Exploring the Processes Underlying Within-Group Homogeneity

Exploring the ProcessesUnderlying Within-GroupHomogeneityClaire M. MasonQueensland University of Technology

Many group-level constructs are based on within-group homogeneity inattitudes, affect, beliefs, and perceptions. In this article, three models for thedevelopment of within-group homogeneity are delineated. These models areshown to have implications both for differences between variables in theirlevel of homogeneity and the conditions under which relatively high and lowhomogeneity should be observed. The models are explored in a small sample(N = 24 groups), where homogeneity in job satisfaction, positive affect,potency beliefs, and task-identity perceptions is examined. The results indi-cate that variables differ in their mean level of homogeneity and suggest thathomogeneity may be the product of a combination of processes. A hierarchi-cal framework for the investigation of homogeneity is suggested for furtherresearch.

Keywords: homogeneity; within-group agreement; job satisfaction; grouppotency; group constructs

Researchers have long been interested in the tendency toward conver-gence of peoples’ perceptions, attitudes, and behavior in groups (e.g.,

Asch, 1951; Festinger, Schachter, & Back, 1950; Sherif, 1936). However,the exploration of homogeneity has tended to occur within specific theoret-ical frameworks (e.g., social identity theory) or in relation to a particulargroup attribute (e.g., group mood; Bartel & Saavedra, 2000). The develop-ment of multilevel theory, which is specifically concerned with relation-ships that cross levels, has fostered systematic analysis of the various formsof group constructs that can emerge when individuals work in groups

Small Group ResearchVolume 37 Number 3

June 2006 233-270© 2006 Sage Publications

10.1177/1046496406288972http://sgr.sagepub.com

hosted athttp://online.sagepub.com

233

Author’s Note: The author would like to thank the reviewers who provided feedback on pre-vious versions of this manuscript. Correspondence concerning this article should be addressedto Claire M. Mason at the Australian Centre for Business Research, School of Management,Queensland University of Technology, GPO Box 2434, Brisbane, Queensland, 4001, Australia.Ph: +61 7 3864 1238. Fax: +61 7 3864 1313. E-mail may be sent to [email protected]

SGR288972.qxd 4/12/2006 4:48 PM Page 233

(Chen, Mathieu, & Bliese, in press; House, Rousseau, & Thomas-Hunt,1995; Kozlowski & Klein, 2000). Nevertheless, there has been relativelylittle theoretical integration of the work that has occurred within eachframework and even less empirical work that has explored homogeneityacross different types of variables. This article begins this integration, firstby delineating alternative models for the development of homogeneity andthen by exploring homogeneity in job satisfaction, positive affect, potencybeliefs, and task-identity perceptions.

Why Investigate Within-Group Homogeneity?

There are several reasons why within-group homogeneity (which, forsimplicity, will be referred to hereafter as homogeneity) deserves investi-gation. First, homogeneity appears to be related to group performance andwell-being. Research into shared mental models is based on the propositionthat if team members share similar mental models of the abilities, skills,and processes of the group, they will be able to communicate and coordi-nate with one another more effectively and, ultimately, perform better (Bar-Tal, 1990; Cannon-Bowers, Salas, & Converse, 1993; Salas, Dickinson,Converse, & Tannenbaum, 1992). The empirical research not only supportsthis proposition (Argote, 1989; Marks, Zaccaro, & Mathieu, 2000; Mathieu,Goodwin, Heffner, Salas, & Cannon-Bowers, 2000; Peterson, Mitchell,Thompson, & Burr, 2000; Walsh, Henderson, & Deighton, 1988) but sug-gests that other forms of homogeneity also are beneficial for group perfor-mance and well-being. For example, homogeneity in attitudes and affect hasbeen found to be associated with lower group conflict, higher member satis-faction, and more prosocial behavior (Barsade, Ward, Turner, & Sonnenfeld,2000; Krebs, 1975). Homogeneous perceptions of the group, that is, agreementabout the level of morale and efficiency within the group, (Georgopoulos,1965), the quality of peer relationships and group leadership (Bliese &Halverson, 1998), and the importance of role clarity (Argote, 1989) also arerelated to measures of group performance and well-being. Bliese and Britt(2001) also found that group members with homogeneous perceptions oftheir group leader were more resilient to the effects of stressors. Therefore,identifying the processes underlying the development of homogeneity shouldenable us to improve group performance and well-being.

Second, as stated earlier, researchers investigating attitudes, affect, beliefs,and perceptions at the group level assume that individuals working in a

234 Small Group Research

SGR288972.qxd 4/12/2006 4:48 PM Page 234

group tend to display homogeneity on these variables. The empirical researchsupports this assumption (George, 1990; Mason & Griffin, 2003b; Totterdell,Kellett, Teuchmann, & Briner, 1998), but at this stage, we do not knowhow such homogeneity develops. An understanding of this process wouldinform this area of research for two reasons. First, published data suggestthat the level of homogeneity tends to vary according to the constructinvestigated, the sample investigated, and even for different groups withinthe same sample (Argote, 1989; Carron et al., 2003; George, 1990). As yet,we do not have an understanding of how or why levels of homogeneityvary. However, it is likely that the effects of group-level constructs willvary according to the level of homogeneity for the construct of interest(George & Bettenhausen, 1990). For example, collective efficacy may have astrong effect on group performance when group members’ efficacy beliefsare very similar but a weaker effect when group members’ efficacy beliefsare only moderately homogeneous. Some research supporting this propo-sition already has appeared in the climate literature. Schneider, Salvaggio,and Subirats (2002) found that climate strength (operationalized as thestandard deviation of employees’ climate ratings) moderated the relation-ship between service climate and customer service ratings. Identifyingcorrelates of homogeneity should help to predict when group-level con-structs are likely to have a relatively strong or weak impact on group andindividual behavior.

The present study begins by delineating alternative models for the devel-opment of homogeneity. These models can explain the various forms ofhomogeneity, encompassing homogeneity in behaviors, attitudes, affect,beliefs, and perceptions. However, in this study, homogeneity in job satis-faction, positive affect, potency beliefs, and task-identity perceptions wasinvestigated empirically. These variables were chosen because they all areknown to exhibit homogeneity—that is, when employees work together,their job satisfaction, affect, beliefs about the group’s effectiveness, andperceptions of the group’s task tend to converge. In addition, group-levelvariance in job satisfaction, positive affect, potency beliefs, and task per-ceptions predicts group performance (Gully, Incalcaterra, Joshi, & Beaubien,2002; Mathieu et al., 2000; Peterson et al., 2000), absenteeism (George,1990; Mason & Griffin, 2003a), prosocial behavior (George, 1990; Mason& Griffin, 2005), and self-development (Pescosolido, 2003). Focusing onthese variables in particular should provide a better understanding of theprocesses that underlie homogeneity in groups, in a set of variables that iscritical to individual well-being and group effectiveness.

Mason / Within-Group Homogeneity 235

SGR288972.qxd 4/12/2006 4:48 PM Page 235

Models Explaining the Developmentof Homogeneity Within Groups

In investigating constructs such as job satisfaction, positive affect,potency beliefs, and task-identity at the group level, researchers assumethat these attitudes, affect, beliefs, and perceptions are sufficiently homo-geneous that a group can be described as a whole in terms of these attrib-utes. Researchers have offered a wide range of explanations to support theexistence of within-group homogeneity on their construct of interest. Forexample, in relation to the investigation of mood as a group-level construct,Totterdell et al. (1998) cited emotional-contagion research to support theinvestigation of team mood effects. George (1990) argued that attraction,selection, and attrition processes (Schneider, 1987) produce similarity inmembers’ personality, and this similarity will lead to a consistent affectivetone within the group. Barsade and Gibson (1998) argued that emotionalcontagion and group norms lead to consistent emotions within groups, andBartel and Saavedra (2000) also referred to emotional-comparison processesto explain convergence of mood within work groups. Thus, just within thedomain of affect and emotion, a wide range of processes have been identi-fied to explain homogeneity. Looking at other group constructs, still moreexplanations for homogeneity can be identified.

The range of explanations for homogeneity that have been offered in theliterature is indicative of the fact that exploration of the processes underly-ing the development of homogeneity lags behind the research investigatingits effects. Multiple explanations have been offered to explain observedhomogeneity for a single construct, and some commonality in these expla-nations can be discerned across different variables. The first step towardunderstanding homogeneity, then, is to develop a theoretical framework.Toward this end, the explanations offered in the literature have been inte-grated, resulting in the delineation of three alternative models for homo-geneity. These three models are presented below.

Shared Experiences

Being part of a group means that a person is likely to share work condi-tions and experiences with his or her fellow group members (Ryan, Schmit, &Johnson, 1996). Considering events such as the announcement of a salarybonus, receipt of positive feedback about group performance, the departureof a group’s supervisor, or a computer network failure illustrates how


SGR288972.qxd 4/12/2006 4:48 PM Page 236

shared experiences occurring in the workplace could lead group membersto experience similar attitudes, affect, beliefs, and perceptions. Moreover,research into job design (e.g., Hackman, 1987), leadership (e.g., Bass &Avolio, 1993), group processes (e.g., Campion, Medsker, & Higgs, 1993;Campion, Papper, & Medsker, 1996), and group climate (e.g., Pritchard &Karasick, 1973) attests to the fact that many conditions that traditionally areshared by group members elicit fairly consistent attitudes, affect, beliefs,and perceptions. These findings provide empirical support for the sharedexperiences model of homogeneity. According to this model, homogeneityis observed because group members tend to have common working condi-tions and experiences, and group members react to these shared experiencesin a homogeneous way.

Attraction-Selection-Attrition Model

Schneider’s (1987) attraction-selection-attrition model provides anotherexplanation for the development of homogeneity. Schneider argues that orga-nizations tend to attract, select, and retain similar personalities, and thatthese similar personalities will tend to react homogeneously. Schneider pro-posed that similar types of people tend to be attracted to the same type oforganization, creating an initial restriction in the range of people coming intoan organization. The organization’s selection processes contribute further torestriction of range, as organizations tend to hire people who fit their needsand culture. Finally, people who do not suit the organizational environmentwill tend to leave, and this attrition will contribute further to the homo-geneity of personality types within an organization. This model is widely citedin the literature and has some empirical support (Schneider, Goldstein, &Smith, 1995; Schneider, Smith, Taylor, & Fleenor, 1998).

Although Schneider’s model focuses on organizational-level attraction,selection, and attrition processes, both Schneider (1987) and George (1990)argued that the model can also be applied at the group level. However, within-group homogeneity means that group members are more homogeneousthan organizational members, in general. Therefore, if attraction, selection,and attrition processes are responsible for within-group homogeneity, it isnecessary to explain how groups come to be made up of individuals that areeven more similar than the already similar personalities that exist within theorganization.

