17
Journal of Computer-Mediated Communication Two Routes Leading to Conformity Intention in Computer-Mediated Groups: Matching Versus Mismatching Virtual Representations Junghyun Kim doi:10.1111/j.1083-6101.2011.01539.x Given that using graphic self-representation (i.e., avatar) has become popular in diverse online services such as network games or virtual life communities (e.g., Second Life), it is important to investigate how visual self-representation can affect people’s behaviors or identities in cyberspace. Among many possible effects of visual cues in cyberspace, the current study focuses on the impact of sharing uniform appearance on virtual group identity. This study predicts that being represented by the same visual cue as others can increase group identity in a virtual group. Such identification with strangers is possible in virtual groups partially because virtual identities can be easily manipulated. Compared to offline settings where it is almost impossible to be disconnected from actual selves, it is much easier for people to be dis- connected from their physical appearances in cyberspace. And people do identify with and attach to their virtual representations that do not necessarily reflect their actual appearances. The current study pushes this supposition of the weak connection between actual self and virtual self-representation to the limit, and asks the following question: Can someone identify with his/her group members sharing the same visual cue as his/hers even when the shared visual cue does not match with his/her actual ethnicity? If answer is yes, it might not be impossible to envision that manipulated virtual self-representation can be used for reducing negative stereotypes existing among different racial groups by temporarily detaching people from their own ethnic group memberships and making them focus on their virtual identities. For instance, members of white supremacy group might be able to lessen their negative attitudes toward other ethnic group members by sharing the same visual cue with and interacting with them in cyberspace. This idea is in accordance with a group of early researchers who perceived cyberspace as a place for augmented equality and Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 271

Two Routes Leading to Conformity Intention in Computer-Mediated Groups: Matching Versus Mismatching Virtual Representations

Embed Size (px)

Citation preview

Journal of Computer-Mediated Communication

Two Routes Leading to Conformity Intentionin Computer-Mediated Groups: MatchingVersus Mismatching Virtual Representations

Junghyun Kim

doi:10.1111/j.1083-6101.2011.01539.x

Given that using graphic self-representation (i.e., avatar) has become popular indiverse online services such as network games or virtual life communities (e.g.,Second Life), it is important to investigate how visual self-representation can affectpeople’s behaviors or identities in cyberspace. Among many possible effects of visualcues in cyberspace, the current study focuses on the impact of sharing uniformappearance on virtual group identity. This study predicts that being represented bythe same visual cue as others can increase group identity in a virtual group. Suchidentification with strangers is possible in virtual groups partially because virtualidentities can be easily manipulated. Compared to offline settings where it is almostimpossible to be disconnected from actual selves, it is much easier for people to be dis-connected from their physical appearances in cyberspace. And people do identify withand attach to their virtual representations that do not necessarily reflect their actualappearances.

The current study pushes this supposition of the weak connection between actualself and virtual self-representation to the limit, and asks the following question:Can someone identify with his/her group members sharing the same visual cueas his/hers even when the shared visual cue does not match with his/her actualethnicity? If answer is yes, it might not be impossible to envision that manipulatedvirtual self-representation can be used for reducing negative stereotypes existingamong different racial groups by temporarily detaching people from their ownethnic group memberships and making them focus on their virtual identities. Forinstance, members of white supremacy group might be able to lessen their negativeattitudes toward other ethnic group members by sharing the same visual cue withand interacting with them in cyberspace. This idea is in accordance with a group ofearly researchers who perceived cyberspace as a place for augmented equality and

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 271

liberation from existing social boundaries (e.g., Dubrovsky, Kiesler, & Sethna, 1991;Kiesler, Siegel, & McGuire, 1984; Sproull & Kiesler, 1991). Thus, the current studyinvestigates whether virtual representations can be used for reducing social influencescoming from one of the social boundaries—ethnicity divisions. In order to achievethis goal, first of all, it starts from investigating whether uniform appearance, whetherthey are matching with people’s actual ethnicity or not, can induce strong groupidentity with others in cyberspace.

Uniform Appearance and Group IdentitySocial identity model of deindividuation effects (SIDE) has been used as one of thetheoretical frameworks suggesting that computer-mediated communication (CMC)accentuates social influences on individuals (Spears, Postmes, Lea, & Watt, 2001).This view is different from a group of earlier CMC studies suggesting that anonymityin CMC could reduce social influences (e.g., Dubrovsky et al., 1991; Kiesler et al.,1984; Sproull & Kiesler, 1991). According to SIDE, the lack of social cues in CMCmakes individuals pay less attention to others’ individuating information but focusmore on a common group membership. This phenomenon is called depersonalization(Reicher, Spears, & Postmes, 1995). The augmented importance of the common groupmembership increases group influence and leads to a higher level of conformity amonggroup members (Lea & Spears, 1991). A group of studies has provided empiricalsupport for SIDE showing that anonymity caused by deprived social cues in CMCactually increases conformity among members of computer-mediated groups (e.g.Postmes & Lea, 2000; Postmes & Spears, 1998; Postmes, Spears, & Cihangir, 2001;Postmes, Spears, & Lea, 1998, 1999; Postmes, Spears, Sakhel, & de Groot, 2001).

