Butler (2001) - Resource-Based Model of Online Social Structures

Membership Size, Communication Activity,and Sustainability: A Resource-Based Model

of Online Social Structures

Brian S. Butler226 Mervis Hall, Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, Pennsylvania 15260

[email protected]

A s telecommunication networks become more common, there is an increasing interest inthe factors underljdng the development of online social structures. It has been proposed

that these structures are new forms of organizing which are not subject to the same constraintsas traditional social structures. However, from anecdotal evidence and case studies it is difficultto evaluate whether online social structures are subject to the same problems as traditional socialstructiires. Drawing from prior studies of traditional social structures and empirical analyses oflongitudinal data from a sample of Internet-based groups, this exploratory work considers therole of size and communication activity in sustainable online social structures.

A resource-based theory of sustainable social structures is presented. Members contributetime, energy, and other resources, enabling a social structure to provide benefits for individ-uals. These benefits, which include information, influence, and social support, are the basisfor a social structure's ability to attract and retain members. This model focuses on the systemof opposing forces that link membership size as a component of resource availability andcommunication activity as an aspect of benefit provision to the sustainability of an onlinesocial structure. Analyses of data from a random sample of e-mail-based Internet social struc-tures (listservs) indicate that communication activity and size have both positive and negativeeffects on a structure's sustainability. These results suggest that while the use of networkedcommunication technologies may alter the form of communication, balancing the opposingimpacts of membership size and communication activity in order to maintain resource avail-ability and provide benefits for current members remains a fundamental problem underlyingthe development of sustainable online social structures.(Online Communities; Electronic Groups; Membership; Dynamics; Social Resources)

Networked social environments that rely on computer-mediated communication systems are increasinglycommon. Public and private sector organizations areinvesting in infrastructures witb the goal of facilitatingcommunication and learning. Empirical and anecdotalevidence indicate that data communications networks,such as the Internet, can create new opportunities forpeople to interact (Baym 1993, Kraut et al. 1996,Rheingold 1993). Whether public or private, data

networks are increasingly sites of social activity. Al-though their functionality varies widely, these infra-structures all provide facilities that enable multipersonsocial communication. Each allows individuals to en-gage in constrained many-to-many communication bybroadcasting and receiving messages within a collec-tion of other people. As a result, each of these systemsprovides the basic communication capabilities neededto support significant social activity.

INFORMATION SYSTEMS RESEARCH, © 2001 INFORMSVol. 12, No. 4, December 2001, pp. 346-362

1047-7047/01/1204/0346$05.001526-5536 electronic ISSN

BUTLER

Membership Size, Communication Activity, and Sustainabitity

To generate many of the expected outcomes, sucb asimproved information sharing and better coordir\ation,telecommunication networks must do more than sim-ply provide facilities for communication; they must alsobe sites of social structures that support ongoing activity(Finholt and SprouU 1990, Sproull and Kiesler 1990).However, information technology only provides tech-nical infrastructures in which social activity may takeplace (Hagel and Armstrong 1998, Hof et al. 1997). Likea park or a conference room, these systems simply pro-vide a context in which people can interact. Just as pro-viding a room for a reception at an academic conferenceencourages but does not ensure social interaction, pro-viding electronic infrastructures supports but does notguarantee the emergence of social activity. The avail-ability of a technical infrastructure does not guaranteethat individuals will be willing to join and participatein online social structures. Likewise, efforts to attractnew members are likely to be wasted if tbese socialstructures fail to maintain tbe membership necessary toprovide valuable benefits over a longer term. Althougbdefinitions of success vary from case to case, a computer-mediated communication system's ability to supportuseful social activity is significantly affected by its abilityto encourage tbe emergence of sustainable online socialstructures; tbat is, structures that are able to continue pro-viding benefits for members over the long term.

Drawing from studies of traditional small groups,voluntary associations, and organizations, this paperpresents a resource-based model of the internal dy-namics of sustainable social structures. At the model'score is a feedback loop linking resource availability,benefit provision, and a social structure's ability to at-tract and retain members. Tbe model is then examinedwith longitudinal data collected from a sample of elec-tronic mail (e-mail)-based Internet listservs. A set oflog-linear, time-series/cross-section regression modelsis estimated to explore tbe relationships proposed inthe model. Implications for researchers and practition-ers interested in online groups and communities arediscussed and areas for future research are described.

Resource Availability, BenefitProvision, and SustainabilityThe core premise of tbe resource-based model of socialstructure sustainability is that, wbether traditional or

online, to be sustainable social structures must main-tain access to a pool of resources and support the socialprocesses tbat convert tbose resources into valuedbenefits for tbe participants. Social structures are sus-tainable wben they are able to provide benefits thatoutweigh the costs of membership (Moreland andLevine 1982). Social structures tbat can provide positivenet benefits are better able to attract and retain mem-bers, and hence survive over tbe long term. Traditionalsocial groups and communities provide many benefitsfor tbeir members, including opportunities for affilia-tion or companionship (McClelland 1985, Roberts 1998,Rubenstein and Shaver 1980); opportunities to influ-ence people (Winter 1973); social support (Wellmanand Whortley 1989,1990; Wellman 1990); access to in-formation; the ability to disseminate ideas rapidly(Kaufer and Carley 1993); and support for collectiveaction (Ostrom 1990). Likewise, online social structuresprovide a variety of benefits by supporting tbe devel-opment of interpersonal relationships, feelings of com-panionship, and perceptions of affiliation (Furlong1995, Hiltz 1985, Meyer 1989, Rbeingold 1993, Waltber1994); encouraging discussion and knowledge sharing(Abbot 1988, Kraut and Attewell 1993, Wellman 1995);allowing individuals to access information and quicklydisseminate their ideas (Constant et al. 1996, Finholtand Sproull 1990, Wbittaker 1996); providing socialand emotional support (King 1994, McCormick andMcCormick 1992, Rice and Love 1987, Waltber 1996);and enabling collective activities such as software de-velopment and political action (Ogan 1993). Whethertraditional or computer-mediated, social structuresprovide a variety of benefits for individual members,enabling them to attract and retain members.

