Why do members contributeknowledge to online

communities?Shih-Wei Chou

Department of Information Management,National Kaohsiung First University of Science and Technology,

Kaohsiung, Taiwan, Republic of China

Abstract

Purpose – This study aims to better understand individuals’ motives for contributing knowledge inan online community.

Design/methodology/approach – An integrated model is developed based on a motivationalmodel and social cognitive theory. To validate the model two online communities: the ElectronicEngineering Times in Taiwan and China were surveyed.

Findings – It was found that both perceived identity verification and performance expectancy arepositively associated with satisfaction, which in turn affects knowledge contribution. Performanceexpectancy is affected by both computer self-efficacy and computer anxiety, and perceived identityverification is influenced by members’ innovativeness in IT.

Originality/value – This is the first study which aims to assess the relationships betweenindividuals’ differences, intrinsic and extrinsic motivation, and knowledge contribution. The findingscan help managers to build an effective community.

Keywords Motivation (psychology), Individual behaviour

Paper type Research paper

IntroductionThe importance of an online community has been recognised by prior studies (Lee et al.,2002; Ma and Agarwal, 2007), because it can help to achieve a wide variety of activitiessuch as knowledge sharing and developing relationships. Online communities are a neworganisational form where people communicate and interact, develop relationships, andcollectively and individually seek to attain some goals. Despite a significant growth inthe number of online communities, studies show that very few are successful at retainingtheir members and motivating their knowledge contribution (Butler, 2001; Wasko andFaraj, 2005). For example most of the communities (91.2 percent) on MSN(www.msn.com) had fewer than 25 members, and the communities averaged betweenone and 20 posts (Ma and Agarwal, 2007). With the recent emergence of new types ofonline communities such as Facebook, Twitter, and LinkedIn, how to increase thenumber of members and posts remains a problem (Lin and Bhattacherjee, 2009; Wangand Chiang, 2009). Communities can be a significant source of value for participantsprovided that members are willing to contribute valuable knowledge to the communities.Thus understanding members’ motives for knowledge contribution and identifying thedeterminants of the motivations are the goals of this study.

Successful knowledge contribution in a virtual environment involves effectivesocial interaction (Chiu et al., 2006; Lin and Bhattacherjee, 2009; Sussman and Siegal,

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/1468-4527.htm

Onlinecommunities

829

Refereed article received17 November 2009

Approved for publication14 March 2010

Online Information ReviewVol. 34 No. 6, 2010

pp. 829-854q Emerald Group Publishing Limited

1468-4527DOI 10.1108/14684521011099360

2003). As prior work has shown (Ma and Agarwal, 2007; Wang and Chiang, 2009),mediated communication suffers from social cue deficiencies because the transmissionof important contextual cues such as body language is difficult to achieve throughcomputer channels. In a virtual environment, the lack of synchronicity and immediacytends to attenuate the influence of social norms on members’ behaviour and to lead tosocial loafing (Chidambaram and Tung, 2005). In addition the establishment of mutualunderstanding to comprehend conversations and knowledge contribution is moredifficult than face-to-face communication in a small group because of members’ diversesocial backgrounds and perspectives.

Despite the above difficulties evidence shows that individuals do engage inprosocial behaviours such as knowledge contribution in online communities (Eastinand LaRose, 2005; Hertel et al., 2003; Huang et al., 2009). Research on online knowledgesharing demonstrates that a variety of drivers motivate this behaviour, including theanticipation of extrinsic benefits (economic rewards, performance expectancy),intrinsic benefits (sense of self-worth, social norms), and social capital (Chiu et al., 2006;Venkatesh et al., 2003; Wasko and Faraj, 2005). We theorise that two factors play a keyrole in knowledge contribution behaviour in an online community: performanceexpectancy and perceived identity verification (PIV). Performance expectancy refers tothe degree to which an individual believes that using an online community will helpthem to improve their job performance (Venkatesh et al., 2003). PIV is defined asindividuals’ beliefs that they are able to successfully communicate with their onlineidentity (i.e. who they are in an online community). Identity refers to an individual’sself-appraisal of a variety of attributes along the dimensions of physical and cognitiveabilities, personal traits and motives, and the multiplicity of social roles includingworker and community citizen (Whitbourne and Connolly, 1999). Individualsparticipating in both offline and online social interaction seek to be understood asthe person they believe themselves to be (Swann et al., 1989). Studies (Berman andBruckman, 2001; Ma and Agarwal, 2007) show that for most participants in atechnology-mediated context, identity – both the establishment of their own reputationand the recognition of others – plays a key role in intrinsic benefits such as anamplified sense of self-worth or recognition. Following Ma and Agarwal (2007), weargue that the extent to which individuals believe they are able to successfullycommunicate their online identity is positively associated – both directly and throughmediation by satisfaction – with knowledge contribution in the community. Improvedjob performance has been recognised as a key extrinsic motivation (Hsu et al., 2007;Venkatesh et al., 2003), i.e. individuals will want to perform an activity if it is perceivedto be instrumental in achieving valued outcomes. An online community is likely toprovide organisational members with an opportunity to gain access to newinformation, expertise, and ideas not available locally, which in turn may lead to higherjob performance (Wasko and Faraj, 2005). We therefore argue that higher performanceexpectancy which serves as extrinsic motivation relates – both directly and throughmediation by satisfaction – to knowledge contribution in the community.

Since technology is the foundation and medium through which communitymembers interact, it is one of the key determinants of the dynamics of the community.The relationships between technology and social relationships have been recognisedby prior research (Walther et al., 2001), arguing that technologies and social systemsevolve together and that technologies may result in different outcomes regarding

OIR34,6

830

member behaviour and ongoing community activities. For example research onhuman-computer interaction reports that information technology (IT) built to supportreal-time conversation (e.g. instant messaging) and social networks allows users tocreate new social relationships (Ma and Agarwal, 2007, Wang and Chiang, 2009) or toincrease their job performance (Venkatesh et al., 2003). However most prior research oncommunity design does not provide a theoretical explanation for these effects. To fillthis gap, this study aims to understand why members’ motivations, in terms ofperformance expectancy and PIV, are affected by their individual characteristics (orindividual differences) based on social cognitive theory (Bandura, 1986; Thatcher andPerrewe, 2002). Individual differences refer to factors such as personality, situational,and demographic variables that influence users’ beliefs about and use of IT. Empiricalstudies show that stable situation-specific individual differences such as personalinnovativeness in IT (PIIT) as well as dynamic, situation-specific individualdifferences such as computer self-efficacy (CSE) and computer anxiety (CA) have animpact on how IT users perceive and use IT (Agarwal and Prasad, 1999; Thatcher andPerrewe, 2002).

We propose and test a theoretical model examining the impact of individualdifferences on community members’ perceived identity verification by others and onperformance expectancy. Although prior research has alluded to the influence of thesefactors, such as individual differences on IT use (Thatcher and Perrewe, 2002), therelationship between PIV and knowledge contribution, and the impact of performanceexpectancy on IT behaviour, no study we are aware of has established and tested anintegrated model for the relationships among individual differences, performanceexpectancy, PIV, and members’ knowledge contributions. This study differs from priorresearch on knowledge contribution in virtual settings in two ways. First based on amotivational model (Davis et al., 1992), the proposed model extends that of Ma andAgarwal (2007) by examining both PIV and performance expectancy, which serve asintrinsic and extrinsic motivation respectively. Second drawing on social cognitivetheory, our model highlights how individual differences affect the foregoingmotivation. The proposed model deepens our understanding of how individuals’knowledge contribution in communities is shaped by their motivations andcharacteristics. Based on primary survey data collected from more than 240 users oftwo distinct online communities, our findings provide important implications for bothresearch and practice.

Theoretical background and research hypothesesKnowledge contribution in online communitiesDrawing on social-psychological perspectives, studies have identified the factorssalient to knowledge contribution in an online community. For example Constant et al.(1996) examine the use of email for helping organisation members to seek and giveinformation and find that citizenship behaviour and the desire to benefit theorganisation are the major motivations for helping behaviour. In studies of onlinecommunities of professionals Wasko and Faraj (2005) point out that reputation,altruism, generalised reciprocity, and community interest play a key role in motivatingmembers to contribute knowledge. Based on usage of electronic knowledgerepositories Kankanhalli et al. (2005) report that both extrinsic benefits (in terms ofreciprocity and organisational reward) and intrinsic benefits (in terms of knowledge

Onlinecommunities

831

self-efficacy and enjoyment in helping others) influence knowledge contribution.Similarly research on individual contributions to tasks conducted on the internet (suchas product reviews of an internet shop) (Ma and Agarwal, 2007; Wasko and Faraj,2005) found that social affiliation, professional self-expression, reputation benefits, andsocial capital play a key role in individuals’ motivations. Other individualcharacteristics, including user experience, recognition from the site, and individualattributes such as being a hobbyist, have been recognised as individual motives forcontribution (Jeppesen and Frederiksen, 2006). The empirical research of Chiu et al.(2006) on virtual communities shows that knowledge sharing is affected by socialcapital and outcome expectations.

