Attitude Change in Computer-Mediated Communication: Effects of Anonymity and Category Norms

http://gpi.sagepub.com/Relations

Group Processes & Intergroup

http://gpi.sagepub.com/content/6/4/405The online version of this article can be found at:

DOI: 10.1177/13684302030064006

2003 6: 405Group Processes Intergroup RelationsKai Sassenberg and Margarete Boos

Category NormsAttitude Change in Computer-Mediated Communication: Effects of Anonymity and

Published by:

http://www.sagepublications.com

can be found at:Group Processes & Intergroup RelationsAdditional services and information for

http://gpi.sagepub.com/cgi/alertsEmail Alerts:

http://gpi.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://gpi.sagepub.com/content/6/4/405.refs.htmlCitations:

What is This?

- Oct 1, 2003Version of Record >>

at KANSAS STATE UNIV LIBRARIES on July 3, 2014gpi.sagepub.comDownloaded from at KANSAS STATE UNIV LIBRARIES on July 3, 2014gpi.sagepub.comDownloaded from

http://gpi.sagepub.com/

http://gpi.sagepub.com/content/6/4/405

http://www.sagepublications.com

http://gpi.sagepub.com/cgi/alerts

http://gpi.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://gpi.sagepub.com/content/6/4/405.refs.html

http://gpi.sagepub.com/content/6/4/405.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml



Attitude Change inComputer-MediatedCommunication: Effectsof Anonymity andCategory Norms

Kai SassenbergFriedrich-Schiller University of Jena

Margarete BoosGeorg-August University of Göttingen

The current research compared the effect of computer-mediated communication (CMC) anddirect communication on attitude change. The social identity model of deindividuation effects(Spears & Lea, 1994) predicts that CMC results in behavior that is more in line with the salientlevel of self-categorization (compared to non-anonymous communication): in CMC salientsocial identity should lead to conformity to group norms whereas salient personal identity wasexpected to result in behavior that fits individual goals. Two experiments showed that whenpersonal identity was salient and when social identity was salient and a category norm wasexplicitly given, CMC led to the predicted effects, whereas the lack of a social category norm ledto lower attitude change in CMC compared to direct communication.

keywords anonymity, computer-mediated communication, group norms, socialinfluence

Group Processes &Intergroup Relations

2003 Vol 6(4) 405–422

WI T H T H E growing number of Internet users anew media for interpersonal and mass com-munication becomes more and more a part ofeveryday life. Computer-mediated communi-cation (CMC) has various effects on communi-cation itself as well as on its outcomes (for asummary see McKenna & Bargh, 2000).Because research on CMC is still comparativelyinfrequent, predictions of the impact of CMChave only been tested in some fields. Studiescomparing attitude change within groups usingface to face (FTF) and CMC produced

inconsistent results: in some studies attitudechange in FTF groups was higher than in groups

GPIR

Copyright © 2003 SAGE Publications(London, Thousand Oaks, CA and New Delhi)

[1368-4302(200310)6:4; 405–422; 035544]

Author’s noteThe studies reported in this article wereconducted while Kai Sassenberg was still at theGeorg-August University.Address correspondence to Kai Sassenberg,Department of Social Psychology,Humboldtstrasse 26, D-07743 Jena, Germany,[email: [email protected]]

06 GPI 6-4 sassenberg (ds) 18/9/03 3:41 pm Page 405

at KANSAS STATE UNIV LIBRARIES on July 3, 2014gpi.sagepub.comDownloaded from

www.sagepublications.com


using CMC (e.g. McGuire, Kiesler, & Siegel,1987). Conversely, in other studies the reversedpattern of results was found (e.g. Kiesler, Siegel,& McGuire, 1984). According to the SocialIdentity Model of Deindividuation Effects(SIDE; Reicher, Spears, & Postmes, 1995; Spears& Lea, 1994) these findings (a) can beexplained as an effect of higher anonymity inCMC compared to FTF communication and (b)this effect of anonymity is moderated by thelevel of salient identity. When social identity issalient, anonymity fosters social influence; thus,attitude change is expected to be higher inCMC compared to FTF communication.However, when personal identity is salient, ano-nymity hinders social influence within groups.Therefore, with personal identity being salientone would expect less attitude change in FTFthan in CMC. The aim of the current researchis to test this explanation of the inconsistentfindings with studies comparing attitude changein groups employing FTF and CMC. At the sametime, the current research examines whetherthe SIDE’s predictions about the impact of ano-nymity can be generalized to the comparisonbetween FTF and CMC.

Social influence in CMC

One of the first models developed to explainthe effects of CMC was the reduced social cuesapproach (RSC; Kiesler et al., 1984; Kiesler &Sproull, 1992) which addresses the effects ofcommunication media on social influencewithin groups. The central assumption of theRSC approach is that in groups using CMC lesssocial contextual information is available andtherefore behavior is less strongly orientedtoward societal norms than in FTF groups.Moreover, Kiesler et al. (1984) argue that CMCembodies some conditions that are also part ofthe deindividuation concept: anonymity, lowerself-awareness, and reduced self-regulation. In astate of deindividuation people show moreextreme behavior (i.e. behavior deviating fromnorms) after being submerged in a crowd (Fes-tinger, Pepitone, & Newcomb, 1952). Thus, thecentral assumptions of the RSC approachparallel deindividuation theory (Diener, 1980):

fewer social contextual cues as well as ano-nymity in crowds lower self-awareness. Result-ing from reduced self-awareness, behavior ispredicted to become deregulated, moreextreme, and less guided by societal norms.Less inhibited behavior leads, according to theassumptions of Kiesler et al. (1984), to moreextreme group decisions and a stronger shift ofgroup members’ attitudes during a discussionin CMC: ‘If [. . .] extreme opinions are lesslikely to be withheld (because behavior is lessinhibited), then we would predict more choiceshift in computer-mediated groups’ (Kiesler etal., 1984, p. 1128). Thus, more polarized atti-tudes become more likely after a computer-mediated than after an FTF discussion. Thisbrief summary does not capture the wholecomplexity of the RSC approach. However, itmight give an impression of the numerousconcepts that were taken into account (for amore extensive summary and critique of theRSC approach see Lea & Spears, 1991; Spears &Lea, 1992).

Studies testing the RSC approach (Kiesler etal., 1984; McGuire et al., 1987; Siegel,Dubrovsky, Kiesler, & McGuire, 1986;Weisband, 1992) usually compared groupscommunicating FTF with groups using CMC.The dependent measure was the attitudechange from the individual decision before thegroup discussion either to the group decisionor to the individual decision after the dis-cussion. The prediction that more attitudechange would occur in CMC was not alwaysconfirmed. McGuire et al. (1987) found lessattitude change and less extreme decisions inCMC groups than in FTF groups and Weisband(1992) observed no differences. However,Siegel et al. (1986) and Kiesler et al. (1984)found larger choice shifts in computer-mediated groups than in FTF groups. Addition-ally, the RSC approach is questioned by severalstudies demonstrating that self-awareness is notlower in CMC, as expected from the model, buthigher than in direct communication(Matheson & Zanna, 1988, 1989; Sassenberg,Rabung, & Boos, 1999). Taken together, theRSC approach is not able to explain these con-tradictory results. Hence, there is a need for a

Group Processes & Intergroup Relations 6(4)

406




theoretical approach that can predict attitudechange in different communication media.

