Testing a model of sense of virtual community

Download Testing a model of sense of virtual community

Post on 05-Sep-2016

220 views

Category:

Documents

2 download

TRANSCRIPT

  • Available online 28 November 2007

    and even blogs. Although the technologies and specic topics vary, what is common to theseonline groups is the members public exchange of information and support (Jones, 1997).

    * Tel.: +1 704 687 4847; fax: +1 704 687 3096.E-mail address: alblanch@uncc.edu

    Available online at www.sciencedirect.com

    Computers in Human Behavior 24 (2008) 21072123

    www.elsevier.com/locate/comphumbeh

    Computers inHuman Behavior0747-5632/$ - see front matter 2007 Elsevier Ltd. All rights reserved.1. Introduction

    Groups of people interacting through computer-mediated communication have becomecommon within organizations and society. These groups can range from the professional(e.g., AoIR, the listserv for association of Internet researchers) to the social (e.g., Hondamotorcycle enthusiasts on Yahoo groups). They can form on a variety of interactive com-munication technologies including listservs, newsgroups and bulletin boards, chatroomsAbstract

    A distinguishing feature of virtual communities is their sense of community, i.e., their partici-pants feelings of membership, identity, inuence, and attachment with each other. This study testsa model in which members perceptions of the groups norms mediate the relationships between sup-porting each other and identifying each other with the members sense of virtual community. Twostudies were conducted providing partial support for the model. The results show that the perceptionof norms mediate the relationship between SOVC and (a) observing and publicly exchanging sup-port, (b) perceiving that others know ones identity, and (c) using technical features to learn and cre-ate identity. Theoretical and practical implications are discussed. 2007 Elsevier Ltd. All rights reserved.

    Keywords: Virtual communities; Sense of virtual community; Social exchange; Social identity; NormsTesting a model of sense of virtual community

    Anita L. Blanchard *

    University of North Carolina Charlotte, Department of Psychology,

    9201 University City Blvd, Charlotte, NC 28223-0001, United Statesdoi:10.1016/j.chb.2007.10.002

  • 2108 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123Some researchers and practitioners refer to these online groups as virtual communities.Virtual communities, as opposed to other online groups, are believed to be particularlyimportant because they are self-sustaining social systems in which members engage andconnect with each other (Rheingold, 1993; Schuler, 1996). But what makes an onlinegroup a virtual community?

    One feature that distinguishes virtual communities from mere virtual groups is theirmembers feelings of community. These feelings of community are more formally knownas a sense of virtual community. Community psychologists have long considered memberssense of community as an important feature of face-to-face (FtF) communities (Chipuer &Pretty, 1999; Fisher, Sonn, & Bishop, 2002; McMillan & Chavis, 1986; Obst & White,2004) and virtual community researchers are beginning to pay attention to these feelingsin virtual groups (Blanchard & Markus, 2004; Koh & Kim, 2003; Obst, Zinkiewicz, &Smith, 2002a; Rheingold, 1993; Roberts, Smith, & Pollock, 2002). The purpose of the cur-rent research is to build upon and expand our knowledge of what contributes to memberssense of virtual community. The next section provides the theoretical and empirical back-ground to this model.

    2. Sense of community

    Sense of community (SOC) has an extended history in the community psychology lit-erature. Sarason (1986) was one of the rst researchers to identify that community mem-bers feelings about each other and the community itself are important to the communityssuccessful functioning. SOC is desired in a community because it leads to satisfaction withand commitment to the community (Burroughs & Eby, 1998), and is associated withinvolvement in community activities and problem-focused coping behavior (McMillan &Chavis, 1986).

    McMillan and Chavis (1986) further developed the SOC construct by dening it asan individuals feelings of membership, identity, belonging, and attachment with agroup. Their SOC descriptive framework has been widely accepted because of its the-oretical base and its qualitative empirical support (Chipuer & Pretty, 1999; Obst &White, 2004).

    Research in FtF communities and organizations has examined the correlates of SOC inan eort to determine what leads to SOC. Za and Devlin (1998) found that the amount ofinteraction between community members and components of the physical environment ledto SOC. Schuster (1998) examined the processes of exchanging support that led to SOC ina writers group in an assisted care home for the elderly. In a qualitative study, Garca andher colleagues (1999) concluded that the communitys history was an important factor increating SOC. Obst and her co-workers (2002) found that among other demographic vari-ables, membership in a local organization, and their view of the neighborhood predictedtheir global measure of SOC.

    In examining SOC at work, Pretty and McCarthy (1991) found both individual (e.g.,gender and hierarchical status) and environmental characteristics such as peer cohesion,supervisor support, and job involvement were positively related to SOC. Burroughs andEby (1998) found that supportive exchanges between employees and the organizationmade a signicant positive contribution to their SOC measure. Clark (2002) found thatsupportive supervision and the intrinsic value of the work lead to her measure of SOC

    in an organization. Royal and Rossi (1999) found that organizational variables (perceived

  • A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2109orderliness of students and support for innovation) and time related variables (employeetenure and time spent with interacting with others) led to SOC in a school.

    In summary, there is a small but growing body of research on what contributes to FtFSOC. For the most part, researchers of FtF SOC have found that member interactionsand, particularly, the exchange of support between members are positively related toSOC. Other consistent correlates of SOC include member involvement with and tenurewithin the community being examined.

    3. Sense of virtual community

    Sense of community is also gaining attention in virtual communities. Roberts and hercolleagues (2002) examined sense of virtual community (SOVC) in a chatroom in whichparticipants use text to create objects with which they interact. Through a qualitativestudy, they found that the virtual communities they examined diered from FtF commu-nities, but members still experienced a sense of community similar to that dened byMcMillan and Chavis. Likewise, Obst, Zinkiewicz, and Smith (2002b) compared thesenses of community between members online science ction fan group and the membersneighborhoods. However, Obst and her colleagues found that the participants feelings ofmembership and inuence were weaker for SOVC compared to SOC.