Attraction, selection, and attrition processes probably operate differentlyat the group level than at the organizational level. Attraction and attrition


SGR288972.qxd 4/12/2006 4:48 PM Page 237

processes are likely to have less of an impact because although individualscan choose to which organizations they apply, once they are in the organi-zation, they usually have less choice about the group to which they areassigned (Ruef, Aldrich, & Carter, 2003). Furthermore, group-level selec-tion processes (the assignment of an individual to a particular group withinthe organization) will be heavily influenced by the knowledge, skills, andabilities required by the group’s task.

Therefore, applied at the group level, the attraction, selection, and attri-tion model suggests that group members will have similar backgrounds,skills, and abilities. This similarity in group members’ backgrounds wouldexplain why there should be more homogeneity within groups than acrossthe organization as a whole. Therefore, this model explains observed homo-geneity as the product of attraction, selection, and attrition processes, whichat the group level will lead to group members having similar backgrounds,skills, and abilities. According to the model, this similarity between groupmembers is the cause of observed homogeneity.

Social Influence Processes

The third model explains homogeneity as the product of social influenceprocesses. Classic studies conducted as far back as the 1930s have demon-strated that the presence of others creates pressures toward uniformity inperceptions, attitudes, and behavior (Asch, 1951; Festinger et al., 1950;Sherif, 1936). Given that group work necessitates the presence of others,social influence processes are likely to operate in the group context and,therefore, may be responsible for homogeneity.

Within this third model of homogeneity, at least four areas of researchcan be identified: research into social information processing, group norms,social identity theory, and emotional contagion. What is common to thisresearch is the proposition that individuals tend to influence one another’sperceptions, attitudes, and behavior. However, within each research area,different mechanisms for social influence are identified; consequently, eachis reviewed briefly here.

Social information processing research (Salancik & Pfeffer, 1978) focuseson the effect of social information (communication between coworkers) onperceptions and attitudes. Social information is thought to affect a recipi-ent’s perceptions and attitudes by focusing attention on particular aspects ofthe environment, constructing meanings of events, providing evaluativeinformation, and influencing how the person interprets his or her needs


SGR288972.qxd 4/12/2006 4:48 PM Page 238

(Salancik & Pfeffer, 1978). Research in this area has demonstrated thatsocial information has an effect on task perceptions (O’Connor & Barrett,1980; H. M. Weiss & Shaw, 1979), job attitudes (Griffin, 1983; Schnake &Dumler, 1985), and levels of anxiety (Miller & Monge, 1985). Belongingto the same social network also has been shown to result in similar levelsof self-efficacy (Burkhardt, 1994). Therefore, the social information sharedby group members could explain homogeneity in attitudes, affect, beliefs,and perceptions.

Norms represent another form of social influence associated with groups.Group norms are assumed to develop, first, because individuals tend toobserve other group members’ behavior to determine what is appropriate(Postmes, Spears, & Lea, 2000), and second, because groups need to regulatethe behavior of their members (Brown, 1988; Hogg & Abrams, 1988; Sherif,1936). Researchers have identified group norms that prescribe what emotionsare appropriate in the workplace (Ashforth & Humphrey, 1995), what levelof productivity individuals should achieve (Seashore, 1954), and the condi-tions under which absenteeism is justified (Johns, 1994). These findings sug-gest that homogeneity might be the product of group norms regulating theattitudes, affect, beliefs, and perceptions of group members.

Social identity and self-categorization research offer another perspectiveon homogeneity. Social identity theorists have proposed that an individual’sself-concept consists of a personal identity and various social identities thatderive from his or her membership in various social categories. When anindividual conceptualizes himself or herself as a member of a particulargroup (that is, adopts a particular social identity), he or she also adopts theattitudes, beliefs, and behaviors that are understood to be prototypical forthat group (see the review by Hogg & Abrams, 1988). According to thismodel, homogeneity is due to the depersonalization of self-conception,which occurs when group members define themselves in terms of a sharedsocial identity and adopt the prototypical characteristics of that identity(Turner, 1985).

The fourth form of social influence that has been posited to account forhomogeneity is emotional contagion. Human beings are known to catch emo-tions from one another through behavioral mimicry, and this process has beenlabelled emotional contagion (Hatfield, Cacioppo, & Rapson, 1994). Ashforthand Humphrey (1995) argued that organizational groups are especially vul-nerable to emotional contagion because of the interdependency, proximity,and shared identity associated with working in groups. Because emotionalcontagion is specifically concerned with the transfer of emotions between


SGR288972.qxd 4/12/2006 4:48 PM Page 239

individuals, this process usually is only cited to account for homogeneity inemotion or affect (Bartel & Saavedra, 2000; Kelly, 2001; Totterdell et al.,1998). However, emotional-contagion processes might indirectly lead to homo-geneity in attitudes, beliefs, and perceptions to the extent that affect influ-ences these variables.

Implications of the Models

Each of these models has implications both for the types of variables thatwill exhibit high or low homogeneity and the conditions under which highor low homogeneity will be observed. Therefore, the predictive utility ofeach model can be tested by exploring differences between variables intheir susceptibility to homogeneity, and identifying the conditions underwhich high versus low levels of homogeneity are observed. These predic-tions are explored in relation to the variables investigated in this study—jobsatisfaction, positive affect, potency beliefs, and task-identity perceptions.

First, how will variables differ in their susceptibility to homogeneity? Theshared experiences model focuses on shared situational factors as the sourceof homogeneity. Therefore, variables that are strongly influenced by sharedconditions and experiences should exhibit higher levels of homogeneity thanvariables that are more strongly influenced by individual-level factors. Task-identity perceptions reflect a shared situational characteristic; consequently,the shared experiences model would predict that this variable would exhibitrelatively high homogeneity. Group potency also represents a group attributeand is affected by shared conditions and experiences, such as the availabil-ity of resources and the group’s success in achieving its deadlines (Guzzo,Yost, Campbell, & Shea, 1993; Lester, Meglino, & Korsgaard, 2002). Incontrast, job satisfaction and positive affect are individually focused, whichmeans that they reflect an individual’s unique experiences at work (e.g., con-flict with someone outside of the group) as well as shared experiences.Furthermore, both of these variables have a significant dispositional compo-nent (Arvey, McCall, Bouchard, Taubman, & Cavanaugh, 1994; Judge,Bono, Ilies, & Gerhardt, 2002; Spector & O’Connell, 1994; Staw, Bell, &Clausen, 1986). Therefore, if shared experiences are solely responsible forobserved homogeneity, job satisfaction and positive affect should exhibitless homogeneity than potency beliefs and task-identity perceptions.

In contrast, the attraction, selection, and attrition model explains homo-geneity as the result of attraction, selection, and attrition processes, which


SGR288972.qxd 4/12/2006 4:48 PM Page 240

result in group members having similar backgrounds, skills, and abilities—and thus similar attitudes, affect, beliefs, and perceptions. Of the variablestested in this study, job satisfaction and positive affect are both known to beaffected by individual characteristics (Arvey et al., 1994; Judge et al., 2002;Spector & O’Connell, 1994; Staw et al., 1986); consequently, this modelwould predict relatively high levels of homogeneity for these variables. Incontrast, task-identity perceptions and group potency beliefs, being groupcharacteristics, would be less affected by individual characteristics and should,therefore, exhibit relatively low homogeneity according to this model.

The predictions for the social influence model vary according to theform of social influence that is assumed to be operating. For example, ifemotional contagion plays a large role in homogeneity, positive affectshould show higher levels of homogeneity than other variables (Ashforth &Humphrey, 1995). In contrast, social information has been shown to havean impact on work environment perceptions (O’Connor & Barrett, 1980;Zalesny & Ford, 1990), job attitudes (Schnake & Dumler, 1985; H. M.Weiss & Shaw, 1979), and affect (Miller & Monge, 1985). Therefore, socialinformation should lead to homogeneity on all of the variables in this study.Finally, both social identity research and research into group norms havedemonstrated that group prototypes or norms can affect a broad range ofvariables, including attitudes, beliefs, values, affect, emotions, behavioralnorms, styles of speech, and language (Hogg & Abrams, 1988). However,prototypes or group norms are most likely to be enforced when the variablein question facilitates group survival, simplifies behavioral expectations,helps avoid embarrassing interpersonal problems, or expresses the centralvalues or identity of a group (Feldman, 1984; Hogg & Abrams, 1988).Within an organizational context, potency beliefs are relevant to the group’ssurvival as they are concerned with the group’s effectiveness. Furthermore,for decision-making purposes, it would be important for group members tohave a shared understanding of the group’s capability. In comparison, homo-geneity in job satisfaction, positive affect, and task-identity perceptionsseems less important for the survival of the group. Therefore, if groupnorms or social identity also contribute to homogeneity, potency beliefsshould be more homogeneous than the other variables.

The models also offer predictions about the conditions under which highversus low levels of homogeneity should be observed. The shared experi-ences model predicts that high levels of homogeneity will be observedwhen there is high commonality in group members’ conditions and experi-ences. One predictor for this model, therefore, should be the amount of


SGR288972.qxd 4/12/2006 4:48 PM Page 241

work time spent on the group task. When individuals only spend some oftheir work time on the group task, and work on separate projects outside ofthe group, there is likely to be more variability in their working conditionsand experiences. More specifically, the work itself and feedback andrewards represent key elements of the work experience. Therefore, homo-geneity should be higher when group members share the same tasks andexperience outcome interdependence (feedback and rewards based on groupperformance rather than individual performance). According to the sharedexperiences model, then, homogeneity should be correlated with indicatorsof shared experiences, such as spending more time in the group, perform-ing similar tasks, and outcome interdependence.

In contrast, the attraction, selection, and attrition model explains homo-geneity in terms of the similarity of group members, which results fromgroup-level attraction, selection, and attrition processes. At the group-level,this similarity should be most apparent in terms of group members’ back-grounds, skills, and experience. However, the original model (explainingorganizational-level homogeneity) predicts that these processes lead to sim-ilarity in group members’ personalities (Schneider, 1987), and there is someevidence from group composition research that suggests organizational groupstend to contain members who are similar in age and gender. Depending onthe effects of attraction, selection, and attrition processes, then homogene-ity might be correlated with similarity in group members’ backgrounds,personality, and demographic characteristics. In this study, similarity ingroup members’ backgrounds, age, and gender were explored as correlatesof homogeneity.