A comparatively simple explanation for thecontradictory findings concerning the effectsof media type on attitude change has beenprovided by SIDE (Spears & Lea, 1992). SIDEbuilds on social identity theory (SIT; Tajfel &Turner, 1979) and self-categorization theory (SCT;Turner, Hogg, Oakes, Reicher, & Wetherell,1987). Both theories suggest that the self-concept is made up by one’s individuality aswell as by one’s group memberships (e.g.gender, school, sports team). The relevance ofthese different aspects of the self depends onthe respective social context. When personalidentity is salient (i.e. self perception is domi-nated by individual characteristics) thebehavior will be guided by personal goals, suchas the need to show individual distinctiveness.However, when a social identity is salient (i.e.self-perception is dominated by the member-ship of a group) the behavior will be guided bythe norms of the respective group. The dualityof behavioral consequences as a function of thelevel of salient identity may also be applied tosocial influence within groups. When socialidentity is salient attitude change after a groupdiscussion is expected to be oriented towardthe group norm. In ad hoc groups the normcorresponds to the prototype of the group,defined as attitude that differs least from theingroup members’ attitudes and the most fromthe outgroup members’ attitudes. In case thecontext does not allow a comparison to anoutgroup (e.g. when an interacting group ismaking a decision in the absence of any othergroup) the range of possible attitudes (i.e. thescale positions which are not held by ingroupmembers) replaces the outgroup as compari-son standard (Turner, 1991). When personalidentity is salient, predictions for attitudechange are less clear: the need for individualdistinctiveness can either be fulfilled by shiftingone’s attitude away from others’ attitudes (i.e.the group norm) or by showing an attitude thatpromotes one’s personal identity (e.g. whenchanging one’s expectations or values leads tosuccess in meeting one’s own standards andmakes one feel better). In sum, SCT explains

group polarization neither by exchange ofarguments (Vinokur & Burnstein, 1974), nor bynormative pressure (Goethals & Zanna, 1979),but instead as an effect of the self-definition asa group member which in turn leads to theintrinsic adherence to the group norm. Thisnorm is either learned in earlier interactionsor, in the case of an interacting group, it is thegroup prototype which is inferred from theindividual attitudes.

Taken together, SIT and SCT provide aframework for the explanation of social influ-ence (beside other group phenomena) andallow predictions of attitude change in twodirections. When social identity is salient atti-tudes should shift toward the group norm (i.e.the given tendency within a group). Conversely,when personal identity is salient attitudesshould shift away from the group norm(allowing individuals to display individual dis-tinctiveness from group norms) or toward aself-serving position. SIDE extends SIT andSCT with respect to the effect of anonymity ongroup processes. Reicher (1984, 1987) firstproposed an explanation for the effect of ano-nymity on behavior in crowds based onSIT/SCT. He suggested that people in anony-mous crowds have an enhanced social identity.Therefore, in such situations people act closelyin line with the norm of their own group. Itshould be recognized that this norm differsfrom general societal norms addressed inearlier deindividuation theory (Diener, 1980).Extending this assumption, SIDE (Reicher etal., 1995) provided a more detailed analysis ofthe effects of anonymity. When social identity issalient, anonymity leads to a homogeneousperception of the ingroup, as anonymityhinders the perception of possible intragroupdifferences. In anonymous groups the socialcategorization becomes even more salient andgroup members show higher conformity to theingroup norm. When personal identity is salient,anonymity limits information concerninginterindividual similarities. Hence, there is nobasis for the development of a common socialidentity. As a consequence, behavior promotingthe uniqueness of the person as an individualwill be more likely in an anonymous as opposed

Sassenberg & Boos attitude change in anonymous groups

407




to a non-anonymous group. Several findingssupport SIDE (for an overview see Postmes &Spears, 1998; Postmes, Spears, & Lea, 1999;Postmes, Spears, Lea, & Reicher, 2000).

Anonymity, the situational featureaddressed in SIDE, is one of the characteristicswith a strong impact on social psychologicalphenomena in which CMC differs from FTFcommunication (McKenna & Bargh, 2000).Thus, the predictions of SIDE may reconcilethe seemingly contradictory findings ofstudies showing more attitude change ingroups using CMC on the one hand (Kiesler etal., 1984; Siegel et al., 1986) and studiesshowing less attitude change in CMC on theother hand (McGuire et al., 1987). Theformer would be expected, according to SIDE,when social identity is salient,1 the latter whenpersonal identity is salient. Even studiesshowing no differences in attitude changebetween both communication media (e.g.Weisband, 1992) can be integrated in thisframework. When neither a personal nor acommon social identity is clearly salient, ano-nymity should not affect the amount ofattitude change. Even though the results fromstudies comparing attitude change in FTF andCMC can be explained based on SIDE, twoquestions remain.

First, in most research on social influencerelated to SIDE (e.g. Spears, Lea, & Lee, 1990;Lea & Spears, 1991; for an overview see Postmeset al., 2000) the degree of anonymity withinCMC rather than the communication mediawas varied. The question yet unanswered iswhether a broader manipulation of anonymity,like the comparison between FTF and CMC,will also lead to the effects predicted by SIDE.This, however, is necessary to conclude that theinconsistent results of studies comparingattitude change in groups using FTF or CMCare due to the level of salient identity in thespecific experimental setting.

Second, and theoretically more important,there is another clear difference betweenstudies on SIDE and studies comparing attitudechange after CMC and FTF discussions. Inalmost all experiments that supported the pre-dictions of SIDE, participants were provided

with information about the group norm. Spearset al. (1990) informed the participating firstyear psychology students explicitly about theattitudes of last year’s freshmen. In otherstudies the norm was induced by a priming pro-cedure (Postmes, Spears, Sakhel, & de Groot,2001) or an initial group task (Postmes &Spears, 1997). In the studies that tested the RSCapproach, neither the norm of the small groupnor the norm of a social category that is super-ordinate to the interacting small group (super-ordinate social category) was explicitly given.Waldzus and Schubert (2000, Experiments1–3) recently addressed the impact of imposinga category norm because they did not succeedin replicating the effect predicted by SIDE in aseries of studies. In these web experimentseither participants’ personal identity or thesocial identity of a virtual group was salient.Participants were informed about the attitudeof the members of their virtual group who wereat the same time members of a common socialcategory. The four group members were eitheranonymous or non-anonymous. Differing fromearlier research on SIDE, no interactionbetween group members was possible. In allthree experiments results contradicted SIDE.When social identity was salient, attitudechange was lower in the anonymous condition,whereas when personal identity was salient,attitude change was higher in the non-anony-mous condition. In an additional experiment,Waldzus and Schubert (2000, Experiment 4)tested the moderating impact of the fit betweenthe norm of the local group and the norm ofthe social category. Participants were informedabout the attitudes of the members of theirvirtual group and the norm of a shared socialcategory. In all conditions the social identitybased on the shared common social categorywas rendered salient. Members of the virtualgroup were either anonymous or non-anony-mous. For half of the items the norm of thelocal group and the social category weresimilar, whereas the two norms differed for therest of the items. As predicted by Waldzus andSchubert, participants in the anonymous con-dition showed more attitude change thanparticipants in the non-anonymous condition


408




(i.e. behaved in line with the predictions ofSIDE) only when there was a fit between thenorm of the social category and the virtualgroup. No effect of anonymity on attitudechange was found when both norms differed.In sum, these experiments indicate that pro-viding participants with a reference norm is aprecondition for the effects of anonymity pre-dicted by SIDE (at least) when participants donot interact with their local group. However, it isnot clear yet whether this is also true for inter-acting groups.

Overview of the current research

The aim of the current research was to testwhether the inconsistent findings from studiescomparing the attitude change in groupscommunicating (non-anonymously) FTF and(anonymously) computer-mediated can beattributed to the moderating effect of thesalient level of identity. In Experiment 1 theeffect of a broader manipulation of anonymity(FTF vs. CMC) on attitude change withingroups, with members having a salient personalvs. a salient social identity, was tested. Experi-ment 2 addressed the second question men-tioned above. Is providing participants with thenorm of a superordinate social category also anecessary precondition for the predictionsmade by SIDE when interaction within thegroup takes place? In other words, the moder-ating role of imposing a norm of the superor-dinate social category was tested.