In most CMC studies, the effect of anonymity on group identity as well asconformity has been investigated by contrasting two conditions: a control conditionin which participants are physically colocated and an experimental condition whereparticipants are isolated and cannot see one another. That is, depersonalization inmost CMC studies has been induced by making experiment participants invisibleto one another. Recently, however, uniform appearance has been used as anothermeans to induce depersonalization in computer-mediated groups (e.g., Kim, 2009;Lee, 2004). Making people look similar to one another is not a new way to inducedepersonalization, but already has been used in face-to-face context studies (e.g.,Maslach, 1974; Zimbardo, 1969). Given that people feel depersonalized and canstrongly identify with others when they are wearing the same outfit in face-to-facecontexts, the current study speculates that uniform appearance might function in thesame way in cyberspace. Thus, sharing the same ‘Avatar’—an electronic image thatrepresents and can be modified by computer users (Merriam-Webster Dictionary,n.d.)—with others should decrease group members’ focus on one another’s individ-ualities, while increasing the level of group identity. On the other hand, if people arevisually distinguished by their unique avatars, they can preserve more distinctivenessthat hinders their identification with one another.

272 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

H1: Individuals who share the same visual cue with others will show greater group identitythan those who are represented by different visual cues from others.

Matching vs. Mismatching Virtual RepresentationsAs indicated by previous studies (e.g., Kim, 2009; Lee, 2004), people do identifywith others who share the same visual representation as theirs’ even when the visualrepresentation is randomly assigned to them. This might be partially based on thefact that the connection between an actual (physical) identity and a virtual identityis weak and the virtual identity can be easily manipulated. Knowing that they canpresent themselves in any way they want in cyberspace, individuals can develop anattachment toward assigned visual cues and even identify with others who are sharingthe same visual cue. Then is it possible for one to identify with others who share thesame virtual cue as his/hers even when the assigned cue conflicts with his/her actualdemographic background? According to Lee’s study (2007), people can identify withone another if they share the same gender in their virtual images, even when there wasa clear mismatch between their actual gender and the gender of their assigned virtualimages. For instance, a male participant who was assigned with a female characteridentified strongly with others who were also represented by female characters.

If individuals can identify with others in cyberspace just because they sharevisual cues of the same gender, sharing exactly the same visual cue might induce evenstronger group identity although ethnicity of the cue does not match with individuals’actual ethnicity. The current study investigates whether sharing the same visual cuemismatching with actual ethnicity induces as strong group identity as sharing thesame visual cue matching with actual ethnicity. Based on the malleability of identity incyberspace and previous findings that sharing the same visual cue accentuates groupidentity (e.g., Kim, 2009; Lee, 2004), the current study predicts that sharing uniformappearance mismatching with individuals’ actual ethnicity can induce group identityas strong as sharing uniform appearance matching with individuals’ actual ethnicity.

H2: Sharing uniform appearance that does not match with people’s actual ethnicity willincrease group identity as much as sharing uniform appearance that matches with their actualethnicity.

The Existence of an Out-GroupAnother factor that can increase group identity other than sharing uniform appear-ance is the existence of an out-group. To know that there is an out-group competingagainst in-group might contribute to the increase of group identity as well asconformity among in-group members (Brewer, 1979). The mere presence of anout-group, even without any interaction or competition between in-group and out-group, can increase in-group identity by enhancing the salience of categorization(Turner, 1987). Especially, a clear visual distinction between two groups can augmentin-group identity through increasing in-group members’ perception of a boundarybetween their group and the other group (Hensley & Duval, 1976). Earlier studies

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 273

have been consistently showing that in-group identification is highly associated within-group members’ perceived dissimilarity between their group and the other group(e.g., Wilson & Kayatani, 1968; Dion, 1973; Billing & Tajfel, 1973). These results areconsistent with social comparison theory (Turner, 1975), which indicates that theexistence of an out-group that is clearly different from in-group can induce strongin-group identity through a contrast effect. Furthermore, the intensified in-groupidentity increases individuals’ willingness to agree with their in-group members, assuggested by SIDE.

H3: Perceived visual distinctiveness between in-group members and out-group members willbe positively associated with individuals’ identification with in-group members.

H4: Stronger identification with in-group members will increase individuals’ willingness toagree with them.

Perceived Threat to Uniqueness & Need for UniquenessWhile uniform virtual appearance leads to stronger group identity, it also makespeople perceive that they are too similar to in-group members. According touniqueness theory (Snyder & Fromkin, 1980) and optimal distinctiveness theory(Brewer, 1991), there are two conflicting motivations that affect people’s conformitybehavior: the need to belong (Baumeister & Leary, 1995; Maslow, 1968) and theneed to be different (Snyder & Fromkin, 1980). The former refers to ‘‘a need forinclusion of the self into larger social collectives,’’ while the latter refers to ‘‘a needfor differentiation of the self from others’’ (Brewer, 1993, p.3). The need to belong isa primitive human need, so to be rejected or ostracized by others is one of the highlynegative experiences for human beings (Williams, 2001). At the same time, the sanctityof self and being true to self is a cultural norm in Western society (Baumeister, 1991).Thus, if people submerge their individualities in order to fulfill group responsibilitiesor maintain peace within their group, such behaviors have been considered asnegative elements to personal growths (Baumeister, 1991; Wallach & Wallach, 1983).