The general proposition underlying the social influence models wouldbe that homogeneity should be high when the influence of the group ishigh. However, more specific predictions are suggested by each area ofresearch. For example, because communication is a critical pathway forsocial information, the social information processing model suggests thathomogeneity should be related to the quality of communication within agroup. Conformity to group norms and prototypes tends to be greater whengroup cohesion is high (Carron, Widmeyer, & Brawley, 1985; Cartwright &Lippitt, 1957) and the group task creates interdependencies among members(Argote, 1989; Bartel & Saavedra, 2000); consequently, this research sug-gests that homogeneity should be related to the level of cohesion and taskinterdependence within a group. Interaction between group members is aprerequisite for emotional contagion; thus, the frequency of group meet-ings might be examined as an indicator of emotional-contagion effects.


SGR288972.qxd 4/12/2006 4:48 PM Page 242

In fact, a wide range of variables have been identified as moderators ofsocial influence effects, meaning that this list might be more extensive.However, as a starting point, the social influence models suggest that highlevels of homogeneity should occur when there is good communicationwithin a group, when cohesion is high, when the group’s task creates inter-dependence between group members, and when group meetings occurfrequently.

Existing Empirical Research

Empirically, then, there are two paths through which the alternativemodels can be differentiated. First, we can explore whether there are sig-nificant differences in the level of homogeneity associated with differenttypes of variables and, if so, whether these differences are more consistentwith the shared conditions and experiences model, the attraction-selection-attrition model, or the social influence models. Second, we can examinewhether homogeneity tends to be observed under conditions that fit theshared conditions and experiences model, the attraction-selection-attritionmodel, or the social influence models.

To date, the empirical research has tended to focus on identifying effectsof homogeneity rather than delineating the processes that contribute tohomogeneity (Bartel & Saavedra, 2000; Bliese & Halverson, 1998; Carronet al., 2003; Gonzalez-Roma, Peiro, & Tordera, 2002; Harrison, Price, &Bell, 1998; Klein, Conn, Smith, & Sorra, 2001; Pfeffer, 1980; Totterdellet al., 1998). Nevertheless, these studies are informative in that they identifycorrelates of homogeneity and, therefore, provide some initial informationabout the conditions under which relatively high and low levels of homo-geneity are observed.

A review of these studies reveals that the correlates identified are con-sistent with the models of homogeneity identified earlier. In line with theshared conditions model, homogeneity has been found to be related tomembership stability (Bartel & Saavedra, 2000) and the level of exposureto a new work environment (Zalesny & Farace, 1986). Consistent with theattraction-selection-attrition model, homogeneity also has been found to berelated to measures of demographic similarity (Harrison et al., 1998) andsimilarity in personalities, values, and interests (Harrison et al., 1998). Thesocial influence models are supported by correlations between measures ofhomogeneity and measures of social interaction (Gonzalez-Roma et al.,


SGR288972.qxd 4/12/2006 4:48 PM Page 243

2002; Klein et al., 2001), task interdependence (Bartel & Saavedra, 2000;Klein et al., 2001), exposure to information in a newsletter (Zalesny &Farace, 1986), leader informing behavior (Gonzalez-Roma et al., 2002),cohesion (Bliese & Halverson, 1998; Carron et al., 2003; Harrison et al.,1998), team commitment (Totterdell et al., 1998), strength of social tiesbetween team members (Bartel & Saavedra, 2000), and mood-regulationnorms (Bartel & Saavedra, 2000).

There was, however, some variability in the pattern of findings observedacross these studies. For example, member similarity (in terms of demo-graphic characteristics) was related to homogeneity in organizational com-mitment (Harrison et al., 1998) but was not related to homogeneity inperceptions of plant innovativeness and financial resources (Klein et al.,2001). Similarly, task interdependence was correlated with homogeneity inperceptions of plant innovativeness and financial resources (Klein et al.,2001) but did not correlate with homogeneity in ratings of group cohesion(Carron et al., 2003). This pattern of findings suggests that the processesunderlying homogeneity may differ from one variable to another.

However, detailed analysis of these findings reveals that they canbe explained equally well by the three models because there is consider-able overlap in the empirical predictions associated with each model. Forexample, shared conditions and experiences tend to also be associated withsocial influence, because they create the sense of common fate or interde-pendence of outcomes, which is one of the conditions underlying the will-ingness of group members to identify themselves as a group and submit togroup influence (Henry, Arrow, & Carini, 1999; Kramer & Brewer, 1984).Similarity (of personal characteristics and backgrounds) also tends to beassociated with high levels of social influence (Festinger, 1954; Shaw &Barrett-Power, 1998) because similarity is known to increase communica-tion (Milliken & Martins, 1996; Zenger & Lawrence, 1989) and interper-sonal attraction (Byrne, Clore, & Worchel, 1966; Tsui & O’Reilly, 1989).Consequently, although the finding that social integration is associated withhomogeneity in perceptions (Bliese & Halverson, 1998) is most consistentwith the social influence models, it also can be explained within the sharedconditions model and the attraction-selection-attrition model. There is alsooverlap between empirical predictions of the attraction-selection-attritionmodel and the shared conditions and experiences model. Although theattraction-selection-attrition model focuses on the trend toward homogeneityof personalities within an organization, Schneider (1987) recognized thathomogeneity reflects the combination of similar personalities experiencingthe same events and conditions. Therefore, the attraction-selection-attrition


SGR288972.qxd 4/12/2006 4:48 PM Page 244

model is able to encompass the finding that commonality of conditions isassociated with greater homogeneity. Similarly, the finding that homogeneityin job satisfaction is correlated with similarity of personality (Harrison et al.,1998) can be explained in terms of social influence because similarity tendsto be associated with stronger social ties and relationships (Byrne et al.,1966; Good & Nelson, 1973; Jehn, Northcraft, & Neale, 1999).

Therefore, one problem with these studies is that they were not designedto differentiate between alternative models for homogeneity and, therefore,are subject to competing explanations. Differentiation of the models is fur-ther complicated by the fact that they are not mutually exclusive. In fact,researchers commonly identify more than one process to explain why the con-struct should exhibit homogeneity (e.g., George, 1990; Mason & Griffin,2003b). Furthermore, as identified earlier, each of the models has empiricalsupport. Consequently, the study hypotheses and design need to be able to dealwith the possibility that observed homogeneity (correlates and levels of homo-geneity for given variables) may be the product of more than one process.

Another issue is that many of these studies have a potential confoundbecause homogeneity (i.e., variability in scores) is interdependent withabsolute-level effects (Bliese & Halverson, 1998). Because these studieswere not concerned specifically with the determinants of homogeneity, meanscores were not controlled for in the relationships reported with measures ofhomogeneity. However, this confound may underlie some of the effects thatwere reported. It explains why, for example, homogeneity in perceptions ofsocial integration correlated with the mean level of social integration (Bliese& Halverson, 1998). The findings from other studies also might reflect thisconfound if both the homogeneity measure and its correlate are related to themean score on the variable of interest. For example, the finding that homo-geneity in perceptions of support was associated with leader-informingbehavior (Gonzalez-Roma et al., 2002) might reflect the fact that such behav-ior is associated with higher levels of support. To identify a correlate ofhomogeneity, it is necessary to establish that the correlation remains aftercontrolling for mean scores on the variable of interest.

Finally, these studies only measured homogeneity for one variable, or, wheremore than one variable was investigated, similar types of variables. There areno studies that report homogeneity statistics for a range of different variables.Consequently, we can only compare levels of homogeneity across differentsamples. In this situation, it is not possible to determine whether differencesin the observed level of homogeneity are because of the sample or the typeof variable under analysis. To investigate whether some variables exhibithigher levels of homogeneity than others, we need homogeneity data on


SGR288972.qxd 4/12/2006 4:48 PM Page 245

a range of different variables from a single sample of groups. Furthermore,if we had homogeneity data for a range of different variables, we alsowould be able to investigate whether the predictors of homogeneity varyaccording to the type of variable under analysis.

Hypotheses

The existing research did not provide a strong basis for making a prioripredictions as to which model or models had the best validity. However, thestudy was designed to address the issues identified earlier. By measuringhomogeneity for a range of different types of variables—attitudes (job sat-isfaction), affect (positive affect), beliefs (potency beliefs), and perceptions(task-identity)—it was possible to compare levels of homogeneity acrossvariables. The models of homogeneity imply that variables will differ in theirsusceptibility to the processes underlying homogeneity. The first hypothesiswas, therefore, that

H1: There will be significant variation in the level of homogeneity exhibitedby job satisfaction, positive affect, potency beliefs, and task-identity.

The models also suggested a range of factors that should be correlatedwith homogeneity. The predictive validity of each model was investigatedby examining which of these factors were associated with homogeneity.Allowing for the possibility that the models would operate in combination,it was hypothesized that

H2: Homogeneity in job satisfaction, positive affect, potency beliefs, andtask-identity perceptions will be related to the proportion of time spent on thegroup task; the similarity of the tasks performed by group members; outcomeinterdependence; similarity of members’ backgrounds, age, and gender; per-ceived communication quality; cohesion; task interdependence; and fre-quency of group meetings.

Method

Participants

The study measures were obtained from a survey carried out in nine differ-ent organizations located in Queensland, Australia. A total of 518 participants


SGR288972.qxd 4/12/2006 4:48 PM Page 246

completed the questionnaire, providing data on 71 work groups. This samplewas diverse and there was wide range in the type of tasks performed by eachwork group, ranging from patient care (in a hospital) to client service (in acall center) to replenishment of stock (on a factory floor) to management(within a fast-food chain).

In studies where homogeneity or variability of responses within groupsis of interest, it is particularly important to have a high within-groupresponse rate. Ideally, to obtain an accurate picture of the variability inresponses, 100% of group members should be represented in the data set.In reality, this criterion is difficult to achieve. There is no established stan-dard in the literature as to what represents an acceptable response rate forstudies investigating homogeneity or variability in responses, but I followedJackson et al. (1991) and analyzed data from groups with a response rate of75% or higher. After applying this criterion, and removing one group of twopeople, 24 groups (representing responses from 160 group members) wereleft in the data set. These groups ranged in size from 3 to 25 groupmembers, with the average group size being 7.66 members (SD = 5.06).

The representativeness of this subsample was assessed by comparingthese groups with the remaining 47 groups from the original sample oneach of the study variables. Independent groups’ t tests revealed that the twosets of groups had similar profiles on all of the study variables, exceptgender and quality of communication. The groups with a high response ratetended to contain more females, (M = 0.35, SD = .29), than groups with alower response rate, (M = 0.53, SD = .37), t(69) = –2.14, p < .05 and wererated as having better communication, (M = 4.17, SD = .42), compared togroups with a lower response rate, (M = 3.90, SD = .53), t(69) = 2.22, p < .05.These two factors need to be borne in mind when assessing the generaliz-ability of the study’s findings.