Underlying a social structure's ability to provide val-ued benefits is the availability of resources such asknowledge, time, energy, money, and material re-sources (Rice 1982). Whether its creation is explicitlymanaged (Fine and Stoecker 1985) or the result of anemergent process, tbe availability of a resource pool isessential if a social structure is to be sustainable. Forexample, to provide member organizations witb accessto new tecbnologies, a research consortium must bavefinancial resources and expertise; to encourage infor-mation sharing, an online interest community must

INFORMATION SYSTEMS RESEARCHVol. 12, No. 4, December 2001 347

BUTLER

Membership Size, Communication Activity, and Sustainability

have members who are knowledgeable about relevanttopics; to provide emotional encouragement and/orcounseling, a support group must have members whoare willing to expend the time and energy to be sup-portive. Whether traditional or online, social structuresare sustainable only if they have access to resourcesthat allow them to provide benefits for their members.The creation and maintenance of social structures is, atleast in part, a process of gathering resources that canbe aggregated to provide valued benefits (Rice 1982).Without resources, it is impossible to provide benefits,and without benefits, it is not possible to attract andretain members.

However, having many sources of resources avail-able is not sufficient to sustain a social structure. It isalso necessary for the pool of potential resources to betransformed, through social activity, into benefits forindividual members. An online community of knowl-edgeable professionals will not continue to exist unlessit also supports discussions that provide informationvalued by the members. A support group with manycaring members who are willing and able to providehelp will not be sustainable if it does not also enablethe communication that is necessary to turn an indi-vidual's time into the supportive contact that troubledpeople need. The amount of money and expertiseavailable to a research consortium is irrelevant if theorganization is unable to coordinate the activitiesneeded to execute development projects. Whether tra-ditional or online, to be sustainable, social structuresmust support the social processes that convert re-sources into valued benefits.

Membership Size, Resource Avaitability, andBenefit ProvisionWhen members are a primary source of resources, thesize of a structure's membership provides a measureof resource availability. Larger voluntary associationstypically have access to more economic resources(McPherson 1983). Likewise, by aggregating theirmembers' knowledge, larger decision-making groupshave access to more information about the problem athand (Wittenbaum and Stasser 1996). In larger socialstructures it is more likely that there is a member whoknows the needed information, has the ability to pro-vide social support, or has the time to coordinate col-lective efforts. When members are able to benefit from

interactions both directly as active participants and in-directly as passive observers, larger groups will be ableto provide greater benefits as a result of the exponen-tial increase in the number of possible interactions(number of possible interactions = n*{n — 1) where nis the number of members). Membership size is also ameasure of the level of "audience resources" that a so-cial structure can provide. For individuals making an-nouncements, seeking visibility, or looking for an au-dience for their ideas, accessible listeners are aresource. Larger audiences are preferred over smallerones with similar members. The value of a structure's"audience resource" is dependent, at least in part, onthe size of its membership (Fulk et al. 1996, Markus1990, Rafaeh and LaRose 1993, Rice 1990). In each ofthese ways, larger social structures will tend to haveaccess to more resources than smaller structures. Be-cause resource availability is an aspect of benefit pro-vision, larger groups are expected to be more able toprovide valuable benefits to members, and hence be sus-tainable, over time (e.g., Haveman 1993, McPherson1983, Rafaeli and LaRose 1993).

While increasing size provides access to more re-sources, it can also have significant adverse effects onthe process of converting those resources into valuedbenefits (Haveman 1993, Moreland et al. 1996, Scott1992). As traditional social structures increase in size,they are subject to increasing logistical problems (Hare1976, Indik 1965). In larger face-to-face social struc-tures, individuals have fewer opportunities to partici-pate and less time to talk (Krech and Crutchfield 1948).As a result, while larger membership leads to greateraudience resources, it also makes it more difficult forindividuals to benefit from that resource. The numberof possible interaction partners increases nonlinearlywith size, making it substantially more difficult toknow the rest of the members (Bossard 1945). This, inturn, may affect the chances that individuals will formpersonal relationships and receive benefits such as so-cial support or information (Feld 1982). It also de-creases the likelihood that individuals will know theentire membership well, increasing the chances thatthey will not be able to fully access the resources thatare available within the structure. These problems cansignificantly hinder the processes by which resources

348INFORMATION SYSTEMS RESEARCHVol. 12, No. 4, December 2001

BUTLERMembership Size, Communication Activity, and Sustainability

are transformed into benefits, ultimately affecting a so-cial structure's ability to attract and retain members.

In addition to causing logistical problems, size mayhave a negative impact on the benefits provision pro-cess because it affects individuals' perceptions and at-titudes (Milgram et al. 1969, Slater 1958). Larger socialstructures are more likely to be subject to free-ridingand social loafing (Markus and Connolly 1990, Rafaeliand LaRose 1993, Thorn and Connolly 1990). Individ-uals will tend to contribute less time, energy, and re-sources because they expect that other members w illprovide enough to achieve the desired benefits (Pettyet al. 1977). Thus, v hile larger structures may havemore potential resource providers, the amount of con-tributions per person (and overall) may be lov^er thanin smaller social collectives (Marwell and Oliver 1993,Olson 1965, Thorn and Connolly 1990). If adequate re-sources are not contributed by the current member-ship, then the social structure wiW not be able to pro-vide the benefits necessary to continue to attract andretain members. The undersupply of resources (andhence lower benefit levels) in larger structures is re-flected in the general finding that individuals in largerstructures tend to be less committed, less satisfied(Cartwright 1968, Indik 1965, Slater 1958), and henceless likely to join or remain members (Baumgartel andSobol 1959, Cleland 1955, Porter and Lawler 1965).