    Koh and Kim (2003) initially draw from McMillan and Chavis conceptualization ofSOC. However, their nal measure was quite dierent, including a scale of immersion,which is not included in any other conceptualizations of SOC or SOVC. Additionally, theyonly used two of McMillan and Chavis components of SOC, membership and inuence,the same two components that Obst, Smith, and Zinkiewiczs (2002) study suggested maynot transfer well from SOC to SOVC. Therefore, it is not certain that their measure ofSOVC overlaps suciently with other researchers conceptualizations of SOC or SOVCto provide useful comparisons to SOVC.

    Blanchard and Markus (2004) examined SOVC in a newsgroup. Consistent with pre-vious researchers, they found similarities to McMillan and Chavis (1986) SOC, includingfeelings of membership, integration of needs, and shared emotional connections. How-ever, Blanchard and Markus found distinct dierences in their groups SOVC. The vir-tual group members did not report feeling that they inuenced or were inuenced byothers which would have been expected by McMillan and Chavis framework. Thismatches with Obst and co-workers (2002) research that inuence may not be as impor-tant in SOVC as in SOC. Also, the virtual group members reported feeling that simplyrecognizing others and relationships with specic other members were important to theirSOVC. Although these could also be components of SOC, they have not been explicitlyrecognized as part of McMillan and Chavis framework. Blanchard and Markus ndingagrees with Obst et al.s (2002) nding that feelings of membership, which include indi-viduals feelings of identity with the community and its members, may not be the samein SOVC as SOC.

    Blanchard and Markus (2004) propose that these dierences in inuence and the impor-tance of recognizing and having relationships with individuals could be due to the specicnature of interacting in the online environment. That is, information and communicationtechnologies (ICT) can be largely anonymous and members may have greater concernsabout the truth of others identities (see Joinson & Dietz Uhler, 2002). Thus, virtual com-

    munity members may have a heightened need to recognize specic individuals and develop

  • 2110 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123relationships with them to feel an SOVC and are less likely to be aware of their own inu-ence or the inuence of others.

    Understanding the similarities and dierences of SOVC compared to SOC is interestingand important; however, it is not as important as understanding how members of groupsexperience a SOVC. The dierences between SOC and SOVC, namely the stronger role ofindividual relationships and the weaker role of individual inuences suggest that the pro-cesses of learning the identity of others and the development of identity of oneself areimportant. Additionally, the previous research on SOC suggests that exchanging supportmay be important. Therefore, this research will draw on the identity and social identitytheories as well as social exchange theories to understand how SOVC develops.

    In addition, this research will examine how the perception of norms in a virtual com-munity mediates the identity and social exchange processes to increase SOVC. Althoughnot tested in FtF SOC research, McMillan and Chavis (1986) predict that norms wouldincrease SOC. They argue that as the community becomes more cohesive, there isgreater pressure on the community members to conform. This pressure creates a consen-sual validation among the community members, essentially a feeling among the membersthat we are alike. This, they argue, is what develops into part of the members SOC.Indeed, as members more closely adhere to the norms of the community, their bond tothe community increases. Thus, development and adherence to norms closely precedeSOC in FtF communities. We suggest that they will similarly lead to SOVC in virtualcommunities.

    4. Identity and social identity theories

    First, in developing a model of SOVC, we will consider identity processes in onlineinteractions. The issues of personal identity and identiability have played central rolesin understanding behavioral and aective outcomes online. It was been well establishedthat the ICT which supports virtual communities has fewer social cues than FtF commu-nication (e.g., Kiesler, Seigel, & McGuire, 1984). However, the presence of fewer cues doesnot necessarily mean that identity is less important in ICT than in FtF interactions (seeCulnan & Markus, 1987). The social information processing (SIP) model argues thatalthough there are fewer personal cues in ICT as compared to FtF interactions, relation-ship development is the same between online and FtF interactions (Walther, 1992, 1995).However, it simply takes a good deal more time for an appropriate amount of cues to beaccumulated in ICT and for the relationships to become similar in scope and magnitude.Research on group aective outcomes in ICT based on SIP has found that over time,feelings between members are about the same in virtual and FtF groups (Walther, 1992;Wilson, Strauss, & McEvily, 2006).

    In contrast, the social identity model of deindividuation eects (SIDE) (Postmes,Spears, & Lea, 1998) proposes that as individuals are more anonymous (i.e., when cuesto individual identity are sparse), this leads to greater immersion in the group andincreased salience of group identity, resulting in several positive group outcomes, suchas greater solidarity and unity. Specically, SIDE suggests that the presence of individu-ating information (e.g., a name or picture) in ICT highlights the individual identity ofonline group members, while making the social identity of the group less salient. In otherwords, ones cognitive representation of the group switches from an emphasis on us to

    an emphasis on you and me.

  • A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2111Although SIDE would suggest that cues to identity negatively inuence group-basedoutcomes, such as group aliation and SOVC, recent research suggests that this maynot always be the case. First, Walther (1996) developed a hyperpersonal model of relation-ship development that partially builds on the SIDE and SIP models. The hyperpersonalmodel argues that when members group identity is salient, members over-interpret theminimal cues that are present and idealize their partners. Thus, minimal cues increaseaect between partners in groups particularly when members identify with the group.Additionally, recent work by Postmes, Spears, Lee, and Novak (2005) suggests that groupand individual identity may coexist, and that expressions of individuality through commu-nication among group members may actually strengthen group identity and solidarity.