Experiment 1

In a 2 � 2 between-subjects design we variedthe communication media and the level of self-categorization. As this experiment testedwhether the inconsistent results of the studieson comparison of social influence between FTFand CMC can be explained by an effect ofsalient personal vs. social identity, the pro-cedure followed as closely as possible thestudies on media comparison referred to above.Groups of three participants either communi-cated FTF or via synchronous text-based CMC.Before the interaction took place, either

personal identity or social identity was madesalient. No reference norm was given or exper-imentally induced, because studies on mediacomparison had not given a reference normeither and SIDE does not state that giving areference norm is a necessary condition withinthe model. Following SCT, the prototype of theinteracting group may also serve as a norm.Turner (1991) argued that in a setting withoutan explicit outgroup, group members perceivethe scale positions which are not occupied bythe ingroup members as possible outgroupopinions.

The overall direction of attitude change waspredicted following the explanation of grouppolarization within the framework of SCT (e.g.Turner, 1991; Wetherell, 1987): When socialidentity is salient, the attitude change wasexpected to be oriented more toward thegroup norm than when personal identity issalient, regardless of the communication mediaused. Moreover, following SIDE it was pre-dicted that this effect would be even stronger inthe CMC condition than in the FTF condition.The salient of a social identity was expected tolead to more adherence to the group normafter CMC than after FTF communication.Personal identity was thought to releasebehavior being more in line with participants’personal goals and deviating from the groupnorm in CMC compared to FTF communi-cation. In contrast to SIDE, the RSC approachwould predict a stronger attitude change awayfrom societal norms in the CMC than in theFTF condition, no matter which identitycategory was salient.

MethodDesign and sample To test these predictions,an experiment with the two independent variables communication media (CMC vs. FTF)and salient level of identity (personal vs. socialidentity) was run. Seventy-two students (47females and 25 males) from the University ofGöttingen (Germany) took part in experi-mental groups of three. One group had to beomitted from the analysis because one partici-pant prematurely left the laboratory. Thestudents’ mean age was 23 years (range 18–29).


409




Participation was voluntary and the partici-pants received no financial compensation.

Procedure The groups were randomlyassigned to one of the four experimental con-ditions. After the arrival of all three partici-pants, they were requested to give a judgmentin terms of percentage in a moral dilemmawhich was adapted from Kohlberg’s medicinedilemma (German version: Hinder, 1987):

Your partner is terminally ill. All the treatmentmethods have failed. A new medicine against thisillness has just been developed. The doctorsbelieve it could save your partner. You know thatthis medication is your partner’s last hope, youwould like to buy it. However, even if you go intodebt the medicine is too expensive for you to payfor. Attempts to bargain over price or to pay ininstallments were unsuccessful. Your only hope ofgetting the medication is to steal it and in doing soto risk getting arrested.

How large would your partner’s recovery chancedue to the medication have to be (in %) for you torisk breaking into the pharmacy?

The answer was measured on a rating scaleranging from 0% for ‘no chance for recovery’to 100% for ‘recovery guaranteed’. The readi-ness to break the law even though there is onlya low chance for the partner to recover wouldindicate the intention to show anti-normativebehavior. One might object that this scenario isnot suitable to measure anti-normativebehavior because it has been developed tomeasure moral development. However, itincludes at least one central element of societalnorms: the law. Thus, this item better addressesthe RSC approach’s prediction that CMCgroups tend to show more anti-normativebehavior than FTF groups than earlier itemsthat were mostly based on political attitudesand risky/cautious behavior.

After the participants made a decision, theywere requested to discuss the same dilemma ingroups of three, either FTF or via CMC, and toreach a decision. After the discussion, partici-pants were again asked to indicate their indi-vidual opinion concerning the dilemma and tofill in a demographic questionnaire. Finally, theparticipants were debriefed and thanked.

Experimental manipulations To vary the com-munication media FTF and CMC were used.Directly upon arrival participants in the CMCcondition were led to separate rooms equippedwith PCs to avoid any contact between themembers of a group. Participants wereinstructed how to use the CMC software.Messages were visible on the screens of theother group members less than 0.5 secondsafter they were sent. Messages scrolled off thescreen as soon as the window was full and a newmessage arrived. The messages were markedonly with numbers assigned to the participantsin order to maintain anonymity while at thesame time allowing participants to follow thediscussion in terms of who said what. Theparticipants were informed that the interactionwould be recorded.2 In every other way the pro-cedure followed the description given above.

In the FTF condition participants gave theirfirst rating at separate desks in the same room.During the group discussion they were placedaround a table. The FTF discussions wererecorded on video. In order to match the dis-cussions for the different amount of timerequired to write and read in comparison tospeaking and listening, a maximum discussiontime of 7: 30 minutes was given in the FTF con-dition, whereas in the CMC condition discus-sions lasted a maximum of 30 minutes (Kiesler& Sproull, 1992).

Salient level of identity was manipulated bydifferent instructions given directly before thediscussion. In the personal identity condition the differing individual characteristics werestressed. Participants received an individualcode number and they were asked to discussthe medication dilemma with two studentsmajoring in other subjects. In the social identitycondition the groups were given group namesand were told that they all had the same major.In order to facilitate this instruction in the FTFcondition, special care was taken to ensure thatthe members of a group did not know oneanother.

Dependent measure The measure of attitudechange was attitude change in the direction ofthe group norm. As the mean scale position


410




preferred by the participants before the dis-cussion was 40.97%, i.e. lower than 50%, it wasassumed that the group norm was to steal forone’s partner. Thus, attitude change toward alower percentage was computed (i.e. attitudechange = attitude before decision – attitudeafter decision). This procedure is in line withthe procedure suggested by SCT as well as byearlier research on group polarization. Highervalues indicate more adherence to the groupnorm and simultaneously attitude change awayfrom the societal norm (i.e. behaving in linewith the law).

ResultsTo test the hypotheses an analysis of covariance(ANCOVA) with the independent factors com-munication media and salient level of identity andthe opinion prior to discussion as a covariate wascomputed.3 The individual opinion before thediscussion covaried with the degree of attitudechange (F(1,64) = 24.91, p <.001).

It was predicted from SIDE that CMCcompared to FTF discussions would result inmore attitude change toward the group normwhen social identity was salient and a strongerstriving for individual distinctiveness whenpersonal identity was salient. A media � salientlevel of identity interaction was found (F(1,64)= 11.73, p = .001). Contrary to the hypotheses,when social identity was salient more attitudechange toward the group norm took place ingroups communicating FTF (M = 20.89, SE =4.46) than in groups using CMC (M = 3.28, SE= 4.86) (F(1,63) = 7.25, p = .009). Conversely,when personal identity was salient there wasstronger polarization in the CMC condition (M= 17.86, SE = 4.44) than in FTF communication(M = 4.30, SE = 4.43) (F(1,63) = 4.66, p = .035)(see Table 1).

Overall, the expected main effect for thesalient level of identity did not occur (F(1,64) =.05). In the FTF condition attitude change wasin line with SIT/SCT: higher when socialidentity was salient than when personal identitywas salient (F(1,63) = 7.26, p = .009). In theCMC condition the opposite effect occurred(F(1,63) = 4.92, p = .030). Finally, in followingthe RSC approach, the main effect of the com-munication condition was not reliable (F(1,64)= .20).