The internal dynamics of sustainable social struc-tures are characterized, in part, by a complex interac-tion between the positive and negative consequencesof membership size. For resources, the available poolwill tend to grow at a constant or declining rate relativeto the size of the social structure. On the other hand,many of the problems of size are linked with the com-plexity of intragroup interaction, and hence tend to riseat an increasing rate as the structure grows (Bossard1945). As a result, the positive and negative impacts ofsize combine in a nonlinear fashion, affecting a socialstructure's ability to provide benefits, and hence attractand retain members. This interaction is further com-plicated by differential salience of costs and benefits inthe processes of member attraction and retention. In-dividuals entering social groups tend to be optimistic,overestimating the potential benefits and underesti-mating the likely costs of involvement (Brinthaupt etal. 1991). Consequently, the positive effects of size are

expected to have a greater effect on member attraction,while member retention is more affected by theproblems.

The interacting effects of size within the benefit pro-vision process combine to limit the size of sustainablesocial structures. In face-to-face contexts, logistical con-straints and free-riding behaviors combine to havenegative effects that overwhelm the positive conse-quences of size in all but the smallest of structures(Moreland et al. 1996). This results in size distributionsthat are skewed towards small groups (James 1953).Although not as extreme, similar results have beenfoimd for social structures that do not rely on continuousinteraction, such as voluntary associations (MacPherson1983), youth gangs (Thraser 1927), and even traditionalorganizations (Simon and Bonini 1958). Another typeof social structure in which the tension between thepositive and negative effects of size can be seen is thecase of brainstorming groups. Proponents of brain-storming argued that, under the right conditions,teams working together would be able to generatemore ideas (i.e., provide more benefits) than individ-uals working alone (Osborn 1957). However, decadesof research examining the operation of brainstorminggroups failed to support this assertion, finding insteadthat the negative logistical and psychological effects ofsize consistently outweigh the gains from increased"resource" availability (Diehl and Strobe 1987).

There are two general approaches to managing thepositive and negative impacts of size on the sustain-ability of a social structure: development of internalstructure or use of alternative communication technol-ogies. Internal structure addresses the negative con-sequences of size by constraining interaction within asocial structure. For example, work teams and orga-nizations both may establish formal roles, structures,and procedures in order to reduce the costs of com-munication, the difficulty of coordination, and the in-cidence of free-riding (Galbraith 1973, Haveman 1993,Indik 1965). Internal structures are attempts to, at somecost, alter the balance between the negative and posi-tive consequences of increased size in an effort to enablethe existence of larger sustainable social structures.

Another way that the impact of size may be affectedis through the use of alternative communication tech-nologies. Underlying many claims about the conse-quences of a computer-mediated social infrastructure



is the belief that information technology has the poten-tial to drastically reduce the negative consequences ofsize, leading to social structures that are either largeror not dependent on internal structure. Technology-supported organizations are expected to be "flatter,"with less internal hierarchy and more fluid processesand communication patterns (Davidow and Malone1992, SprouU and Keisler 1990). Through the use ofgroup support systems, which make use of a varietyof communication technologies, virtual work teams areexpected to be able to draw together the knowledgeand efforts of more people (Lipnack and Stamps 1997,Turoff 1991). Studies of online brainstorming havefound that, unlike in traditional groups, larger onlinebrainstorming groups are able to generate more ideasthan smaller ones (Connolly 1997). Computer-mediatedcommunication infrastructures, which provide fea-tures such as communication buffering and archiving,have the potential to drastically reduce the logisticalproblems that occur in traditional social structures(Nunamaker et al. 1991). Other features, such as mem-ber anonymity and the general invisibility of individ-uals (Finholt and Sproull 1990) may lower the salienceof a structure's membership, and hence reduce the neg-ative psychological effects of size. Whether synchronousor asynchronous, archived or temporary, computer-mediated communication systems are expected to re-duce the negative consequence of size, potentiallyshifting the balance from the negative to the positive,even in the absence of internal structure (Rice 1982,1987).

Communication Activity, Benefit Provision, andMembership CostsThe processes by which social structures provide bene-fits are based on communication activity. As a result,communication activity is a factor in the dynamics ofsustainable social structures. No matter what resourcesare available within a structure, without communica-tion activity those resources will remain dormant, andno benefits will be provided for individuals. The im-portance of communication activity is reflected in theo-retical definitions of small groups and communitiesthat highlight the importance of interaction (Bonner1959, Hare 1976, Homans 1950, Stogdill 1959, Shaw1981). Without some form of communication activity.

influence, social support, coordination, or informationsharing cannot occur. Thus, in the absence of com-munication activity, a social structure will fail to pro-vide valued benefits for individuals.

To the degree that communication activity is at thecore of the social processes underlying provision ofbenefits for individuals, there is expected to be a posi-tive relationship between the volume of communica-tion activity and the amount of benefit provided. Atthe extreme, a social structure in which there is nocommunication at all cannot provide benefits for itsmembers. Even nominal or minimal structures rely onsome basic communication activity to support the for-mation of an identity among their members. Morecommunication activity is expected to enable moreinformation sharing, development of strong relation-ships among members, and coordination of more com-plex activities—all of which correspond to the provi-sion of more benefits for individual members.