Procedure

A letter was sent to 33 managers from organizations operating in thelocal metropolitan area to recruit groups for the study. These managerswere identified either because they belonged to a network for practitionerswith an interest in teamwork issues, or because their organization was knownto employ work groups. In the mail-out, the study was described as inves-tigating team effectiveness. Approximately 27% of the organizations thatoriginally were contacted agreed to participate in the study. Each organiza-tion was asked to identify at least three work groups to participate in thestudy. Work groups were defined for organizational representatives as


SGR288972.qxd 4/12/2006 4:48 PM Page 247

groups consisting of three or more employees who met on a regular basisand were jointly responsible for one or more tasks.

In return for participating in the study, organizations were provided withfeedback describing how their work groups compared with work groupsfrom other organizations on the questionnaire measures. Out of the nineorganizations that participated in the study, two elected to survey all of theirwork groups; the remaining seven organizations chose to survey between3 and 13 work groups. The survey was marketed to organizations as anopportunity to benchmark against work groups from other organizations.The organizational contacts were encouraged to identify a representativesample of work groups to participate in the survey to ensure that this bench-marking exercise would be meaningful. Unless specifically asked for guid-ance, I allowed organizational contacts to determine what a representativesample for their organization was.

The questionnaire was distributed and collected by a contact personwithin the organization, but employees were informed that the sealed ques-tionnaires would be sent to the research team for analysis and that no indi-vidual data would be reported back to the organization. Response ratesvaried between organizations, from 50% to 89%; the response rate acrossthe total sample was 69%.

Measures

Job satisfaction. Job satisfaction was measured with the short-form versionof the Minnesota Satisfaction Questionnaire (D. J. Weiss, Dawis, England,& Lofquist, 1967). This scale consists of 20 items, assessing satisfactionwith various aspects of work, which are summed to produce a general sat-isfaction score (Cook, Hepworth, Wall, & Warr, 1981). The lead-in for eachitem begins with “On my present job this is how I feel about,” and each itemrepresents a specific job attribute, such as “The way my job provides forsteady employment.” Respondents indicate their level of agreement witheach statement, using a Likert-type scale from 1 (very dissatisfied) to 5 (verysatisfied).

Positive affect. Burke, Brief, George, Roberson, and Webster’s (1989)Job Affect scale was used to measure positive affect. The scale consists of20 items, of which 10 items measure individuals’ level of positive affect andthe other 10 items measure their negative affect. Respondents are asked toindicate how they have felt at work during the past week by circling a


SGR288972.qxd 4/12/2006 4:48 PM Page 248

response from 1 (very slightly or not at all) to 5 (very much) to each item.Each item represents a particular emotion, such as “excited.” Burke et al.(1989) reported that the structure of this scale was improved when onlypositive indicators of the job affect constructs were used. A factor analysison the data from the current study confirmed that the factor structure wasmore consistent when the positive indicators were used; consequently, thereduced scale structure was employed in this study.

Potency beliefs, task-identity, task variety, task interdependence, out-come interdependence, and member heterogeneity. Campion et al.’s (1993)Work Team Characteristics scale was included in the questionnaire, and thisprovided measures of potency beliefs, task-identity, sharing of tasks withinthe group (task variety), member heterogeneity, task interdependence, andoutcome interdependence. Each subscale consisted of three items, forwhich respondents were asked to indicate their level of agreement, on aLikert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree).Campion et al.’s three-item measure of task variety captures the extent towhich group members share their tasks: “Most members of my team get achance to learn the different tasks the team performs,” “Most everyone onmy team gets a chance to do the most interesting tasks,” and “Task assign-ments often change from day to day to meet the workload needs of theteam.” This measure, therefore, was expected to be positively related tomeasures of homogeneity.

Cohesion. Group cohesion was measured from the group integration-social subscale of the Group Environment Questionnaire (Widmeyer, Brawley,& Carron, 1985). This measure (based on four items) assesses group members’perceptions of the closeness of the group as a whole, focusing on closenessbased on members’ liking for one another. Responses are obtained on a9-point Likert-type scale with anchors of 1 (strongly disagree) and 9 (stronglyagree). Widmeyer et al. (1985) provide a detailed account of the reliabilityand validity studies for the questionnaire and report a Cronbach’s alpha of.72 for this measure. Because Widmeyer et al.’s scale was developed for usewith sports teams, some of the items were reworded slightly to suit theorganizational context; for example, “Our team would like to spend timetogether outside of work hours.”

Perceived communication quality. The measure of perceived communi-cation quality was made up of four items: “Our team is satisfied with the


SGR288972.qxd 4/12/2006 4:48 PM Page 249

quality of communication within this group,” “Members of my team arevery willing to share information with other team members about ourwork,” “Most members of my team get a chance to participate in decisionmaking,” and “Teams enhance the communication among people workingon the same product.” Two of these items came from the communication/cooperation within the group subscale of Campion et al.’s (1993) WorkTeam Characteristics scale. However, because the third item of this scaleassessed cooperation rather than communication, this item was replacedwith two additional items that focused on communication. Items were mea-sured on a Likert-type scale that ranged from 1 (strongly agree) to 5 (stronglydisagree).

Group time. The measure of group time was developed for this study.Respondents were asked, “What proportion of your time at work is spenton tasks that come under the responsibility of this team?” The responseoptions were 0 to 15%, 15 to 30%, 30 to 45%, 45 to 60%, 60 to 75%, 75 to90%, or 90 to 100%.

Meeting frequency. Frequency of meetings was assessed by askingmembers to report how often their group met as a whole. Individuals respondedto this item by selecting one of the following options: daily, twice a week,weekly, fortnightly, monthly, and yearly.

Similarity in member age and gender. Respondents reported their ageand gender on the questionnaire. Similarity in age was measured with thecoefficient of variation (Allison, 1978), a scale invariant measure of disper-sion that is appropriate for interval-level variables with a theoretically fixed0. This measure is widely used in studies of demographic variability (Kleinet al., 2001). Given that gender represents a categorical variable, Blau’s(1977) index of heterogeneity was used to measure similarity in gender,which also is widely employed (e.g., Harrison, Price, Gavin, & Florey,2002; Jackson et al., 1991; Klein et al., 2001). Because of the way in whichthese measures were calculated, high values were indicative of diversityrather than similarity.

Group size. Group members were asked to identify the number of peoplein their team (including themselves). These estimates (which did not showmuch variability) were averaged to obtain a measure of the size of thegroup.


SGR288972.qxd 4/12/2006 4:48 PM Page 250

Analyses

In other studies where homogeneity or consensus has been assessed forLikert-type scale measures, researchers have used the rWG(j) statistic (Bartel& Saavedra, 2000), variance statistics (Bliese & Halverson, 1998), andstandard deviations (Schneider et al., 2002) to measure homogeneity.Whereas standard deviation and variance are directly related and, therefore,exhibit exactly the same pattern of relationships, the rWG(j) statistic is basedon a different approach. Put simply, the rWG(j) statistic represents the pro-portional reduction in error variance over that which would be expected ifresponses were completely random (James, Demaree, & Wolf, 1993). It iscalculated by comparing the obtained variance among raters and the vari-ance expected under a theoretical distribution where there is no true agree-ment among raters. Because the rWG(j) statistic provides slightly differentinformation than the variance statistic, the decision was made to measurehomogeneity through both rWG(j) and the variance. Bliese and Halverson’s(1998) procedure was employed to transform the variance statistic from ameasure of dispersion to a measure of homogeneity. This procedureinvolves calculating item-level variance statistics, averaging the item vari-ances to calculate the averaged scale variance, and then multiplying theaveraged scale variance by –1. This procedure creates a transformed vari-ance statistic where a high score represents a high level of homogeneity anda low score represents a low level of homogeneity.

However, for the sake of simplicity, the results of the analyses arereported for the transformed variance only. The rWG(j) statistics and thetransformed variance statistics were strongly correlated (correlations rangedfrom r = .93 to .98 when the two measures were compared for the samevariable), and in the analyses, very similar results were obtained for the twomeasures.

Finally, the criterion that groups needed a within-group response rate of75% or higher to be included in the study meant that the sample size wasrelatively small (N = 24). G*Power (Erdfelder, Faul, & Buchner, 1996) wasused to determine what effect sizes could be detected with a sample of thissize, using the conventional two-tailed alpha level of .05 and a power levelof .80. The analysis revealed that effect sizes would need to be more than.50 to be identified as statistically significant, meaning that even moderateeffects would not be detected with an alpha level of .05. Based on thisanalysis, the alpha level was set at .10, which meant that the probability ofa Type I error was increased but made the detection of effects more likely(Cohen, 1988).


SGR288972.qxd 4/12/2006 4:48 PM Page 251

Results

Before testing the hypotheses, the reliability of each measure waschecked and rWG(j) statistics and intraclass correlations (ICC[1]) were cal-culated to ensure that it was appropriate to analyze the data at the grouplevel. The alpha coefficients (reported in Table 1) ranged from .65 to .88,indicating that most of the scales had good internal consistency reliability.The reliabilities of the scales measuring outcome interdependence (α = .65)and task-identity (α = .66) were below the recommended cut-off of .70(Nunnally & Berstein, 1994) but given the fact that these constructs are rel-atively broad in scope, these lower reliabilities are to be expected (Van deVen & Ferry, 1980). The mean rWG(j) statistics (also reported in Table 1)ranged from .95 (for individual job satisfaction) to .60 (for group time). TherWG(j) statistic is sensitive to the number of raters, and with an average of6.67 respondents in each group, these statistics are indicative of moderateto high agreement (cf. Kozlowski & Hattrup, 1992). More than half (57%)of the intraclass correlations (see Table 2) were more than the median valueof .12 reported by James (1982) and all but one (task variety) was eitherwithin or above the range of .05 to .20, which Bliese (2000) reported astypical for applied field research involving groups.


Table 1Measures of Scale Reliability and Within-Group Agreement

rWG(j)

Construct Alpha Coefficient M SD

Individual job satisfaction .88 .95 .05Positive affect .83 .79 .21Potency .78 .87 .11Task-identity .66 .76 .24Group time n.a. .60 .38Task variety .76 .63 .33Outcome interdependence .65 .70 .23Heterogeneity in backgrounds .72 .81 .26Communication quality .71 .74 .30Cohesion .71 .72 .22Task interdependence .73 .62 .34Meeting frequency n.a. .70 .36

n.a. = Not applicable because the measure was based on one-item only.