A common theme underlying the above literature is that the design of thecommunity is assumed to be immutable. A second stream of research focuses onmanipulating the social-psychological factors underlying knowledge contribution andintegrating them into the management of the community to promote contribution (Linget al., 2005). Most of the prior studies emphasise the establishment of special featuresby which members’ contributions can be recognised, including design factorsreminding users about the uniqueness of their contribution, and the manner in whichpostings are organised and disseminated. Extending this we contend that in order tofacilitate the efficiency of the above features, it is important to understand therelationship between individual differences and users’ motivations.

The reasons for the above argument are twofold. First as noted previously,individual differences affect users’ beliefs about and use of IT. In addition to thefeatures provided by IT, the individual differences regarding IT use may also affect theability of IT to facilitate communication and interaction in an online community. Forexample users with higher computer self-efficacy are more likely to form positiveperceptions of IT and are more likely to use IT frequently (Venkatesh et al., 2003). Thisin turn may lead to users’ positive beliefs about benefits derived from using IT (ormotivation). Second a mediated environment can assist knowledge accumulation byprocessing and presenting information in new and flexible ways, and by changing ordistorting the social context to promote equal conversation (Connolly et al., 1990; Maand Agarwal, 2007). The extent to which members establish such a mediatedenvironment may depend on their individual differences. For example individuals withhigher personal innovativeness in IT are more likely to make good use of thecommunity artefacts so that they may contribute knowledge to the community moreefficiently and effectively (Thatcher and Perrewe, 2002). As a result they may be moremotivated, in terms of performance expectancy or PIV, to contribute knowledge thanthose who are less innovative. However despite the ubiquitous use of many communityartefacts, few of the prior studies aim to understand why and how individualdifferences affect knowledge contribution. Our research seeks to address this gap.

Drawing on social cognitive and social psychology theory rooted in the concept ofidentity, we develop a theoretical model delineating the relationships among individualdifferences, outcome expectations, PIV, and knowledge contribution. In the discussionthat follows we first establish the role of motivations, in terms of outcome expectations(or performance expectancy) and PIV, in online contexts. Second we describe howindividual differences, in terms of CSE, CA, and PIIT, affect community users’motivation. Our research model is shown in Figure 1.

OIR34,6

832

Knowledge contribution, satisfaction, and community users’ motivationThe key dependent variable that this study emphasises is knowledge contribution.Membership in a community is fundamentally a social relationship. Studies have foundthat satisfaction with social relationships is positively associated with relationshipcontinuance and commitment (Givertz and Segrin, 2005). The more individuals aresatisfied with the use of a community, the more likely they are to affectively andnormatively commit to the relationship and engage in behaviours that will maintain ahealthy relationship, such as providing help or accommodating others’ needs (Rusbultand Buunk, 1993). The more members are satisfied with the community, the more theyhelp maintain a better relationship and extend the resources available to all communitymembers (Ma and Agarwal, 2007). This in turn may lead to more knowledgecontribution.

H1. An online community member’s satisfaction with the community is positivelyrelated to their knowledge contribution.

Studies suggest that post-consumption expectation plays a key role in satisfaction withIT use (Davis et al., 1989). The above expectation has been conceptualised as ex postusefulness (Bhattacherjee, 2001) or performance expectancy (Venkatesh et al., 2003).Research on IT behaviour also points out that outcome expectations are beliefs salientto IT acceptance behaviours across a broad range of end-user computing technologiesand user populations (Chiu et al., 2006; Taylor and Todd, 1995) including members ofonline communities. As noted previously performance expectancy refers to thecapabilities of a system that enhances an individual’s job performance. The moreperformance expectancy members have when using a community, the more likely theyare to have higher post-acceptance affect such as satisfaction (Chiu et al., 2006;Venkatesh et al., 2003). Thus,

H2. An online community member’s performance expectancy is positively relatedto their satisfaction with the community.

Figure 1.Research model

Onlinecommunities

833

According to Davis et al. (1989) people form intentions toward behaviours they believewill increase their job performance, over and above whatever positive or negativefeelings may be evoked toward the behaviour. Improving performance is helpful forachieving various rewards that are extrinsic to the task context, such as promotions ormonetary gains (Bhattacherjee, 2001). IT use such as participating in an onlinecommunity aimed at knowledge exchange is often viewed as the means to that end.Davis et al. (1989) also suggest that the above means-end behaviour is largely based oncognitive decision rules, indicating that even without necessarily activating thepositive affect related to performance-contingent rewards, the heuristics of IT users areinvoked without conscious thought whenever they are faced with similar behaviouralcontexts. Thus in an online community context the more members feel that engaging inthe community focused on knowledge exchange tends to increase their jobperformance, the more likely they are to devote their time and effort to knowledgecontribution.

H3. An online community member’s performance expectancy is positively relatedto their knowledge contribution.

As noted previously, perceived identity verification refers to the perceivedconfirmation from other community members of a focal person’s belief about theirown identity. Following Ma and Agarwal (2007) PIV is conceptualised as a perceptualconstruct. The reasons are twofold. First although it is possible that people mayperceive more self-confirmation than actually exists (Swann et al., 2004), empiricalstudies show that individuals’ actual behaviour is affected by their perceptionsregardless of accuracy. For example research on cognitive dissonance in interpersonalcontexts (Matz and Wood, 2005) found that the extent of consonance depends mainlyon the individual’s beliefs about what group members think. Thus how the individual’ssalient referents actually perceive them is less important; what is important is theindividual’s perceptions of others’ assessments.

Second, this study measures the perceived confirmation of identities rather than theobjective agreement between an individual’s self-view and others’ appraisal, becausethe operational requirements for assessing the latter would be overwhelming. Researchon self-verification has typically investigated dyads or small groups of three or fourmembers, but an online community can be relatively large. Thus it is difficult if notimpossible for an individual to determine the true beliefs of salient referents (Whittakeret al., 1998). In addition from the perspective of measurement and operationalisation,the extant concept of self-verification in face-to-face settings cannot be applied directlyto the online context. For example in the study by Swann et al. (2000) perceived identityis conceptualised by asking subjects to rate each other to obtain a measure of actualagreement. It is not feasible to request each subject to rate individually the hundreds ofother respondents in online settings.

Extending Ma and Agarwal’s (2007) research we explore the role of PIV from twoperspectives: the influence of PIV on performance expectancy and on knowledgecontribution. For the former we expect that members’ PIV is positively related to theirperformance expectancy. The reasons are twofold. First information acquisition ismore efficient when the expert is identifiable. Knowing the identity of knowledgecontributors helps knowledge seekers recognise the source credibility of a community(Sussman and Siegal, 2003), which in turn facilitates members’ knowledge adoption

OIR34,6

834

and knowledge exchange between members. Therefore the more an individual’sidentity is recognised by others in a community, the more likely they are to feel thatthey have more knowledge than those who have lower PIV. Since individuals’knowledge is related to their job performance, high PIV is likely to lead to highperformance expectancy. Second PIV facilitates relationship-building, because themore PIV members perceive, the more likely they are to feel that members of thecommunity possess similar interests or attitudes and that better relationships betweenmembers can be built ( Jensen et al., 2002; Wasko and Faraj, 2005). Thus an individualtends to believe that other members would provide useful information, helping them tocomplete a task and increase their own job performance. Based on the abovearguments we expect that individuals’ perceived identity verification is positivelyassociated with their performance expectancy.

H4. An online community member’s PIV is positively related to their performanceexpectancy.