The optimal distinctiveness theory argues that the existence of such opposingforces creates tension in constructing social identities when individuals are locatedin a group. Because individuals always try to look for an optimal balance pointbetween these two needs, they are aware of and try to compensate for any externalfactor that disturbs the balance. For instance, extreme visual similarity shared amonggroup members might make them feel that their uniqueness as distinctive individualsdiminished or even threatened. On the other hand, if individuals are allowed torepresent themselves using unique visual cues, they might not perceive as muchthreat to their uniqueness as those who are sharing the same visual cue. Regardless ofwhether the shared visual cue matches with their actual ethnicity or not, people areexpected to perceive strong threat to their uniqueness if they share the same visualrepresentation with others.

H5: Individuals who share the same visual cue with others will perceive a higher degree ofthreat to their uniqueness than those who are represented by different visual cues from others.

274 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

H6: Sharing uniform appearance that does not match with their actual ethnicity will be asmuch threat to people’s uniqueness as sharing uniform appearance that matches with theiractual ethnicity.

People’s desires to maintain optimal levels for both similarity and distinctivenessmight not tolerate the state of diminished uniqueness. Thus, perceived threat touniqueness augmented by sharing the same visual cue with others triggers people’scraving for uniqueness. Such heightened need for uniqueness (NFU) (Fromkin &Snyder, 1980; Snyder, 1992) urges people to get satisfaction from being unique,special, and separable from the masses (Fromkin & Snyder, 1980). Therefore, NFUstirred up by sharing the same visual cue with others will lead people to actively seekways to restore their uniqueness, for example, by deviating from a dominant opinionshared by the rest of their group members (Kim, 2009; Maslach, 1974; Simonson &Nowlis, 2000).

H7: A strong threat to uniqueness intensified by sharing the same visual cue with others willmake individuals desire for more uniqueness.

H8: Stronger need for uniqueness will lead individuals to be less willing to agree with theirgroup members as a way to restore their uniqueness.

As indicated by theory of reasoned action (Ajzen & Fishbein, 1980), individuals’willingness to agree with others is expected to have a significant positive associationwith their actual conformity to a dominant opinion promoted by their in-groupmembers.

H9: Stronger willingness to agree with in-group members will lead to individuals’ actualconformity to a dominant opinion promoted by the in-group members.

Integrating Hypotheses 3, 4, 7, 8, and 9, the following path model is set forth:Finally, the current study investigates whether this hypothesized path model can

be applied to a condition where individuals are represented by virtual representations

Figure 1 A hypothesized path model with two routes leading to conformity intention andconformity

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 275

matching with their actual ethnicity and to another condition where individuals arerepresented by virtual representations not matching with their actual ethnicity. Basedon the proposition that sharing uniform appearance with others can have the sameeffect regardless of the distance between actual self and virtual self-representation, thisstudy expects that the hypothesized model will not show any significant differenceacross the two conditions—a condition where people’s actual ethnicity and virtualethnicity match, and the other condition where people’s actual ethnicity and virtualethnicity do not match (H10).

Research Methods

SampleThe sample included 258 undergraduate students from a variety of majors enrolledin a multisection course required as a part of the liberal education requirement alarge university in the Midwestern United States. Exactly half of the participants werefemale students. Racially, 87.5% of the sample was White, 8.9% of the sample wasBlack, and the rest 3.6% was Asian. Participants were asked to visit a website andparticipate in a series of online discussions during assigned time slots.

ProcedureOnce participants logged into the experiment website, they were asked to give theirethnicity information. Before they could see their own avatars and other discussants’avatars, they had to read three scenarios of dilemma situations and made theirdecisions (see Appendix). Then the participants were randomly assigned to one ofthe three conditions: 1) a condition in which discussants shared the same ethnicavatar, and their actual ethnicity matched with their virtual ethnic avatars (i.e.,matched-avatar condition), 2) another condition in which discussants shared thesame ethnic avatar, but their actual ethnicity mismatched with their virtual ethnicavatars (i.e., mismatched-avatar condition), and 3) a control condition in whichdiscussants were represented by diverse ethnic avatars that matched with their actualethnic backgrounds. Ninety-six participants were assigned to the matched-avatarcondition, 72 were assigned to the mismatched-avatar condition, and 90 wereassigned to the control condition.

