofessionals, 4 percent; sales personnel, 2 percent;engineers, 2 percent; R & D staff members, 2 per-cent; paralegals, 2 percent; and other, 2 percent.

Respondents were from 35 work groups in 20different organizations. The groups ranged in sizefrom 4 to 10 employees, with the average size being5.34 employees. Industries and the numbers of or-ganizations in them were as follows: consumerproducts, 7; consulting and accounting, 6; financialservices, 2; real estate, 2; social services, 1; insur-ance, 1; and gambling, 1.

Close to equal numbers of men (52%) andwomen (48%) participated in the study. Age rangedfrom 21 to 65 years, with an average of 32. Fifty-onepercent were married. Educational level varied: 9.4percent had high school educations, 29 percent hadsome college experience, 29.2 percent had bache-lor's degrees, and 32.7 percent possessed graduatedegrees.

Procedures. Participating organizations were se-lected from a list of firms recruiting from a largemidwestern business school. Thirty-two percentagreed to participate. Organizationally definedboundaries were used to define work groups. Mostorganizations provided multiple and varied workgroups for the study. For example, a manufacturerof aircraft brake linings provided five work groupsin administration, engineering, and sales, and anoptical and eye care product firm provided workgroups in manufacturing, accounting, and admin-istration. We used this sampling procedure becauseit enabled us to obtain a sample that was morediverse in terms of industries, organizations, andoccupational groups. This diversity was valuable in

that it enabled us to more confidently generalizeour results to the larger employee population.

All employees within each work group wereasked to participate. They received surveys by mailand were asked to mail completed surveys back tous. Response rates within the work groups rangedfrom 37 to 100 percent, and the overall responserate was 67 percent.

Design. Oui design was cross-level and primarilyinvolved contextual independent and moderatingvariables and individual-level dependent variables.We measured the contextual variables by askingemployees to report on characteristics of theirgroups and aggregating those reports for eachgroup. Aggregating employee perceptions is a com-mon and valid means by which to assess contextualvariables (Rousseau, 1985). Aggregation of individ-ual perceptions is more useful than using individ-ual perceptions alone to measure contextual vari-ables because it reduces error by averaging outrandom individual-level errors and biases.

Where possible, we used a version of a splitsample design that allowed us to reduce many ofthe problems associated with common methodbias. Common method bias can pose problems forsurvey research that relies on self-report data, es-pecially if the data are provided by the same personat the same time (Campbell & Fiske, 1959). Oneimportant concern in such cases is that commonmethod bias may artificially inflate observed rela-tionships between variables. This problem may beavoided or reduced by using different respondentsfor reports of the independent and dependent vari-ables. Our split sample design allowed us to do justthat.

When analyzing the impact of contextual vari-ables on an individual-level variable such as anemployee's antisocial behavior, we excluded thefocal employee's report of the independent variablefrom the aggregation of the contextual variable. Wedeveloped a program in SPSS that (1) averaged theemployees' reports for each work context variableacross each group and (2) assigned, to each em-ployee within a particular group, the average scoreof each work context variable for that group exclud-ing the employee's own assessment of the workcontext variable. Hence, the respondents who pro-vided information on the independent variableswere different from the respondents who providedinformation on the dependent variable.

Dependent Variables

Individual antisocial behavior. The dependentvariable, individual antisocial behavior, was mea-sured with a scale composed of the following items:


"damaged property belonging to my employer,""said or did sometiiing to purposely hurt someoneat work," "did work badly, incorrectly or slowly onpurpose," "griped with coworkers," "deliberatelybent or broke a rule(s)," "criticized people at work,""did something that harmed my employer or boss,""started an argument with someone at work," and"said rude things about my supervisor or organiza-tion."

The behaviors included in this scale were de-rived from two sources. The primary source wasother scales found in the literature that measuresimilar behaviors (e.g., Robinson & Bennett, 1995;Skarlicki & Folger, 1997). A secondary source was asurvey of 50 employees on the university campuswho were asked to provide examples of antisocialbehavior at work that they had witnessed. Fromthese two sources, we derived a large pool of be-haviors. Five coders who were blind to the purposeof the study independently categorized the behav-iors. Inappropriate behaviors and redundant de-scriptions were removed. Initially, we selected tenbehaviors on the basis of their frequency, general-izability, and distinctiveness. Many of the behaviordescriptions were then modified to ensure that theywere clear, consistent, and not compound. Giventhe sensitive nature of these behaviors, we modi-fied many descriptions in order to reduce sociallydesirable response bias; for instance, innocuous-sounding terms were employed where possible.

Respondents used a one-to-five scale to indicatethe extent to which they had engaged in each be-havior in the year prior to data collection. We usedthe retrospective time frame of one year to addressthe expected low base rate of self-reported sociallyundesirable behaviors (Hulin & Rousseau, 1980). Ahigh score indicated a high level of antisocial be-havior.

Face validity assessment of individual antiso-cial behavior. We used a sample of 70 eveningmaster of business administration (M.B.A.) stu-dents to assess face validity. All worked full-time,and most had at least two years of work experience.Their average age was 28.12 years, and 29.4 percentwere women. They were provided with a list of theten behaviors and asked to indicate whether theyperceived each behavior as potentially harmful toan organization and/or its members.

With the exception of one item, the behaviorswere indicated to be potentially harmful by the vastmajority of respondents. The one behavior that wasnot perceived as potentially injurious was "usedproperty belonging to my employer for my ownbenefit," which only 47.1 percent of the sampleidentified as potentially harmful. Thus, this itemwas dropped from the scale. For the remaining nine

items, 71 to 100 percent of respondents said thebehavior described was potentially harmful to or-ganizations and/or their members. In sum, this va-lidity analysis ensured that the items captured inthe scale closely reflected our definition. The re-vised scale was composed of nine items.