SGR288972.qxd 4/12/2006 4:48 PM Page 252

The first hypothesis was that there would be significant differences in thelevel of homogeneity associated with measures of job satisfaction, positiveaffect, potency beliefs, and task-identity perceptions. The descriptive statis-tics reported in Table 3 show that potency beliefs exhibited the highest levelof homogeneity, followed by task-identity perceptions and job satisfaction,with positive affect displaying the lowest level of homogeneity. To determinewhether these differences were significant, a repeated measures ANOVA wasconducted. The analysis was significant, F(3, 21) = 17.53, p < .001, indicat-ing that, as predicted, there were significant differences between the variablesin their level of homogeneity. To identify the source of these differences, pair-wise comparisons were carried out with a Bonferroni adjustment. Theseanalyses revealed that potency beliefs (M = –.56, SD = .34) were significantlymore homogeneous than job satisfaction (M = –.89, SD = 37) and positiveaffect (M = –1.07, SD = .42), although they were not significantly morehomogeneous than task-identity perceptions (M = –.85, SD = .69). The otherpairwise comparisons were not significant.

Table 3 also shows the correlations among the variables. The first pointof interest is the relationship between homogeneity in job satisfaction, pos-itive affect, potency beliefs, and task-identity perceptions. These correla-tions ranged from weak, r = .25, p > .10, to moderate in strength, r = .66,p > .01. The weaker correlations provide an indication that the processesunderlying homogeneity may differ from one variable to another.


Table 2Proportion of Between-Group Variance Exhibited by Variables

Construct ICC(1)

Job satisfaction .14Positive affect .09Potency beliefs .24Task-identity .06Group time .09Task variety .03Outcome interdependence .07Membership homogeneity .22Age .23Gender .27Communication quality .12Social cohesion .23Task interdependence .11Meeting frequency .50

SGR288972.qxd 4/12/2006 4:48 PM Page 253

254

Tabl

e 3

Mea

ns,S

tand

ard

Dev

iati

ons,

and

Inte

rcor

rela

tion

sA

mon

g M

easu

res

(N=

24)

Var

iabl

eM

SD1

23

45

67

89

1011

1213

1415

1617

1. H

omog

enei

ty in

job

–0.8

9.3

7sa

tisfa

ctio

n (T

V)

2. H

omog

enei

ty in

–1.0

7.4

2.3

1po

sitiv

e af

fect

(T

V)

3. H

omog

enei

ty in

–0.5

6.3

4.5

8**

.27

pote

ncy

(TV

)4.

Hom

ogen

eity

in–0

.85

.69

.59*

*.6

6**

.25

task

-ide

ntity

(T

V)

5. J

ob s

atis

fact

ion

(M)

3.65

.31

.38†

–.05

.06

.11

6. P

ositi

ve a

ffec

t (M

)3.

09.3

9.3

2.2

3–.

09.4

7*.3

8†7.

Pot

ency

(M

)4.

02.4

3–.

11–.

39†

.19

–.34

.16

.04

8. T

ask-

iden

tity

(M)

3.70

.34

.37†

.19

.01

.41†

.57*

*.2

0.0

59.

Gro

up ti

me

(M)

6.00

.83

.01

.37†

.42†

.18

.04

–.15

.31

.24

10. T

ask

vari

ety

(M)a

3.38

.47

.17

.21

.01

.01

.48*

.35

.40†

.29

.12

11. O

utco

me

3.00

.52

.15

.26

–.03

.13

.35

.45*

.42†

.29

.18

.81*

**in

terd

epen

denc

e (M

)12

. Per

ceiv

ed h

eter

ogen

eity

3.96

.51

.07

.29

–.26

.52*

.17

.39†

–.09

.23

.07

–.03

.31

in b

ackg

roun

ds (

M)

13. D

iver

sity

in g

ende

r (I

H)

0.23

.09

–.04

–.05

.11

.22

.14

–.10

–.02

.07

.20

–.27

–.37

.19

14. D

iver

sity

in a

ge (

CV

)0.

29.2

0–.

19–.

12.0

4–.

05.0

7–.

10.0

8.0

7.2

4–.

24–.

16–.

08.1

415

. Com

mun

icat

ion

4.17

.42

–.17

–.25

.23

–.44

*.1

9–.

1084

***

.04

.34

.28

.28

–.15

.08

.08

qual

ity (

M)

(con

tinu

ed)

SGR288972.qxd 4/12/2006 4:48 PM Page 254

255

Tabl

e 3

(con

tinu

ed)

Var

iabl

eM

SD1

23

45

67

89

1011

1213

1415

1617

16. S

ocia

l coh

esio

n (M

)4.

11.6

7–.

00–.

01.0

8–.

19.1

2.0

1.4

6*.1

7–.

02.3

2.2

9.0

4.0

9–.

05.5

7**

17. T

ask

inte

rdep

ende

nce

(M)

3.41

.50

.31

.31

.04

.55*

*.3

7†.2

5–.

16.5

3*.2

3.2

5.2

3.4

6*.1

8–.

22–.

35–.

0518

. Mee

ting

freq

uenc

y (M

)3.

641.

20.4

1†.5

4*.4

1†.3

5–.

08.2

2.0

0–.

10.1

5.0

0.1

8.2

2–.

25–.

12.0

7.3

0–.

06

Not

e:T

V =

Blie

se a

nd H

alve

rson

’s (

1998

) tr

ansf

orm

ed v

aria

nce

(hig

h sc

ores

rep

rese

nt g

reat

er h

omog

enei

ty);

IH

= B

lau’

s (1

977)

Ind

ex o

f H

eter

ogen

eity

(hig

h sc

ores

rep

rese

nt g

reat

er d

iver

sity

); C

V =

Alli

son’

s (1

978)

Coe

ffic

ient

of

Var

iatio

n (h

igh

scor

es r

epre

sent

gre

ater

div

ersi

ty).

a. T

he m

easu

re o

f ta

sk v

arie

tyca

ptur

es th

e ex

tent

to w

hich

task

s ar

e sh

ared

acr

oss

the

grou

p. I

t was

ther

efor

e ex

pect

ed to

hav

e a

posi

tive

rela

tions

hip

with

hom

ogen

eity

.†p

<.1

0. *

p<

.05.

**p

<.0

1. *

**p

<.0

01.

SGR288972.qxd 4/12/2006 4:48 PM Page 255

As the first step toward identifying these processes, the correlationsbetween the predictor variables (group time, task variety, outcome inter-dependence, member heterogeneity, gender diversity, age diversity, qualityof communication, social cohesion, task interdependence, and meetingfrequency) and the homogeneity measures were examined. All of the mea-sures were calculated such that higher scores were indicative of higherlevels on the construct being assessed. The correlations revealed that homo-geneity in job satisfaction was related to meeting frequency. Homogeneityin positive affect and potency beliefs was related to group time and meet-ing frequency. Homogeneity in task-identity perceptions was related toheterogeneity of members’ backgrounds, the quality of group communica-tion (this correlation was negative and, therefore, ran counter to the secondhypothesis), and task interdependence. Task variety (used as an indicatorof task sharing), outcome interdependence, similarity in gender and age, andcohesion did not show any significant correlations with the homogeneitymeasures.

These correlational analyses suffer from the same potential confound asthe correlations reported in other studies because mean scores on the depen-dent variables have not been controlled for. Table 3 shows that mean scoreson the dependent variables were correlated with measures of homogeneity,and they also correlated with some of the predictor variables. Consequently,semipartial correlations (controlling for mean scores on the dependent vari-ables) were also used to examine the relationship between the homogene-ity measures and the predictor variables.

The semipartial correlations (reported in Table 4) showed that theserelationships remained significant when controlling for mean scores on thedependent variables. Meeting frequency remained the strongest correlate,and once mean scores were controlled for, it showed significant correlationswith measures of homogeneity on all of the dependent variables. Grouptime also retained its correlations with homogeneity in positive affect andpotency beliefs but remained unrelated to homogeneity in individual jobsatisfaction and perceptions of task-identity. The significant bivariate corre-lations associated with homogeneity in task-identity perceptions also werereplicated by the semipartial correlations in that task interdependence waspositively related to homogeneity in task-identity perceptions, and theunexpected negative relationship with communication quality and positiverelationship with member heterogeneity also appeared in the semipartialcorrelations, indicating that these effects were not because of a confoundwith mean scores on the dependent variables.


SGR288972.qxd 4/12/2006 4:48 PM Page 256

Discussion

Because this study investigated homogeneity across multiple variables, itoffers some new findings. First, it demonstrates that the level of homogene-ity differs from one variable to another, with potency beliefs exhibiting sig-nificantly greater homogeneity than job satisfaction and positive affect.Second, the correlations among measures of homogeneity for different vari-ables were weak to moderate in strength, which suggests that differentprocesses may underlie homogeneity for different variables. Finally, the lowsample size meant that failure to observe relationships potentially could beattributed to lack of statistical power. However, those relationships that wereobserved (e.g., the consistent effects for meeting frequency and group time)supported both the shared experiences and social influence models.


Table 4Semipartial Correlations Controlling for Mean Scores

on the Homogeneity Variables (N = 24)

Homogeneity Measures

Job Positive Potency Task-IdentityModel Variables Satisfaction Affect Beliefs Perceptions

Indicators of sharedexperiences

Group time .00 .42† .38† .09Task variety –.01 .14 –.07 –.13Outcome interdependence .02 .18 –.13 .01

Indicators of attraction,selection, and attritioneffects

Heterogeneity in .00 .23 –.25 .48*backgrounds

Diversity in gender –.10 –.03 .11 .21Diversity in age –.23 –.10 .02 –.09

Indicators of social influenceCommunication quality –.27 –.23 .12 –.50*Cohesion –.05 –.01 –.02 –.29Task interdependence .20 .27 .07 .44*Frequency of meetings .48* .51* .42† .43†

†p < .10 *p < .05. **p < .01. ***p < .001.

SGR288972.qxd 4/12/2006 4:48 PM Page 257

Meeting frequency emerged as the most reliable correlate of homogene-ity, and yet members’ perceptions of communication quality were associ-ated with lower homogeneity (in task-identity perceptions). These findingscould be considered contradictory in that meetings should foster discussionand dialogue between group members. However, where homogeneity isconcerned, members’ perceptions of the quality of communication may notbe as important as the uniformity of the information being exchangedwithin the group. Although the measure of meeting frequency was concep-tualized as a measure of social influence, it may be the fact that it also rep-resents a shared experience for group members (by exposing them to thesame information), which made it such a reliable predictor of homogeneity.