    Thus, although SIDE originally suggested that cues to others identity would decreasegroup-based outcomes, more recent theoretical and empirical evidence from SIP, thehyperpersonal model and even SIDE suggests that learning cues of other peoples identi-ties can accentuate certain intra-group outcomes, in this case, SOVC. However, this is onlytrue to a point. The SIDE model and the hyperpersonal model both argue that as cues tothe personal identity increase, deindividuation is likely to decrease and the social inuenceis likely to decrease (Postmes et al., 1998; Walther, 1996). It is not clear at what level oflearning others identity this occurs. That is, cues to others identity should increaseSOVC, but at some point, too much individuality will begin to decrease the members feel-ings of social unity and will decrease members SOVC. At this point in the model devel-opment, we can acknowledge this, but hypothesize that generally, cues to othersidentity increase SOVC.

    Although previous research has focused primarily on learning others identity, creatingones own identity in the group and believing that others understand it also plays animportant role in ICT outcomes (Ma & Agarwal, 2007). Ma and Agarwal focus on howdeveloping ones own identity aects participation in and satisfaction with the virtual com-munity. However, their attention to creating ones own identity highlights a neglected areaof the personal and social identities approach. As members learn of others identities, forexample, through the use of technological features (e.g., signature les), they also presentinformation about themselves using these same features. As they learn others identities,they perceive that others are learning theirs. Thus, as members perceive others individualcharacteristics as providing important cues as to the groups characteristics of solidarity(Postmes, Spears, & Lea, 2000; Tanis & Postmes, 2005), they may perceive that theirown identity cues could do the same. Therefore, we hypothesize that both learning othersand creating ones own identity are potentially related to SOVC.

    5. Social exchange theory

    Although learning the identity of other members is an important by product of inter-acting in virtual communities, it is not the main function of member participation. Indeed,the exchange of support is a very important reason for the existence of many virtual com-munities (Baym, 1997; Rothaermel & Sugiyama, 2001; Wellman & Guilia, 1999). Thisresearch adds to understanding SOVC by examining the support exchange behavior bygroup members through social exchange theories.

    There are a variety of ways in which members exchange support in virtual communities.Support may be exchanged publicly in posts for the entire group to read or may occur pri-

    vately through emails exchanged behind the scenes. Wellman and Guilia (1999) have

  • 2112 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123argued that the public exchange of support may increase members perceptions of being asupportive group when in fact, few people are actually involved in the supportiveexchange. Thus, there is a perception that the group is very supportive, even if only afew of the members actually help each other. However, because everyone can read themessage, all group members benet from the support exchange even if they were not activein creating it.

    Social exchange theory is one of the fundamental theories for understanding behaviorbetween individuals and within groups. It explains why people help each other, why theyexchange information, support, and love among other commodities (Cropanzano &Mitchell, 2005). Social exchange theory is based on the near universal norm of reciprocity(Goulder, 1960). This reciprocity can either be direct as in the help exchanged between twopeople or indirect when help is exchanged with an entire group (Flynn, 2005).

    Additionally, social exchange theory argues that peoples aective attachment is gov-erned by the entity with which they are exchanging support (Flynn, 2005). That is, ifthe exchange is dyadic, the attachment remains between the two social exchange partners.But if the exchange occurs indirectly within a group or organization, the attachment is tothe group or the organization. Therefore, in this model, we hypothesize that exchangingsupport in a virtual community is positively related to SOVC.

    6. Development and adherence to group norms

    Thus far, the model proposes that both identity processes and social exchange processescontribute to the development of SOVC. However, what is the mechanism by which theseprocesses lead to SOVC? The development and adherence to group norms may serve asone important mediator of this relationship. In support of this proposal, past research sug-gests that identity and social identity processes, as well as social exchange processes, leadto the formation of group norms. Members of groups, such as those in naturally formingvirtual communities, have been shown to create and then observe norms of behavior (Post-mes et al., 2005). In particular, through learning others members identity, they inductivelycreate a social identity, and subsequently develop norms about what this group does andwhat its particular characteristics are (Postmes et al., 2005).

    Similarly, Cropanzano and Mitchell (2005) argue that one of the basic tenets of socialexchange theory is that people develop and then are constrained by certain rules ofexchange, norms that serve as guidelines for peoples interactions. These norms of behav-ior can develop as people participate in the exchange (Cropanzano & Mitchell, 2005) or bymerely watching other people interact (Postmes et al., 2005). As discussed previously, vir-tual community members can observe others exchanging support and can also participatein it privately through email or publicly through posting messages to the group. Thus asmembers observe and also participate in the exchange of support, they are developingnorms of behavior. However, this is likely to happen only with publicly exchanging sup-port and not with privately exchanging support through email. Email connotes a dyadicrelationship and therefore should not have an eect on group norms.

    Finally, norms are believed to be an antecedent to FtF SOC (McMillan & Chavis,1986). As members perceive norms of behavior and adhere to these norms, their bondsto the group increase. It is likely that the same processes in virtual groups and SOVC,too. Therefore, this model proposes that norms mediate the relationships between identity

    and social exchange with SOVC.

  • Fig. 1 represents the model to be tested in this research.

    7. Study 1

    7.1. Methods

    7.1.1. Participants

    Participants were 216 members of ve online groups. The groups were non-randomlypicked from a list of listservs and usenet newsgroups. To be considered for this study,the groups had to be active with messages posted daily. Additionally, the topic aroundwhich the group had formed had to be similar on both the newsgroup and the listserv, withone exception to be discussed later.

    Two of the groups were listservs and three were newsgroups. All of the online groupshad social purposes (i.e., two food lover groups, two pet lover groups, and one age related

    Email Support

    Observing Support

    Learning Identity

    Creating Identity

    Post Support

    Norms Sense of Virtual Community

    Fig. 1. Proposed model of SOVC for Study 1.