In the matched-avatar condition and the mismatched-avatar condition, a par-ticipant could see that five other discussants and the participant him/herself werecategorized into two groups of three during discussion sessions—‘‘my group’’ and‘‘the other group.’’ ‘‘My group’’ composed of three discussants including the par-ticipant, and they shared the same ethnic avatar as that of the participant. ‘‘Theother group’’ consisted of the remaining three discussants who were sharing adifferent ethnic avatar from that of the members of ‘‘my group.’’ For example, aBlack female participant who was assigned to the matched-avatar condition wouldobserve that ‘‘my group’’ composed of three Black female avatars, including herself,

276 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Figure 2 A snapshot of the matched-avatar condition when a participant is a Black woman

while ‘‘the other group’’ composed of three White female avatars (Figure 2). In themismatched-avatar condition, on the other hand, the ethnic avatar representing ‘‘mygroup’’ members did not represent the participant’s actual ethnicity. Instead, ‘‘theother group’s’’ ethnic avatar matches with the participant’s actual ethnicity. So, ifa Black female participant was assigned to the mismatched-avatar condition, shewould be assigned with a White female avatar and see that ‘‘my group’’ is composedof three White female avatars including herself, while ‘‘the other group’’ is composedof three Black female avatars (Figure 3). Through such manipulation, it is possible toexamine whether the participant would identify with others sharing the same virtual

Figure 3 Snapshot of the mismatched-avatar condition when a participant is a Black woman

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 277

self-representation as his/hers, even with the presence of discussants represented bythe ethnic avatar that actually matched with the participant’s true ethnicity. In thecontrol condition, all six discussants were represented by diverse ethnic avatars andthey matched with participants’ actual ethnicity.

After looking at the way their in-group members were represented, participantswere asked to score how much they were able to identify with their in-group members(i.e., group identity). They were also asked how much they felt that their uniqueness asdistinctive individuals was diminished by being visually indistinguishable from theirin-group members (i.e., perceived threat to uniqueness). Both in the matched-avatarcondition and the mismatched-avatar condition, participants were asked how muchdifference they could perceive from the way ‘‘my group’’ members were representedand the way ‘‘the other group’’ members were represented (i.e., perceived in-groupand out-group distinction).

After seeing avatar representations, participants proceeded to real-time onlinediscussion sessions with other discussants. They had to read the same dilemmasituation scenarios presented previously and were asked to make their secondchoices. Although all participants were told that they were going to have real-timediscussions with other discussants, the other five discussants were programmed ones.Such control was necessary to create the same in-group consensus norm for allparticipants. That is, whatever initial decisions were made by participants, two othermembers of ‘‘my group’’ were programmed to pick the opposite choice, while threemembers of ‘‘the other group’’ were programmed to choose the same opinion asthe one chosen by the participants. After reading all five discussants’ opinions, theparticipants had to choose whether they would stick with their original decisionsor change them. If they would change their decisions closer toward the opinionpromoted by ‘‘my group’’ members, it means that the participants conformed to theconsensus norm of their in-group members. Once participants were done with threediscussions, they had to fill in a group of questionnaires.

MeasuresSince there has not been any previous study that specifically measured perceived threatto uniqueness, five newly created items were used (M = 4.85, SD = 1.77, α = .92).These five items were analyzed with a principle component factor analysis usingvarimax rotation and turned out to be unidimensional (see Table 1). Group identitywas measured by seven items from Cheney’s (1983) organizational identification scale(M = 5.17, SD = 1.07, α = .87). Perceived difference between in-group and out-groupmembers was measured by four original items (M = 5.54, SD = 1.07, α = 1.26).Another principle component factor analysis with varimax rotation was performedto see if these four items were unidimensional (Table 2). Conformity intention wasmeasured by six items asking participants’ willingness to agree with their owngroup members (i.e., ‘‘my group’’ members) for upcoming discussions (M = 4.18,SD = 1.26, α = .89). Need for uniqueness (NFU) was measured by 10 items from thescale developed by Snyder and Fromkin (1980) (M = 4.83, SD = .80, α = .77). All

278 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Table 1 Primary Factor Loadings of Perceived Threat to Uniqueness

Items Factor loading

I am represented as a distinctive individual in my group. (reversed) .922I can be easily distinguished from others in my group. (reversed) .914I am represented different from other members of my group. (reversed) .892My group members see me as a distinctive individual in this group. (reversed) .872My individuality is not represented well in my group. .764

Note: Eigen value = 3.83, explained variance = 76.52%

Table 2 Primary Factor Loadings of Perceived Difference Between In-Group and Out-GroupMembers

Items Factor loading

I can see the difference between my group members and other groupmembers.

.872

I find that my group members are different from other group members. .827I can clearly distinguish my group members from other group members. .822I feel closer to my group members than other group members. .678