Construct validity assessment of individual an-tisocial behavior. We used a sample of 133 full-time employees from the Midwest. Forty-nine per-cent were women, and their average age was 43.58years (s.d. = 10.94). The education level of therespondents varied; the average years of work ex-perience was 23.52 (s.d. = 10.57); and the modalannual income was between $15,000 and $30,000.

First, we wanted to examine convergent validityby demonstrating that our measure of antisocialbehavior was highly correlated with measures pur-ported to assess similar constructs. Thus, we exam-ined the relationship between our scale and twoother scales. Lehman and Simpson's (1992) mea-sure of antagonistic behaviors (r = .49, p < .01) andRusbult, Farrell, Rogers, and Mainous's (1988) mea-sure of neglect behaviors. We found that our mea-sure was positively related to the antagonistic be-haviors scale (r = .49, p < .01) and the neglectbehaviors scale (r = .46, p < .01).

A demonstration that a focal measure is moder-ately correlated with measures of theoretically re-lated constructs in a predicted fashion also pro-vides evidence of convergent validity. Thus, weexamined the relationship between our measure ofantisocial behavior and two theoretically relatedconstructs: intentions to leave a firm (Rusbult et al.,1988) and lack of job involvement. Those with highscores on these measures should also have highscores on our measure of antisocial behavior atwork. The correlations were .30 (p < .01) for inten-tions to leave and .20 (p < .05) for job involvement.As expected, these correlations between our anti-social behavior measure and measures of theoreti-cally related constructs were significant and in thepredicted direction but not as high as the correla-tions between our measure of antisocial behaviorand measures purporting to assess similar con-structs.

We also examined discriminant validity, assess-ing the relationship between our measure of anti-social behavior and constructs expected to beconceptual opposites of antisocial behavior: con-sciousness, a type of citizenship or prosocial be-havior (Podsakoff, MacKenzie, Moorman, & Fetter,1990), and loyalty behaviors (Rusbult et al., 1988).As expected, our antisocial behavior measure wasmoderately and negatively related to both the mea-sure of citizenship behavior (r = -.24, p < .05) aridthe measure of loyalty (r = - .46, p < .05).


In summary, the above results lent support to theconstruct validity of our antisocial behavior mea-sure. The results revealed that the measure wasrelated, in expected ways, to measures purportingto measure similar constructs, to measures of theo-retically related constructs, and to measures of con-ceptually opposing constructs.

Reliability assessment of individual antisocialbehavior. We collected additional data from thesample of 102 evening M.B.A. students. These re-spondents were asked to complete the above nine-item scale measuring individual antisocial behav-ior at two points in time two weeks apart. Thetest-retest reliability for this scale was .87.

The internal reliability (Cronbach's alpha) of thisnine-item scale was .68 for the primary respon-dents used in this study. However, the alphas forthe evening M.B.A. students were .75 at time 1 and.81 at time 2.

Other dependent variables. Satisfaction withcoworkers was measured with one facet of the JobDescriptive Index (JDI; Smith, Kendall, & Hulin,1969). Respondents used "yes," "no," or "uncer-tain" to indicate whether each of 18 different ad-jectives described different aspects of the peoplewith whom they worked. The higher the score, thehigher the satisfaction with coworkers. Reliabilityfor this scale was .84.

We measured employees' intentions to leavetheir organizations with the following question:"How long do you intend to remain with yourcompany?" Respondents were given an open re-sponse format and all responses were converted toyears. These were then reverse-coded so that a highvalue reflected a high level of intentions to leave.

Independent, Moderator, and Control Variables

Group antisocial behavior. The self-report scaleused to measure the dependent variable individualantisocial behavior was also used to measure groupantisocial behavior, a contextual variable. The dis-tinction between these two variables was that thedependent variable, individual antisocial behavior,was each employee's self-report of his or her anti-social behavior, whereas the independent variable,group antisocial behavior, was a value assigned toeach person that reflected an aggregation of all em-ployees' self-reports of antisocial behavior acrossthe group, excluding the employee whose antiso-cial behavior was being predicted.

Moderator variables. To measure individualtenure in group, we asked respondents to indicate,in years, how long they had worked in the samegroup.

To measure task interdependence, we used Van

de Ven, Delbecq, and Koenig's (1976) index. In thismeasure diagrams represent tasks as either inde-pendent, sequentially dependent, reciprocally de-pendent, or teamwork. Respondents were asked toindicate the percentages of their work that con-formed to each of these types of interdependence.The higher the rating, the greater the degree of taskinterdependence. Measuring internal reliabilitywas not appropriate for this scale because eachitem (a diagram) represents an increased level ofinterdependence. However, Van de Ven and col-leagues (1976), using two alternative measures ofthe construct, reported the correlation between themeasures to be .72.

Likelihood of punishment was measured by giv-ing respondents a list of four antisocial behaviorsand asking them to use a one-to-five scale to indi-cate the extent to which they would be rewarded orpunished by their superiors for engaging in eachbehavior. The behaviors were "engaging in destruc-tive activities such as stealing or damaging prop-erty belonging to our employer," "doing things thatcould hurt other people in the organization," "do-ing things that could hurt the department or theorganization we work for," and "doing work badly,slowly or incorrectly on purpose." A high scoreindicated a greater likelihood of punishment forengaging in these behaviors. Reliability was .73.

To measure close supervision, we used a modi-fied version of Dewar and Werbel's (1979) measureof surveillance and enforcement; this measure pri-marily reflects the degree to which employees aresupervised closely by their superiors. Respondentswere asked to indicate, using a one-to-five scale,the extent to which they agreed or disagreed withsix statements. Examples of those statements in-clude "We are constantly being watched to see thatwe obey all rules" and "Our superiors are aware ofeverything that we do on the job." The higher thescore, the closer the supervision. The reliability ofthis scale was .88.