On the other hand, benefits are not valued equallyby all individuals. Communication activity seen by oneindividual as providing valuable benefits may be seenby another as noise. As a result, the diversity of contentpresent in a group's communication is expected to af-fect the sustainability of the social structure (Rafaeliand LaRose 1993). Information that is useful to onemember may be distracting to another. Interaction thatprovides social support for one individual may be per-ceived as unimportant by others. It is rarely possibleto provide benefits that are valued equally by all mem-bers. Different types of communication activity pro-vide different benefits, which are, in turn, valued dif-ferently by various subsets of a social structure'smembers and potential members. Thus, communica-tion volume and variation are related to the benefitsthat arise from that communication activity. Volumeand variation interact to determine, for a given mem-bership, the overall net benefits, and hence the sustain-ability of the structure.

However, while communication activity is an im-portant factor in the provision of benefits, it is also amajor source of costs for the members of a social struc-ture. Individuals incur costs when they contribute re-sources to a social structure. When individuals chooseto actively participate in the communication, they are



explicitly deciding to contribute their time, energy, at-tention, and knowledge. However, members also im-plicitly contribute resources to a social collective whenthey choose to remain a member, and hence remainpart of the audience that is exposed to the communi-cation activity. Attendees at conferences and meetingsincur costs, in terms of time, energy, and financial re-sources, whether or not they choose to explicitly con-tribute to the communication activity. Simply by beingpart of an audience, individuals contribute resourcesto a social structure, incur costs, and hence receivelower net benefits.

While more and more diverse communication activ-ity is likely to be associated with more benefits, it alsoimposes higher costs. Longer meetings, more issues ofa newsletter, or more electronic mail messages all havethe potential to support higher levels of informationsharing, social support, and other benefits. However,they also result in higher costs, and hence lower newbenefit levels. Longer meetings require that everyonespend more time and energy. Higher message volumesforce members to expend more time and attention toprocess the communication, even if they do not per-sonally benefit. For an individual, more (and more di-verse) communication activity is an improvement onlyif the benefits provided by that communication out-weigh the costs of being exposed to it. For a socialstructure overall, higher volume and diversity of com-munication activity enhances its sustainability only ifthe number of members who are attracted or retainedbecause of the additional benefits outnumber thosewho are lost due to the increased cost.

As with size, there are two approaches to managingthe positive and negative impacts of communicationon the provision of benefits: internal structure or useof alternative communication technologies. Both inter-nal structure and technology facilitate sustainable so-cial structure by altering the costs of communicationand/or constraining the content of communication ac-tivity. The use of jargon or special symbols reduces thecosts of communication activity, making it possible formembers to communicate complex ideas more effi-ciently. Formal summaries, meeting agendas, andstructured presentations enable members to more se-lectively participate in the audience, reducing the costsof being part of a social structure's audience. Editorial

control of content also facilitates sustainable socialstructures by screening out communication activitythat imposes costs on large segments of a structure'smembership that are not consistent with the benefitsprovided. Internal structures for managing the effectsof communication are attempts to, at a cost, mitigatethe negative effects of communication activity in theprocess of benefit creation.

Using different technologies can also moderate theimpact of communication activity on the sustainabilityof a social structure. In discussions of small groups andcommunities, it has often been assumed that face-to-face communication must be the basis for these socialcollectives' communication infrastructures. Underly-ing this assumption is the observation that physicalmeeting spaces can be an effective "technology" forsupporting communication activity that contributes tothe sustainability of a social structure. However, it isan oversimplification to assume that only face-to-faceinteraction can support sustainable social structures.Kaufer and Carley (1993) argue that the application ofprint provides many social structures with an alter-native to physical space/face-to-face communicationinfrastructures. Print-enhanced infrastructures, theyargue, supported the development of large-scale, long-term social structures, such as professions and aca-demic disciplines. Similarly, as other communicationtechnologies have developed, there has been increasedinterest in the idea of "virtual" social structures. It hasbeen expected that the use of new communicationtechnologies will reduce or even eliminate the costs ofcommunication activity (Davidow and Malone 1992,Malone et al. 1987). However, it remains unclearwhether the use of new communication technologiesis itself enough to overcome the negative consequences(i.e., costs) of communication activity within the ben-efit provision process and hence fundamentallychange the internal dynamics of sustainability.

A Resource-Based Model ofSustainable Social StructuresAs described above, the resource-based model focuseson a feedback process at the core of the internal dy-namics of social structure sustainability (Figure 1).


BUTLER

Membership Size, Communication Activity, and SustainabiUty

Figure 1 A Resource-Based Model of Sustainable Social Structures

ResourceAvailability

MembersbipSize

Benefit Creation Process

CommunicationActivity

Current members act as key providers of resources.Through communication processes, those resourcesare transformed into benefits for individuals which, inturn, enable the social structure to attract and retainmembers, hence developing and sustaining its re-source base. Features of the communication activitymediate the transformation of resources into valuedbenefits, a process that is posited to underlie the sus-tainability of a social structure.

The resource-based model of social structure sus-tainability frames social structure development interms of a single dynamic process: benefit provision.Other approaches to the study of groups and com-munities highlight different aspects of these socialstructures (see Arrov ^ et al. 2000 for summary and dis-cussion). It is not our intent to suggest that these ap-proaches are incorrect or inappropriate. Rather, thegoal of this model is to describe the dynamics of a widevariety of social structures in a way that allows us tobetter understand the general impact of informationtechnology. Toward this end, the resource-basedmodel focuses on benefit provision as a process that isa necessary condition for sustainability in a wide va-riety of structures. Whether the goal is to support pro-fessional development, provide social support, de-velop the market for a product, or engage in collectiveaction, social structures rely on the continued involve-ment of individuals. Although the relevant resources,the important benefits, and the form of communicationmay differ from case to case, benefit provision remains

a core process in the ongoing existence of a wide va-riety of social structures.