Satisfaction indicates whether a member is content with the use of the community orwith their access to the community resources. Self-verification theory suggests thatpeople seek confirmation of their identities (Swann et al., 2000). The more PIV amember perceives, the more likely that they develop a sense of coherence andunderstanding, promoting positive attitudes such as satisfaction (Hertel et al., 2003).The longitudinal study by Swann et al. (2000) confirms this by showing that personalidentity (academic ability and social competence) verification heightens participants’feelings of connection to their group and amplifies satisfaction with the groupinteraction. When individuals’ identities are recognised and verified by other membersof an online community, they feel better understood and are more likely to believe theywill be treated in desired ways (Ma and Agarwal, 2007). In addition higher PIVindicates that individuals are able to interact with others more easily with fewermisunderstandings, resulting in low differences between interaction partners’expectations and the focal person’s self-view. In other words as individuals’ PIVincreases, they believe that they can better predict and control how the socialinteraction proceeds, leading to more satisfaction.

H5. An online community member’s perceived identity verification is positivelyrelated to their satisfaction with the community.

As noted previously the more PIV an individual has, the more likely that they are ableto develop a sense of understanding and coherence (Swann et al., 2000, 2004). Researchon self-verification shows that people prefer interacting with partners who verify theiridentities than with those who do not (Swann et al., 1989). Studies also report thatacknowledgement from group members increases a person’s contribution (Hertel et al.,2003). PIV also positively affects an individual’s belief: the more the individual’sself-view is confirmed by other members of a community, the more likely they are tobelieve that they are able to achieve better interpersonal discourse (Matz and Wood,2005), which motivates the focal individual to continue the interaction and contribute tothe community.

H6. An online community member’s perceived identity verification is positivelyrelated to their knowledge contribution.

Onlinecommunities

835

Individual differencesIn addition to the impact of individuals’ motivations (PIV and performance expectancy)on knowledge contribution, this study delineates the relationship between individualdifferences and the above motivations. As noted previously individual differences referto the situational variables that affect users’ beliefs about and use of IT. Specificallythis study emphasises dynamic individual differences (CSE and CA) and stableindividual differences (PIIT).

Computer self-efficacy refers to individuals’ judgment of their capabilities of usingIT in diverse situations (Compeau and Higgins, 1995). The theory suggests thatindividuals who possess high CSE are more likely to form positive perceptions ofinformation systems (IS) and to use IS more frequently (Compeau et al., 1999; Thatcherand Perrewe, 2002), showing a strong link between CSE and individual reactions to IT,both in terms of adoption and use of IS, and in terms of learning to use IS. Our beliefsabout our capabilities to use technology successfully are associated with our decisionsabout whether and how much to use technology, and the extent to which we are able tolearn from training (Compeau et al., 1999). Social cognitive theory (Bandura, 1986)suggests that the more computer self-efficacy IS users have, the more likely it is thatthey have the requisite skills and confidence in their skills to be successful in their use,leading to higher performance expectancy. The study by Venkatesh et al. (2003)identifies the similarity between outcome expectations and performance expectancy,and evidence shows that CSE has a positive influence on outcome expectations(Compeau et al., 1999). Based on the above arguments, we propose H7.

H7. An online community member’s computer self-efficacy is positively related totheir performance expectancy.

Computer anxiety refers to fears about the implications of computers such as the lossof important data or fear of other possible mistakes or malfunctions (Thatcher andPerrewe, 2002). The theory suggests that IS users with high computer anxiety tend touse IS less, because computer anxiety represents the negative side of IS users’ affectiveresponses to using computers (Bandura, 1986; Marakas et al., 2000). Individuals withhigh computer anxiety have less confidence in their computer capabilities. Thus theytend to demonstrate a weaker proclivity to use computers (Compeau et al., 1999) whichin turn may reduce the performance expectancy. Thus we test

H8. An online community member’s computer anxiety is negatively related totheir performance expectancy.

Personal innovativeness refers to individuals’ willingness to change, and it is believedto be a function of individuals’ tolerance of risk (Lu et al., 2008; Thatcher and Perrewe,2002). PIIT is defined as the willingness of an IS user to try out any new IT and isexpected to have a stable influence across situations involving IS (or a stable trait)(Agarwal and Prasad, 1998). Research suggests that highly innovative individualsmore frequently seek out new, mentally or sensually stimulating experiences (Uray andAyla, 1997). According to self-presentation theory identity communication reflects anindividual’s effort to express and present their identity to others with the goal ofachieving a shared understanding (Goffman, 1967). This can be regarded as innovatingand stimulating activities when using communities and participating in socialdiscourse. We therefore posit that the more PIIT members of a community have, the

OIR34,6

836

more likely they are to have higher perceived identity verification because they tend totreat PIV as an important tool through which they can make better use of thecommunity.

H9. An online community member’s personal innovativeness in IT is positivelyrelated to their perceived identity verification.

MethodologySample and data collectionWe tested the hypotheses using primary survey data collected from two well-knownonline communities, Electronic Engineering Times located in Taiwan(www.eettaiwan.com) and China (www.eetchina.com). These communities provide aforum for electronic products’ daily news, technical papers and application notes fordesign, test and production engineering. Both communities fall into the category of “acommunity of common interest or information exchange” (Armstrong and Hagel,1996). Although these two communities have not existed for long, their membership isgrowing at a fast rate. This is so because these communities aim to cover a diverserange of subjects related to electronic engineering, including Lynx design systems, 3Dvisual design, instrument drivers and so on, so that members can either acquireknowledge or they may have an exchange of views about a specific subject that theyare interested in learning about. Through the inclusion of two communities, we aim toextend the generalisability of the findings regarding how individuals’ extrinsic andintrinsic motivations, conceptualised as performance expectancy and PIV respectively,affect their knowledge contribution.

MeasuresThe survey items are listed in the Appendix. They were adapted from measures thathad been validated by related studies, and have been modified to fit the context of thisstudy. To develop the scales for this study, we conducted exploratory interviews withten community members from five different communities. They are not in our sample.In addition a pilot test with 30 individuals was performed to validate the instrument.

Seven constructs were measured in this study: knowledge contribution, satisfaction,performance expectancy, PIV, computer self-efficacy, computer anxiety, and PIIT. Theconstructs were measured using multiple-item scales, drawn from pre-validatedmeasures in IS use or knowledge management (KM) research (wherever possible), andreworded to relate specifically to the context of KM-related community. We adoptedthe knowledge contribution measures from prior studies (Ma and Agarwal, 2007;Wasko and Faraj, 2005). Following Bhattacherjee (2001) satisfaction items were basedon seven-point semantic differential scales. The scale items and the description of theabove variables are given in the Appendix.

Regarding users’ extrinsic motivation, the four items of performance expectancyidentified by Venkatesh et al. (2003) were adapted to develop a four-item scale forassessing the extent of performance expectancy. Intrinsic motivation was conceptualisedas perceived identity verification, which was adapted from Kuhn and McPartland’s(1954) modified Twenty Statements Test (TST), aiming to capture the salient identitiesof each community member. The TST asks respondents to complete 20 statements suchas “I am___.” It is an open-ended identity measure that has been used and validated byprior studies (Kuhn and McPartland, 1954; Ma and Agarwal, 2007). The TST specifically

Onlinecommunities

837

acknowledges that an individual can have multiple identities and that different identitiesmay become dominant in different contexts. Following Ma and Agarwal’s (2007) study,we reduced the number of items from 20 to five to minimise the effects of fatigue. Aftercompleting the TST, the respondents were asked to rate their perception of othercommunity members’ verification of those five identities separately, using two items (seeAppendix). We used two items for each solicited identity, yielding ten items in total.Factor analysis showed that three of the five identities load on their correspondingfactors, indicating they were salient for the subjects. Thus six items used to measureperceived identity verification are retained.

As to the individual differences that may affect individuals’ motivations, they wereoperationalised based on related studies. Computer self-efficacy was measured usingten items developed by Thatcher and Perrewe (2002), and computer anxiety wasmeasured using four items based on the computer anxiety rating scale. Both of theabove constructs have been validated by prior work (Thatcher and Perrewe, 2002;Thatcher et al., 2007). Finally PIIT was measured using four items validated by Yi et al.(2006). The items in the questionnaire were measured using seven-point scalesanchored from “strongly disagree” to “strongly agree.”