Control variables. We controlled for severalvariables in most of our analyses. These were vari-ables perceived to be common predictors of antiso-cial behavior in the workplace. To the extent wewere able to control for alternative predictors ofantisocial behavior, we were able to demonstratethe unique influence of group antisocial behavior

• on individual antisocial behavior.First, we controlled for both close supervision

and likelihood of punishment, except in direct testsof these variables. In traditional explanatory mod-els of destructive behavior, such behavior has beenviewed as largely dependent upon the presence orabsence of formal constraints (e.g., Hirschi, 1969),such as monitoring employee behavior and apply-


ing sanctions (e.g.. Tittle, 1973). Indeed, a host ofresearch has suggested that the use of monitoringand sanctions can reduce employee deviance (e.g.,HoUinger & Clark, 1983), workplace theft (e.g.,Greenberg & Scott, 1996), and cheating (e.g.. Covey,Saladin, & Kilen, 1989).

We also controlled for level of job satisfaction,including the facets of job satisfaction found withinthe JDI: pay satisfaction, promotion satisfaction,work satisfaction, and supervisor satisfaction. Inanother common explanatory model of destructiveworkplace behavior, such behavior is viewed asemanating from job dissatisfaction, or perceiveddiscrepancies between what has been received andwhat is desired or expected (Ditton, 1977; Merriam,1977). Indeed, a common assumption is that dis-gruntled employees are the most likely to engage inharmful or vengeful acts (Skarlicki & Folger, 1997).A high score on this general satisfaction measureindicated high general satisfaction.

We also controlled for perceived control, sinceanother theoretical explanation for antisocial work-place behavior is that it is the result of a lack ofcontrol on the part of an employee. Indeed, someresearch has suggested that such behavior is anunconventional means by which employees seek togaih more control over their environment whenconventional means are unavailable (Bennett,1998; Storms & Spector, 1987). This variable wasassessed with Ashford, Lee, and Bobko's (1989)three-item measure, which uses a five-point re-sponse scale. A high score on this variable indi-cated high perceived control.

We also controlled for a number of demographicvariables, including gender, age, tenure in the or-ganization, and educational level, for several rea-sons. First, these variables provide distal proxiesfor an individual predisposition toward antisocialbehavior. Second, numerous studies have founddemographics to be significant predictors of antiso-cial behavior. For example, gender has been linkedto aggression (Baron & Richardson, 1994), absentee-ism (Johns, 1997), and drug use at work (Hollinger,1988); education has been linked to alcohol andsubstance abuse on the job (Mensch & Kandel,1988); and age has been associated with absentee-ism (Martocchio, 1994), substance abuse at work(Steffy & Laker, 1991), and organizational theft(Murphy, 1993).

Gender was measured with the question "Areyou male or female?" (male was coded 1 and fe-male, 0). Age was measured with the open-endedquestion "How old are you?" Educational level wasmeasured with "What is your educational level?"Respondents could check one of four categories asrepresenting the highest level they had completed:

high school (coded 1), bachelor's degree (2), somecollege or advanced training from a college (3), orgraduate degree (4).

RESULTS

Justification of Aggregation

We assessed the appropriateness of aggregatingour contextual variables to the group level usingtwo different procedures: between-group analysisof variance (ANOVA) and r^^ Qames, Demaree, &Wolf, 1984, 1993; George & James, 1993), a meaning-ful indicator of within-group agreement (Kozlowski &Hattrup, 1992). The results of the ANOVAs indicatedthat there were significant between-group differencesfor group antisocial behavior (F34 5̂2 = 1.96, p < .01),task interdependence (F34 4̂5 = 3.27, p < .01), closesupervision (F34 4̂4 = 1.39, p < .09), and likelihoodof punishment (̂ 34,151 = 2.82, p < .01).

The results of r̂ ,̂ analyses (mean r^^'s) using arectangular uniform null distribution yielded val-ues for all the variables that indicated high levels ofwithin-group agreement. Values above 0.70 are de-sirable (George, 1990; Nunnally, 1978). Values forgroup antisocial behavior ranged from 0.92 to 0.99,with median of 0.98. Values for the measure of taskinterdependence ranged from 0.80 to 0.97, with amedian of 0.89. For the measure of close supervi-sion, values ranged from 0.26 to 0.99, but the me-dian was 0.90, and 86 percent of the groups hadvalues above 0.70. And values for the measure oflikelihood of punishment ranged from 0.75 to 1.0,with a median of 0.98.

Tests of Hypotheses

The correlation matrix and descriptive statisticsof our variables are presented in Table 1. Hypoth-esis 1 predicted that group antisocial behaviorwould be positively related to the antisocial behav-ior of individual group members. To test this hy-pothesis, we performed hierarchical regressionanalyses. Table 2 presents these results. In the firststep, only the control variables were entered intothe equation: tenure in the organization, age, edu-cation, gender, perceived control, general satisfac-tion, likelihood of punishment, and close supervi-sion. In the second step, group antisocial behaviorwas added to the equation. Group antisocial behav-ior significantly predicted individual antisocial be-havior (j3 = .29, p < .01; Fgias = 3.36, p < .01;R^ = .19). Thus, as predicted, group antisocial be-havior explained considerable unique variance inindividual antisocial behavior beyond that whichwas accounted for by the control variables.