Examining online social structures in terms of aresource-based model of sustainability provides a ba-sis for considering whether new communication tech-nologies alter a fundamental aspect of social structuredevelopment. Most prior studies relating membershipsize and communication activity have treated size as afixed characteristic and asked how size affected thecommunication activity within the group (Bonito andHollingshead 1996). On the other hand, studies inwhich group, club, or organization size is treated as anoutcome have generally abstracted away from the in-ternal activities of the structure (e.g., McPherson 1990,McPherson and Rotolo 1996). The resotirce-based modelof sustainable structure is proposed as a bridge be-tween these two general approaches. It suggests thatsocial structures are faced with the fundamental prob-lem of balancing the positive and the negative conse-quences of size and communication activity. In tradi-tional social contexts, the negative consequences ofsize and communication activity tend to quickly over-shadow the positive effects, creating what appear to befundamental limits. In response to these limits, tradi-tional social structures adopt various internal struc-tures that mitigate the negative effects of size and ac-tivity by constraining membership and limitingcommunication. However, computer-mediated infra-structures are expected to reduce the negative effectsof size and communication activity, resulting in neworganizational forms that are subject to different dy-namics (Daft and Lewin 1993). In the analysis that fol-lows, the resource-based model will be applied to datafrom a random sample of e-mail-based social struc-tures (listservs) to consider whether communicationtechnologies, such as e-mail, alter the dynamics of so-cial structure sustainability.

Data, Measures, and MethodsThe possibility that telecommunication technologiesfundamentally alter the dynamics of sustainability wasexplored using data from a sample of e-mail-based In-ternet listservs. These online social structures useInternet-based e-mail and a server (i.e., a list server orlistserv) to centrally maintain a mailing list that en-ables individuals to broadcast text messages to the



other members. E-mail-based social structures werechosen for this study because of their prevalence, avail-ability, and ability to support the necessary data col-lection. E-mail-based social structures are known to beprevalent in both private and public networked envi-ronments (Finholt and Sproull 1990). As a result, e-mail-based Internet communities are both representative ofa large class of naturally occurring online social struc-tures and were available for study. Also, unlike otherdecentralized communications infrastructures, thecentralized architecture of a listserv supports measure-ment of membership size and change.

A stratified random sample was selected to ensurethat the included listservs represented a variety of top-ical focuses and member populations. From a censusof approximately 70,000 structures, a stratified basesample of 1,066 listservs was created. While one-thirdof the initially selected collectives focused on work-related topics, one-third focused on personal topics(hobbies, lifestyles, etc.), and the remaining listservsconsidered topics that mixed work-related and per-sonal interests (e.g., geographic locations). A multiple-stage confirmation process was then used to constructan analysis sample in which the listservs had compa-rable technology infrastructures and minimal internalstructure. Online structures that integrate different net-work technologies such as e-mail, WWW conferencingtools, or USENET newsgroups were eliminated, focus-ing the sample on listservs that relied solely on e-mailfor supporting online communication. Social collec-tives that made use of identifiable internal structures,such as moderated listservs, newsletters, or formalnew-member screening, were also removed from thesample. Each listserv was checked to ensure that it wasmechanically functional, able to provide the neededdata, and available for inclusion in the study. The re-sult of this process was a set of 284 unstructured list-servs that relied on e-mail as a basis for online com-munication. Membership and communication activityrecords were collected daily for these listservs fromAugust 1, 1997 to November 30, 1997. As data werecollected and measures were constructed, the analysissample was reduced to 206 as listservs ceased opera-tion, changed structure, or restricted access to mem-bership data.

Measures of size, communication activity, and mem-bership change serve as the basis for examining theresource-based model of sustainability (Figure 1). Thesample includes only minimally structured electronicgroups. This focuses the analysis on the question ofwhether, in the absence of a strong internal structure,communities that rely on computer-mediated com-munication technologies remain subject to the funda-mental interactions discussed above. Membership sizeis included as a central characteristic of the commu-nity's resource base and as a factor in the benefit pro-vision process. Communication activity, an importantaspect of benefit provision, is characterized in terms ofvolume and topic variation. A social structure's abilityto attract and retain members is measured in terms ofmember gain and loss. The dynamics of resource sus-tainability are then explored by examining the rela-tionships between size, communication activity, andsubsequent changes in membership.

Membership size was measured by counting thenumber of individuals in a listserv's mailing list at thebeginning of each month during the observation pe-riod. The raw data consisted of a listing of member'se-mail addresses, which was acquired by sending acommand to each listserv's server software. This mea-sure is based on the premise that an online social struc-ture's members are those people who are exposed toits internal communication activity. The impact of asocial structure on an individual is limited if she is notexposed to the communication activity. Likewise, theimpact of an individual is limited if she is not exposedto or contributing to the provision of benefit throughcommunication. Thus, in a listserv, the relevant mem-bers are those individuals who receive the broadcastmessages. While this measure may overestimate sizeby including individuals who receive messages but donot read them, it is conceptually equivalent to countingthe number of meeting attendees, a common measureof size in studies of traditional groups and organiza-tions. To simplify interpretation of the empirical re-sults, the size measure was divided by 100.