To ensure the equivalence of the questionnaires in two languages, backwardtranslation was used (with the material translated from English into Chinese then backinto English, the versions compared, and discrepancies resolved) as suggested by priorstudies (Bock et al., 2005). An English version of the questionnaire was first compiledand modified to suit the context of an online community and then translated intoChinese by a bilingual research associate. The Chinese version of the questionnairewas verified and refined for its accuracy of translation by one MIS professor and twosenior doctoral students, who were familiar with and had done extensive research on ISbehaviour and the KM of virtual communities. To ensure face and content validity, thedraft questionnaire was pre-tested by two industry experts and two students. Theexperts had expertise in electronic engineering and experience in participation inKM-related communities. Both pairs of students and experts had participated inElectronic Engineering Times for more than three months. These procedures led tosome modifications of the wording of a few questionnaire items.

Data analysis and resultsStructured equation modelling (SEM) with partial least squares (PLS) analysis allowsempirical assessment of the measurement model used in this study. PLS was selectedbecause it is neither contingent on data having multivariate normal distributions norrequires the large sample sizes of other methods (Chin et al., 2003). Using ordinary leastsquares as its estimation technique, PLS performs an iterative set of factor analysisand a bootstrap approach to assessing the significance (t-values) of the paths (Chinet al., 2003). This study used PLS-Graph Version 3.01 to verify the measurement andtest hypotheses.

Characteristics of samplesTable I lists the demographic profile of the respondents. The members of ElectronicEngineering Times (EET) in Taiwan and China returned 152 and 143 questionnaires tothe researchers respectively. Responses with incomplete data were eliminated fromfurther analysis. Thus 128 and 113 useful responses from EET of Taiwan and of China

OIR34,6

838

were used in the data analysis. Both of them were dominated by males. The averageage of Taiwan’s EET members was lower than that of China’s EET. Respondents fromboth communities were well educated – approximately 90 percent of the respondentshave university degrees – and they have significant internet experience (more thanfive years for both communities). Finally all respondents have been members of theirrespective online communities for a substantial period of time (0.8 year for TaiwanEET and 1.3 years for China EET).

To test for possible non-response bias, we compared means for all the majorvariables and demographics for early respondents and late respondents. The results oft-tests for the demographic profiles, perceived identity verification, performanceexpectancy, satisfaction, PIIT, CA, CSE, and community tenure were not significant.The only significantly different construct is knowledge contribution, suggesting, notsurprisingly, that participants who are more active tend to respond earlier.

Measurement modelFollowing recommended two-stage analytical procedures (Chin et al., 2003)confirmation factor analysis was first conducted to assess the measurement model,then the structural relationships were examined.

Reliability and validityTo validate our measurement model, three types of validity were examined: contentvalidity, convergent validity, and discriminant validity. Content validity wasestablished by ensuring that the measurement items are consistent with the extantliterature. This was done by both interviewing senior members of the communities andpilot-testing the instrument. Regarding convergent validity we examined bothcomposite reliability and average variance extracted (AVE) from the measures

EET (Taiwan) EET (China)(n ¼ 128) % (n ¼ 113) %

GenderMale 98 77 92 81Female 30 23 21 19

Age,20 7 5 1 121-30 95 74 50 4431-40 19 15 32 2841-50 6 5 24 21.51 1 1 6 6

EducationHigh school 1 1 11 10Undergraduate 72 56 75 66Postgraduate 55 43 27 24

Membership (months),12 85 66 14 1313-17 22 17 24 21.18 21 17 75 66

Table I.Respondents’

demographic information

Onlinecommunities

839

(Hair et al., 1998). As recommended by Chin et al. (2003), 0.7 refers to the reliabilitythreshold of a construct. As indicated in Table II the composite reliability values ofEET Taiwan range from 0.892 to 0.956, and those of EET China range between 0.895and 0.951. For the AVE a score of 0.5 indicates acceptability (Fornell and Larcker,1981). Table II shows that the AVEs of Taiwan’s Electronic Engineering Times rangefrom 0.621 to 0.844, and those of China’s Electronic Engineering Times range from0.572 to 0.822, indicating the acceptability of AVE. Tables III and IV list the weightsand loadings of Taiwan’s and China’s EET respectively. As expected all measureshave significant path loadings at the level of 0.01. Finally we verified the discriminantvalidity of our instrument by looking at the square root of the AVE as recommendedby Fornell and Larcker (1981). The discriminant validity is confirmed by the resultsshown in Table V: the square root of the AVE for each construct is greater than thelevel of correlations involving the construct. Further as shown in Tables VI and VII theresults of the inter-construct correlations also demonstrate that each construct hasmore variability within its own measures than with other measures. The loadings andcross-loadings of these tables confirm the discriminant validity. In addition to thevalidity assessment we also examined the multicollinearity due to the relatively highcorrelations between some variables (e.g. a correlation of 0.809 between SAT and PE).The values of the variance inflation factor (VIF) indicate acceptability, ranging from1.321 to 4.016. Based on the results shown in these tables, there is confidence in thediscriminant and convergent validity of the constructs with evident cross-loadings.

Addressing common method varianceIn addition to validity assessment we considered the common method variance (CMV),which refers to a potential threat to internal validity, especially to research using surveysthat collect responses in a single setting. To deal with CMV we used the followingapproaches. First we collected data in two separate stages, with dependent andindependent variable measurement separated in time. Second we used factor analysis toexamine the CMV in the data set. According to Harman’s one-factor test CMV is high if asingle factor accounts for a majority of covariance in the independent and dependentvariables. Our factor analysis did not detect any single factor explaining a majority of thecovariance. Thus CMV is unlikely to occur in this study.

Items Composite reliability AVEConstruct Taiwan China Taiwan China Taiwan China

CSE 10 10 0.941 0.931 0.621 0.572CA 4 4 0.953 0.951 0.844 0.822PIIT 3 3 0.882 0.895 0.732 0.743PE 4 4 0.914 0.912 0.711 0.736PIV 6 6 0.943 0.912 0.752 0.658SAT 4 4 0.956 0.934 0.844 0.793KC 3 3 0.892 0.905 0.721 0.757

Notes: CSE: Computer self-efficacy; CA: Computer anxiety; PIIT: Personal innovativeness in IT; PE:Performance expectancy; PIV: Perceived identity verification (PIV); SAT: Satisfaction; KC: Knowledgecontribution

Table II.Reliability and AVE

OIR34,6

840

Structural modelWith an adequate measurement model the proposed hypotheses were tested with PLS.Our findings are shown in Figure 2 and summarised in Table VIII. We describe theresults in the following sequence: relationship between individuals’ motivations andknowledge contribution (H1-H6) and the relationship between individual differencesand members’ motivation (H7-H9). As shown in Figure 2 exogenous variables explainconsiderable proportions of the variance: 43.8 percent and 26.8 percent for performanceexpectancy, 30.8 percent and 28.3 percent for perceived identity verification, and 36.9percent and 41.7 percent for continuance intention.

Item Weight Loading SD t-value

CSECSE1 0.107 0.765 0.052 14.605CSE2 0.113 0.789 0.051 15.416CSE3 0.121 0.770 0.043 18.142CSE4 0.130 0.812 0.047 17.462CSE5 0.118 0.784 0.041 19.195CSE6 0.144 0.840 0.034 24.817CSE7 0.150 0.832 0.04 20.691CSE8 0.120 0.639 0.091 7.047CSE9 0.131 0.764 0.08 9.559CSE10 0.134 0.864 0.033 26.161

CACA1 0.182 0.905 0.117 7.703CA2 0.258 0.887 0.135 6.558CA3 0.369 0.963 0.101 9.58CA4 0.276 0.911 0.124 7.372

PIITPIIT1 0.462 0.937 0.014 66.153PIIT2 0.191 0.674 0.128 5.272PIIT3 0.473 0.927 0.016 56.744

PEPP1 0.325 0.871 0.024 36.524PP2 0.266 0.822 0.077 10.688PP3 0.304 0.906 0.024 37.076PP4 0.284 0.787 0.052 15.102

PIVPIV1 0.189 0.875 0.033 26.878PIV2 0.183 0.841 0.046 18.19PIV3 0.173 0.864 0.04 21.754PIV4 0.211 0.883 0.029 30.736PIV5 0.200 0.891 0.026 34.095PIV6 0.194 0.864 0.033 26.21

SATSAT1 0.266 0.904 0.021 42.271SAT2 0.273 0.940 0.014 69.904SAT3 0.276 0.906 0.022 41.323SAT4 0.272 0.928 0.016 57.794

Table III.Weights and loadings of

the measures (TaiwanEET)

Onlinecommunities

841

H1 and H2 are supported as expected, but H3 is partially supported: the pathcoefficient of China is 0.371 ( p , 0.05), but that of Taiwan EET (b ¼ 20:009) isinsignificant. Regarding perceived identity verification, as expected both H4 and H5are supported, but the findings of H6 are mixed. PIV is positively associated withcontinuance intention (b ¼ 0:171, p , 0.1) in Taiwan EET, but the above relationshipis insignificant in China EET.