666 Academy of Management Journal December

TABLE 1Correlations and Descriptive Statistics for All Variables^

Variables Mean s.d. 10 11 12 13

Control1, Perceived control2, Gender''3, Age4, Tenure in firm5, Education6, General satisfaction

Dependent7, Individual antisocial

behavior8, Satisfaction with

coworkers9, Intentions to leave

Independent10, Group antisocial

behavior"Moderators

11, Individual tenure ingroup

12, Task interdependence"^13, Likelihood of

punishment"14, Close supervision"

2,72 0,841,46 0,50 -,15*

32,54 10,13 -.11 ,095,93 6.34 ,08 -,01 ,74**2,84 1.00 -,12 -,23** -,35** -.39**

167.02 32,76 .40** -,14 -.03 -.03 ,19*

1.45 0.38 -.07 -.04 -.01 -,04 ,12

42.85 9.67 .24** -.04 -.06 -.11 .28**

9.62 12.13 .16 -.05 .59** .61** - .51**

-.24**

.56** -.05

.10 -.03 -.15

1.82 0.14 .04 -.22** - .21*

3.87 5,07 .07 .06 ,51*

-,19*

.59*

.11 -.05

-.26** -.07

,17* .02 -.18*

.13 -.12 .50* -.11

2.29 0,52 -.03 -.10 -.03 .024,34 0.51 -.10 .01 -.03 -.00

.19** .17* .09

.10 -.00 -.01.14* -.17 .15* -.11,08 .06 -.01 .00 .15

3.08 0.53 .03 .10 .00 .09 -.25** -.23** -.10 -.15* .00 -.02 .16* .07 .03

"Arranges from 164 to 187.'• Coded as male = 1, female = 0." Values are group means (each individual in a group was assigned the group niean for this variable).

* p < .05** p < .01

Hypotheses 2 through 4 each predicted a moder-ating effect for the relationship between group an-tisocial behavior and individual antisocial behav-ior. We used moderated regression analysis to testthese hypotheses; Table 3 has results. In order toreduce the multicollinearity associated with theuse of interaction terms, we centered the indepen-dent variables around zero before creating the in-teraction terms (Aiken & West, 1991).

Hypothesis 2 predicted that the degree of simi-larity in the levels of group members' antisocialbehavior would moderate the relationship betweengroup antisocial behavior and individual antisocialbehavior in such a way that the more similar thegroup antisocial behavior (the lower the variance),the stronger the relationship. The interaction ofgroup antisocial behavior and variance in groupantisocial behavior was significant (j3 = - .48, p <.01; Fii, 121 = 6.60, p < .01; R^ = .37). To interpretthe direction of this interaction effect, we followedAiken and Wests's (1991) recommendations. First,the regression equation was restructured algebra-ically to express the regression of individual anti-social behavior on group antisocial behavior at dif-ferent levels of variance in the group measure.Next, we derived a series of simple regression equa-

tions by substituting in three different values ofvariance, as recommended by Cohen and Cohen(1983). Next, we examined the simple slopes todetermine if they differed from zero and each other.The interaction term was found to be in the pre-dicted direction: variance in group antisocial be-havior moderated the relationship between groupand individual antisocial behavior in such a waythat the lower the group variance (the more similarthe levels of antisocial behavior in the group), thestronger the relationship. Hence, Hypothesis 2 wassupported.

Hypothesis 3 predicted that individual tenure ina work group would moderate the relationship be-tween group and individual antisocial behavior insuch a way that the longer an employee has been inhis or her group, the stronger the relationship willbe. The interaction of group antisocial behavior andindividual tenure was significant in the moderatedregression analysis (j3 = .25, p < .01; F^^ ^^l - 3-54,p < .01; R^ = .24). To test the direction of thissignificant relationship, we performed the samecalculus described above and found that the direc-tion of the interaction effect was as predicted.Hence, Hypothesis 3 was supported.

Hypothesis 4 predicted that the greater the task


TABLE 2Results of Hierarchical Regression Analysis

Examining the Influence of Group AntisocialBehavior on Individual Antisocial Behavior"

Variable

ControlTenure in firmAgeEducationGender*"Perceived controlGeneral satisfactionLikelihood of punishmentClose supervision

Independent effectGroup antisocial behavior

Adjusted R^Change in R^Fdf

Step 1:Control

.22**-.21

.26**-.11-.01-.24*

.01-.14

.21

.14

3.26**10, 124

Step 2:Independent Effect

.29**

.19

.14

.023.36**9, 125

" Entries are betas.^ Goded as male = 1, female = 0.

* p < .05**p < .01

interdependence of a work group's members, thestronger the relationship between group antisocialbehavior and individual antisocial behavior. Whenthis interaction term was entered into the equationwith the control variables and the main effects, itwas significant (/3 = .20, p < .05; f^ 121 = 3-51' P <.01; R^ = .24) and in the predicted direction. Thus,Hypothesis 4 was supported.

Hypotheses 5 and 6 predicted that to the extentthat an individual group member engaged in lessantisocial behavior than his or her group, the focalindividual's satisfaction with group memberswould be lower and his or her intentions to leavewould be greater. To test this prediction, we cre-ated a variable representing the difference betweengroup and individual antisocial behavior; thehigher the value of this variable, the higher thelevel of group antisocial behavior relative to theindividual level. Separate hierarchical regressionanalyses were performed for each of the dependentvariables. Table 4 shows results. In the first step,we entered the control variables age, education,gender, tenure in group, and tenure in organization.In the second step, the variable reflecting the dif-ference between group and individual antisocialbehavior was entered. As predicted, the higher thegroup's antisocial behavior relative to the individ-ual's antisocial behavior, the less satisfied an indi-vidual was with his or her coworkers (j3 = - .25,

p < .01; F7 148 = 3.04, p < .01; R^ = .13). However,the difference between group and individual anti-social behavior was not significantly related to in-tentions to leave (/3 = - .09, n.s.; F7 0̂6 ~ 15.69, p <.01; R^ = .51). Thus, support was found for Hypoth-esis 5 but not Hypothesis 6.