A listserv's communication activity consists of thetext messages that are distributed to all members. Thecommunication volume is measured by counting thenumber of messages distributed with the mailing listduring each month. Communication variation is in-ferred from the dialog structure of messages. In e-mail


BUTLER


communication, participants typically link relatedmessages by labeling later messages as replies to ear-lier ones. This linking creates sequences of related mes-sages known as discussion threads. Topic variation ina listserv's communication is reflected by the concen-tration of messages within the discussion threads. Dis-cussion threads are identified by removing the "re:"marker, a subject line tag commonly used to label mes-sages as replies, and grouping messages based on thefirst 40 characters of the remaining subject line text.Concentration of messages across the discussionthreads is characterized by computing a normalizedHerifindal-Hirschman index [HHI]^ (Hirschman 1964)for each day's communication activity. The HHI valuewas set to 0 in cases where there were no messages,indicating that there is no topic variation when thereis no communication activity. The HHI, a measure ofconcentration with a range from 0 to 1, was reversed(1 - HHI) to create a measure of variation, and themean value for each month was calculated to deter-mine topic variation of a listserv's communicationactivity.

This measure provides an indication of participants'assessments of topic variation in the listserv's com-munication activity. The HHI was chosen as the basisfor the topic variation measure because it capturesboth concentration within a set of categories (i.e., dis-cussion threads) and the number of categories (Davies1988). Messages are identified as being similar or dif-ferent based on the labels assigned by the participantsthemselves. Individuals label messages as replieswhen they expect their contribution to be of interest tothe same individuals who read the earlier messages.Thus, this measure of topic variation is roughly theequivalent of using knowledgeable coders to clustermessages based on the subset of the listserv's membersthat would be interested in them. While it is subject tonoise arising from flawed labeling and alternative uses

'The topic variation measure is computed according to the followingformula;

Topic variation = (1 — HHI) =

(1 - ([S? + S^]/MsgCount)),

where S, is the percentage of messages [0. . 1] which are part ofdiscussion thread i and MsgCount is the total number of messagesdistributed that day.

of the "re:" tag, this measure provides a general indi-cation of whether the communication activity has thepotential to provide benefits to few or many subsets ofthe listserv's membership. In this context, communi-cation, or topic, variation does not indicate how dif-ferent topics are from one another; rather, it refers tothe relative variation in the content of the communi-cation. Low topic variation indicates that the messagesin a listserv have focused on a small number of topics,while high topic variation means that many topicswere considered.

An online social structure's ability to attract and re-tain members is reflected in the inflow of new mem-bers and the loss of current members over time. Indi-viduals enter and leave listservs by requesting to beadded to or removed from the mailing list. Hence,membership gain and loss rates can be determined byrecording changes in the mailing list records. Membergain is measured by counting the individuals whosenames are added to a listserv's mailing list eacb month.This count also includes individuals who are comingback to a listserv after a formal absence. However, be-cause across the entire sample returning individualsrepresent fewer tban 10% of all entering members, noadjustment was made. Member loss is measured bycounting the people wbo are removed from a listserv'smailing list during each month. The raw data used tocalculate member gain-and-loss measures consisted ofa daily list of tbe individuals on each listserv's mailinglist. Daily member loss (gain) was calculated by com-paring each day's mailing list to the previous day's todetermine tbe number of individuals who bad beenremoved (or added) in the ensuing 24-hour period. Tbedaily data was then aggregated to create a monthlymeasure that ensured that individuals who are mem-bers of a listserv for less tban a montb were also in-cluded. Monthly measures were used for all constructsbecause tbey supported the focus on internal dynamicsof tbese structures. Comparable results were foundwhen analyses were conducted at the weekly and dailylevels, but use of tbe montbly data allow^ed us to makeuse of minimal assumptions about external influencessuch as day of the week, weekend vs. weekday, andholidays.

These measures of size, communication activity.


BUTLERMembership Size, Communication Activity, and Sustainabitity

Table 1 Construct and Measure Summary

Construct Label Measure Units

Wlennbership SizeCommunication Volume

Topic Variation

Member GainMember Loss

Size, The number of peopie on a iistserv's mailing iist at the beginning of month t. 1OO's of PeopleVolume, The number of messages distributed to individuais on a listserv's maiiing list during Messages/Month

month tVariation, The mean value of a reversed, normalized HHi which reflects fhe number of topics and (Unitless)

the disfribufion of messages among those fopics each day during month tMemberGain, The number of people who are added to a lisfserv's maiiing list during month t. People/MonthMemberLoss, The number of people who are removed from a listserv's maiiing lisf during month f. Peopie/Month

Table 2 Descriptive Statistics for Listserv Data

Membership SizeCommunication VolumeTopic VariationMember GainMember Loss

Mean

1.6833.29

0.0610.207.45

Std Deviation

2.8794.770.18

26.3620.84

Minimum

0.030.000.000.000.00

Maximum

23.921084.00

0.90277.00208.33

member loss, and member gain were constructed foreach of tbe listservs in the sample, resulting in a 206(listservs) X 4(montbs) X 5(variables) panel data set(Table 1). Tbe sample contains listservs covering arange of sizes and levels of communication activity(Table 2). Tbe descriptive statistics for the dataset alsoindicate that there is significant variation in the mem-ber gain-and-loss measures.

Analysis and ResultsA set of time series, cross-section, random effects, log-linear regression models was used to estimate the re-lationships in the resource-based model of sustain-ability. Tbe error structure for eacb model included atime-period-dependent component, a cross-section-(i.e., listserv) dependent component, and a componentthat was assumed to be normally distributed and in-dependent of tbe time period and listserv. Two-wayrandom-effects models were used because both thesampled listservs and tbe sampled time periods wererepresentative of a large "population" (Greene 1993).Log-linear models were selected because they focus oninteraction of size and communication activity, wbileminimizing problems witb nonnormal data. Concep-

tually, it is tbe interaction of communication activityand size that is expected to play a role in the sustain-ability of online social structures. Furtbermore, distri-butions of tbe various measures are skewed. Applyingthe log transformation reduces tbe impact of tbis non-normality. Because tbere were a significant number ofcases in which tbe measures bad zero values, a smallconstant (0.000001) was added to allow the log (base10) transformation to be applied. The constant wascbosen based on tbe smallest nonzero value in the dataset. Tbis minimized tbe transformation's impact of tbeordering of tbe data.