Note: Eigen value = 2.58, explained variance = 64.48%

of these measures were measured by 7-point Likert-type scales (7 = Strongly Agree,1 = Strongly Disagree). Conformity was operationalized in the same way as used in agroup of previous studies (e.g., Kim, 2009; Lee, 2005). It was measured by the extentto which participants changed their initial decisions closer to the dominant opinionpromoted by their in-group members who were programmed to agree on the oppositeopinion to those of the participants. Participants had to choose one out of six possibleanswers when they made their decisions on dilemma situations: ‘‘Definitely shoulddo A,’’ ‘‘Should do A,’’ ‘‘Probably should do A,’’ ‘‘Probably should do B,’’ ‘‘Shoulddo B,’’ and ‘‘Definitely should do B.’’ Conformity score is the size of the gap betweenthe participant’s first answer and his/her final answer. For instance, if a participantchose ‘‘Definitely should do A’’ in the first choice and changed his/her answer into‘‘Probably should do B’’ after seeing his/her in-group members support option B,3 points were given (because three intervals existed between these two replies). Inthe same way, if another participant changed his/her answer from ‘‘Should do B’’ to‘‘Probably should do A’’, 2 points were given. These two examples indicate that bothparticipants conformed to ‘‘my group’’ members’ conformity pressure by switchingtheir initial choices into ones closer to their in-group members’. Meanwhile, ifparticipants stuck with the same decision as their initial ones or even moved furtheraway from the one promoted by their in-group members, 0 point was given. All ofthe scores collected from the three discussions were summed and used as conformityscores (M = 1.58, SD = 1.92).

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 279

Results

Hypothesis 1 predicted that participants who were sharing the same avatar withothers, whether the avatar matched with their actual ethnicity or not, would showstronger group identity than those who were represented by different avatars. Ona related note, Hypothesis 2 indicated that participants in the mismatched-avatarcondition would identify with their in-group members as strongly as those inthe matched-avatar condition. In order to test these two hypotheses, a one-wayANOVA was performed with two planned sets of contrasts: [+1 (the matched-avatarcondition), +1 (the mismatched-avatar condition), −2 (the control condition)] and[+1 (the matched-avatar condition), −1 (the mismatched-avatar condition), 0 (thecontrol condition)]. There was a significant difference in group identity scores acrossthe three avatar conditions [F(2, 256) = 12.77, p < .001, η2 = .09]. ANOVA resultwith the first set of contrast (+1, +1, −2) indicated that participants in either thematched-avatar condition or the mismatched-avatar condition showed a significantlyhigher group identity (M = 5.17, SD = 1.07) than those in the control condition(M = 4.52, SD = .82) [t (256) = 5.04, p < .001]. Application of the second set ofcontrast (+1, −1, 0) revealed that there was no significant difference in group identitybetween the matched-avatar condition (M = 5.17, SD = 1.07) and the mismatched-avatar condition (M = 5.17, SD = 1.07) [t (256) = .003, p = .997]. Thus, bothHypotheses 1 and 2 were supported.

The same sets of contrasts were applied to test Hypothesis 5 and Hypothesis6. Hypothesis 5 predicted that participants who were sharing the same avatar withtheir in-group members, whether the avatar matched with their actual ethnicityor not, would perceive a stronger threat to their uniqueness than those who wererepresented by different avatars. Application of the first set of contrast (+1, +1,−2) revealed that participants from either the matched-avatar condition or themismatched-avatar condition showed a significantly higher perceived threat touniqueness (M = 4.85, SD = 1.77) than those in the control condition (M = 2.82,SD = .80) [t(256) = 10.36, p < .001]. Meanwhile, Hypothesis 6 predicted thatparticipants in the mismatched-avatar condition would perceive as much threat totheir uniqueness as those in the matched-avatar condition. Application of the secondset of contrast (+1, −1, 0) showed that there was no significant difference in perceivedthreat to uniqueness between the matched-avatar condition (M = 4.82, SD = 1.77)and the mismatched-avatar condition (M = 4.88, SD = 1.77) [t (256) = .25, p =.81]. Thus, both Hypotheses 5 and 6 were supported.

In order to test Hypotheses 3, 4, 7, 8, and 9, the hypothesized path model(Figure 1) was tested with AMOS 6.0 (Arbuckle, 2005). The model showed a goodfit with the data χ2 (9) = 12.55, p = .18, CFI = .96, RMSEA = .05 (Figure 4).As suggested by Hypothesis 3, a high perceived distinctiveness between in-groupmembers and out-group members increased participants’ identification with theirin-group members (β = .28, p < .01). As indicated by SIDE, the intensified groupidentity led to a high level of willingness to agree with participants’ in-group members

280 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Figure 4 The path analysis results for the hypothesized model∗p > .05, ∗∗p > .01

(β = .21, p < .01). Thus, Hypothesis 4 was supported. As predicted by Hypothesis7, a stronger perceived threat to uniqueness led participants to desire for moreuniqueness (β = .27, p < .01), and reduced willingness to agree with their in-groupmembers (β = −.24, p < .01). There was also a significant negative associationbetween NFU and conformity intention (β = −.36, p < .01). Conformity intentionhad a significant positive association with conformity (β = .24, p < .01).

Finally, a multigroup analysis was performed to test Hypothesis 10. Multigroupanalysis is an analysis to check whether the same path model can be applied acrossdifferent data sets. The multigroup analysis was restricted to the matched-avatarcondition and the mismatched-avatar condition, because the current study wasinterested in investigating whether any path in the hypothesized model would besignificantly different between these two conditions. In order to figure out whichspecific path was significantly different between these two conditions, each causalpath was constrained once at a time, and contrasted against the model without anyconstraint. The χ2 of each model with one path coefficient constrained to equalitywas contrasted against that of the unconstrained model. If any of the χ2 of the modelwith one constrained path became significantly worse than that of the unconstrainedmodel, it indicates that the constrained path coefficient is significantly different acrossthe two conditions (Kline, 1998). The multigroup analysis result indicated that therewas no significant difference in any path of the hypothesized model between thematched-avatar condition and the mismatched-avatar condition. In other words,the same model can be applied to both conditions. The path coefficients from thematched-avatar condition and from the mismatched-avatar condition are shown inTable 3.