Finally, we used moderated hierarchical regres-sion analysis to test Hypothesis 7 and Hypothesis 8,which predicted that likelihood of punishment bymanagement and degree of close supervisionwould moderate the relationship between groupand individual antisocial behavior. As Table 5shows, when the interaction of group antisocialbehavior and likelihood of punishment was en-tered into the equation, it was significant (/3 =- . 2 0 , p < .05; Fio 124 = 3-76, p < .01; R^ = .23).Following the same procedure previously de-scribed, we found that the relationship was in thepredicted direction. However, group antisocial be-havior by close supervision was not significantwhen it was entered ()3 = .02, n.s.; F^Q, 124 = 3.00;p < .01; R^ = .19). Hence, support was found forHypothesis 7 but not for Hypothesis 8.

DISCUSSION

In this research, we explored the extent to whichwork group context influenced the antisocial be-havior of individual employees. We found that,even with many other explanatory variables con-trolled, the antisocial behavior exhibited by a workgroup was a significant predictor of an individual'santisocial behavior at work. These findings providepreliminary evidence that a group-level focus isappropriate and important for understanding suchbehavior in work settings.

Our findings also suggest that several conditionsmoderate this group influence. Consistent withattraction-selection-attrition theory (Schneider,1975, 1987; Schneider & Reichers, 1983), groupswith stronger antisocial climates appeared to havegreater ability to influence individual members' an-tisocial actions. In addition, we found that the in-fluence of a group's antisocial behavior on an indi-vidual's antisocial behavior became stronger as theindividual's time in the group increased. Finally,our results indicated that when group members hadto rely upon each other for task accomplishment(task interdependence was high), individual behav-ior was more strongly related to the level of antiso-cial behavior exhibited by the group. When consid-ered together, these findings present a consistentpattern, suggesting that as the richness of the groupexperience increases, members become more likelyto match their level of antisocial behavior to that ofthe group.


TABLE 3Results of Hierarchical Regression Analysis Examining Moderators of the Group Antisocial

Behavior-Individual Antisocial Behavior Relationship"

Variable

ControlTenure in firmAgeEducationGender''Perceived controlGeneral satisfactionLikelihood of punishmentClose supervision

Independent effectsGroup antisocial behaviorVariance in group antisocial behaviorIndividual tenure in groupTask interdependence

InteractionsGroup antisocial behavior X variance in group antisocial behaviorGroup antisocial behavior X tenure in groupGroup antisocial behavior X task interdependence

Adjusted R^Change in R^Fdf

Step 1: Control andIndependent Effects

.25**-.15

.19

.14-.00- .21*

.01-.06

.24**

.14

.06

.11

.20

.14

3.17**12, 122

Steps

-.48**

.37

.32

.176.60**13, 121

2-4: Interactions

.25**.20*

.24 .24

.17 .17

.04 .043.54** 3.51**13,121 13,121

" Entries are betas.^ Coded as male = 1, female - 0.

* p < .05**p < .01

Another interesting finding of this study is that tothe extent that an individual employee exhibitedless antisocial behavior than his or her group, he orshe was less satisfied with coworkers. This findingis consistent with the attraction-selection-attritionperspective, suggesting that prosocial individualsworking in a sea of antisocial individuals may ex-perience discomfort or dissonance that in turn maylead to attrition among those who do not fit. How-ever, contrary to our prediction, we did not findthat intentions to leave an organization were higherfor those whose antisocial behavior was lower thanthat of the group. Possibly our single-item measureof intentions to leave was inadequate. Another ex-planation is that a misfit with one's work groupresults in attrition through leaving the group (avariable we did not assess in this study) rather thanthrough leaving the organization.

Implications for Future Research

These results have interesting implications forthe groups literature, for research on antisocial be-havior in organizations, and for practicing manag-

ers. In light of the groups literature, the generalfinding that groups influence their members is notunexpected. Since the time of the Hawthorne stud-ies (Roethlisberger & Dickson, 1939), managementscholars have noted the effects of work groups onindividuals. What has changed over time, however,is the manner in which these effects have beenviewed. Early research depicted groups as sourcesof management problems, but more recent researchhas conceptualized groups as solutions to manywork-related problems (Shea & Guzzo, 1987). Ofcourse, in reality group effects are both positive andnegative. Our results indicate that groups display-ing high levels of antisocial behavior may influencemembers to perform antisocial actions. The alter-native, however, should not be overlooked—thatgroups displaying high levels of prosociai behaviorencourage such behavior by members (George,1990). Therefore, the primary implication of ourresearch is not that groups have negative effects onindividuals but, rather, that group effects are sig-nificant and should be better understood. Thegroups literature suggests that the group influenceswe found were driven by the development of anti-



Examining the Impact of Differences betweenGroup and Individual Antisocial Behavior^


Examining the Moderating Effects of Punishmentand Close Supervision on the Group Antisocial

Variable

ControlAgeEducationGender*"Tenure in firmTenure in groupGroup antisocial behavior

Independent effectsGroup antisocial behaviorIndividual antisocial behaviorDifference

Adjusted fl^Change in R^F

df

° Entries are betas.'' Coded as male = 1, female

* p < .05* * p < . 0 1

Satisfactionwith

CoworkersIntentionsto Remain

Step 1 Step 2 Step 1 Step 2

-.01.32**.03.10

-.07.15

-.25**

.09 .13

.06 .08.04

2.56* 3.04**ie6, 149 7, 148 6

= 0.

social norms within the studied groups.tant future research direction will be

.27**

.34**

.17*

.29**

.09

.06

-.09

.50 .51

.48 .48

1.07**15.69**, 107 7, 106

An impor-to identify

cenavior-inaiviauai AntisocialRelationship"