Overall, the findings support an integrated model of homogeneity, incor-porating both shared experiences and social influence as drivers of homo-geneity. Both group time (an indicator of shared experiences) and meetingfrequency (an indicator of social influence effects) were found to be reli-able correlates of homogeneity. Furthermore, potency beliefs had beenidentified as potentially affected by shared experiences, social information,and group norms; and they exhibited higher homogeneity than the othervariables in the study. If these effects operate in combination, it wouldexplain why potency beliefs were highly homogeneous. Some of the otherindicators of shared experiences and social influence (e.g., task variety, out-come interdependence, and social cohesion) did not emerge as correlates ofhomogeneity, but this may be because of the low sample size and lack ofstatistical power. Further empirical research with a larger sample is neededto clarify these nonsignificant results. In the case of the measure of taskvariety, the lack of effects could also be attributed to the fact that this mea-sure had relatively little between-group variance, ICC(1) = .03; conse-quently, an alternative measure of similarity in task characteristics shouldbe employed in future research.

Lack of statistical power also might be the reason why none of theattraction-selection-attrition model predictor variables were found to bepositively correlated with homogeneity. However, the finding that hetero-geneity in members’ backgrounds, skills and abilities was associated withhigher (rather than lower) homogeneity in task-identity perceptions rancounter to the predictions of this model, and deserves some explanation.The positive effect for member heterogeneity was specific to homogeneityin task-identity perceptions. Research has found that when group memberscome from different educational or functional backgrounds, the group tendsto engage in more task-related conflict and debate (Jehn et al., 1999). These


SGR288972.qxd 4/12/2006 4:48 PM Page 258

debates about the task ultimately might create a shared understanding of thetask, which is reflected here by homogeneity in task-identity perceptions.This explanation also provides a reason for the negative relationship betweencommunication quality and homogeneity in task-identity perceptions. Thehigh level of task conflict that creates homogeneity in task-identity percep-tions may lead to negative evaluations of the quality of communicationwithin the team.

A feature that differentiated this study from existing research on homo-geneity (Bartel & Saavedra, 2000; Gonzalez-Roma et al., 2002; Harrisonet al., 1998; Klein et al., 2001) was that it tested whether the effect of the pre-dictor variables were observed after controlling for the mean score on thedependent variable. The fact that the same correlates of homogeneity wereobserved, both before and after controlling for mean scores, gives weightboth to these findings and those from other studies where mean scores werenot controlled for. However, in terms of being able to definitely attributethese effects to specific models, this study suffers from the same problemas the other studies. As noted earlier, the effect of meeting frequency couldbe interpreted to support either the social influence model or the sharedexperiences model. By arguing that member similarity leads to greaterattraction and, therefore, more frequent meetings between group members(Tsui & O’Reilly, 1989), this finding can even be encompassed by theattraction-selection-attrition model.

A hierarchical approach might offer the solution for differentiating thealternative models for homogeneity. Rather than seeking to differentiatebetween competing models, it may be more useful to conceive of the modelsin terms of a hierarchy, based on the parsimony of the processes they pro-pose. The shared experiences model represents the most parsimoniousmodel because it suggests that homogeneity can occur solely as a result ofcommonality in group members’ conditions and experiences. This modelimplies that we would observe similar levels of homogeneity between inde-pendent individuals as we would observe within groups, as long as they hadthe same degree of commonality in their conditions and experiences. Theattraction-selection-attrition model is less parsimonious because it assumesthat common conditions and experiences are not sufficient by themselvesto explain homogeneity. This model proposes that we will observe higherlevels of homogeneity within groups (than amongst independent individuals)because attraction, selection, and attrition processes result in groups consist-ing of individuals with similar backgrounds. Although the attraction-selection-attrition model invokes specific processes associated with the formation


SGR288972.qxd 4/12/2006 4:48 PM Page 259

of groups to explain homogeneity, this model still can be considered moreparsimonious than the social influence model in that it still explains homo-geneity in terms of individuals’ reactions to the environment. In contrast,the social influence models introduce the additional element of groupprocesses, suggesting that processes such as communication, emotionalcontagion, norms, and prototypes lead to levels of higher levels of homo-geneity than would be expected solely on the basis of the commonality ingroup members’ work environment and backgrounds.

The benefit of conceiving of the models in terms of a hierarchy is that itprovides a structure for differentiating the models empirically. The sharedconditions and experiences model suggests that variation in homogeneityshould be able to be explained solely on the basis of the degree of com-monality in group members’ work environment. However, the attraction-selection-attrition model suggests that if we control for commonality inconditions and experiences, it should be possible to explain additional vari-ance in homogeneity based on the similarity in group members’ back-grounds and experiences. Finally, the social influence models suggest thateven after controlling for commonality in work environments and groupmember similarity, we should be able to explain additional variance inhomogeneity by taking into account the strength of social influence withingroups. The sample for this study was not large enough to conduct a hier-archical regression analysis but this approach offers promise for furtherresearch in this area.

Limitations

The sample size for the analyses limited the conclusions that could bedrawn from this study. By restricting the analyses to groups with a responserate of .75 or higher, the sample size was reduced significantly, and conse-quently, the statistical power of the analyses was low. Not only did thismean that the explanatory power of the alternative models could not becompared, it also meant that some of the nonsignificant results could poten-tially be attributed to low statistical power. The contribution of this study,therefore, lies more in the effects that were identified than in the ability torule out alternative models.

A second limitation of this study relates to the way in which the attraction-selection-attrition model was operationalized. The results of the analyses didnot offer much support for this model, in that measures of member similarity


SGR288972.qxd 4/12/2006 4:48 PM Page 260

(in backgrounds, age, and gender) were not associated with greater homo-geneity. In addition, the variables that were expected to reflect attraction,selection, and attrition process (job satisfaction and positive affect) exhib-ited lower levels of homogeneity. However, member similarity was assessedin terms of age, gender, and perceived similarity in backgrounds. In theirreview of research testing the attraction-selection-attrition model at the orga-nizational level, Schneider et al. (1998) recommended using a five-factorpersonality measure to operate this model. Ideally, to ensure that the modelwas tested adequately, measures of personality should have been includedto assess member similarity. For these reasons, we should not rule outthe possibility that attraction, selection, and attrition effects contribute tohomogeneity. Hopefully, future research will address this limitation of thisexploratory study.

A third limitation of the study is the fact that the predictor variableswere all assessed with self-report measures, which are vulnerable to vari-ous biases. Apart from using nonself-report measures of variables, thislimitation could be addressed in future research by investigating homo-geneity in an experimental setting, where exposure to shared conditionsand experiences, group member similarity, and opportunities for socialinfluence could actually be manipulated. An experimental approach wouldalso be useful in terms of being able to differentiate the effects of the alter-native models.

Finally, although data were obtained from groups that performed a rangeof different functions and worked in different organizations and industries,the analyses were based on data from groups with a high response rate,which represented less than 50% of the original sample. The groups ana-lyzed contained more females and had higher ratings of communicationquality than the groups with a lower response rate. These characteristicsmay have reduced the impact of some of the predictor variables. For example,positive effects for communication quality might have been observed ifgroups with more variability on this construct had been included in theanalyses.

Practical Implications

There is growing awareness of the effects of diversity (or homogeneity)in demographic characteristics on group functioning (Harrison et al., 2002;Jehn et al., 1999; McLeod & Lobel, 1996) but less awareness of the other,


SGR288972.qxd 4/12/2006 4:48 PM Page 261

more subtle ways in which groups can be homogeneous or heterogeneous.The theoretical framework presented in this study suggests that many of theconditions associated with group work (in particular, shared experiencesand social influence) are likely to lead to homogeneity in group members’attitudes, affect, beliefs, and perceptions. Existing research suggests thathomogeneity has implications for group well-being and performance (Argote,1989; Barsade et al., 2000; Bliese & Halverson, 1998; Georgopoulos, 1965;Krebs, 1975). Consequently, it would be useful to know how to influence thistrend. Under some circumstances (such as when creativity and innovationare sought), it may be equally useful to know how to minimize homogene-ity in attitudes, affect, beliefs, and perceptions.

The empirical findings from this study offer some suggestions for thoseseeking to either minimise or maximise homogeneity. They suggest that, ingeneral, higher levels of homogeneity can be achieved by organizing regu-lar meetings between group members and ensuring that group membersspend a significant proportion of their time at work on the group task. In sit-uations where homogeneity is not desirable, one might deliberately createa group comprised of people who normally work in different areas but arebrought together as required to work on the group’s project. For example,in a jury situation, each group member’s unique perspective and indepen-dent evaluation is of particular importance. In this context, it makes senseto create juries by bringing together people from different walks of life,who work together on a short-term basis only.

These findings also have application beyond task-based groups operat-ing in an organizational environment. There are many other situationswhere groups need either high or low levels of homogeneity in perceptionsor attitudes to function effectively. For example, crew members of a sailingboat often need to be able to coordinate their responses quickly. When crewmembers are located in different areas of the boat and environmental noise(e.g., a storm) makes verbal communication difficult, it would be vital foreach crew member to have a shared perception of the environment and tobe in agreement as to the appropriate response to those environmental con-ditions. Research that explores the differences in, and correlates of, homo-geneity in perceptions and attitudes, can inform these groups. For example,the findings suggest that when establishing a crew (e.g., for a race), therewould be value in bringing the whole crew together before the race, on aregular basis, so that they develop shared experiences and are exposed tothe same social information. With this preparation, the crew should be morelikely to respond homogeneously when it encounters a crisis situation.


SGR288972.qxd 4/12/2006 4:48 PM Page 262

The findings from this study also point to the importance of understand-ing the different forms of homogeneity within the group. Externally, a groupmight appear very homogeneous, due, for example, to homogeneity in affec-tive reactions. This external appearance might lead one to assume that groupmembers are similarly homogeneous in terms of their understanding of thegroup’s task. However, this study revealed that a group may be homogeneouson one variable and relatively heterogeneous on another. Similarly, althoughsome general prescriptions for managing homogeneity can be offered, thefindings suggest that the determinants of homogeneity may vary from onevariable to another. With further research into homogeneity, it should bepossible to offer more specific prescriptions for increasing or decreasinghomogeneity for a given outcome.