DiscussionThe present study examined whether a broadermanipulation of anonymity leads to the patternof attitude change predicted by SIDE. It wasexpected that CMC, in being more anonymousthan FTF communication, fosters the orien-tation toward the ingroup norm when socialidentity is salient, while strengthening thestriving for individual distinctiveness whenpersonal identity is salient. Additionally, as pre-dicted by SCT, salient social identity wasexpected to result in more conformity to thegroup norm than personal identity. These pre-dictions were only partially confirmed. In theFTF condition, group polarization was greaterwhen social identity was salient than whenpersonal identity was salient. This pattern is inline with the predictions of SCT. In CMC,attitude change showed the reversed patterninstead of the same pattern in a more extremefashion (as predicted by SIDE). In fact, theinfluence of social categorization disappearedin CMC. In the personal identity/CMC con-dition attitude change occurred. Participantswere more willing to steal the medicine for theirpartner after the discussion than prior to it. Thisattitude change was directed toward the groupnorm. This is an unexpected finding, because


411

Table 1. Estimated marginal means and standard errors for attitude change within ANCOVA (communicationmedia � level of salient identity, covariate decision before discussion) (N = 69)

FTF CMC

M SE M SE

Social identity 20.89 4.46 3.28 4.86Personal identity 4.30 4.44 17.86 4.44




salient personal identity was predicted to causeattitudes that contradict the interacting group’snorm.4 However, as the participants in this con-dition were more willing to break the law fortheir partner after than prior to the discussion,this shift can also be interpreted as showing anintention that corresponds with the individualgoal to care for one’s partner. Therefore, thisfinding indicates a stronger fulfillment ofpersonal needs in the CMC than in the FTF con-dition and confirms the predictions of SIDE.Taken together, the findings in the personalidentity condition were unexpected on the onehand, because higher anonymity brings aboutmore conformity to the group norm. On theother hand, they were in line with the predic-tions of SIDE, because participants seem tobehave more in line with their personal goals inthe personal identity condition. This summaryshows that it is hard to deduce clear cut predic-tions for individuals with salient personalidentity from SCT and SIDE, unless one knowstheir personal goals. This is not surprising, asSCT mainly focuses on social identity. Due tothis, we focus on group members with salientsocial identity in Experiment 2, as has beendone in various other studies on SIDE (e.g. Lea,Spears, & de Groot, 2001; Postmes, Spears, &Lea, 1998, 2001).

In the personal identity condition the RSCapproach was supported as participants usingCMC were more willing to violate the law thanparticipants discussing FTF. However, in thesocial identity condition the results contradictthe RSC approach, because the reversed effectof the personal identity condition occurred. Atthe same time, the results in the personalidentity condition also contradict SIDE,because in the anonymous compared to thenon-anonymous condition more instead of lessadherence to the group norm was observed.These results are opposite to the patternreported by Spears et al. (1990).

The reversed pattern of results for partici-pants with salient social identity in Experiment1 compared to the predictions of SIDE mighthave been caused by one of the two essentialdifferences of the current compared to priorSIDE research. These differences are (a) the

lack of a given group norm and (b) themanipulation of communication media insteadof varying visual anonymity in the presentresearch. As mentioned above, anonymity (orthe lack of social context cues) is considered asone of the most striking differences betweenCMC and FTF. Additionally, Waldzus andSchubert (2000, Experiments 1–3) also found areversed pattern of results compared to SIDE’sprediction in their studies with a manipulationof visual anonymity. In these studies the partici-pants were not informed about the norm of asuperordinate category either. Thus, it seemsmore likely that the crucial differing character-istic of our Experiment 1 is the absence of areference norm.

The Impact of Category Norms onAttitude Change

Earlier studies on SIDE showed that when acategory norm was given, group membersbehaved in line with this norm as long as thesalience of social identity was not underminedin a non-anonymous condition (e.g. Lea &Spears, 1991; Postmes et al., 1998; Spears et al.,1990). As reported above, Waldzus andSchubert (2000, Experiment 4) found that thefit between the prototype of the local groupand the norm of the superordinate category isa necessary precondition for the predictedeffect of anonymity on conformity to the normpertaining to the superordinate category. Fromthis finding the authors conclude that theprocess of inferring the group norm from theinformation that is available (norm induction)is the crucial mediating cognitive processwithin SIDE.

The perceived norm might differ dependingon whether interaction within the local groupsis given (as in our Experiment 1 and earlierresearch on SIDE) or not (as in Waldzus &Schubert’s Experiment 4). Under which con-dition is the interacting small group a meaning-ful social entity to its members? It is very unlikelythat the feeling of a common local groupemerges solely from reading statements ofother persons once. However, in an ongoinginteraction one might increasingly feel like a


412




member of the interacting group. Thus, in asetting where the reference norm of a superor-dinate social category is given to the interactingsmall group, both the superordinate socialcategory and the interacting small group canbe sources of social influence and attitudechange. We assume that anonymity does notonly moderate the influence of the norm of thesuperordinate social category as predicted bySIDE (more influence in the anonymous thanin the non-anonymous conditions), but also theeffect of the interacting small group and itsmembers. The interacting group may be lessrelevant in anonymous communication than innon-anonymous communication, because aninteracting small group (without past andfuture) only becomes meaningful when one hasformed an impression about this group and itsmembers. In anonymous communication asgiven in CMC, impression formation is muchharder and takes longer than in FTF communi-cation (Walther, 1992). A meaningless groupdoes not provide the basis for social identifi-cation (Simon, Hastedt, & Aufderheide, 1997),whereas a small group that becomes a subjec-tively important subcategory within the super-ordinate group can be a source of influence.

Which pattern of attitude change results fromthe influence of the interacting small groupand the superordinate social category depend-ing on anonymity and the knowledge about anorm of the social category? The more salientthe superordinate social categor y, the moreattitude change toward a category norm isexpected when the norm is socially shared (e.g.explicitly given as reference norm). When thenorm is not socially shared inconsistent or noattitude change at all should result. This shouldbe the case even when a strong social categoriz-ation exists, because adherence to perceivednorms that differ between participants of astudy do not result in a consistent pattern ofattitude change. Thus, in anonymous groupscompared to non-anonymous groups moreattitude change toward a socially sharedcategory norm should result from the superor-dinate category. Without a socially sharednorm, no (consistent) effect of anonymity canbe predicted.

The impact of the interacting small group,which is stronger in the non-anonymous thanin the anonymous condition, will also dependon the existence of a socially shared norm ofthe superordinate social category. In general,intragroup influence will lead to attitudechange toward the group prototype which, inmost cases, resembles a shift toward the direc-tion that is dominant in the group prior to dis-cussion (Turner, 1991). When a referencenorm is given, this norm might act as anattribute of the superordinate category fromwhich the small group distances itself (e.g. inthe experiment of Spears et al., 1990 theattitude held by first year psychology students ofthe preceding year compared to the attitude ofthe small group members). Therefore, giving areference norm can affect the prototype of thesmall group (i.e. the norm of the small group)similarly to outgroup attitudes and the proto-type should shift toward the opposite pole ofthe scale (Hogg, Turner, & Davidson, 1990).Attitudes will shift away from the norm of thesuperordinate category.5 In other words: whenthe norm of the superordinate category isimposed, non-anonymous group memberscreate a local group identity based on theirinterpersonal bonds which they can explicitlycontrast with the superordinate category and itsnorm—hence the interacting small group initi-ates a tendency to depolarize. When there is noexplicit reference norm, however, the differen-tiation process in the interacting group has noclear anchor to work against. For this reason,prototype-based social influence will lead topolarization (i.e. an attitude change in thedirection of the mean individual decision madebefore the discussion).

Taken together, we assume that two processesunderlie the attitude change in small groups.The first is described by SIDE and is based onthe adherence to the norm of a social categorythat is superordinate to the interacting groupand shared by all group members. The secondis the process of adherence to the small groupnorm (i.e. its prototype) which is based onnorm induction. The first one is stronger whenanonymity fosters the salience of the superordi-nate category, whereas the second one is


413




stronger when non-anonymous interactionsallow group members to form a meaningfulsmall group. Postmes and Spears (2000) reportfindings that fit our predictions. They foundthat social influence in groups based on acommon identity was stronger when they wereanonymous (compared to non-anonymousgroups), whereas social influence in groups thatare based on interpersonal bonds was strongerwhen they were non-anonymous.

Based on this line of argumentation wepredict more attitude change toward a givenreference norm in anonymous groups than innon-anonymous groups. The attitude change inthe non-anonymous condition with a givenreference norm might even be directed awayfrom the norm because the interacting groupdistances itself from the category norm. If noreference norm is given, the attitude change inthe non-anonymous condition should begreater than in the anonymous condition whenthe influence of the local group is strongerthan the influence of the social category.6 Theattitude change will be directed toward the pro-totype of the local group. When the referencenorm fits the norm within the sample (as in ourExperiment 2) the attitude change will also bedirected toward this norm even without givinga reference norm, because the attitudes of thegroup members before the discussion (andtherefore also the group prototype) are mostlikely in line with the norm. The group dis-cussion will amplify this tendency. These pre-dictions on the one hand correspond to theresults of earlier research on SIDE (with a givennorm) and on the other hand correspond tothose of Waldzus and Schubert (2000, Experi-ments 1–3) and Experiment 1 (without a givennorm). Experiment 2 tested these hypotheses.