Variable

ControlTenure in firmAgeEducationGender''Perceived controlGeneral satisfaction

Independent effectsClose supervisionGroup antisocial behaviorLikelihood of punishment

InteractionsGroup antisocial behavior X

punishmentGroup antisocial behavior X

surveillance

RAdjusted R^Change in R^F

df

" Entries are betas.'' Coded as male = 1, female

* p < .05**p < .01

Step 1:Independent

Effects

.21*-.13

.21*

.12

.02-.23**

-.07.29**.01

.19

.14

3.36**9,125

= 0.

juenavio

Step

ir

2:Interactions

-.20*

.23

.17

.043.76**10, 124

.02

.19

.13

.003.00**10, 124

how work group expectations regarding antisocialbehavior are developed, communicated, and en-forced.

Our results also are informative for the work-place deviance literature. The findings reportedhere provide preliminary evidence that a groupeffect does occur, suggesting that future research atthe group level will be beneficial to an enhancedunderstanding of why, when, and how people be-have aggressively at work. Previous research ondeviance has emphasized the importance of bothindividual and environmental factors in predictingindividuals' antisocial workplace actions (O'Leary-Kelly et al., 1996). To date, however, environmen-tal factors have been described primarily as organi-zation-level phenomena. Of course, the work groupitself is also a component of an individual's workenvironment. However, we would argue that thework group's influence should be conceptualizedas more than just an additional environmental vari-able to be explored. Because of its centrality to theindividual-organization interface, an understand-ing of the group may be critical to an understanding

of other antecedents, both environmental and indi-vidual.

For example, with regard to environmental ante-cedents to aggression, previous research has ex-amined factors such as adverse treatment (forinstance, an individual's perception that an orga-nization has blocked his or her personal goals;O'Leary-Kelly et al., 1996). Of course, an individ-ual's belief that he or she has been adverselytreated by an organization depends on how thisterm is defined. The individual's definition willbe influenced not only by the actual treatmentthat is received, but also by his or her interpre-tation of this treatment, with that interpretationbeing affected by information received from thesocial environment—for instance, from the workgroup (Salancik & Pfeffer, 1978; Schneider,1987). A strong work group can be conceptual-ized as a setting that exists at the intersection ofthe individual and the organization. The groupprovides a social context that is critical to theindividual's interpretation of organization-level


systems. This social context, therefore, can havea significant effect on individuals' antisocial be-havior.

Implications for Managers

Finally, our research has implications for manag-ers. Perhaps most obviously, it highlights the ne-cessity for strong actions to be taken to counterantisocial behavior in organizations. The group ef-fect we found here might be described more simplyas a contagion effect. This effect may help to ex-plain why it appears that acts of aggression seem tospread or occur in clusters within a given industryor organization. The message for managers seemsclear—antisocial groups encourage antisocial indi-vidual behavior. It is crucial to nip behaviorsdeemed harmful in the bud so as to avoid a socialinfluence effect. Managers who expect that isolat-ing or ignoring antisocial groups will encouragethem to change are probably mistaken.

The findings related to the influence of likeli-hood of punishment and closeness of supervisionalso have interesting implications for practice. Al-though we found that the likelihood of punishmentweakened the relationship between group antiso-cial behavior and individual antisocial behavior,we did not find that closeness of supervision madeany difference. This suggests that common forms ofdeterrence, such as monitoring employee behavior,will not necessarily impact individuals' antisocialbehavior if perceptions of punishment are unaf-fected.

Limitations

As always, limitations of our study should beconsidered in its interpretation. First, our designwas cross-sectional, not longitudinal. Research thatassesses the influence of group context over timewould provide additional and stronger support forthe existence of these effects. Another possible lim-itation is the sampling method we employed. Weobtained a diverse sample of occupational groupsfrom different organizations and industries so thatour results would be generalizable to the overallpopulation of employees. This method, however,may have introduced unnecessary noise, deflatingour observed relationships. In addition, we couldnot be certain that we obtained representativegroups from each organization. For example, cer-tain types of employees within each group mayhave been more likely to respond than others, thusbiasing our results. However, we contend that oursample is more representative of the overall popu-

lation of employees than a sample drawn from allemployees within only one organization.

Also, we were unable to reduce all commonmethod bias. Our split-sample design was useful inreducing most of the method variance because, forthe most part, different respondents were used toassess the dependent and independent variables.Nevertheless, this design cannot eliminate allforms of common method variance. For example,perhaps the observed relationships were somewhatinflated because all of the variables were assessedwith survey measures.

Finally, we were unable to control for the role ofindividual predispositions in explaining antisocialbehavior at work. Although our control variablesincluded individual demographic characteristics asproxies for predispositions, ideally we would havemore directly measured and controlled for antiso-cial predispositions, such as prior learning of ag-gressive behavior, family background, personality,and child-rearing practices. Future researchersshould consider this possibility.

Conclusion

Despite these limitations, the implications of thisstudy are significant. Antisocial behavior is notsimply an individual-level phenomenon. As theresults of this study show, the social context of thework group has an extensive influence overwhether and when individuals will behave in an-tisocial ways at work. From a research perspective,these findings suggest numerous useful directionsfor future investigation. From a managerial per-spective, these findings indicate that organizationshave both the ability and responsibility to influ-ence antisocial behavior by shaping work groupdynamics. Our results confirm that antisocial be-havior at work is not only prevalent—it is alsocontagious.

REFERENCES

Aiken, L. S., & West, S. G. 1991. Multiple regression:Testing and interpreting interactions. ThousandOaks, CA: Sage.