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2113group). The groups were chosen to match each other in terms of topic and type of group.For example, there was one pet lover listserv and one pet lover newsgroup. The only groupthat did not match was the age related newsgroup in which members chatted about a vari-ety of topics. This group was included even though a comparable listserv group could notbe located because it was highly active. The mean age of the participants was 48(SD = 11.9). Sixty-two percentage of the participants were female. Fifty-ve percentageof the participants were from the newsgroups.

    7.1.2. Measures

    Sense of virtual community. Eighteen items were used to assess sense of community(Blanchard, 2007). Sample items include I think this is a good group for me to be a mem-ber, I anticipate how some members will react to certain questions or issues in thisgroup, and I feel obligated to help members of this group. Responses ranged from1 = strongly disagree to 4 = strongly agree.

    Learning and creating identity. To assess creating and learning identities, one itemwas used for each measure. I learn about people by reading their posts was usedto assess learning about others. People learn about me by reading my posts was used

  • 2114 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123to assess creating identity. Responses ranged from 1 = strongly disagree to 4 = stronglyagree.

    Observing and exchanging support. To assess exchanging support, three measures werecreated: one to assess members observation of support and two measures to assess theirparticipation in exchanging support through email and posting to the group. Two itemswere used to assess how much people observed support: Members of this group supporteach other and Members of this group help each other. Responses ranged from1 = strongly disagree to 4 = strongly agree.

    Emailing support was assessed from four items about the participants email behavior.Participants were asked how often they emailed a question to other members, emailed ananswer to other members, emailed a comment to other members or emailed a response.Responses ranged from 1 = never to 4 = quite a lot.

    Posting support was assessed from four items about the participants behavior on thegroup. Participants were asked how often they posted a question to other members, postedan answer to other members, posted a comment to other members or posted a response.Responses ranged from 1 = never to 4 = quite a lot.

    Norms. Members perceptions of the groups norms were measured by asking partici-pants one item People in this group know what the proper behavior is. Responses ran-ged from 1 = strongly disagree to 4 = strongly agree.

    7.2. Results

    7.2.1. Validating measures

    The rst step in analyzing the data is to determine that the measures used in this studyare valid and that there is not a serious problem with mono-method bias (Podsako, Mac-Kenzie, Lee, & Podsako, 2003). This was achieved by conducting an exploratory factoranalysis using principal axis factoring and a promax rotation (Fabrigar, Wegener, Mac-Callum, & Straham, 1999). Promax rotation was used as an oblique rotation because itis expected that the factors are likely to correlate with each other.

    Four factors were extracted to assess each of the multi-item scales in the study (i.e.,SOVC, observing support, emailing support, and posting support). The initial solutionindicated that four SOVC items loaded inappropriately on other factors, an issue notedin other research (Blanchard, 2007). These items were therefore eliminated from the scale.A nal exploratory factor analysis resulted in all items loading onto their appropriatescales and with commonalities in the appropriate range (Fabrigar et al., 1999).

    7.2.2. Hypothesis testingDescriptive analyses are in Table 1. The studys hypotheses were tested using structural

    equation modeling and path analyses on AMOS 6 (Kline, 2005) followed by calculatingSobel tests and condence intervals to determine the reliability of the mediation eect(MacKinnon, Lockwood, Homan, West, & Sheets, 2002; Shrout & Bolger, 2002).

    The original model was a poor t to the data with v2 (14) = 69.35, p < .001,RMSEA = .16, CFI = .83 and RMR = .04 (Kline, 2005). Modication indices suggestthat including direct relationships from observing and posting support to SOVC couldimprove the model. These relationships are theoretically valid indicating partial insteadof complete mediation by the norms variable. An additional analysis was, therefore, run

    including these paths.

  • The resulting modied model is much improved with v2 (9) = 16.184, p = .06,RMSEA = .07, CFI = .96 and RMR = .03. Even though the RMSEA is greater than.05,it is less than.08 and therefore represents reasonable error of approximation (Kline, 2005).Therefore, this modied model represents a reasonable approximation of the data.

    As expected, creating identity (b = .26, p < .001) and observing support (b = .38,p < .001) were related to norms. However, learning identity (b = .11, p = .32) and postingsupport (b = .02, p = .81) were not. Also, as expected, norms (b = .28, p < .001) were relatedto SOVC. Additionally, observing support (b = .20, p < .01) posting support (b = .25,p < .001) and emailing support (b = .29, p < .001) were also directly related to SOVC.

    Two Sobel tests were calculated to determine if norms partially mediated the relation-ship between observing support, creating identity and SOVC. The other Sobel tests werenot calculated because these variables (posting support and learning identity) failed theinitial test of being related to norms (Baron & Kenny, 1986). The perception of normsmediates the relationship between creating identity and SOVC (z = 2.13, p < .05, con-dence interval from.03 to.08) and also mediates the relationship between observing sup-port and SOVC (z = 3.12, p < .001, condence interval .06.11). Thus, although there isa direct relationship between creating ones own identity and observing support to SOVC,additional variance can be explained by examining how creating ones own identity andobserving support aects norms.

    7.3. Discussion

    Table 1Descriptive analyses of Study 1 variables

    Variable Mean SD 1 2 3 4 5 6

    1. SOVC 3.19 .41 (.93)2. Identify others 3.29 .53 .24** ()3. Identity of self 3.18 .53 .31*** .73*** ()4. Observe support 3.63 .51 .40*** .34*** .28*** (.85)5. Email support 2.41 .66 .45*** .06 .12 .19* (.88)6. Post support 2.91 .52 .43*** .15 .22** .16* .46*** (.85)7 Norms 3.25 .57 .40*** .22** .29*** .42*** .07 .12

    Note: N = 216. *p < .05, **p < .01, ***p < .001. Reliabilities are in the diagonal.