Concerning the influence of individual differences, our findings support H7-H9 asexpected. Computer self-efficacy is positively associated with performance expectancy(bs of Taiwan and China are 0.454 and 0.294 respectively), but it is negatively affectedby computer anxiety (bs of Taiwan and China are 20.178 and 20.176 respectively).Personal innovativeness in IT correlates positively with perceived identity verification:b of Taiwan is 0.555 and that of China is 0.532.

Construct Item Weight Loading SD t-value

CSE CSE1 0.618 0.582 1.147 4.404CSE2 0.127 0.767 1.078 7.537CSE3 0.084 0.712 1.083 6.370CSE4 0.091 0.804 1.106 10.593CSE5 0.165 0.790 1.039 14.128CSE6 0.161 0.839 0.899 14.990CSE7 0.154 0.724 1.026 9.155CSE8 0.132 0.824 0.984 12.204CSE9 0.128 0.755 1.090 9.602CSE10 0.124 0.729 0.976 8.583

CA CA1 0.002 0.838 1.591 4.393CA2 0.258 0.917 1.537 5.491CA3 0.288 0.937 1.537 5.923CA4 0.519 0.949 1.590 4.914

PIIT PIIT1 0.463 0.920 1.034 62.231PIIT2 0.321 0.819 1.092 11.620PIIT4 0.368 0.848 1.033 12.152

PE PP1 0.280 0.813 1.085 13.374PP2 0.308 0.905 0.995 30.104PP3 0.319 0.918 0.931 54.912PP4 0.256 0.782 1.071 12.673

PIV PIV1 0.239 0.839 1.220 27.581PIV2 0.173 0.793 0.990 11.621PIV3 0.200 0.841 1.063 19.854PIV4 0.237 0.822 1.221 14.768PIV5 0.189 0.723 1.176 7.198PIV6 0.200 0.816 1.202 11.021

SAT SAT1 0.312 0.810 0.956 10.083SAT2 0.288 0.925 0.954 46.075SAT3 0.264 0.916 0.946 44.218SAT4 0.265 0.903 0.975 23.305

Table IV.Weights and loadings ofthe measures (ChinaEET)

OIR34,6

842

To further explore the mixed findings of H3 and H6, post hoc analyses wereconducted, aiming to investigate the intermediate effect of satisfaction. Specifically, weexamined two types of mediating effect:

(1) the mediating effects of satisfaction on the relationship between performanceexpectancy and knowledge contribution in Taiwan EET; and

(2) the mediating effects of satisfaction on the relationships between PIV andknowledge contribution in China EET.

For the former we first tested the direct relationships between performance expectancyand knowledge contribution: b ¼ 0:501, p , 0.01, accounting for 29.2 percent variance.We then proceeded to see if there is a mediation effect by adding the interveningconstructs (satisfaction). As shown in Figure 2 satisfaction completely mediated therelationship between performance expectancy and continuance intention in TaiwanEET, because the indirect paths were significant and the direct effect becomesinsignificant. Moreover the increase in R2 (i.e. from 0.292 to 0.369) perhaps shows thatperformance expectancy and PIV are not the only variables predicting satisfaction.Using the same method to test the mediating role of satisfaction for China EET, thedirect relationship between PIV and knowledge contribution is significant (b ¼ 0:323,p , 0.01), accounting for 32.7 percent of variance. We then verified the mediation roleof satisfaction by adding it to the above direct relationships. Figure 2 shows thatsatisfaction fully mediated the relationship between PIV and knowledge contributionbecause the indirect paths were significant and the direct path was insignificant.Further the increase of R 2 (i.e. from 0.327 to 0.417) perhaps implies that PIV andperformance expectancy are not the only predictors of satisfaction.

Mean SD CSE CA PIIT PE PIV SAT KC

Taiwan EETCSE 5.049 0.973 0.788CA 4.963 1.631 0.097 0.917PIIT 5.018 1.014 0.736 20.052 0.855PE 4.844 1.005 0.598 20.126 0.523 0.848PIV 5.228 1.081 0.599 0.032 0.555 0.535 0.870SAT 5.012 1.009 0.674 20.034 0.602 0.809 0.607 0.920KC 2.693 0.753 0.540 20.147 0.489 0.484 0.467 0.592 0.791

China EETCSE 4.690 0.825 0.756CA 4.398 1.455 0.137 0.910PIIT 4.947 0.962 0.526 0.112 0.863PE 4.947 0.833 0.420 20.108 0.451 0.857PIV 4.938 0.909 0.512 0.095 0.532 0.427 0.807SAT 5.062 0.879 0.525 0.080 0.582 0.704 0.505 0.890KC 3.699 1.936 0.365 20.048 0.420 0.599 0.289 0.590 0.762

Note: The italic figures in the diagonal row are square roots of the average variance extracted (AVE)

Table V.Correlations between

constructs

Onlinecommunities

843

Discussion, implications, and limitationsSummary of results and implicationsDespite the rapid growth of online communities and the wide use of variouscommunity technologies, a systematic model relating members’ motivations andcharacteristics to knowledge contribution behaviour has been lacking. The goal of thisstudy was to propose and test such a model. To this end we developed a theoreticalmodel centring on the key motivating roles of performance expectancy and perceivedidentity verification in knowledge contribution, and argued why and how the above

CSE CA PIIT PE PIV SAT

CSECSE1 0.771 0.028 0.586 0.396 0.378 0.478CSE2 0.795 0.028 0.649 0.42 0.45 0.516CSE3 0.777 0.096 0.609 0.450 0.474 0.509CSE4 0.818 0.055 0.604 0.482 0.492 0.533CSE5 0.791 0.155 0.507 0.438 0.473 0.56CSE6 0.847 0.228 0.606 0.535 0.519 0.591CSE7 0.838 0.107 0.572 0.557 0.547 0.579CSE8 0.644 20.185 0.471 0.447 0.333 0.451CSE9 0.77 0.079 0.541 0.485 0.433 0.534CSE10 0.871 0.131 0.697 0.498 0.614 0.579

CACA1 0.186 0.912 0.15 20.074 0.122 0.039CA2 20.004 0.894 20.005 20.105 0.053 20.041CA3 0.111 0.97 0.071 20.15 0.053 20.041CA4 0.088 0.918 0.002 20.112 20.009 20.042

PIITPIIT1 0.69 0.134 0.944 0.464 0.546 0.59PIIT2 0.462 20.165 0.679 0.349 0.226 0.401PIIT3 0.708 0.048 0.934 0.52 0.559 0.544

PEPE1 0.568 20.04 0.533 0.878 0.509 0.784PE2 0.447 20.117 0.444 0.829 0.41 0.635PE3 0.536 20.059 0.43 0.913 0.523 0.709PE4 0.481 20.027 0.349 0.794 0.375 0.624

PIVPIV1 0.52 0.064 0.465 0.493 0.882 0.552PIV2 0.559 0.04 0.49 0.462 0.848 0.528PIV3 0.472 0.023 0.457 0.406 0.871 0.471PIV4 0.565 0.056 0.522 0.513 0.89 0.553PIV5 0.543 0.024 0.497 0.487 0.898 0.545PIV6 0.487 20.038 0.486 0.447 0.871 0.538

SATSAT1 0.587 20.02 0.507 0.764 0.557 0.912SAT2 0.61 20.044 0.562 0.739 0.573 0.947SAT3 0.668 20.011 0.593 0.75 0.577 0.914SAT4 0.635 20.051 0.567 0.75 0.543 0.935

Notes: Italics indicate that figures can be grouped into one factor

Table VI.Matrix of loadings andcross-loadings (TaiwanEET)

OIR34,6

844

motivations are affected by individual characteristics. Our model extends Ma andAgarwal’s (2007) research by arguing that knowledge contribution is affected not onlyby PIV (intrinsic) but also by performance expectancy (extrinsic motivation). Themotivation is in turn influenced by individual characteristics, in terms of CSE, CA, andPIIT. Surveys were conducted in two online communities (Taiwan EET and ChinaEET) to provide empirical support for the proposed model. The study represents one ofthe first attempts to quantitatively measure the impact of individual characteristicsand motivation on knowledge contribution behaviour.