Discussion

The current study tried to expand CMC research, first of all, by investigating therole of visual similarity shared among group members as another way to induce

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 281

Table 3 Multigroup Analysis Results of the Matched-Avatar Condition and theMismatched-Avatar Condition

Standardized β

Path Matched Mismatched

Perceived difference between in-group andout-group → Group identity

.21∗ .39∗∗

Group identity → Conformity intention .22∗ .18Perceived threat to uniqueness → Need for uniqueness .23∗ .32∗∗

Need for uniqueness → Conformity intention – .42∗∗ – .27∗

Perceived threat to uniqueness → Conformity intention – .23∗∗ – .29∗∗

Conformity intention → Conformity .21∗ .30∗∗

Note: None of the standardized path coefficients is significantly different between the matched-avatar condition and the mismatched-avatar condition∗p < .05, ∗∗p < .01

depersonalization other than removing individuating cues—a dominant way toinduce depersonalization in most CMC research. Consistent with a previous study(Kim, 2009), individuals who shared the same visual cue with their group membersidentified more strongly with their group members and perceived a higher level ofthreat to their uniqueness, compared to those who were represented by unique visualcues in their group. Founded on these findings, the current study came up with ahypothesized path model indicating that people are affected by two opposing factorswhen making decisions: Group identity augmented by uniform appearance increasesconformity intention, while perceived threat to uniqueness induced also by uniformappearance interferes with their conformity intention (Kim, 2009).

The current study built up a hypothesized path model with two different routesconnecting uniform appearance and conformity intention. On the route throughgroup identity perceived distinctiveness between in-group and out-group was addedas an element augmenting people’s identification with their in-group members. Thisresult replicates previous studies (e.g., Wilson & Kayatani, 1968; Dion, 1973; Billing& Tajfel, 1973) based on social comparison theory (Turner, 1975). It also attests tothe strength of simple categorization between in-group and out-group, which canincrease in-group identity even without any intergroup interaction or competition.

Meanwhile, on the route through perceived threat to uniqueness, NFU was addedas a factor connecting perceived threat to uniqueness and conformity intention. Aprevious study has suggested that too much visual similarity could make people cravefor uniqueness as a backlash to their diminished uniqueness (Lee, 2004). In orderto examine whether perceived threat to uniqueness could provoke people’s desireto be distinctive from others, the current study measured NFU right after groupdiscussions. As expected, the data analysis results showed that perceived threat touniqueness led to NFU, which eventually hindered participants’ willingness to agreewith their group members.

282 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Given that the connection between actual and virtual identities is weak, the currentstudy examined whether sharing uniform appearance mismatching with actualethnicity have the same effect on group identity and perceived threat to uniquenessas that of uniform appearance matching with their actual ethnicity. Interestinglyenough, there was no significant difference in group identity and perceived threatto uniqueness between those two conditions. This finding indicates the robustnessof sharing uniform visual cue in cyberspace, even to the degree that the mismatchbetween actual ethnicity and virtual images’ ethnicity does not disturb the virtualgroup identification process.

Moreover, the current study investigated the effect of sharing uniform appearanceon group identity even in a more conservative way. The fact that participants in themismatched-avatar condition identified strongly with their in-group members, evenwith the presence of out-group members who were represented by an avatar of theparticipants’ actual ethnicity, indicates that sharing a randomly assigned uniformvirtual cue can overpower people’s actual ethnicity group memberships. This resultsuggests that people can identify with others regardless of their sociodemographic orcultural differences, as long as they are assigned with the same virtual cue.

Sharing the same visual cue would help people identify with and even understandone another better, but might not always increase assimilation in virtual groups.This is because a high level of visual similarity makes people feel their uniquenessthreatened, which leads them to be less willing to conform to others. This indicatesthat people can be influenced by the way they are represented in cyberspace otherthan their own personal opinions when they make decisions. For instance, considera case in which members of a virtual community must make a decision and they arerepresented by the same avatar in an effort to increase group identity. An individualmember might feel compelled to offer a divergent opinion as an expression ofhis/her uniqueness, rather than a representation of his/her actual feelings. A similarphenomenon can be found in text-based online groups. Most text-based onlinecommunities present members in standardized formats that do not allow muchroom for individuation, so members might feel that their uniqueness as distinctiveindividuals is suppressed. Thus, we have to be cautious in interpreting conformityor dissent in virtual communities, since visual representations might affect people’sdecisions more than we think.