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2115The results of Study 1 partially support the research model. Norms mediates the rela-tionship between creating identity, observing support, and SOVC. However, learning iden-tity was not related to norms. SOVC was directly related to observing the exchange ofsupport as well as privately emailing and publicly posting support.

    Thus, in Study 1, believing that others know who one is through reading posts (i.e., cre-ating identity) is related to perceptions of norms of behavior, but knowing others identityis not. The latter nding supports the original SIDE model (Postmes et al., 1998) thatargues that as people are perceived as individuals online, unifying group processes (e.g.,norms, SOVC) break down. The former nding is more supportive of the SIP model (Wal-ther, 1995) and the more recent SIDE formulations (Postmes et al., 2000, 2005) even sug-gesting that as people perceive they are known in the group, they are likely to perceivegroup norms of behavior and feel a SOVC. Perhaps they are feel they are accountable fortheir behavior and feel more accepted in the group.

  • 2116 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123All three ways of exchanging support were related to SOVC, thus suggesting, as withFtF SOC, that support is important in SOVC. Only observing support was related tonorms, which additionally partially mediated its relationship to SOC. This suggests thatobserving others exchanging support increases perceptions of norms of behavior andincreases members attachment to the group as proposed by Flynn (2005).

    7.4. Limitations

    Although Study 1 partially supports the model, there are several limitations that sug-gest a second study. First, several of the measures used in Study 1 were 1-item measures(e.g., norms) and involved perceptions of behavior instead of actual behavior. For exam-ple, learning and creating identity focused on perceptions that members have that theyknow others identity and that others know theirs. Although this approach agrees withBlanchard and Markus (2004) study, participants in Blanchard and Markus study addi-tionally reported using the technical features of the communication medium (e.g., signa-ture les) to get to know other people as well as create their own identities.

    Additionally, the exchanging support items do not cover the full range of online behav-iors that can occur in observing as well as exchanging support. In particular, in this study,measures of observing others exchanging support is more of a perception that othersexchange behavior than a report of others behavior. Study 2 will address the limitationsby adding behavioral items to the measures of exchanging support, perceptions of norms,creating and learning identities.

    8. Study 2

    8.1. Methods

    8.1.1. Participants

    Participants were 277 members of 11 bulletin boards from Babycenter.com, a veryactive online information, support and commercial center for parents. The groups werenon-randomly chosen to reect stage of life-stage of parenting (pregnancy versus earlyparenthood). All the bulletin boards met the same minimal level of activity during theobservation period as Study 1. Members were recruited through a message posted to eachgroup. In exceptionally active boards, a message was added to the survey recruitmentthread to bump the post to the rst page of the groups messages. Average age of the par-ticipants was 29 (SD = 4.29) and 99% of the respondents were women, which is typical forBabycenter.com.

    8.2. Measures

    Sense of virtual community. The same 18 items from Study 1 were used to assess sense ofcommunity. However, the response scale was changed from 1 = strongly disagree to7 = strongly agree to allow for more scale sensitivity.

    Learning and creating identity. Six items were developed to assess peoples perceptionsof learning others identity. These items include I know the screen names of other peoplein this group and I know the real names of other people in this group. Three items were

    developed for perceptions creating ones identity Other people in this group know my

  • screen name, Other people in this group know my real name, and Other people know

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2117my personality in this group. Participants were asked how much they agree with theseitems and responses ranged from 1 = strongly disagree to 7 = strongly agree.

    To assess how participants used the technology features to learn and create identity,participants were asked how often they engaged in four items including do people in thisgroup put personal information about themselves or their families at the end of their mes-sage? and do you put personal information about you or your family at the end of yourmessages (like names & ages)? Responses ranged from 1 = never to 6 = all the time.

    Observing and exchanging support. Study 2 expanded on the measures developed inStudy 1 to assess observing and exchanging support. These additional items related tothe online behaviors that could be observed or enacted (e.g., asked a question, askedfor help, asked for support, provided information, shared experiences). Sixteen items weredeveloped for observing support and posting support; 14 items were developed for email-ing support because two of the public support items (e.g., posting a short comment andposting a message not related to the topic) were not appropriate in email. Participantswere asked how often they engaged in the behaviors and responses ranged from 1 = neverto 6 = all the time.

    Norms. Four items were used to assess perceptions of the groups norms including Iunderstand what appropriate behaviors are for this group and People generally behaveappropriately on this group. Responses ranged from 1 = strongly disagree to7 = strongly agree.

    8.3. Results

    Again, the rst step of the Study 2s analysis was measurement validation. Although thenumber of participants in the study represents an acceptable level of power in testing themodel once items have been collapsed into their respective scales, the 81 items present aproblem in validating our measures through a factor analysis. Estimates are for ve timesas many observations as there are variables (Stevens, 2001) which would call for over 405observations.

    To address this issue, two factor analyses were tested for the measurement model,1 opt-ing to assess the measures as rigorously as possible with the behavior measures evaluatedin one analysis (53 items, requiring 265 observations) and the aective and perceptionmeasures evaluated in the second (28 items, requiring 140 observations). This strategy isrigorous because mono-method bias is more likely within measures of similar type (behav-ior, aect) than between dierent types of measures (Podsako et al., 2003). Thus, if wecan determine appropriate loading of measures in these analyses, we can be less concernedabout mono-method biases in the analysis of our model. In all of the analyses, we usedprincipal axis factoring with a promax rotation (Costello & Osborne, 2005; Fabrigaret al., 1999).

    In the rst factor analysis assessing exchanging support (emailing support, observingsupport, posting support) and the adherence to norms factors, four factors were extractedwith the factor structure generally corresponding to our intended variables; however, threeitems from the observing others exchange support scale loaded inappropriately on the1 Variations on the analyses presented here yield very similar results.

  • norms factor and were thus deleted. In the last steps, the communalities of the remainingitems were examined and were adequate to reect the items reliability within the factorstructure.