Experiment 2

In the following experiment participants eitherreceived or did not receive information aboutthe norm of their common superordinatesocial category. Additionally, anonymity wasvaried during CMC within groups of three.Instead of a manipulation of communicationmedia a manipulation of visual anonymity was

chosen, because this experiment sought to testthe impact of a given group norm in thecontext of SIDE. The impact of giving a groupnorm should generalize to the comparison ofdifferent communication media.

MethodDesign and participants An experiment with a2 (anonymity: anonymous vs. non-anonymous)� 2 (reference norm: given vs. not given)design was conducted. Participants were 81students (23 male, 58 female) of the Universityof Göttingen with a mean age of 23 years.Participants were recruited during lectures andfrom a dining hall and received DM 10 (about€5) for participation. Five participants wereexcluded from further analysis because of a studentized deleted residual greater than 2.65(� < .01) in multiple criteria outlier analysis(Neter, Kutner, Nachtschiem, & Wasserman,1996).

Procedure and experimental manipulationsThe experimental procedure was similar toExperiment 1. Directly upon arrival, partici-pants in the anonymous condition were led tothree different rooms equipped with com-puters, whereas participants in the non-anony-mous condition were led to one roomequipped with three computers. All questionsand instructions were presented on thecomputer screen. The participants in the non-anonymous condition were not allowed to talkto one another. This manipulation of ano-nymity was successfully applied by Spears et al.(1990).

First, the participants had to indicate theiropinion about two dilemmas, taken from aGerman version (Witte, 1971) of the choicedilemma questionnaire (Kogan & Wallach,1964):

25 years old Mrs. B. has a large scar close above herleft eye. This scar disfigures her face and causes hera great deal of distress. Mrs. B. has the option ofplastic surgery. However, with this interventionMrs. B. would run the risk of losing her sight in herleft eye.

After an injury to his spine Mr. G. now has tospend his whole life in bed. His physician suggests


414




surgery that might improve his physical state. Thissurgery is not always successful. When it fails thereis a high risk of fatality.

Participants were asked to decide at which levelof chance for healing they would give theadvice to decide in favor of the respectiveoperations. These tasks were chosen to avoidthe fulfillment of personal needs from theattitude indicated as in Experiment 1. There-fore, the decision task was more comparable tothe task employed in earlier studies on SIDE.

Having read these dilemmas, the vantage ofgroup decisions in extreme situations likeillness was stated as a reason for the group dis-cussion that followed. Moreover, the partici-pants were informed that the other groupmembers were also students of the University ofGöttingen. Both statements were made in orderto reinforce participants’ social identity. Partici-pants in the reference norm given condition wereadditionally informed (ostensibly as a stimu-lation of the discussion) that students of theiruniversity in earlier studies tended to decideagainst the surgery. Exact means from pilotdata for both items were reported. Before the discussion started, the CMC software wasexplained to the participants. The participantswere asked to discuss each dilemma for about15 minutes. After the discussion, the partici-pants again had to indicate their opinion con-cerning the dilemma and to work onquestionnaires presented on the screen. Finally,they were debriefed, thanked, and paid.

Dependent measures The main dependentmeasure attitude change was assessed as meanattitude shift on the two dilemma items towardthe cautious pole of the scale. The pretest aswell as the mean individual decisions before thediscussion indicated a norm for cautiousbehavior. Hence, the score was computed in away that higher values indicate a strongerattitude change toward the cautious norm.

Additionally, social identification with theinteracting small group/the students of theUniversity of Göttingen were assessed with fouritems each (e.g. ‘I am glad to be a member of this group/a student of the University of Göttingen’, � = .75/.70).

To check for the effect of the manipulationstwo items asked for the anonymity of the self andthe group. As both items were significantly cor-related (r(75) = .797 p = .001), the mean of bothitems was computed as a measure of anonymity.Two additional items measured the norm to becautious within the interacting small group(e.g. ‘From the very beginning there was atendency in our group to meet a cautiousdecision.’). As these items were also highly cor-related (r(76) = .70, p = .001), they wereaveraged. Finally, one item was added tomeasure the perceived norm of the students ofthe University of Göttingen (‘Most of thestudents of the University of Göttingen would,in both cases, have selected a high percentageand therefore a low level of risk as a precondi-tion for an operation.’). All items had to beanswered on a 6-point scale ranging from 1 for‘I do not agree’ to 6 for ‘I entirely agree’.

ResultsManipulation checks To check for the effectsof the independent manipulations ANOVAswith anonymity and reference norm as factorswere computed. Participants in the anonymouscondition felt more anonymous (M = 3.63, SD =1.14) than participants in the non-anonymouscondition (M = 2.68, SD = 1.34) (F(1,71) =11.47, p = .001). Participants who receivedinformation about a reference norm of thesuperordinate social category gave an estimateof this norm that was more cautious and there-fore more in line with the given norm (M =4.37, SD = 1.24) than participants who were notinformed (M = 3.81, SD = 1.20) (F(1,68) = 4.01,p = .049). The perceived norm of the inter-acting small group was not affected by themanipulation of the reference norm (F(1,72) =0.27).

The identification with the local group (M =3.73, SD = 0.89) and with the university (M =4.04, SD = 0.86) were both significantly abovethe midpoint of the scale (ts > 2.25, ps < .03).Therefore, it seems reasonable to assume thatparticipants identified both with the localgroup and with their university. Takentogether, the experimental manipulations weresuccessful.


415




Attitude change It was predicted that theattitude change toward the group norm wouldbe stronger in the anonymous than in the non-anonymous condition when a reference normof the superordinate social category was given(higher scores in the anonymous than in thenon-anonymous condition). Conversely, theattitude change should be less strong in theanonymous condition when participants werenot informed about the reference norm (lowerscores in the anonymous than in the non-anonymous condition). To test this predictiona 2 (anonymity) � 2 (reference norm)ANCOVA with mean attitude prior to dis-cussion as a covariate and attitude changetoward the group norm as the dependentmeasure was computed. The attitude prior todiscussion was a reliable predictor for theattitude change (F(1,71) = 28.11, p<.001).Moreover, there was a main effect of the refer-ence norm (F(1,71) = 6.63, p = .012). Although,the category norm was perceived as morecautious when the reference norm was given,the attitude change toward a cautious decisionwas lower for participants who were informedabout a group norm (M = –4.40, SE = 1.96) thanfor participants who were not informed aboutthis norm (M = 2.55, SE = 1.85). This maineffect was qualified by the predicted anonymity� reference norm interaction (F(1,71) = 5.90,p = .018) (see Table 2). Participants whoreceived a reference norm showed an attitudechange away from the group norm in the non-anonymous condition (M = –8.04, SE = 2.88)that was greater than in the anonymous con-dition (M = –0.75, SE = 2.71) (F(1,71) = 3.34, p= .072). Conversely, participants who did notreceive a reference norm showed slightly moreattitude change toward the group norm in the

non-anonymous (M = 5.58, SE = 2.63) than inthe anonymous condition (M = –0.48, SE =2.62) (F(1,71) = 2.65, p = .108).8

In additional analyses the manipulationcheck for perceived group norms was used totest the assumed processes. More precisely,partial correlations within the anonymous andthe non-anonymous condition between attitudechange and the students’ norm as well as thenorm of the interacting group controlled forthe norm manipulation and the attitude priorto discussion were computed. This was done totest the assumption that the interacting grouphas more impact in the non-anonymous con-dition than in the anonymous conditions andthe social category has more influence in theanonymous than in the non-anonymous con-dition. As predicted, in the non-anonymouscondition a partial correlation between the per-ceived norm of the local group and the attitudechange was found (p = .30, p = .041, one-tailed),whereas this correlation was not found in theanonymous condition (p = .08, p = .329, one-tailed). In a multiple regression the anonymity� local group norm interaction was a signifi-cant predictor of attitude change (� = .24, p =.016) indicating that the difference betweenthe two partial correlations is reliable. Thus, asexpected, anonymity moderates the influenceof the local group on attitude change.