In summary, the current study has shown that uniform virtual appearanceaccentuates individuals’ identification with group members as well as perceivedthreat to uniqueness. These two factors inconsistently affect individuals’ willingnessto agree with their in-group members. Perceiving clear visual difference betweenin-group and out-group helps individuals identify with their in-group members, andmakes them willing to agree with them. Meanwhile, perceived threat to uniquenessarouses NFU, which drives individuals to disagree with their in-group membersas a way to restore their uniqueness. Moreover, this study has demonstrated therobustness of sharing the same visual cue in cyberspace surpassing individuals’ socialor cultural differences by showing that 1) the effect of sharing uniform appearance on

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 283

group identity as well as on perceived threat to uniqueness was constant regardlessof its matching or mismatching with individuals’ actual ethnicity, and 2) the samehypothesized path model was applicable to both the matched-avatar condition andthe mismatched-avatar condition.

Limitations & Future ResearchOne major limitation of the current study pertains to unnatural environments foronline discussions created by 1) assigning avatars to participants without giving themany choice, and 2) using programmed discussants instead of actual ones. Althoughthese controls might not have created natural settings similar to participants’ dailyonline interaction environments, assigning avatars to individuals was necessaryto make sure that they were either represented by avatars matching with theiractual ethnicity or mismatching with their actual ethnicity. They were also requiredto create the same conformity pressure coming from the participants’ in-groupmembers’ agreeing on an opinion that was always opposite to the one chosen byparticipants.

One way to make experimental settings more natural is to let actual participantstake part in discussions and ask to reach a decision within limited time. This mightallow more time and interactions for participants to naturally affiliate with theirown avatars and to recognize their similarity to (i.e. the same-avatar condition)or difference from (i.e. the different-avatar condition) avatars of other discussants.Considering that there are many virtual communities that are using only textualidentities, to design an experiment with three different conditions - a group in whichmembers are represented by uniform appearance, a group in which members arerepresented by whatever visual cues they choose, and a group in which members arerepresented by textual identities - might be useful. Adding a text-only condition willallow us to contrast the impacts of uniform graphic representation and standardizedtextual representation on group identity. Finally, provided that there are many virtualcommunities in which users’ real identities are available (e.g., WELL), it would beinteresting to contrast group identification processes of perfectly anonymous onlinegroups and those of identifiable online groups.

References

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior.Englewood Cliffs, NJ: Prentice-Hall.

Arbuckle, J. L. (2005). Amos 6.0. Spring House, PA: Amos Development Corporation.Billing, M., & Tajfel, H. (1973). Social categorization and similarity in intergroup behavior.

European Journal of Social Psychology, 3, 27–52.Baumeister, R. F. (1991). Meanings of life. New York: Guilford.Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal

attachments as a fundamental human motivation. Psychological Bulletin, 117, 497–529.Brewer, M. B. (1979). In-group bias in the minimal intergroup situation: A cognitive-

motivational analysis. Psychological Bulletin, 86, 307–324.

284 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Brewer, M. B. (1991). The social self: On being the same and different at the same time.Personality and Social Psychology Bulletin, 17, 475–482.

Brewer, M. B. (1993). The role of distinctiveness in social identity and group behavior. InM. Hogg & D. Abrams (Eds.), Group motivation: Social psychological perspectives(pp. 1–16). New York: Harvester-Wheatsheaf.

Cheney, G. (1983). On the various and changing meanings of organizational membership:Field study of organizational identification. Communication Monographs, 50, 342–362.

Dion, K. L. (1973). Cohesiveness as a determinant of ingroup-outgroup bias. Journal ofPersonality and Social Psychology, 28, 163–171.

Dubrovsky, V. J., Kiesler, S., & Sethna, B. N. (1991). The equalization phenomenon: Statuseffects in computer-mediated and face-to-face decision-making groups. HumanComputer Interaction, 6, 119–146.

Hensley, V., & Duval, S. (1976). Some perceptual determinants of perceived similarity, liking,and correctness. Journal of Personality and Social Psychology, 34, 159–168.

Kiesler, S., Siegel, J., & McGuire, T. (1984). Social psychological aspects ofcomputer-mediated communication. American Psychologist, 39, 1123–1134.

Kim, J-H. (2009). ‘‘I want to be different from others’’: The role of visual similarity in virtualgroup identity. Computers in Human Behavior, 25, 88–95.

Kline, R. B. (1998). Principles and practice of structural equation modeling. New York:Guilford.

Lea, M., & Spears, R. (1991). Computer-mediated communication: Deindividuation andgroup decision-making. International Journal of Man-Machine Studies, 34, 283–301.

Lee, E. (2004). Effects of visual representation on social influence in computer-mediatedcommunication: Experimental tests of the social identity model of deindividuationeffects. Human Communication Research, 30, 234–259.

Lee, E. (2005). When and how does depersonalization increase conformity to group norms incomputer-mediated communication? Communication Research, 33, 423–447.

Lee, E. (2007). Deindividuation effects on group polarization in computer-mediatedcommunication: The role of group identification, public-self-awareness, and perceivedargument quality. Journal of Communication, 57, 385–403.

Maslach, C. (1974). Social and personal bases of individuation. Journal of Personality andSocial Psychology, 29, 411–425.