Research Implications

One of the goals of this study was to guide those who are conductingresearch looking at group constructs. The findings of this study indicate thatthere is systematic variation in the level of homogeneity for attitudes, affect,beliefs, and perceptions. Researchers are likely to be faced with some datafrom groups that are not homogeneous on the variable of interest, particularlyif group members do not meet regularly or spend relatively little time work-ing on the group task (as in most laboratory studies). By including thesegroups in studies, we may underestimate the effects of group-level constructs.Furthermore, given that the findings of this study indicate that a group mayexhibit relatively high levels of homogeneity on one variable at the sametime that it exhibits relatively low levels of homogeneity for another variable,it is necessary to determine on a case-by-case basis whether a group hasdeveloped shared affect, job satisfaction, task perceptions, or potency beliefs.In environments where the researcher has a relatively high level of experi-mental control, it may be possible to minimize such loss of data by ensuringthat group members are exposed to shared conditions and experiences andfostering high levels of social influence.

One means of improving our understanding of psychologically basedgroup constructs may be to investigate the processes underlying homo-geneity for the construct in question. The processes underlying homogene-ity may even have implications for the way in which group constructsshould be conceptualized. For example, Chan (1998) differentiated betweendirect consensus models and referent-shift consensus models for unit-levelconstructs. Applied to group constructs, the direct consensus model assumes


SGR288972.qxd 4/12/2006 4:48 PM Page 263

that a group-level construct is functionally isomorphic to the correspondingindividual-level construct and that it is appropriate to operationalize thisconstruct by simply aggregating individual-level measures. In contrast, thereferent-shift consensus model assumes that a group-level construct is con-ceptually distinct from the individual-level construct and that the group-level construct should be measured directly, using group-referenced measures.If observed homogeneity is due solely to commonality in group members’conditions and experiences, or similarity of group members, it may beappropriate to conceptualize the group-level construct in terms of the directconsensus model, because the observed homogeneity is not the product ofgroup-level processes. However, when homogeneity is the product of socialinfluence processes, such as group norms and social identification, itmay be more appropriate to use a referent-shift consensus model. Socialinfluence processes can cause group-level attributes or outcomes to differfrom the average of individual-level attributes and outcomes (Hogg &McGarty, 1990; Myers & Lamm, 1976) and this would make the group-level construct conceptually and empirically distinct from the individual-level construct.

Conclusion

The issue of homogeneity seems likely to continue to attract attentionfrom researchers, both in the organizational context and in other settings.The indications are that homogeneity is likely to be an important predictorvariable (for group viability and performance) and a moderator of othergroup-level effects. However, the full potential of this research can only berealized if we continue to investigate the processes that are responsible forthe observed homogeneity. This research is fundamental to the question ofwhen and why group members develop shared attitudes, affect, beliefs, andperceptions.

References

Allison, P. D. (1978). Measures of inequality. American Sociological Review, 43, 865-880.Argote, L. (1989). Agreement about norms and work-unit effectiveness: Evidence from the

field. Basic and Applied Psychology, 10, 131-140.Arvey, R. D., McCall, B. P., Bouchard, T. J., Taubman, P., & Cavanaugh, M. A. (1994). Genetic

influences on job satisfaction and work values. Personality and Individual Differences,17, 21-33.


SGR288972.qxd 4/12/2006 4:48 PM Page 264

Asch, S. E. (1951). Effects of group pressure on the modification and distortion of judg-ment. In H. Guetzow (Ed.), Groups, leadership, and men (pp. 177-190). Pittsburgh, PA:Carnegie.

Ashforth, B. E., & Humphrey, R. H. (1995). Emotion in the workplace: A reappraisal. HumanRelations, 48, 97-125.

Barsade, S. G., & Gibson, D. E. (1998). Group emotion: A view from top and bottom. InD. H. Gruenfeld (Ed.), Composition: Research on managing groups and teams (Vol. 1,pp. 81-102). Stamford, CT: JAI.

Barsade, S. G., Ward, A. J., Turner, J. D. F., & Sonnenfeld, J. A. (2000). To your heart’s con-tent: A model of affective diversity in top management teams. Administrative ScienceQuarterly, 45, 802-836.

Bar-Tal, D. (1990). Group Beliefs: A conception for analyzing group structure, processes, andbehavior. New York: Springer-Verlag.

Bartel, C. A., & Saavedra, R. (2000). The collective construction of work group moods.Administrative Science Quarterly, 45, 197-231.

Bass, B. M., & Avolio, B. J. (1993). Transformational leadership: A response to critiques. InM. M. Chemers & R. Ayman (Eds.), Leadership theory and research: Perspectives anddirections (pp. 49-80). San Diego, CA: Academic.

Blau, P. M. (1977). Inequality and heterogeneity: A primitive theory of social structure. New York:Free Press.

Bliese, P. D. (2000). Within-group agreement, non-independence, and reliability: Implicationsfor data aggregation and analysis. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multileveltheory, research, and methods in organizations: Foundations, extensions and new direc-tions (pp. 348-381). San Francisco: Jossey-Bass.

Bliese, P. D., & Britt, T. W. (2001). Social support, group consensus and stressor-strain rela-tionships: Social context matters. Journal of Organizational Behavior, 22, 425-436.

Bliese, P. D., & Halverson, R. R. (1998). Group consensus and psychological well-being:A large field study. Journal of Applied Social Psychology, 28, 563-580.

Brown, R. (1988). Group processes: Dynamics within and between groups. Oxford, UK: BasilBlackwell.

Burke, M. J., Brief, A. P., George, J. M., Roberson, L., & Webster, J. (1989). Measuring affectat work: Confirmatory factor analyses of competing mood structures with conceptuallinkage to cortical regulatory systems. Journal of Personality and Social Psychology, 57,1091-1102.

Burkhardt, M. E. (1994). Social interaction effects following a technological change: A longi-tudinal investigation. Academy of Management Journal, 37, 869-898.

Byrne, D., Clore, G. L., & Worchel, P. (1966). Effect of economic similarity-dissimilarity oninterpersonal attraction. Journal of Personality and Social Psychology, 4, 220-224.

Campion, M. A., Medsker, G. J., & Higgs, A. C. (1993). Relations between work group char-acteristics and effectiveness: Implications for designing effective work groups. PersonnelPsychology, 46, 823-850.

Campion, M. A., Papper, E. M., & Medsker, G. J. (1996). Relations between work teamcharacteristics and effectiveness: A replication and extension. Personnel Psychology, 49,429-452.

Cannon-Bowers, J. A., Salas, E., & Converse, S. (1993). Shared mental models in expert teamdecision making. In J. N. John Castellan (Ed.), Individual and group decision making:Current issues (pp. 221-246). Hillsdale, NJ: Lawrence Erlbaum.


SGR288972.qxd 4/12/2006 4:48 PM Page 265

Carron, A. V., Brawley, L. R., Eys, M. A., Bray, S., Dorsch, K., Estabrooks, P., et al. (2003).Do individual perceptions of group cohesion reflect shared beliefs? An empirical analysis.Small Group Research, 34, 468-496.

Carron, A. V., Widmeyer, W. N., & Brawley, L. R. (1985). The development of an instrumentto assess cohesion in sport teams: The Group Environment Questionnaire. Journal of SportPsychology, 7, 244-266.

Cartwright, D., & Lippitt, R. (1957). Group dynamics and the individual. International Journalof Psychotherapy, 7, 86-102.

Chan, D. (1998). Functional relations among constructs in the same content domain at differ-ent levels of analysis: A typology of composition models. Journal of Applied Psychology,83, 234-246.

Chen, G., Mathieu, J. E., & Bliese, P. D. (in press). A framework for conducting multilevelconstruct validation. In F. J. Dansereau & F. Yamarino (Eds.), Research in multi-level issues:The many faces of multi-level issues (Vol. 3). Oxford, UK: Elsevier Science.

Cohen, J. (1988). Statistical power analyses for the behavioral sciences (2nd ed.). New York:Academic Press.

Cook, J. D., Hepworth, S. J., Wall, T. D., & Warr, P. B. (1981). The experience of work: A com-pendium and review of 249 measures and their use. London: Academic Press.

Erdfelder, E., Faul, F., & Buchner, A. (1996). GPOWER: A general power analysis program.Behavior Research Methods, Instruments, & Computers, 28, 1-11.

Feldman, D. C. (1984). The development and enforcement of group norms. Academy ofManagement Review, 9(1), 47-53.

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140.Festinger, L., Schachter, S., & Back, K. (1950). Social pressures in informal groups: A study

of human factors in housing. New York: Harper.George, J. M. (1990). Personality, affect, and behavior in groups. Journal of Applied Psychology,

75, 107-116.George, J. M., & Bettenhausen, K. (1990). Understanding prosocial behavior, sales performance,

and turnover: A group-level analysis in a service context. Journal of Applied Psychology,75, 698-709.

Georgopoulos, B. S. (1965). Normative structure variables and organizational behavior.Human Relations, 18, 155-169.

Gonzalez-Roma, V., Peiro, J. M., & Tordera, N. (2002). An examination of the antecedents andmoderator influences of climate strength. Journal of Applied Psychology, 87, 465-473.

Good, L. R., & Nelson, D. A. (1973). Effects of person-group and intragroup attitude simi-larity on perceived group attractiveness and cohesiveness. Psychological Reports, 33,551-560.

Griffin, R. W. (1983). Objective and social sources of information in task redesign: A fieldexperiment. Administrative Science Quarterly, 28, 184-200.

Gully, S. M., Incalcaterra, K. A., Joshi, A., & Beaubien, J. M. (2002). A meta-analysis of team-efficacy, potency, and performance: Interdependence and level of analysis as moderatorsof observed relationships. Journal of Applied Psychology, 87, 819-832.

Guzzo, R. A., Yost, P. R., Campbell, R. J., & Shea, G. P. (1993). Potency in groups:Articulating a construct. British Journal of Social Psychology, 32, 87-106.

Hackman, J. R. (1987). The design of work teams. In J. W. Lorsch (Ed.), Handbook of orga-nizational behavior (pp. 315-342). Englewood Cliff, NJ: Prentice-Hall.


SGR288972.qxd 4/12/2006 4:48 PM Page 266

Harrison, D. A., Price, K. H., & Bell, M. P. (1998). Beyond relational demography: Time and theeffects of surface- and deep-level diversity on work group cohesion. Academy of ManagementJournal, 41, 96-107.

Harrison, D. A., Price, K. H., Gavin, J. H., & Florey, A. T. (2002). Time, teams, and task perfor-mance: Changing effects of surface- and deep-level diversity on group functioning. Academyof Management Journal, 45, 1029-1045.

Hatfield, E., Cacioppo, J. T., & Rapson, R. L. (1994). Emotional contagion. Paris: CambridgeUniversity Press & Editions de la Maison des Sciences de l’Homme.

Henry, K. B., Arrow, H., & Carini, B. (1999). A tripartite model of group identification:Theory and measurement. Small Group Research, 30, 558-581.