There was no evidence for a moderation ofthe impact of the social category on attitudechange by anonymity. In both conditions theperceived category norm and the attitudechange correlated similarly (anonymous: p =.31, non-anonymous: p = .32, both p < .05, one-tailed, partial correlation controlled fordecision prior to discussion and norm manipu-lation). Thus, we did not find evidence for the


416

Table 2. Estimated marginal means and standard errors for attitude change within ANCOVA (anonymity �reference norm, covariate decision before discussion) (N = 76)

Non-anonymous Anonymous

M SE M SE

Reference norm given –8.04 2.88 –0.75 2.71No reference norm given 5.58 2.63 –0.48 2.62




mechanism proposed by SIDE: the impact ofthe social category was not stronger in theanonymous than in the non-anonymous con-dition.

DiscussionExperiment 2 was conducted to test the impactof informing the members of an interactingsmall group about the norm of a superordinatecategory before they discuss a topic within theirsmall group. It was predicted that informedmembers of anonymous groups show a greaterattitude change toward the group norm thaninformed members of non-anonymous groups,whereas the opposite pattern was expected forgroup members who did not receive any infor-mation about the group norm. The results con-firmed this prediction. The overall interactionwas statistically reliable and the simple maineffects were close to the conventional level ofsignificance.

Even though the predicted reference norm �anonymity interaction occurred and the relativeattitude change was perfectly in line with thepredictions, it is somewhat unexpected thatparticipants in the anonymity/reference normgiven condition did not show any attitudechange toward the group norm. When theattitude change was not computed as a shifttoward the end point of the scale (as it is usuallydone in research on group polarization), whichwas given as verbal description of the norm ofthe superordinate social category, but as a shifttoward the actual values that were additionallygiven as category norms, the effect of anonymitycould also be found (F(1,33) = 7.93, p = .008).However, the parameter values differed. In theanonymous condition the attitude change wasdirected more toward the explicitly given norm(M = 10.00, SE = 2.22) than in the non-anony-mous condition (M = 0.77, SE = 2.35). Theseresults demonstrate that participants in theanonymous/reference norm given conditionchanged their attitude toward the group normstated in the instruction. However, the valuesgiven were more influential than the generaldirection that was stated verbally. Additionally,in this analysis the simple main effect that failedto be significant in the analysis reported above

is significant. This result provides additionalsupport for our prediction.

The index applied in the above analysis onlymakes sense for the condition with referencenorm, because in the condition without refer-ence norm participants were not informedabout this norm. Thus, attitude change in thiscondition was directed neither toward nor awayfrom the numerical norm (t(39) = .69). Furtherresearch should address the question of whichconcrete norm is inferred by group memberswhen they are informed about the generaldirection of a group norm and/or the exactmean of the ingroup norm.

The hypothesis that the attitude change isdirected away from the given reference norm inthe non-anonymous condition was supported.Taken together, the results for those groupsthat were informed about the reference normmatch the findings of earlier studies on SIDE,whereas the results in the conditions withoutreference norm coincide with the results ofExperiment 1 and of Waldzus and Schubert(2000, Experiments 1–3).

Finally, Experiment 2 provided initialevidence for the assumed underlying processes.The impact of the interacting group and itsnorm was found to be higher in the non-anonymous condition than in the anonymouscondition. At the same time, we did not findevidence for the process suggested by SIDE.The impact of the superordinate socialcategory was not stronger in the anonymousthan in the non-anonymous condition butpresent in both conditions. One possiblereason for this lack of evidence is the measurethat was employed in this study: the perceivednorm. SIDE does not suggest a change in influ-ence of the group norm but a difference insalience of the social categorization. The latterhas been found to mediate the impact of anonymity in a context where group normswere primed (Postmes et al., 2001). Thus, thereis evidence for this aspect of the process notfrom the current, but from earlier research.However, it should also be noted that Lea et al.(2001) argued that categorizing oneself interms of the interacting group is fostered byanonymity and thus they suggest that SIDE


417




effects are based on the interacting group. Thisassumption is contradicted by the currentfindings in which anonymity leads to lessattitude change driven by the small groupnorm.

General discussion

The aim of the present paper was to examinethe explanation offered by SIDE for the incon-sistent findings from research comparingattitude change in FTF and CMC. SIDE wouldpredict that the differing degree of anonymitycomparing both communication conditionswill interact with the salient level of categoriz-ation in its impact on attitude change. If theexperimental setting initiates a salient personalidentity, higher attitude change in non-anonymous FTF communication is expected,whereas a setting that initiates a salient socialidentity will lead to higher attitude change inCMC. To test the validity of this explanationtwo assumptions had to be tested:

• Varying visual anonymity as well as imple-menting a broader manipulation of ano-nymity, namely the comparison of FTF andanonymous CMC, result in the effects pre-dicted by SIDE.

• The knowledge about the norm of the socialcategory moderates the effects predicted bySIDE.

Experiment 1 addressed the assumption thata broader manipulation of anonymity leads tothe effects predicted by SIDE. The results con-firmed this prediction at least partly whenpersonal identity is salient. After a computer-mediated discussion participants with salientpersonal identity behaved more in line withpersonal goals and needs in terms of theattitude they expressed than participants after aFTF discussion. At the same time, they did notdistance themselves from the group norm asone could also predict based on SCT. Whensocial identity was salient greater attitudechange after FTF communication compared toCMC was found. This unexpected pattern ofresults does not necessarily question therelevance of SIDE for the explanation of

behavioral differences between groups com-municating FTF and computer-mediated if onetakes into account the results of Experiment 2.

Experiment 2 was conducted to test whetherthere is an impact of imposing a referencenorm for a superordinate category. Thefindings of Experiment 2 indicate that inform-ing participants explicitly about a referencenorm moderates the effect of anonymity onattitude change. When the category norm wasexplicitly given, the effect of anonymity was inline with the prediction of SIDE, whereas it wasreversed when no category norm was given.Thus, information about a reference norm is acrucial factor for the influence of a discussionon group members’ attitudes. In sum, theeffects of category norms are twofold: on theone hand, attitude change guided by themembership in a social category can only occurwhen knowledge about the group norm issocially shared. Therefore, a socially sharedgroup norm is mandatory for self-categoriz-ation based attitude change. On the otherhand, when the salience of the superordinatesocial category is undermined, as is the case innon-anonymous interacting small groups, themeaning of the category norm changes. Non-anonymous interacting small groups candistance themselves from the superordinatecategory and its norm and thus depolarize.

The current findings fit the pattern ofattitude change predicted based on the twoprocesses described above. However, theresearch reported here provides only littleevidence for the assumed processes themselves.Only the moderation of the correlationbetween the norm of the local group andattitude change by anonymity points to the factthat the local group becomes more importantwhen anonymity is low. Hence, both processesand their interplay have to be studied. Furtherresearch should specifically address the processof norm induction as suggested by Waldzus andSchubert (2000) to obtain a clearer picture ofthe cognitive processes that underlie therespective type of attitude change.

Following this reasoning, in a situation whereseveral members of a social category interact,anonymity does not only moderate the salience


418




of the superordinate social category as stated bySIDE, but also the impact of the interactingsmall group on their members’ attitudes. Ano-nymity enhances the salience of the socialcategory, whereas it undermines the impact ofthe interacting small group. Both processesmight have been at work in studies on SIDEthat used a paradigm including both types ofgroups (Postmes, Spears, & Lea, 2002; Spears etal., 1990). An increasing level of anonymityfosters the conformity to the norm of the super-ordinate category and hinders the intragroupdynamic that would lead to an attitude changeaway from the category norm.