Ashford, S. J., Lee, C, & Bobko, P. 1989. Content, causes,and consequences' of job insecurity; A theory-basedmeasure and substantive test. Organizational Be-havior and Human Decision Processes, 43: 207-242.

Bandura, A. 1973. Aggression: A social learning anal-ysis. Englewood Cliffs, NJ: Prentice-Hall.

Bandura, A. 1977. Social learning theory. EnglewoodCliffs, NJ: Prentice-Hall.


Bandura, A. 1986. Social foundations of thought andaction: A social cognitive theory. Englewood Cliffs,NJ: Prentice-Hall.

Bandura, A. 1990. Selective activation and disengage-ment of moral control. Joumal of Social Issues,46(1): 27-46.

Bandura, A. 1991. Social cognitive theory of moralthought and action. In W. M. Kurtines & J. L. Gewirtz(Eds.), Handbook of moral behavior and develop-ment: Theory, research, and application, vol. 1:71-129. Hillsdale, NJ: Erlbaum.

Bandura, A., Barbaranelli, G., Gaprara, G. V., & Pastorelli,G. 1996. Mechanism of moral disengagement in theexercise of moral agency. Journal of Personalityand Social Psychology, 71: 364-374.

Baron, R. A., & Richardson, D. R. 1994. Human aggres-sion (2nd ed.). New York: Plenum.

Bennett, R. 1998. Perceived powerlessness as a cause forworkplace deviance. In R. Griffin, A. O'Leary-Kelly,& J. Gollins (Eds.), Dysfunctional behavior in orga-nizations: 229-240. Greenwich, GT: JAI Press.

Brief, A. P., & Motowidlo, S. J. 1986. Prosocial organiza-tional behaviors. Academy of Management Re-view, 11: 710-725.

Gampbell, D. T., & Fiske, D. W. 1959. Gonvergent anddiscriminant validation by the multitrait-multi-method matrix. Psychological Bulletin, 56: 81-105.

Gohen, J., & Gohen, P. 1983. Applied multiple regres-sion/correlation analyses for the behavioral sci-ences (2nd ed.). Hillsdale, NJ: Erlbaum.

Govey, M. K., Saladin, S., & Kilen, P. J. 1989. Self-moni-toring, surveillance, and incentive effects on cheat-ing. Joumal of Social Psychology, 129: 673-679.

DePaulo, P. J., & DePaulo, B. M. 1989. Gan deception bysalespersons and customers be detected throughnonverbal behavioral cues? Joumal of Applied So-cial Psychology, 19: 1552-1577.

Dewar, R. D., & Werbel, J. 1979. Universalistic and con-tingency predictions of employee satisfaction andconflict. Administrative Science Quarterly, 24:426-448.

Ditton, J. 1977. Part-time crime: An ethnography offiddling and pilferage. London: Macmillan.

George, J. M. 1990. Personality, affect and behavior ingroups. Joumal of Applied Psychology, 75: 107-116.

George, J. M., & James, L. R. 1993. Personality, affect, andbehavior in groups revisited: Gomment on aggrega-tion, levels of analysis, and a recent application ofwithin and between analysis. Joumal of AppliedPsychology, 78: 798-804.

Giacalone, R. A., & Greenberg, J. 1996. Antisocial behav-ior in organizations. Thousand Oaks, GA: Sage.

Greenberg, J. 1990. Employee theft as a reaction to un-

derpayment inequity: The hidden cost of pay cuts.Joumal of Applied Psychology, 75: 561-568.

Greenberg, J. 1993. Stealing in the name of justice: Infor-mational and interpersonal moderators of theft reac-tions to underpayment inequity. Organizational Be-havior and Human Decision Processes, 54: 81-103.

Greenberg, J., & Scott, K. S. 1996. Why do workers bitethe hands that feed them? Employee theft as a socialexchange process. In B. M. Staw & L. L. Gummings(Eds.), Research in organizational behavior, vol.18: 111-156. Greenwich, GT: JAI Press.

Gruber, J. E. 1990. How women handle sexual harass-ment: A literature review. Social Science Research,74: 3-9.

Harper, D. 1990. Spotlight abuse—save profits. Indus-trial Distribution, 79(3): 47-51.

Hirschi, T. 1969. Causes of delinquency. Berkeley: Uni-versity of Galifornia Press.

Hollinger, R. G. 1988. Working under the influence(WUI): Gorrelates of employees' use of alcohol andother drugs. Joumal of Applied Behavioral Sci-ence, 24: 439-454.

Hollinger, R. G., & Glark, J. P. 1983. Deterrence in theworkplace: Perceived certainty, perceived severity,and employee theft. Social Forces, 62: 398-418.

Hulin, G., & Rousseau, D. M. 1980. Analyzing infrequentevents: Once you find them, your troubles begin. InK. H. Roberts & L. Burstein (Eds.), Issues in aggre-gation: 65-75. San Francisco: Jossey-Bass.

James, L. R., Demaree, R. G., & Wolf, G. 1984. Estimatingwithin-group interrater reliability with and withoutresponse bias. Joumal of Applied Psychology, 69:431-444.

James, L. R., Demaree, R. G., & Wolf, G. 1993. r^^: Anassessment of within-group interrater agreement.Joumal of Applied Psychology, 78: 306-309.

Johns, G. 1997. Gontemporary research on absence fromwork: Gorrelates, causes, and consequences. In G. L.Gooper & I. T. Robertson (Eds.), International re-view of industrial and organizational psychology:115-174. London: Wiley.

Kidwell, R. E., & Bennett, N. 1993. Employee propensityto withhold effort: A conceptual model to intersectthree avenues of research. Academy of Manage-ment Review, 18: 429-456.