    The second factor analysis assessed the SOVC and identity measures (creating identity,learning identity and using the identity technologies). Four factors were extracted thatwere consistent with the proposed structure of our variables. However, the same fourSOVC items from Study 1 again loaded onto inappropriate factors and were thus elimi-nated from the SOVC measure. In the last step, the communalities were examined anddetermined adequate. A nal analysis was run using only the top loading variables fromthe exchanging support items and all the remaining items from the other scales and con-rmed that the items loaded onto their respective constructs.

    2118 A.L. Blanchard / Computers in Human Behavior 24 (2008) 210721238.4. Hypothesis testing

    Table 2 contains the descriptive analyses of this study. The analysis strategy for testingmediation in this study is the same as Study 1, a SEM path analysis. As with Study 1, themodel was rst run testing for full mediation, which was a poor t to the data v2

    (15) = 86.11, p < .001, RMSEA = .13, CFI = .92 and RMR = .14. Modication indicesagain support a direct link to SOVC from posting and observing support as well as creat-ing identity. Because partial mediation still falls within the studys theoretical argument,these paths were included.

    This resulted in a much better t, v2 (8) = 15.80, p < .05, RMSEA = .06, CFI = .99 andRMR = .08. As predicted, observing support (b = .13, p < .05) and posting support one-self (b = .15, p = .06) were related to norms. In addition, creating ones own identity(b = .24, p < .05) and using the technologies that support identity (b = .14, p < .05) werealso related to norms. Unlike predicted, knowing others identity (b = .15, p = .07)was negatively and marginally related to norms.

    Also as predicted, the perception of norms (b = .46, p < .001) is strongly related toSOVC. However, emailing support (b = .04, p = .29) was not related to SOVC. Theadded direct relationships of observing support (b = .09, p < .05), participating in support(b = .19, p < .001), and creating identity (b = .28, p < .001) were related to SOVC.

    The variables which showed a direct relationship to norms were then tested for a medi-ation eect on SOVC. The relationships of SOVC to observing support (z = 2.15, p < .05,CI from .05 to .13), participating in support (z = 1.84, p < .06, CI from .03 to .10), creating

    Table 2Descriptive analyses of Study 2 variables

    Variable Mean SD 1 2 3 4 5 6 7 8

    1. SOVC 5.22 1.08 (.91)2. Identify others 4.92 1.65 .40*** (.97)3. Identity of self 4.21 1.98 .55*** .71*** (.99)4. ID technology 3.97 .94 .27*** .26*** .28*** (.65)5. Observe support 5.39 .69 .21*** .17*** .05 .27*** (.99)6. Email support 1.76 1.51 .04 .13* .10 .09 .02 (.95)7. Post support 3.03 1.14 .54*** .48*** .72*** .30*** .17** .16** (.97)8. Norms 5.99 .82 .61*** .15* .29*** .24*** .17** .01 .32*** (.81)

    Note: N = 272. *p < .05, **p < .01, ***p < .001. Reliabilities are in the diagonal.

  • identity (z = 2.31, p < .05, CI from .03 to .09) and using the identity technologies(z = 2.19, p < .05, CI from .04 to .11) are partially mediated by norms. Therefore, themodel is partially supported.

    9. Discussion

    Study 2 further supports the research model (see Fig. 2). Like previous research, theexchange of support positively aects SOVC. In Study 2, both observing support andpublicly participating in exchanging support had strong direct relationships to SOVC.However, unlike Study 1 and other previous research (cf., Blanchard & Markus,2004), the private exchange of support through email was not related to SOVC. Oneexplanation for this is the low level of using email to exchange support as reported bythe participants in Study 2. Study 1s participants reported exchanging support throughemail at only a slightly lower rate than through posts. But Study 2s participants reporteda substantially lower level of exchanging support through email. This suggests than whenemail is used, it positively aects SOVC. However, emailing support is not necessary fora SOVC. Future research should pursue how exchanging support through email aects

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2119virtual community functioning. For instance, do members of the group have individualrelationships with each other or not? Do new members use email was a way to ease intothe group or not?

    Participants perceptions that others know their identity positively aects SOVC. Thesendings agree with previous research that developing ones identity leads to more virtualcommunity involvement and SOVC (Blanchard & Markus, 2004; McKenna, Green, &Gleason, 2002).

    A signicant contribution of this research is that it goes beyond the direct relationshipsof antecedents and SOVC to understand the process by which these antecedents work.This study demonstrates that norms, the perception by group members that there are rulesthat govern behavior in the group, mediate the relationship between the antecedents andSOVC.

    Email Support

    Observing Support

    Learning Identity

    Creating Identity

    Post Support

    Norms Sense of Virtual Community

    ID Technology

    -.11,-.15

    .26*, .24*

    .14*

    .38***, .13*

    .02, .15t

    .29***, -.04

    .28***

    .20**,.09*

    .25***,.19***

    .20***,.46***

    Fig. 2. Results from Studies 1 and 2. Note: results from Study 2 are in bold and italicized. Numbers are

    standardized b coecients. tp = .06, *p < .05, **p < .01, ***p < .001.

  • 2120 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123Social identity theory predicted that members learning the identity of others anddeveloping an identity for themselves would lead to an increased SOVC through thedevelopment of norms. Both Study 1 and Study 2 support that members perceptionsthat others understand them leads to norms and then to SOVC. However, neither studyfound that perceiving the identities of others was related to group norms. Study 2 addi-tionally added in a behavioral measure, the use of technical features to create and learnidentity, for which norms also mediated the relationship to SOVC. These results suggestthat perceptions of that others know you is the more important perception in devel-oping norms. Perhaps it indicates a feeling of acceptance into or accountability in thegroup.