This conclusion does not question the mainassumptions of SIDE, as there is also researchsupporting the predictions of SIDE in whichonly one level of categorization was present(Postmes et al., 1998; Sassenberg, 2000) and inwhich no category norm was given (Wilder,1990). However, the effect of anonymity on theimpact of the interacting small group has beenunderestimated and thus received insufficientattention until now. Exceptions are the work byPostmes and Spears (2000) who comparedgroups having strong interpersonal bonds withgroups whose members share a commonidentity, and also the work by Sassenberg andPostmes (2002) who addressed the effect ofanonymity on intragroup dynamics.

The assumption that a broader manipulationof anonymity, such as manipulating the com-munication media, results in the same patternof attitude change as a manipulation of visualanonymity used in earlier studies on SIDE wasnot supported by the results of Experiment 1 inthe salient social identity conditions. However,in this study no reference norm was given andthe results in Experiment 1 parallel the resultsof Experiment 2 for groups that did not receiveany information about the category norm.Thus, the effect of communication media (FTFversus CMC) does not differ from the effect ofanonymity. The inconsistent results fromresearch on media comparison can beexplained based on SIDE when the effect ofanonymity on intragroup processes is takeninto account. In studies with a social contextthat fosters the salience of personal identity atti-

tudes are supposed to be more stronglyoriented toward individual goals in the anony-mous condition. This might lead to less attitudechange in CMC groups than in groups dis-cussing FTF when personal goals are notrelated to the content of the discussed topic. Instudies that foster the salience of social identityand suggest an ingroup norm, attitude changeis supposed to be more strongly orientedtoward this norm in the anonymous condition.When no norm is given attitude change issupposed to be stronger when anonymity islow.

Taken together, the results suggest that highanonymity (compared to low anonymity) as wellas CMC (compared to FTF communication)cause more conformity to individual needs orgoals when personal identity is salient. Higherconformity to a socially shared superordinatecategory norm is the result when social identityis salient. In the case of a socially sharedcategory norm being absent, attitude change isdirected toward the norm of the local group.This intragroup influence is stronger when ano-nymity is low.

Notes1. One might object that the publicity in FTF

communication and in a non-anonymous con-dition leads to more conformity toward the groupnorm when social identity is salient. However,following SCT the adherence to group norms is not a result of public compliance but an outcome of perceiving oneself as a group member and thus,considering the group norm as the way one wantsto be. In line with this assumption Noel, Wann, andBranscombe (1995) have found that groupmembers show the same amount of conformity toingroup norms no matter whether they state theirattitudes publicly or privately except when theybecame a member of the group very recently.Moreover, there is evidence for strong conformitytoward group norms in CMC that is based on socialidentification with the ingroup (e.g. Sassenberg,2002).

2. Participants in the FTF and the CMC conditionwere informed that their behavior was recorded.This might affect the anonymity of the self,whereas it does not affect the anonymity of thegroup (i.e. the amount of information a


419




participant has about his/her group members),which is the crucial variable within the cognitiveaspect of SIDE addressed here (see Spears & Lea,1994, for an extensive discussion of the differingeffects of the anonymity of the self and the anonymity of the group). Thus, informing theparticipants about the recording of their behavior, which is necessary for ethical reasons,should not affect the results.

3. This calculation follows the procedure applied bySpears et al. (1990) and thus ensures the com-parability of the results. In reference to Campbelland Stanley (1963) the authors argued that withthis procedure, even with randomized assignmentof the participants to the conditions, the remain-ing error variance in the analysis should be con-trolled.

4. The personal identity condition can also be seenas an intergroup context made up by the differentmajors of the group members (we wish to thankMartin Lea for this alternative interpretation). Inthis case an attitude shift in the predominantdirection can be interpreted as a stronger con-formity toward the group norm that is inferredfrom the individual’s attitude via self-anchoring(Cadinu & Rothbart, 1996). However, in thisstudy it does not matter if differences betweengroup members were established based oninterindividual differences or intergroup differ-ences within the interacting group. The import-ant fact is that the social identity of the smallgroup was salient in one and not salient in theother condition.

5. This would also explain why in the non-anony-mous / social identity condition in Spears et al.’s(1990) experiment a slight attitude change awayfrom the reference norm occurred and not a shifttoward the reference norm as expected from SCT.

6. The strength of the influence of the local groupand the social category can be determined byseveral factors such as the salience of the respec-tive level of categorization and the social identifi-cation with the group.

7. Due to technical problems some questionnairedata from four participants are missing.

8. Maxwell and Delaney (1990, p. 145) suggest that Ftests for factors with two levels can be computedone-tailed. Hence, reporting two-tailed tests inour case is very conservative given that ourhypothesis was directed. A less conservative one-tailed test reveals that one p reaches the con-ventional level of statistical significance and theother one is very close to if (p = .054).

AcknowledgmentsWe are grateful to Kai J. Jonas, Thomas Kessler,Martin Lea, Russell Spears, Thomas Schubert, SvenWaldzus, and K. Andrew Woltin for their extensiveand valuable comments on an earlier version of thisarticle. We also extend our appreciation to HazelWillis for help with manuscript preparation andMelanie Appel, Lars Gargulla, and Kai Heutelbeckfor their assistance in data collection.

ReferencesCadinu, M. R., & Rothbart, M. (1996). Self-

anchoring and differentiation processes in theminimal group setting. Journal of Personality andSocial Psychology, 70, 661–677.

Campbell, D. T., & Stanley, J. C. (1963). Experimentaland quasi-experimental designs for research. Chicago:Rand-McNally.

Diener, E. (1980). Deindividuation: The absence ofself-awareness and self-regulation in groupmembers. In P. B. Paulus (Ed.), The psychology ofgroup influence (pp. 209–242). Hillsdale, NJ:Erlbaum.

Festinger, L., Pepitone, A., & Newcomb, T. (1952).Some consequences of de-individuation in agroup. Journal of Abnormal and Social Psychology, 47,382–389.

Goethals, G. R. & Zanna, M. P. (1979). The role ofsocial comparison in choice shifts. Journal ofPersonality and Social Psychology, 37, 1469–1476.

Hinder, E. (1987). Grundlagenprobleme bei der Messungdes sozial-moralischen Urteils [Basic problems in themeasurement of moral judgement]. Frankfurt:Lang.

Hogg, M. A., Turner, J. C., & Davidson, B. (1990).Polarized norms and social frames of reference: Atest of the self-categorization theory of grouppolarization. Basic and Applied Social Psychology, 11,77–100.

Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Socialpsychological aspects of computer-mediatedcommunication. American Psychologist, 39,1123–1134.

Kiesler, S., & Sproull, L. (1992). Group decisionmaking and communication technology.Organizational Behavior and Human DecisionProcesses, 52, 96–123.

Kogan, N., & Wallach, M. A. (1964). Risk taking:A study in cognition and personality. New York:Holt.

Lea, M., & Spears, R. (1991). Computer-mediatedcommunication, de-individuation and group


420




decision making. International Journal of Man-Machine Studies, 39, 283–301.

Lea, M., Spears, R., & de Groot, D. (2001). Knowingme, knowing you: Anonymity effects on socialidentity processes within groups. Personality andSocial Psychology Bulletin, 27, 526–537.

Matheson, K., & Zanna, M. P. (1988). The impact ofcomputer-mediated communication on self-awareness. Computers in Human Behavior, 4,221–233.

Matheson, K., & Zanna, M. P. (1989). Persuasion as afunction of self-awareness in computer-mediatedcommunication. Social Behaviour, 4, 99–111.

Maxwell, S. E., & Delaney, H. D. (1990). Designingexperiments and analyzing data. Pacific Grove, CA:Brooks/Cole.

McGuire, T. W., Kiesler, S., & Siegel, J. (1987).Group and computer-mediated discussion effectsin risk decision making. Journal of Personality andSocial Psychology, 52, 917–930.