The presence of mediation, as proposed in the modeldescribed above, can be tested by estimating a seriesof regression models (Baron and Kenny 1986). First,models^ of size and tbe membersbip cbange variables(gain and loss) are estimated:

LOG(MemberGain,) =

LOG(MemberLoss,) =

+ BiLOG(Size,);

+ B^ LOG(Size,).

Then, models are estimated to assess tbe relationshipbetween size, communication volume, and topicvariation:

LOG(Volume,) = BQ -(- BiLOG(Size,);

LOG(Variation,) = BQ -i- B,LOG(Size,).

These log-linear models correspond to the following multiplicativemodels:

(a) MemberGain, = lO**" Sizef' Volumef Variation,"';

(b) MemberLoss, = 10"" Sizef' Volume," Variation,"':

Hence, the log-linear models also capture the interaction betweensize and communication activity described in Figure 1.


BUTLER


Finally, models of both the primary (size) and medi-ating (volume and variation) factors are considered:

LOG(MemberGain,) =

- I -

BiLOG(Size,)

LOG(MemberLoss,) = BQ + BiLOG(Size,)

+ B2LOG(Volume,) + B3LOG(Variation,).

The results are then examined to determine whetherthere is support for the proposed mediation effect. To-gether these analyses examine the relationships thatcomprise the core of the resource-based model of sus-tainable social structures (Figure 1). The remaining re-lationships in Figure 1, the links between member gain,member loss, and subsequent changes in size, are de-terministic and hence need not be statisticallyestimated.

The statistical models described above were esti-mated with the TSCSREG procedure in SAS (v6.12)(Table 3). The regression results suggest that size andcommunication activity have a significant impact onthe ability of online social structures to attract and re-tain members. The significant coefficients for size andcommunication activity in the model predicting mem-ber gain suggest that these features positively impacta listserv's ability to attract members. Larger and more

Table 3 Mediated Model Results

Dependent

Variable

Intercept

Size, (Log)

Volume, (Log)

Variation, (Log)

*: p < 0.05,

/V = 206 X

Size&

Communication

Activity

(a) (b)Volume Variation

(Log) (Log)

- 1 . 5 2 " * - 3 . 7 4 * * *

2.94*** 1.95***

0.10 0.09

, * * : p < 0 . 0 1 , *** :

4

Member

Attraction

(c)Member

Gain

(Log)

- 1 . 0 1 * * *

3.25***

0.24

(d)Member

Gain

(Log)

-0.40

2.49***

0.22***

0.07

0.31

p < 0.001

Member

Retention

(e)Member

Loss

(Log)

-2 .03** '

3.21**'

0.21

(t)Member

Loss

(Log)

' - 0 . 5 4 * *

' 1.78***

0.31***

0.27***

0.44

active listservs see higher rates of member gain. How-ever, the significant coefficients for size, communica-tion activity, and topic variation in the member lossmodel imply that these features also have a negativeeffect on a listserv's ability to retain members. Largerlistservs and those with more and more varied com-munication activity have more member loss. Thus, sizeand communication activity have both positive andnegative impacts on the sustainability of the online so-cial structures.

It is possible that the negative relationship betweensize and a listserv's ability to retain members is an ar-tifact of the measure of member loss. Member loss isassessed in terms of an absolute count of number ofpeople who leave a listserv in a given month. Thismeasure is subject to a variable ceiling. The number ofindividuals who could potentially leave is limited tothe number of individuals who are present (size) andthe number who have entered (member gain). Thus, itis possible that relationship between size and memberloss is a trivial consequence of larger networked socialstructures having the potential for more people toleave.

To determine whether the effect of size on member-ship loss extended beyond the trivial ceiling effect, themember loss model was estimated using proportionalmember loss as an alternative measure of a listserv'sability to retain members. Proportionate member losswas assessed by dividing the absolute member loss bythe listserv's size. This measure captures a social struc-ture's ability to retain members, relative to its size. Theresults of this model indicate that the impact of size ona listserv's ability to retain members is not simply dueto the variable ceiling for absolute membership loss(Table 4). The coefficient of size remains statisticallysignificant and positive. These results suggest thatlarger listservs will tend to have a higher relative lossrate. Not only do larger online structures lose moremembers, they also lose a larger percentage of theirmembership than smaller structures.

In addition, the original analysis also provides sup-port for the proposition that communication activityserves as a partial mediator of the effects of size on astructure's ability to attract and retain members. Thedifference between the size coefficients in the directmodels (c and e) and the comparable coefficient n in

356INFORMATION SYSTEMS RESEARCH

Vol. 12, No. 4, December 2001


Table 4 Proportional Member Loss Model Results

Proportional Member Loss (Log)

interceptSize, (Log)Volume, (Log)Variation, (Log)

-0.74*1.34*0.33*0.25*0.39

*: p < 0.05, **: p < 0.01, ***: p < 0.001A/ = 206 X 4

the full models (d and f) indicates that for both membergain and loss, there is a mediation structure involvingsize and communication activity. For member gain,there is a reduction of 0.77 in the size coefficient whencomparing the size model (c) with the full model (d)(one-tailed test: t = 1.85, p < 0.05). This reduction im-plies that 23% of the impact of size on member attrac-tion is accounted for by the indirect link between sizeand membership gain that is mediated by communi-cation activity. In the case of member loss, the size co-efficient drops by 1.43 (or 44%) when communicationvolume and variation are included (one-tailed test: t =3.51, p < 0.001). Overall these results indicate thatsome portion of the impact of size on an online socialstructure's ability to attract and retain members is me-diated by communication activity. However, in bothcases there remains a significant direct relationship be-tween size and the membership change variables.Thus, while the link between size and communicationactivity may play a role in the sustainability of onlinesocial structures, size has other more direct conse-quences as well.