CSE CA PIIT PE PIV SAT

CSECSE1 0.587 20.034 0.373 0.384 0.358 0.391CSE2 0.774 20.069 0.306 0.290 0.317 0.350CSE3 0.718 20.033 0.296 0.191 0.249 0.307CSE4 0.811 0.109 0.245 0.207 0.278 0.338CSE5 0.797 0.178 0.415 0.376 0.414 0.470CSE6 0.846 0.146 0.401 0.367 0.416 0.512CSE7 0.731 0.170 0.590 0.352 0.519 0.542CSE8 0.831 0.162 0.511 0.301 0.457 0.386CSE9 0.762 0.178 0.325 0.293 0.386 0.265CSE10 0.735 0.179 0.396 0.284 0.365 0.281

CACA1 0.016 0.840 0.169 0.000 0.114 0.187CA2 0.128 0.926 0.109 20.078 0.085 0.064CA3 0.055 0.945 0.111 20.075 0.066 0.072CA4 0.172 0.957 0.101 20.135 0.106 0.083

PIITPIIT1 0.534 0.190 0.928 0.447 0.550 0.617PIIT2 0.456 20.073 0.827 0.439 0.381 0.467PIIT4 0.374 0.132 0.856 0.292 0.437 0.413

PEPE1 0.258 0.095 0.391 0.820 0.403 0.606PE2 0.458 20.088 0.386 0.913 0.405 0.654PE3 0.381 20.140 0.429 0.927 0.353 0.636PE4 0.348 20.249 0.350 0.789 0.315 0.532

PIVPIV1 0.421 0.083 0.492 0.436 0.846 0.443PIV2 0.316 0.100 0.320 0.346 0.800 0.301PIV3 0.462 0.073 0.411 0.368 0.849 0.399PIV4 0.550 0.079 0.485 0.376 0.830 0.510PIV5 0.359 0.045 0.374 0.289 0.729 0.342PIV6 0.356 0.088 0.482 0.247 0.823 0.435

SATSAT1 0.346 0.011 0.440 0.638 0.381 0.817SAT2 0.495 0.100 0.554 0.648 0.461 0.934SAT3 0.507 0.061 0.522 0.617 0.474 0.925SAT4 0.549 0.123 0.576 0.612 0.505 0.911

Notes: Italics indicate that figures can be grouped into one factor

Table VII.Matrix of loadings and

cross-loadings (ChinaEET)

Onlinecommunities

845

Figure 2.Results of PLS analysis

ResultHypothesis Taiwan China

H1. An online community member’s satisfaction with the community ispositively related to their knowledge contribution

U U

H2. An online community member’s performance expectancy is positivelyrelated to their satisfaction with the community

£ U

H3. An online community member’s performance expectancy is positivelyrelated to their knowledge contribution

U U

H4. An online community member’s perceived identity verification ispositively related to their performance expectancy

U £

H5. An online community member’s perceived identity verification ispositively related to their satisfaction with the community

U U

H6. An online community member’s perceived identity verification ispositively related to their knowledge contribution

U U

H7. An online community member’s computer self-efficacy is positively relatedto their performance expectancy

U U

H8. An online community member’s computer anxiety is negatively related totheir performance expectancy

U U

H9. An online community member’s personal innovativeness in IT is positivelyrelated to their perceived identity verification

U U

Notes: “U”indicates that the hypothesis is supported; “ £ ”indicates that the hypothesis is notsupported

Table VIII.Results of hypothesistesting

OIR34,6

846

Overall, our findings provide strong empirical support for the proposed hypotheses.The results of this study demonstrate that compared with individuals’ motivations,their satisfaction with the community is the strongest predictor of knowledgecontribution. Members’ motives for participating in communities (identity verificationand performance expectancy) are not directly associated with knowledge contribution;rather they affect it indirectly through satisfaction. In addition satisfaction fullymediates the relationships between performance expectancy and knowledgecontribution in Taiwan EET, and the relationships between PIV and knowledgecontribution in China EET. Coupled with a strong intention/behaviour associationtheorised and validated in prior IS use research (Bhattacherjee, 2001) our findingsconfirm that satisfaction and motivation (indirect and mediated by satisfaction) play akey role in predicting knowledge contribution.

Comparing our findings with prior research on human behaviour and KM (Ma andAgarwal, 2007; Venkatesh et al., 2003), some interesting patterns emerge. First bothperformance expectancy and PIV refer to individual beliefs, and satisfaction reflectstheir post-acceptance affect. Performance expectancy is the strongest predictor ofbehavioural intention among the related antecedents such as effort expectancy andsocial influence (Venkatesh et al., 2003). However satisfaction is a stronger predictor ofknowledge contribution than performance expectancy based on our findings.Extending prior work (Chiu et al., 2006; Ma and Agarwal, 2007) this study suggeststhat individuals’ motivations and satisfaction are helpful in explaining knowledgecontribution in virtual communities. In addition the effect of performance expectancyon users’ intentions in both acceptance and continuance contexts confirms therobustness and significance of this association across temporal stages of IS use.However the size of this effect, compared to that of affect (satisfaction), seems todecrease over time. This phenomenon also applies to the community context.

From a KM perspective our findings extend previous research (Chiu et al., 2006;Wasko and Faraj, 2005) by showing that KM depends not only on social capital, butalso on individual motivation. Our findings conceptualise motivation by showing thatmembers’ knowledge contributions are motivated not only by whether they receiveuseful information from the community (i.e. improving performance), but also bywhether the technology is capable of helping them create social relationshipseffectively (through identity verification). As suggested by prior studies (Ma andAgarwal, 2007; Venkatesh et al., 2003), because both identity verification andperformance expectancy are more crucial for acceptance intention and satisfaction ismore dominant for continuance intention and behaviour, a community manageraiming to retain or attract more members should adopt a two-fold strategy: informingnew (potential) users of the potential benefits of using the community and giving old(continued) users advice on how to use the community effectively so as to maximisetheir satisfaction. Because performance expectancy and PIV play a key role in affectingsatisfaction, a community manager should consider how to assist members providinguseful information so that they can attain gains in job performance. Regarding PIVstudies (Ma and Agarwal, 2007) have lent credence to the use of community artefactssupporting identity communication. In addition because both PIV and performanceexpectancy are related to individual characteristics, understanding the relationshipsamong individual characteristics, their beliefs (motivation) and affective state (i.e.satisfaction), and their KM behaviour in online settings helps a community manager

Onlinecommunities

847

design an effective training programme (Thatcher and Perrewe, 2002; Yi et al., 2006). Inorder to retain community members we suggest that for individuals with highinnovativeness in IT, improving their identity verification in an online communitybecomes very important. For those with low computer self-efficacy and high computeranxiety, how to improve their job performance becomes critical.

The above findings have practical implications for implementing computer trainingprogrammes. Understanding the antecedents salient to members’ motivations mayhelp community managers or company managers aiming to leverage the capability of acommunity to design and place employees in training programmes. As notedpreviously members with higher CSE and lower anxiety tend to have higher externalmotivation (performance expectancy). Thus community managers may provide thesemembers with more knowledge-intensive training. Individuals with higher PIIT aremore concerned with effective identity communication and relationship-building. Thusmanagers should focus more on providing new or interesting features in a community,such as an attractive interface and a user-friendly web navigation system

The findings of this study also have implications for academics. Because of thepopularity of online communities and KM, exploration of continuance behaviour inKM-related communities is worthwhile. Understanding the members’ motives forparticipating in the communities and how their motivations are related to users’individual differences will be critical to companies in the “e” century. This studycontributes to theory development by integrating IS users’ behaviour, self-presentationtheory, and KM, and uses them to explain knowledge contribution behaviour in virtualsettings. In addition we theorised the associations between motivation and individualdifferences based on social cognitive theory. Since the proposed model describes KMbehaviour from multiple perspectives based on sound theories, future research mayeasily extend our findings to explaining KM in other contexts.