Kozlowski, S. W., & Hattrup, K. 1992. A disagreementabout within-group agreement: Disentangling issuesof consistency versus consensus. Joumal of AppliedPsychology, 77: 161-167.

Latane, B. 1981. The psychology of social impact. Amer-ican Psychologist, 36: 343-356.

Lehman, W., & Simpson, D. 1992. Employee substanceabuse and on-the-job behaviors. Joumal of AppliedPsychology, 77: 309-321.

Martinko, M. J., & Zellars, K. L. 1996. Toward a theory of

672 Academy of Management Journal December

workplace violence: A social learning and attribu-tional perspective. Paper presented at the annualmeeting of the Academy of Management, Cincinnati.

Martocchio, J. J. 1994. The effects of absence culture onindividual absence. Human Relations, 47: 243-262.

McGurn, M. 1988. Spotting the thieves who work amongus. Wall Street Journal, March 8: A16.

Mensch, B. S., & Kandel, D. B. 1988. Do job conditionsinfluence the use of drugs? Journal of Health andSocial Behavior, 29; 169-184.

Merriam, D. H. 1977. Employee theft. Criminal JusticeAbstracts, 9: 375-406.

Murphy, K. R. 1993. Honesty in the workplace. Belmont,CA: Brooks/Cole.

Nunnally, J. C. 1978. Psychometric theory (2nd ed.).New York: McGraw-Hill.

O'Leary-Kelly, A. M., Griffin, R. W., & Glew, D. J. 1996.Organization-motivated aggression: A researchframework. Academy of Management Review, 21:225-253.

Podsakoff, P. M., MacKenzie, S., Moorman, R. H., & Fet-ter, R. 1990. Transformational leader behaviors andtheir effects on followers' trust in leader, satisfac-tion, and organizational citizenship behaviors.Leadership Quarterly, 1: 107-142.

Robinson, S. L., & Bennett, R. J. 1995. A typology ofdeviant workplace behavior: A multidimensionalscaling study. Academy of Management Journal,38: 555-572.

Robinson, S. L., & Bennett, R. J. 1997. Workplace devi-ance: Its nature, its causes, and its manifestations. InR. J. Lewicki, R. J. Bies, & B. H. Sheppard (Eds.),Research on negotiation in organizations, vol. 6:3-28. Greenwich, GT: JAI Press.

Robinson, S. L., & Greenberg, J. 1998. Employees behav-ing badly: Dimensions, determinants and dilemmasin the study of workplace deviance. In D. M. Rous-seau & G. Gooper (Eds.), Trends in organizationalbehavior, vol. 5: Forthcoming. New York: Wiley.

Roethlisberger, F. J., & Dickson, W. J. 1939. Managementand the worker. Gambridge, MA: Harvard Univer-sity Press.

Rousseau, D. M. 1985. Issues of level in organizationalresearch. In L. L. Gummings & B. M. Staw (Eds.),Research in organizational behavior, vol. 7: 1-37.Greenwich, GT: JAI Press.

Rusbult, G., Farrell, D., Rogers, G., & Mainous, A. 1988.Impact of exchange variables on exit, voice, loyalty,and neglect: An integrative model of responses todeclining job satisfaction. Academy of Manage-ment Journal, 31: 599-627.

Salancik, G. J., & Pfeffer, J. 1978. A social informationprocessing approach to job attitudes and task design.Administrative Science Quarterly, 23: 224-253.

Schneider, B. 1975. Organizational climates: An essay.Personnel Psychology, 28: 447-480.

Schneider, B. 1987. The people make the place. Person-nel Psychology, 40: 437-454.

Schneider, B., & Reichers, A. E. 1983. On the etiology ofclimates. Personnel Psychology, 36: 19-39.

Shea, G. P., & Guzzo, R. A. 1987. Groups as humanresources. In K. M. Rowland & G. R. Ferris (Eds.),Research in personnel and human resources man-agement, vol. 5: 323-356. Greenwich, GT: JAI Press.

Skarlicki, D. P., & Folger, R. 1997. Retaliation in theworkplace: The roles of distributive, procedural, andinteractional justice. Journal of Applied Psychol-ogy, 82: 416-425.

Smith, P. G., Kendall, L. M., & Hulin, G. L. 1969. Themeasurement of satisfaction in work and retire-ment. Ghicago: Rand McNally.

Steffy, B. D., & Laker, D. R. 1991. Workplace and personalstresses antecedent to employee's alcohol use. Jour-nal of Social Rehavior and Personality, 6: 115-126.

Storms, P., & Spector, P. 1987. Relationships of organi-zational frustration with reported behavioral reac-tions: The moderating effect of locus of control. Jour-nal of Occupational Psychology, 60: 227-234.

Tittle, G. R. 1973. Sanction fear and the maintenance ofsocial order. Social Forces, 55: 579-596.

Van de Ven, A. H., Delbecq, A. L., & Koenig, R. 1976.Determination of coordination modes within organi-zations. American Sociological Review, 41: 322-338.

Vardi, Y., & Wiener, Y. 1996. Misbehavior in organiza-tions: A motivational framework. OrganizationalScience, 7: 151-165.

Sandra L. Robinson (Ph.D., Northwestern University) iscurrently an assistant professor of organizational behav-ior at the University of British Golumbia. Her researchinterests include psychological contracts and contractbreach in employment relationships, workplace devi-ance, and workplace aggression.

Anne M. O'Leary-Kelly is an associate professor of man-agement at the University of Arkansas. She received herPh.D. in organizational behavior and human resourcemanagement from Michigan State University. Her cur-rent research interests include aggressive work behavior,motivational issues in groups, and psychological con-tracts.

Documents

monkey see monkey do the influence of work groups on the antisocial behavior of employees (Robinson & O'leary Kelly, 1998).pdf