Hogg, M. A., & Abrams, D. (1988). Social identifications. London: Routledge.Hogg, M. A., & McGarty, C. (1990). Self-categorization and social identity. In D. Abrams &

M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp. 10-27).Hemel Hempstead, UK: Harvester Wheatsheaf.

House, R., Rousseau, D. M., & Thomas-Hunt, M. (1995). The meso paradigm: A framework forthe integration of micro and macro organizational behavior. Research in OrganizationalBehavior, 17, 71-114.

Jackson, S. E., Brett, J. F., Sessa, V. I., Cooper, D. M., Julin, J. A., & Peyronnin, K. (1991). Somedifferences make a difference: Individual dissimilarity and group heterogeneity as correlatesof recruitment, promotions, and turnover. Journal of Applied Psychology, 76, 675-689.

James, L. R. (1982). Aggregation bias in estimates of perceptual agreement. Journal of AppliedPsychology, 67, 219-229.

James, L. R., Demaree, R. G., & Wolf, G. (1993). rwg: An assessment of within-group interrateragreement. Journal of Applied Psychology, 78, 306-309.

Jehn, K. A., Northcraft, G. B., & Neale, M. A. (1999). Why differences make a difference:A field study of diversity, conflict, and performance in workgroups. Administrative ScienceQuarterly, 44, 741-763.

Johns, G. (1994). Absenteeism estimates by employees and managers: Divergent perspectivesand self-serving perceptions. Journal of Applied Psychology, 79, 229-239.

Judge, T. A., Bono, J. E., Ilies, R., & Gerhardt, M. W. (2002). Personality and leadership:A qualitative and quantitative review. Journal of Applied Psychology, 87, 765-780.

Kelly, J. R. (2001). Mood and emotions in small groups and work teams. OrganizationalBehavior and Human Decision Processes, 86, 99-130.

Klein, K. J., Conn, A. B., Smith, D. B., & Sorra, J. S. (2001). Is everyone in agreement?An exploration of within-group agreement in employee perceptions of the work environ-ment. Journal of Applied Psychology, 86, 3-16.

Kozlowski, S. W. J., & Hattrup, K. (1992). A disagreement about within-group agreement: Disen-tangling issues of consistency versus consensus. Journal of Applied Psychology, 77, 161-167.

Kozlowski, S. W. J., & Klein, K. J. (2000). A multilevel approach to theory and research inorganizations: Contextual, temporal, and emergent processes. In K. J. Klein & S. W. J.Kozlowski (Eds.), Multilevel theory, research, and methods in organizations: Foundations,extensions, and new directions (pp. 3-90). San Francisco: Jossey-Bass.

Kramer, R. M., & Brewer, M. B. (1984). Effects of group identity on resource use in a simu-lated commons dilemma. Journal of Personality and Social Psychology, 46, 1044-1057.

Krebs, D. (1975). Empathy and altruism. Journal of Personality and Social Psychology, 32,1134-1146.


SGR288972.qxd 4/12/2006 4:48 PM Page 267

Lester, S. W., Meglino, B. M., & Korsgaard, M. A. (2002). The antecedents and consequencesof group potency: A longitudinal investigation of newly formed work groups. Academy ofManagement Journal, 45, 352-369.

Marks, M. A., Zaccaro, S. J., & Mathieu, J. E. (2000). Performance implications of leaderbriefings and team-interaction training for team adaptation to novel environments. Journalof Applied Psychology, 85, 971-986.

Mason, C. M., & Griffin, M. A. (2003a). Group absenteeism and positive affective tone: A lon-gitudinal study. Journal of Organizational Behavior, 24, 667-687.

Mason, C. M., & Griffin, M. A. (2003b). Identifying group task satisfaction at work. SmallGroup Research, 34, 413-442.

Mason, C. M., & Griffin, M. A. (2005). Group task satisfaction: The group’s shared attitude tothe task and work environment. Group and Organization Management, 30, 625-652.

Mathieu, J. E., Goodwin, G. F., Heffner, T. S., Salas, E., & Cannon-Bowers, J. A. (2000). Theinfluence of shared mental models on team process and performance. Journal of AppliedPsychology, 85, 273-283.

McLeod, P. L., & Lobel, S. A. (1996). Ethnic diversity and creativity in small groups. Small GroupResearch, 27, 248-270.

Miller, K. I., & Monge, P. R. (1985). Social information and employee anxiety about organi-zational change. Human Communication Research, 11, 365-386.

Milliken, F. J., & Martins, L. L. (1996). Searching for common threads: Understanding themultiple effects of diversity in organizational groups. Academy of Management Review,21, 402-433.

Myers, D. G., & Lamm, H. (1976). The group polarization phenomenon. Psychological Bulletin,83, 602-627.

Nunnally, J. C., & Berstein, I. H. (1994). Psychometric theory. New York: McGraw-Hill.O’Connor, E. J., & Barrett, G. V. (1980). Informational cues and individual differences as

determinants of subjective perceptions of task enrichment. Academy of ManagementJournal, 23, 697-716.

Pescosolido, A. T. (2003). Group efficacy and group effectiveness: The effects of groupefficacy over time on group performance and development. Small Group Research, 34,20-42.

Peterson, E., Mitchell, T. R., Thompson, L., & Burr, R. (2000). Collective efficacy and aspectsof shared mental models as predictors of performance over time in work groups. GroupProcesses and Intergroup Relations, 3, 296-316.

Pfeffer, J. (1980). A partial test of the social information processing model of job attitudes.Human Relations, 33, 457-476.

Postmes, T., Spears, R., & Lea, M. (2000). The formation of group norms in computer-mediatedcommunication. Human Communication Research, 26, 341-371.

Pritchard, R. D., & Karasick, B. W. (1973). The effects of organizational climate on manage-rial job performance and satisfaction. Organizational Behavior and Human Performance,9, 126-146.

Ruef, M., Aldrich, H. E., & Carter, N. M. (2003). The structure of founding teams: Homophily,strong ties, and isolation among U.S. entrepreneurs. American Sociological Review, 68,195-222.

Ryan, A. M., Schmit, M. J., & Johnson, R. (1996). Attitudes and effectiveness: Examiningrelations at an organizational level. Personnel Psychology, 49, 853-882.


SGR288972.qxd 4/12/2006 4:48 PM Page 268

Salancik, G. R., & Pfeffer, J. (1978). A social information processing approach to job attitudesand task design. Administrative Science Quarterly, 23, 224-253.

Salas, E., Dickinson, T. L., Converse, S. A., & Tannenbaum, S. I. (1992). Toward an under-standing of team performance and training. In R. W. Swezey & E. Salas (Eds.), Teams:Their training and performance (pp. 3-29). Norwood, NJ: Ablex.

Schnake, M. E., & Dumler, M. P. (1985). The social information processing model of taskdesign. Group and Organization Studies, 12, 221-240.

Schneider, B. (1987). The people make the place. Personnel Psychology, 40, 437-453.Schneider, B., Goldstein, H. W., & Smith, D. B. (1995). The ASA framework: An update.

Personnel Psychology, 48, 747-773.Schneider, B., Salvaggio, A. N., & Subirats, M. (2002). Climate strength: A new direction for

climate research. Journal of Applied Psychology, 87, 220-229.Schneider, B., Smith, D. B., Taylor, S., & Fleenor, J. (1998). Personality and organizations: A test

of the homogeneity of personality hypothesis. Journal of Applied Psychology, 83, 462-470.Seashore, S. F. (1954). Group cohesiveness in industrial work groups. Ann Arbor: Survey

Research Centre, University of Michigan.Shaw, J. B., & Barrett-Power, E. (1998). The effects of diversity on small work group processes

and performance. Human Relations, 51, 1307-1325.Sherif, M. (1936). The psychology of social norms. New York: Harper.Spector, P. E., & O’Connell, B. J. (1994). The contribution of personality traits, negative affec-

tivity, locus of control, and Type A to the subsequent reports of job stressors and jobstrains. Journal of Occupational and Organizational Psychology, 67, 1-11.

Staw, B. M., Bell, N. E., & Clausen, J. A. (1986). The dispositional approach to job attitudes:A lifetime longitudinal test. Administrative Science Quarterly, 31, 56-77.

Totterdell, P., Kellett, S., Teuchmann, K., & Briner, R. B. (1998). Evidence of mood linkagein work groups. Journal of Personality and Social Psychology, 74, 1504-1515.

Tsui, A. S., & O’Reilly, C. A. (1989). Beyond simple demographic effects: The importance ofrelational demography in superior-subordinates dyads. Academy of Management Journal,32, 402-429.

Turner, J. C. (1985). Social categorization and the self-concept: A social cognitive theory ofgroup behavior. In E. J. Lawler (Ed.), Advances in group processes: Theory and research(Vol. 2, pp. 77-122). Greenwich, CT: JAI Press.

Van de Ven, A. H., & Ferry, D. L. (1980). Measuring and assessing organizations. New York:John Wiley.

Walsh, J. P., Henderson, C. M., & Deighton, J. (1988). Negotiated belief structures and deci-sion performance: An empirical investigation. Organizational Behavior and HumanDecision Processes, 42, 194-216.

Weiss, D. J., Dawis, R. V., England, G. W., & Lofquist, L. H. (1967). Manual for the MinnesotaSatisfaction Questionnaire. Minneapolis, MI: Industrial Relations Center, University ofMinnesota.

Weiss, H. M., & Shaw, J. B. (1979). Social influences on judgements about tasks. OrganizationalBehavior and Human Performance, 24, 126-140.

Widmeyer, W. N., Brawley, L. R., & Carron, A. V. (1985). The measurement of cohesion insport teams: The Group Environment Questionnaire. London: Sports Dynamics.

Zalesny, M. D., & Farace, R. V. (1986). A field study of social information processing. HumanCommunication Research, 13, 268-290.


SGR288972.qxd 4/12/2006 4:48 PM Page 269

Zalesny, M. D., & Ford, J. K. (1990). Extending the social information processing perspective:New links to attitudes, behaviors and perceptions. Organizational Behavior and HumanDecision Processes, 47, 205-246.

Zenger, T. R., & Lawrence, B. S. (1989). Organizational demography: The differential effectsof age and tenure distributions on technical communication. Academy of ManagementJournal, 32, 353-376.

Claire M. Mason is a postdoctoral research fellow within the School of Management atQueensland University of Technology in Australia. Her research interests include group con-structs and leadership development.


SGR288972.qxd 4/12/2006 4:48 PM Page 270

Exploring the Processes Underlying Within-Group Homogeneity

Documents