    The behavioral component of Study 2s measure of identity, using the identity technicalfeatures, is related to norms and reliably and indirectly aect SOVC. This behavioral mea-sure supports social identity theorys inductive development of norms. This nding is par-ticularly interesting because at the time of the study, participants had to type in theiridentity creating features (e.g., signature les) every time they posted a message. Clearly,using these features was important to the participants. In general, the use of technical fea-tures in online social processes tends to be a neglected area of study (Markus, 2005).Future research should compare groups with dierent forms of technical features that cre-ate identity to determine how they are used and which are more eective.

    Social exchange theory predicts that norms should mediate the relationships betweenthe public exchange of support and SOVC. The results from both studies suggest thatobserving others exchange support is important in this mediation process. Wellman andGuilia (1999) have argued that observing others exchange support is important, but thatit may promote slacking from other participants because these lurkers never have to con-tribute to the community. However, these results indicate that observing others exchangesupport helps create norms of behavior which lead to SOVC.

    Participating in the public exchange of support oneself had a weaker mediated relation-ship as indicated by the lower condence interval in Study 2, even though it had a strongerdirect relationship to SOVC. Along with Study 1s lack of a mediated relationship, thissuggests that active involvement in publicly exchanging support has less of a relationshipin developing group norms than would have been expected (e.g., Flynn, 2005).

    9.1. Limitations

    Like Study 1, Study 2 has limitations. Babycenter.com is composed primarily ofwomen. Therefore, the results of this study may be limited. However, Study 1s ndingsare generally similar. Therefore, generalizability may not be compromised. However,two studies increase the external validity of this model by demonstrating that it holds indierent populations.

    Of more concern for the generalizability of this model is that these are social groups andnot professional groups. Virtual community research needs to establish and examineSOVC in professional virtual communities. As with social online groups, some profes-sional groups may be more or less likely to have a SOVC. For example, professional list-servs that are more informational distributing (e.g., ISWORLD) may be less likely to havea SOVC because help and information may be less likely to be exchanged. However, otherprofessional groups may engage in more interactive exchanges and, thus, this model may

    be appropriate.

  • Chipuer, H. M., & Pretty, G. H. (1999). A review of the sense of community index: Current uses, factor structure,

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2121reliability and further development. Journal of Community Psychology, 27, 643658.Clark, S. C. (2002). Employees sense of community, sense of control, and work/family conict in native

    American organizations. Journal of Vocational Behavior, 61, 92108.Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Four recommendations for

    getting the most out of your data. Practical Assessment, Research and Evaluation, 10(7), 19.Cropanzano, R., & Mitchell, M. S. (2005). Social exchange theory: An interdisciplinary review. Journal of

    Management, 31(6), 874900.Culnan, M. J., & Markus, M. L. (1987). Information technologies: Electronic media and

    interorganizational communication. In F. M. Jablin, L. L. Putnam, K. H. Roberts, & L. W.Porter (Eds.), Handbook of organizational communication: An interdisciplinary perspective. NewburyPark, CA: Sage.

    Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Straham, E. J. (1999). Evaluating the use of exploratoryfactor analysis in psychological research. Psychological Methods, 3, 272299.

    Fisher, A. T., Sonn, C. C., & Bishop, B. J. (2002). Psychological sense of community: Research, applications andimplications. New York: Kluwer Academic/Plenum Publishers.

    Flynn, F. J. (2005). Identity orientation and forms of social exchange in organizations. Academy of ManagementReview, 30(4), 737750.

    Garca, I., Giuliani, F., & Wiesenfeld, E. (1999). Community and sense of community: The case of an urbanbarrio in Caracas. Journal of Community Psychology, 27, 727740.

    Goulder, A. W. (1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25,9.2. Conclusion

    Members of virtual communities experience a sense of virtual community which comesfrom the exchange of support within the community, the identity of other members andthemselves, and interacting with others outside of the virtual community. Additionally,exchanging support and creating identity help create norms of behavior in the group whichin turn increase SOVC.

    Virtual communities can provide a pro-social environment in which individuals andgroups benet from their interactions. Virtual communities and other online social sys-tems are likely to become integral parts of peoples lives. Our understanding of whatmakes them experienced as communities can help us develop and manage them.

    Acknowledgements

    The author would like to thank Lynne Markus, Jennifer Welbourne, Marla Boughtonand Chris Henle for helpful comments on earlier versions of this paper.

    References

    Baron, R. M., & Kenny, D. A. (1986). The moderatormediator variable distinction in social psychologicalresearch: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology,51, 11731186.

    Baym, N. (1997). Interpreting soap operas and creating community: Inside an electronic fan culture. In S. Keisler(Ed.), Culture of the Internet. Manhaw, NJ: Lawrence Erlbaum Associates.

    Blanchard, A. L. (2007). Developing a sense of virtual community measure. CyberPsychology & Behavior, 10(6),827830.

    Blanchard, A. L., & Markus, M. L. (2004). The experienced sense of a virtual community: Characteristics andprocesses. The DATA BASE for Advances in Information Systems, 35(1), 6579.

    Burroughs, S. M., & Eby, L. T. (1998). Psychological sense of community at work: A measurement system andexplanatory framework. Journal of Community Psychology, 26, 509532.161178.

  • 2122 A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123Joinson, A. N., & Dietz Uhler, B. (2002). Explanations for the perpetration of and reactions to deception in avirtual community. Social Science Computer Review, 20(3), 275289.

    Jones, Q. (1997). Virtual-communities, virtual settlements & cyber-archaelogy: A theoretical outline. Journal ofComputer-Mediated Communication, 3(3), 24.

    Kiesler, S., Seigel, J., & McGuire, T. (1984). Social psychological aspects of computer-mediated communication.American Psychologist, 39, 11231134.

    Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York: Guilford.Koh, J., & Kim, Y.-G. (2003). Sense of virtual community: A conceptual framework and empirical validation.

    International Journal of Electronic Commerce, 8(2), 75.Ma, M., & Agarwal, R. (2007). Through a glass darkly: Information technology design, indentity verication,

    and knowledge contribution in online communities. Information Systems Research, 18(1), 4267.MacKinnon, D. P., Lockwood, C. M., Homan, J. M., West, S. G., & Sheets, V. (2002). A comparison of

    methods to test the signicance of the mediated eect. Psychological Methods, 7, 83104.Markus, M. L. (2005). Technology-shaping eects of e-collaboration technologies: Bugs and features.

    International Journal of e-Collaboration, 1(1), 123.McKenna, K. Y. A., Green, A. S., & Gleason, M. E. J. (2002). Relation formation on the Internet: Whats the big

    attraction? Journal of Social Issues, 59(1), 923.McMillan, D. W., & Chavis, D. M. (1986). Sense of community: A denition and theory. Journal of Community

    Psychology, 14, 623.Obst, P., & White, K. M. (2004). Revisiting the sense of community index: A conrmatory factor analysis.

    Journal of Community Psychology, 32(6), 691705.Obst, P., Smith, S. G., & Zinkiewicz, L. (2002). An exploration of sense of community, Part 3: Dimensions and

    predictors of psychological sense of community in geographical communities. Journal of CommunityPsychology, 30(1), 119133.

    Obst, P., Zinkiewicz, L., & Smith, S. G. (2002a). Sense of community in science ction fandom, part 1:Understanding sense of community in an international community of interest. Journal of CommunityPsychology, 30(1), 87103.

    Obst, P., Zinkiewicz, L., & Smith, S. G. (2002b). Sense of community in science ction fandom, part 2:Comparing neighborhood and interest group sense of community. Journal of Community Psychology, 30(1),105117.

    Podsako, P. M., MacKenzie, S. B., Lee, J. Y., & Podsako, N. P. (2003). Common method biases in behavioralresearch: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88,879903.

    Postmes, T., Spears, R., & Lea, M. (1998). Breaching or building social boundaries? Side eects of computer-mediated communication. Communication Research, 25, 689718.

    Postmes, T., Spears, R., & Lea, M. (2000). The formation of group norms in computer-mediated communication.Human Communication Research, 26(3), 341371.

    Postmes, T., Spears, R., Lee, A. T., & Novak, R. J. (2005). Individuality and social inuence in groups: Inductiveand deductive routes to group identity. Journal of Personality and Social Psychology, 89(5), 747763.

    Pretty, G. H., & McCarthy, M. (1991). Exploring psychological sense of community among women and men ofthe corporation. Journal of Community Psychology, 19, 351361.

    Rheingold, H. (1993). The virtual community: Homesteading on the electronic frontier. Reading, MA: Addison-Wesley.

    Roberts, L. D., Smith, L. M., & Pollock, C. M. (2002). MOOing till the cows come home: The sense ofcommunity in virtual environments. In C. C. Sonn (Ed.). Psychological sense of community: Research,applications, implications. New York: Kluwer Academic/Plenum.

    Rothaermel, F. T., & Sugiyama, S. (2001). Virtual internet communities and commercial success: Individual andcommunity-level theory grounded in the atypical case of TimeZone.com. Journal of Management, 27(3),297312.

    Royal, M. A., & Rossi, R. J. (1999). Predictors of within-school dierences in teachers sense of community.Journal of Educational Research, 92, 259267.

    Sarason, S. B. (1986). Commentary: The emergence of a conceptual center. Journal of Community Psychology, 14,405407.

    Schuler, D. (1996). New community networks: Wired for change. New York: ACM Press.Schuster, E. (1998). A community bound by words: Reections on a nursing home writing group. Journal ofAging Studies, 12(2), 137148.

  • Shrout, P. E., & Bolger, N. (2002).Mediation in experimental and non-experimental studies: New procedures andrecommendations. Psychological Methods, 7, 422455.

    Stevens, J. P. (2001). Applied multivariate statistics for the social sciences. Lawrence Erlbaum.Tanis, M., & Postmes, T. (2005). A social identity approach to trust: Interpersonal percepetion, group

    membership and trusting behaviour. European Journal of Social Psychology, 35, 413424.Walther, J. B. (1992). Interpersonal eects in computer-mediated interaction: A relational perspective.

    Communication Research, 19(1), 5291.Walther, J. B. (1995). Relational aspects of computer-mediated communication: Experimental observations over

    time. Organization Science, 6(2), 186203.Walther, J. B. (1996). Computer mediated communication: Impersonal, interpersonal and hyperpersonal

    interaction. Communication Research, 22, 3343.Wellman, B., & Guilia, M. (1999). Net surfers dont ride alone: Virtual communities as communities. In B.

    Wellman (Ed.), Networks in the global village: Life in contemporary communities. Westview.Wilson, J. M., Strauss, S. G., & McEvily, B. (2006). All in due time: The development of trust in computer-

    mediated and face-to-face teams. Organizational Behavior and Human Decision Processes, 99, 1633.Za, J., & Devlin, S. (1998). Sense of community in housing for the elderly. Journal of Community Psychology, 26,

    381398.

    A.L. Blanchard / Computers in Human Behavior 24 (2008) 21072123 2123

    Testing a model of sense of virtual communityIntroductionSense of communitySense of virtual communityIdentity and social identity theoriesSocial exchange theoryDevelopment and adherence to group normsStudy 1MethodsParticipantsMeasures

    ResultsValidating measuresHypothesis testing

    DiscussionLimitations

    Study 2MethodsParticipants

    MeasuresResultsHypothesis testing

    DiscussionLimitationsConclusion

    AcknowledgementsReferences