Maslow, A. H. (1968). Toward a psychology of being. New York: Van Nostrand.Merriam-Webster’s dictionary. (n.d.). Retrieved November 15, 2007, fromhttp://www.merriam-webster.com/dictionary/avatar

Postmes, T., & Lea, M. (2000). Social processes and group decision-making: Anonymity ingroup decision support systems. Ergonomics, 43, 1252–1274.

Postmes, T., & Spears, R. (1998). Deindividuation and anitnormative behavior: Ameta-analysis. Psychological Bulletin, 123, 238–259.

Postmes, T., Spears, R., & Cihangir, S. (2001). Quality of decision-making and group norms.Journal of Personality and Social Psychology, 80, 918–930.

Postmes, T., Spears, R., & Lea, M. (1998). Breaching or building social boundaries?:SIDE-effects of computer-mediated communication. Communication Research, 25,689–715.

Postmes, T., Spears, R., & Lea, M. (1999). Social identity, normative content and‘‘deindividuation’’ in computer-mediated groups. In N. Ellemers, R. Spears, & B. Doosje(Eds.), Social identity: Context, commitment, content (pp. 174–183). Oxford: Blackwell.

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 285

Postmes, T., Spears, R., Sakhel, K., & de Groot, D. (2001). Social influence incomputer-mediated communication: The effects of anonymity on group behavior.Personality and Social Psychology Bulletin, 27, 1243–1254.

Reicher, S., Spears, R., & Postmes, T. (1995). A social identity model of deindividuationphenomena. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology(pp. 161–198). Chichester: Wiley.

Simonson, I., & Nowlis, S. M. (2000). The role of explanations and need for uniqueness inconsumer decision making: Unconventional choices based on reasons. Journal ofConsumer Research, 27, 49–68.

Snyder, C. R., & Fromkin, H. L. (1980). Uniqueness: The human pursuit of difference. NewYork: Plenum.

Spears, R., Postmes, T., Lea, M., & Watt, S. E. (2001). A SIDE view of social influence. InJ. P. Forgar & K. D. Williams (Eds.), Social influence: Direct and indirect processes. TheSydney Symposium on Social Psychology Sseries (Vol. 3, pp. 331–350). Philadelphia:Cambridge University Press.

Sproull, L. & Kiesler, S. (1991). Connections: New ways of working in the networkedorganization. Cambridge, MA: MIT Press.

Turner, J. C. (1975). Social comparison and social identity: some prospects for intergroupbehavior. European Journal of Social Psychology, 5, 5–34.

Turner, J. C. (1987). A self-categorization theory. In J. C. Turner, M. A. Hogg, P. J. Oakes,S. D. Reicher, & M. S. Wetherell (Eds.), Rediscovering the social group: A self-categorizationtheory (pp. 42–67). Oxford, England: Blackwell.

Wallach, M. A., & Wallach, L. (1983). Psychology’s sanction for selfishness: The error of egoismin theory and therapy. San Francisco: Freeman.

Williams, K. D. (2001). Ostracism: The power of silence. New York: Guilford.Wilson, W., & Kayatani, M. (1968). Intergroup attitudes and strategies in games between

opponents of the same or of a different race. Journal of Personality and Social Psychology,9, 24–30.

Zimbardo, P. (1969). The human choice: Individuation, reason, and order versusdeindividuation, impulse, and chaos. In W. Arnold & D. Levine (Eds.), Nebraskasymposium on motivation (pp. 237–307). Lincoln: University of Nebraska Press.

Appendix

Scenario 1‘‘Ms. E, a college senior, has studied the piano since childhood. She has won amateurprizes and given small recitals, suggesting that she has considerable musical talent.As graduation approaches, she has the choice of taking a medical school scholarshipto become a physician, a profession which would bring certain financial rewards, orentering a conservatory of music for advanced training with a well-known pianist.She realizes that even upon completion of her piano studies, success as a concertpianist would not be assured.’’

Option A: She should study music.Option B: She should go to medical school.

286 Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association

Scenario 2‘‘Mr. G is a surgeon with a well-established surgical practice. He is married and hasthree children, one of which is just starting college. During a backyard session offootball, he seriously dislocated his shoulder. Although the shoulder was properlyreset at the time, the dislocation produced some nerve damage and he has beenexperiencing a great deal of pain ever since. An operation is available that will relievethe pain if completely successful, but the operation also poses a risk of producing apermanent reduction in manual dexterity. The decrement in dexterity is normallyinconsequential, but in his case, it could prevent him from continuing his surgicalpractice.’’

Option A: Receive the shoulder surgery.Option B: Do not receive the shoulder surgery.

Scenario 3‘‘Ms. K is a successful businesswoman who has participated in a number of civicactivities of considerable value to the community. She has been approached by theleaders of her political party as a possible congressional candidate in the next election.Her party is a minority party in the district, though the party has won occasionalelections in the past. She would like to hold political office, but to do so would involvea serious financial sacrifice, since the party has insufficient campaign funds. Shewould also have to endure the attacks of her political opponents in a hot campaign.’’

Option A: She should run for the election.Option B: She should not run for the election.

Journal of Computer-Mediated Communication 16 (2011) 271–287 © 2011 International Communication Association 287