LimitationsPrior to discussing the implications of our findings, we acknowledge the limitations ofthis study. First although the proposed model was tested using a more than adequatesample size, it is difficult to generalise this model and extend the results to othercommunity settings because members from only two communities were surveyed.Further it is possible that the variables (e.g. identity communication and computer-selfefficacy) of our model are somewhat context dependent. As shown in our findings, pathcoefficients and significance are different for the two sites studied. To generalise ourfindings to other types of communities, this study suggests that culture or social norms(patterns of accepted behaviours) should be taken into consideration because theseplay a key role in affecting members’ behavioural expectations (Wang and Chiang,2009; Wasko and Faraj, 2005).

Second, due to the cross-sectional design of this study, no causation can be determined.The significant paths between constructs can only be interpreted as correlational; thecausal inferences are based solely on theories. The recursive relationships between ourvariables are recognised. For example it is possible that members with higherperformance expectancy tend to have less computer anxiety and higher computerself-efficacy. Future studies may use longitudinal or experimental designs to analyse thecausal relationship between constructs. A longitudinal study aiming to relate perceivedidentity verification to continuance intention or behaviour would enrich our findings.

OIR34,6

848

ConclusionsThe goal of this research was to identify the salient determinants of knowledgecontribution in the context of online communities and to delineate the relationshipsbetween the above antecedents and individual differences. Toward this goal anintegrated model was proposed based on IS use, social cognitive theory, andKM-related theory. Data collected from field surveys of two online communitiesfocused on knowledge exchange provided empirical support for the proposed model.The results show that while both performance expectancy and PIV continue to affectmembers’ knowledge contribution, their satisfaction with prior use has a relativelystronger impact on knowledge contribution: in certain situations, satisfaction mayeven mediate the relationships between members’ motivations (performanceexpectancy or PIV) and knowledge contribution. The above motivation, in turn, wasinfluenced by individual differences. Noteworthy contributions of this study includedrawing attention to the subsequent differences between members’ motivations andknowledge contribution, theorising and validating one of the earliest models ofknowledge contribution in the context of online KM-related communities, andconsidering knowledge contribution from dual perspectives (motivation and individualdifferences).

References

Agarwal, R. and Prasad, J. (1998), “A conceptual and operational definition of personalinnovativeness in the domain of information technology”, Information Systems Research,Vol. 9 No. 2, pp. 204-15.

Agarwal, R. and Prasad, J. (1999), “Are individual differences germane to the acceptance of newinformation technology?”, Decision Sciences, Vol. 30 No. 2, pp. 361-91.

Armstrong, A.G. and Hagel, J. (1996), “The real value of online communities”, Harvard BusinessReview, Vol. 74 No. 3, pp. 134-41.

Bandura, A. (1986), Social Foundations of Thought and Action: A Social Cognitive Theory,Prentice-Hall, Englewood Cliffs, NJ.

Berman, J. and Bruckman, A.S. (2001), “Exploring identity in an online environment”,Convergence, Vol. 7 No. 3, pp. 83-102.

Bhattacherjee, A. (2001), “Understanding information systems continuance: an expectation-confirmation model”, MIS Quarterly, Vol. 25 No. 3, pp. 351-70.

Bock, G.W., Zmud, R.W., Kim, Y.G. and Lee, J.N. (2005), “Behavioral intention formation inknowledge sharing: examining the roles of extrinsic motivators, social-psychologicalforces, and organizational climate”, MIS Quarterly, Vol. 29 No. 1, pp. 87-111.

Butler, B. (2001), “Membership size, communication activity, and sustainability: a resource-basedmodel of online social structures”, Information Systems Research, Vol. 12 No. 4, pp. 346-62.

Chidambaram, L. and Tung, L.L. (2005), “Is out of sign, out of mind? An empirical study of socialloafing in technology-supported groups”, Information Systems Research, Vol. 16 No. 2,pp. 149-68.

Chin, W.W., Marcolin, B.L. and Newsted, P.R. (2003), “A partial least squares latent variablemodeling technique for measuring interaction effects: results from a Monte Carlosimulation study and an electronic-mail emotion/adoption study”, Information SystemsResearch, Vol. 14 No. 2, pp. 189-217.

Onlinecommunities

849

Chiu, C.M., Hsu, M.H. and Wang, E.T.G. (2006), “Understanding knowledge sharing in virtualcommunities: an integration of social capital and social cognitive theories”, DecisionSupport Systems, Vol. 42 No. 3, pp. 1872-88.

Compeau, D.R. and Higgins, C.A. (1995), “Computer self-efficacy: development of a measure andinitial test”, MIS Quarterly, Vol. 19 No. 2, pp. 189-211.

Compeau, D., Higgins, C.A. and Huff, S. (1999), “Social cognitive theory and individual reactionsto computing technology: a longitudinal study”, MIS Quarterly, Vol. 23 No. 2, pp. 145-58.

Connolly, L.A., Jessup, L. and Valacich, J. (1990), “Effect of anonymity and evaluative tone onidea generation in computer-mediated groups”, Management Science, Vol. 36 No. 6,pp. 689-703.

Constant, D., Sproull, L. and Kiesler, S. (1996), “The kindness of strangers: the usefulness ofelectronic weak ties for technical advice”, Organization Science, Vol. 7 No. 2,pp. 119-35.

Davis, F.D., Bagozzi, R.R. and Warshaw, P.R. (1989), “User acceptance of computer technology:a comparison of two theoretical models”, Management Science, Vol. 35 No. 8, pp. 982-1103.

Davis, F.D., Bagozzi, R.R. and Warshaw, P.R. (1992), “Extrinsic and intrinsic motivation to usecomputers in the workplace”, Journal of Applied Social Psychology, Vol. 22 No. 14,pp. 1111-32.

Eastin, M.S. and LaRose, R. (2005), “Alt.support: modeling social support online”, Computers inHuman Behavior, Vol. 21 No. 6, pp. 977-92.

Fornell, C. and Larcker, D.F. (1981), “Structural equation models with unobservable variablesand measurement errors”, Journal of Marketing Research, Vol. 18 No. 2, pp. 39-50.

Givertz, M. and Segrin, C. (2005), “Explaining personal and constraint commitment in closerelationships: the role of satisfaction, conflict responses, and relational bond”, Journal ofSocial and Personal Relationships, Vol. 22 No. 6, pp. 757-75.

Goffman, E. (1967), Interaction Ritual: Essays on Face-to-face Behavior, Doubleday, Garden City,NY.

Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W.C. (1998), Multivariate Data Analysis,5th ed., Prentice-Hall, Englewood Cliffs, NJ.

Hertel, G., Niedner, S. and Hermann, S. (2003), “Motivation of software developers in open sourceproject: an internet-based survey of contributors to the Linux Kernel”, Research Policy,Vol. 32 No. 7, pp. 1159-77.

Hsu, M.H., Ju, T.L., Yen, C.H. and Chang, C.M. (2007), “Knowledge-sharing behavior in virtualcommunities: the relationship between trust, self-efficacy, and outcome expectations”,International Journal of Human-Computer Studies, Vol. 65 No. 2, pp. 153-69.

Huang, C.C., Lin, T.C. and Lin, K.J. (2009), “Factors affecting pass-along e-mail intentions(PAEIs): integrating the social capital and social cognition theories”, Electronic Commerceand Applications, Vol. 8 No. 3, pp. 160-9.

Jensen, C., Davis, J. and Farnham, S. (2002), “Finding others online: reputation systems for socialonline spaces”, Proceedings of CHI (Computer Human Interaction) 2002 Conference ofHuman Factors in Computation Systems, Vol. 2002, ACM Press, New York, NY, pp. 447-54.

Jeppesen, L.B. and Frederiksen, L. (2006), “Why do users contribute to firm-hosted usercommunities? The case of computer-controlled music instruments”, Organization Science.,Vol. 17 No. 1, pp. 45-63.

Kankanhalli, A., Tan, B. and Wei, K.K. (2005), “Contributing knowledge to electronic knowledgerepositories: an empirical investigation”, MIS Quarterly, Vol. 29 No. 1, pp. 113-43.

OIR34,6

850

Kuhn, M.H. and McPartland, T.S. (1954), “An empirical investigation of self-attitudes”, AmericanSocial Review, Vol. 19 No. 1, pp. 68-76.

Lee, F., Vogel, D. and Limayem, M. (2002), “Virtual community informatics: what we know andwhat we need to know”, Proceedings of the 35th Hawaii International Conference onSystem Sciences, IEEE Computer Society, Los Alamitos, CA, pp. 2863-72.

Lin, C.P. and Bhattacherjee, A. (2009), “Understanding online social support and its antecedents:a socio-cognitive model”, The Social Science Journal, Vol. 46 No. 4, pp. 724-37.