McKenna, K., & Bargh, J.A. (2000). Plan 9 fromCyberspace: The implications of the Internet forpersonality and social psychology. Personality andSocial Psychology Review, 4, 57–75.

Neter, J., Kutner, M. H., Nachtschiem, C.J., &Wasserman, W. (1996). Applied linear statisticalmodels (4th ed.). Boston: McGraw-Hill.

Noel, J. G., Wann, D. L., & Branscombe, N. R.(1995). Peripheral ingroup membership status andpublic negativity toward outgroups. Journal ofPersonality and Social Psychology, 68, 127–137.

Postmes, T., & Spears, R., (1997). Quality of decisions,group norms and social identity: Biased informationsampling or sampled biases. Paper presented at the5th Münster Workshop on the Social IdentityApproach, Rothenberge, Germany.

Postmes, T., & Spears, R., (1998). Deindividuationand antinormative behavior: A meta-analysis.Psychological Bulletin, 123, 238–259.

Postmes, T., & Spears, R., (2000). Refining thecognitive redefinition of the group:Deindividuation effects in common bond vs.common identity groups. In T. Postmes,R. Spears, M. Lea, & S. Reicher (Eds.), SIDE effectscentre stage: Recent developments in studies of de-individuation in groups (pp. 63–78). Amsterdam:KNAW.

Postmes, T., Spears, R., & Lea, M. (1998). Breachingor building social boundaries? SIDE-effects ofcomputer-mediated communication.Communication Research, 25, 689–715.

Postmes, T., Spears, R., & Lea, M. (1999). Socialidentification, normative content, and‘deindividuation’ in computer-mediated groups.

In N. Ellemers, R. Spears, & B. Doosje (Eds.),Social identity (pp. 164–183). Oxford: Blackwell.

Postmes, T., Spears, R., & Lea, M. (2002). The effectsof anonymity in intergroup discussions:Bipolarization in computer-mediated groups.Group Dynamics: Theory, Research and Practice, 6,3–16.

Postmes, T., Spears, R., Lea, M., & Reicher, S. (Eds.)(2000). SIDE effects centre stage: Recent developments instudies of de-individuation in groups. Amsterdam:KNAW.

Postmes, T., Spears, R., Sakhel, K., & de Groot, D.(2001). Social influence in computer-mediatedgroups: The effects of anonymity on groupbehavior. Personality and Social Psychology Bulletin,27, 1243–1254.

Reicher, S. D. (1984). Social influence in the crowd:Attitudinal and behavioural effects of de-individuation in conditions of high and low groupsalience. British Journal of Social Psychology, 23,341–350.

Reicher, S. D. (1987). Crowd behavior as socialaction. In J. C. Turner, M. A. Hogg, P. J. Oakes, S.D. Reicher, & M. S. Wetherell (Eds.), Rediscoveringthe social group. A self-categorization theory (pp.171–202). New York: Basil Blackwell.

Reicher, S. D., Spears, R., & Postmes, T. (1995). Asocial identity model of deindividuation phenomena. In W. Stroebe & M. Hewstone (Eds.),European Review of Social Psychology (Vol. 6,pp. 161–198). Chichester, UK: Wiley.

Sassenberg, K. (2000). The two faces of anonymity:Effects on group decision making. In T. Postmes,R. Spears, M. Lea, & S. Reicher (Eds.), SIDE issuescentre stage: Recent developments in studies of de-individuation in groups (pp. 93–106). Amsterdam:KNAW.

Sassenberg, K. (2002). Common bond and commonidentity groups on the Internet: Attachment andnormative behavior in on-topic and off-topic chats.Group Dynamics: Theory, Research, and Practice, 6,27–37.

Sassenberg, K., & Postmes, T. (2002). Cognitive andstrategic processes in small groups: Effects ofanonymity of the self and anonymity of the groupon social influence. British Journal of SocialPsychology, 41, 463–480.

Sassenberg, K., Rabung, S., & Boos, M. (1999).Selbstaufmerksamkeit in dercomputervermittelten Kommunikation [Self-awareness in computer-mediated communication].In U.-D. Reips, B. Batinic, W. Bandilla, M. Bosnjak,L. Gräf, K. Moser et al. (Eds.), Aktuelle OnlineForschung—Trends, Techniken, Ergebnisse. [Current


421




internet science—trends, techniques, results].Zurich: Online Press. [Retrieved September 152002 from the World Wide Web]. Available URL:http: //dgof.de/tband99/.

Siegel, J., Dubrovsky, V., Kiesler, S., & McGuire, T. W.(1986). Group processes in computer-mediatedcommunication. Organizational Behavior andHuman Decision Processes, 37, 157–187.

Simon, B., Hastedt, C., & Aufderheide, B. (1997).When self-categorization makes sense: The role ofmeaningful social categorization in minority andmajority members’ self-perception. Journal ofPersonality and Social Psychology, 73, 310–320.

Spears, R., & Lea, M. (1992). Social influence andthe influence of the ‘social’ in computer-mediatedcommunication. In M. Lea (Ed.), Contexts ofcomputer-mediated communication (pp. 30–65). NewYork: Harvester Wheatsheaf.

Spears, R., & Lea, M. (1994). Panacea orpanopticum? The hidden power of computer-mediated communication. Communication Research,21, 427–459.

Spears, R., Lea, M., & Lee, S. (1990). De-individuation and group polarization in computer-mediated communication. British Journal of SocialPsychology, 29, 121–134.

Tajfel, H., & Turner, J. C. (1979). An integrativetheory of intergroup conflict. In W. G. Austin & S.Worchel (Eds.), The social psychology of intergrouprelations (pp. 33–47). Monterey, CA: Brooks/Cole.

Turner, J. C. (1991). Social influence. Milton Keynes,UK: Open University Press.

Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S.D., & Wetherell, M. S. (Eds.). (1987). Rediscoveringthe social group. A self-categorization theory. New York:Basil Blackwell.

Vinokur, A., & Burnstein, E. (1974). Effects ofpartially shared persuasive arguments on group-induced shifts. Journal of Personality and SocialPsychology, 29, 305–315.

Waldzus, S., & Schubert, T. (2000). Group norm andcategory norm in anonymous situations: Twosources of social influence. In T. Postmes, R.Spears, M. Lea, & S. Reicher (Eds.), SIDE effectscentre stage: Recent developments in studies of de-individuation in groups (pp. 31–46). Amsterdam:KNAW.

Walther, J. B. (1992). Interpersonal effects incomputer-mediated interaction. CommunicationResearch, 19, 52–90.

Weisband, S. P. (1992). Group decision and firstadvocacy effects in computer-mediated and face-to-face decision making groups. OrganizationalBehavior and Human Decision Process, 53, 352–380.

Wetherell, M. S. (1987). Social identity and grouppolarization. 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. 142–170). New York: Basil Blackwell.

Wilder, D. A. (1990). Some determinants of thepersuasive power of in-group and out-group:Organization of information and attribution ofinterdependence. Journal of Personality and SocialPsychology, 59, 1202–1213.

Witte, E. H. (1971). Das ‘risky-shift’-Phänomen: Einekritische Untersuchung der bestehendenHypothesen [The ‘risky shift’ phenomenon: Acritical survey of the existing hypotheses].Psychologie und Praxis, 15, 104–117.

Paper received 7 February 2002; revised version accepted 2 January 2003.

Biographical notesK A I S A S S E N B E R G is an assistant professor at the

Department of Social Psychology at Friedrich-Schiller University of Jena (Germany). He receivedhis PhD from the Georg-August University ofGöttingen (Germany). His current researchinterests include social discrimination, self-regulation in intergroup behavior, and groupnorms.

M A R G A R E T E B O O S is a professor in the Departmentof Psychology at Georg-August University ofGöttingen (Germany). She received her PhD fromthe Friedrich-Wilhelms University of Bonn(Germany). Her current research interests includecomputer-mediated communication, groupproblem solving, group decision making, andcoherence in dialogues.


422




Documents

Attitude Change in Computer-Mediated Communication: Effects of Anonymity and Category Norms