Ling, K., Beenen, G., Ludford, P., Wang, X., Chang, K., Li, X., Cosley, D., Frankowski, D., Terveen, L.,Rashid, A.M., Resnick, P. and Kraut, R. (2005), “Using social psychology to motivatecontributions to online communities”, Journal of Computer-mediated Communication, Vol. 10No. 4, pp. 38-42.

Lu, J., Liu, C., Yu, C.S. and Wang, K. (2008), “Determinants of accepting wireless mobile dataservices in China”, Information and Management, Vol. 45 No. 1, pp. 52-64.

Ma, M. and Agarwal, R. (2007), “Through a glass darkly: information technology design, identityverification, and knowledge contribution in online community”, Information SystemsResearch, Vol. 18 No. 1, pp. 42-67.

Marakas, G.M., Johnson, M.D. and Palmer, J.W. (2000), “A theoretical model of differential socialattributions towards computing technology: when the metaphor becomes the model”,International Journal of Human-Computer Studies, Vol. 45 No. 3, pp. 529-52.

Matz, D.C. and Wood, W. (2005), “Cognitive dissonance in groups: the consequences ofdisagreement”, Journal of Personality and Social Psychology, Vol. 88 No. 1, pp. 22-37.

Rusbult, C.E. and Buunk, B.P. (1993), “Commitment processes in close relationships:an interdependence analysis”, Journal of Social and Personal Relationships, Vol. 10 No. 2,pp. 175-204.

Sussman, S.W. and Siegal, W.S. (2003), “Informational influence in organizations: an integratedapproach to knowledge adoption”, Information Systems Research, Vol. 14 No. 1, pp. 47-65.

Swann, W.B., Milton, L.P. and Polzer, J.T. (2000), “Should we create a niche or fall in line? Identitynegotiation and small group effectiveness”, Journal of Personality and Social Psychology,Vol. 79 No. 2, pp. 238-50.

Swann, W.B., Pelham, B.W. and Krull, D.S. (1989), “Agreeable fancy or disagreeable truth?Reconciling self-enhancement and self-verification”, Journal of Personality and SocialPsychology, Vol. 57 No. 5, pp. 782-91.

Swann, W.B., Polzer, J.T., Seyle, D.C. and Ko, S.J. (2004), “Finding value in diversity: verificationof personal and social self-view in diverse groups”, Academic Management Review, Vol. 29No. 1, pp. 9-27.

Taylor, S. and Todd, P.A. (1995), “Understanding information technology usage: a test ofcompeting models”, Information Systems Research, Vol. 6 No. 2, pp. 144-76.

Thatcher, J.B. and Perrewe, P.L. (2002), “An empirical examination of individual traits asantecedents to computer anxiety and computer self-efficacy”, MIS Quarterly, Vol. 26 No. 4,pp. 381-96.

Thatcher, J.B., Loughry, M.L., Lim, J. and McKnight, D.H. (2007), “Internet anxiety: an empiricalstudy of the effects of personality, beliefs, and social support”, Information andManagement, Vol. 44 No. 4, pp. 353-63.

Uray, N. and Ayla, D. (1997), “Identifying fashion clothing innovators by self-report method”,Journal of Euro-marketing, Vol. 6 No. 3, pp. 27-46.

Venkatesh, V., Morris, M.G., Davis, G. and Davis, F.D. (2003), “User acceptance of informationtechnology: toward a unified view”, MIS Quarterly, Vol. 27 No. 3, pp. 425-78.

Onlinecommunities

851

Walther, J.B., Slovacek, C.L. and Tidwell, L.C. (2001), “Is a picture worth a thousand words?”,Communication Research, Vol. 19 No. 1, pp. 105-34.

Wang, J.C. and Chiang, M.J. (2009), “Social interaction and continuance intention in onlineauctions: a social capital perspective”, Decision Support Systems, Vol. 47 No. 4, pp. 466-76.

Wasko, M.M. and Faraj, S. (2005), “Why should I share? Examining social capital and knowledgecontribution in electronic networks of practice”, MIS Quarterly, Vol. 29 No. 1, pp. 35-57.

Whitbourne, S.K. and Connolly, L.A. (1999), “The developing self in midlife”, in Willis, S.L. andReid, J.D. (Eds), Life in the Middle: Psychological and Social Development in Middle Age,Academic Press, New York, NY, pp. 25-45.

Whittaker, S., Terveen, L., Hill, W. and Cherny, L. (1998), “The dynamics of mass interaction”,Proceedings of the ACM Conference on Computation Supported Cooperative Work, ACMPress, New York, NY, pp. 09-18.

Yi, M.Y., Jackson, J.D., Park, J.S. and Probst, J.C. (2006), “Understanding information technologyacceptance by individual professionals: toward an integrative view”, Information andManagement, Vol. 43 No. 2, pp. 350-63.

Appendix. Questionnaire itemsComputer self-efficacyI could complete my job using the technology, if . . .

1. . . . there was no one around to tell me what to do.

2. . . . I had never used a package like it before.

3. . . . I had the software manuals for reference.

4. . . . I had seen someone else using it before trying it myself.

5. . . . I could call someone for help if I got stuck.

6. . . . someone else helped me get started.

7. . . . I had a lot of time to complete the job for which the software was provided.

9. . . . I had just the built-in help facility for assistance.

9. . . . someone showed me how to do it first.

10. . . . I had used similar packages like this one before to do the job.

Computer anxiety

1. I feel apprehensive about using computers.

2. It scares me to think that I could cause the computer to destroy a large amount of

information by hitting the wrong key.

3. I hesitate to use a computer for fear of making mistakes that I cannot correct.

4. Computers are somewhat intimidating to me.

OIR34,6

852

Personal innovativeness in IT

1. If I heard about new IT, I would look for ways to experiment with it.

2. Among my peers, I am usually the first to try out new IT.

3. In general, I am hesitant to try out new IT. (deleted)

4. I like to experiment with new IT.

Performance expectancy

1. I would find the system useful in my job.

2. Using the system enables me to accomplish tasks more quickly.

3. Using the system increases my productivity.

4. If I use the system, I will increase my chances of getting a raise.

Perceived identity verificationBelow are five fill-in-blank areas for you to answer the question “In this community, who am I?”Simply type in an answer next to the numbered item and make each answer different (e.g. high,smart, happy, antisocial, dependable, conservative, student, computer geek, Linux expert, father,board master, Catholic, woman, engineer etc.). Answer as if you were giving the answers to yourself,not to somebody else. Write the answers to yourself, not to somebody else. Write the answers in theorder that they occur to you. There are no right or wrong answers.In this community, I am _____

(1) Please think about your interactions with people in this community and indicate theextent to which other know that you define yourself as . . . (list the five responses justanswered by the respondent one by one)

(2) Other members in this community understand that I am . . . (list the five items justanswered by the respondent one by one).

SatisfactionI am___with my use of the community (IS)

1. Extremely displeased . . . Extremely pleased

2. Extremely frustrated . . . Extremely contented

3. Extremely angry . . . Extremely delighted

4. Extremely dissatisfied . . . Extremely satisfied

Knowledge contribution

1. I often help other people in this community who need help/information from othermembers.

2. I have contributed knowledge to this community.

3. I have contributed knowledge to other members that resulted in their development ofnew insights.

Note: All measures employ a seven-point Likert scale from “very frequent” to “very rarely” or“extremely likely” to “extremely unlikely”.

Onlinecommunities

853

About the authorShih-Wei Chou is a Professor in the Management of Information Systems Department atNational Kaohsiung First University of Science of Technology in Taiwan, where he teachesknowledge management, software engineering, object-oriented information systems, andmanagement of information systems. He received his PhD in computer science from IllinoisInstitute of Technology, and he holds a Master’s degree in computer science from MississippiState University. Prior to getting his doctorate, he spent two years working in thetelecommunication industry. His research interests include knowledge management, ISdevelopment and design, ERP systems, and e-learning. His work has appeared in the Journalof Information Science, Journal of Computer-aided Learning, and Decision Support Systems, andhas been presented at PACIS, HICSS, and other international conferences. Shih-Wei Chou can becontacted at: swchou@ccms.nkfust.edu.tw

OIR34,6

854

To purchase reprints of this article please e-mail: reprints@emeraldinsight.comOr visit our web site for further details: www.emeraldinsight.com/reprints