DiscussionThe resource-based model builds on prior studies ofgroups, associations, and organizations to consider therole of technology in a fundamental problem of orga-nizing viable online social structures. As discussedabove, the component constructs have been used inprior studies of traditional social structures. For ex-ample, in applications of public goods theories to com-munication systems, the critical mass model (Markus

1990) is discussed in terms of resources that are avail-able while discretionary database models (Thorn andConnolly 1990) consider the processes by which re-sources are converted to benefits for members. Whileit has been noted elsewhere that these models addressdifferent aspects of online social structure (Markus andConnolly 1990), issues of resource availability and ben-efit provision have not been well integrated (Rafaeliand LaRose 1993). One contribution of this paper is toexplicitly link resource availability, benefit provision,and member attraction and retention in a model ofsustainability.

However, the resource-based model presented herealso differs from prior work in a more fundamentalfashion. Although it draws on prior studies for boththeoretical and operational constructs the resource-based model is structured around what Robey andBoudreau (1999) have called "logics of opposition."That is, instead of characterizing changes in online so-cial structures in terms of outcomes predicted by a setof factors, the resource-based model frames the studyof online structure sustainability in terms of sets of op-posing forces that serve to simultaneously promoteand hinder the processes of change (Robey andBoudreau 1999). It is in the identification of these op-posing forces, and their implications for the develop-ment of online social structures, that the primary valueof the resource-based model lies.

At the lowest level, the resource-based model ischaracterized not by unidirectional links between re-source availability, benefit provision, and sustainabil-ity, but rather by links that involve opposing effects.The empirical relationships between size and membergain and loss show that the role of size in the devel-opment of online social structures is not as a single,positive effect, as it is often characterized in criticalmass analyses (Markus 1990, Rafaeli and LaRose 1993,Rice 1990), but rather the result of the interaction ofopposing effects with greater size being associatedwith both more member gain and more member loss.This result is significant in part because it suggeststhat, at least within the context of Internet listservs, theimplications of increasing size for online social struc-tures are more complex than typically posited in priorstudies. The development of online communities is nota process which proceeds smoothly at an increasing


BUTLER


rate as the value of the available resource increases(Markus 1990), but rather it is a process which involvessignificant "churn," with members coming and goingeven as the magnitude of available resources increases.Practically, the association of increasing size with bothincreased ability to attract members and a decreasedability to retain members is important because it in-dicates that increased rates of member loss may not bea bad sign. While it is clearly a problem if member lossoutpaces member gain, this study suggests that devel-opers of online social structures should expect to seeincreasing member loss as part of the developmentprocess of a growing online community.

Similarly, the empirical results indicate that the im-pact of communication activity on sustainability is alsonot a unidirectional effect, but rather the aggregateconsequence of opposing forces of communication ac-tivity on member gain and loss. Although communi-cation activity is a primary mechanism for benefit pro-vision, it is an activity that is not costless or uniformlybeneficial for all individuals. Hence, characterizationsof online social activities that consider only the benefitsof communication fail to accurately reflect the com-plete role of this activity in the development of sus-tainable online structures. For practitioners, this im-plies that efforts to manage message activity, whetherthrough technological change, editorial control, or so-cial intervention, are double-edged, each having bothpositive and negative impacts on a group's ability tomaintain a resource base and provide benefits in thefuture.

The low-level structures of opposition underlyingthe impact of resource availability (i.e., size) and ben-efit provision (i.e., communication activity) are furthercomplicated in the resource-based model when theytoo are linked—creating yet another set of opposingeffects. The coefficient of size in the member gainmodel (Model d in Table 3) is larger than the coefficientof size in the member loss model (Model e) (one-tailedf-test: t = 1.76, p < 0.05) indicating that the net effectof increased size on a listserv's sustainability is posi-tive. However, comparison of the communication ac-tivity effects suggests that these are in the oppositedirection. The negative impact of more varied com-munication activity on the retention of members is sig-nificantly greater than the positive impact on member

attraction (one-tailed f-test: t = 1.717, p < 0.05). In ad-dition, while the difference is not statistically signifi-cant, the negative impact of communication volume isalso greater than its positive effects. Thus, while thedirect impact of size on a listserv's ability to attract andretain members is generally positive, the net impact ofcommunication features, specifically variation, is neg-ative. This is further complicated by the presence of arelationship between size, communication volume,and communication diversity. Size is positively asso-ciated with communication volume and variation,which in turn negatively impact sustainability (Table3, Models a and b). Together these results describe amultilayered structure of opposition in which resourceavailability in the form of member presence, and ben-efit provision in the form of communication activity,combine through a series of opposing effects to ulti-mately impact an online group or community's futureresource sustainability.

The interplay of size and communication features inthe empirical results, mirroring the linked constructsof resource availability and benefit provision, brings tothe fore limitations inherent in conceptualizations ofonline social structure that are not balanced in theirconsideration of the development process. For exam-ple, critical mass models (Markus 1990, Rice 1990)highlight the number of members, in either absoluteor relative terms, as a significant positive factor in thedevelopment of communication systems and onlinesocial structures. Likewise, discussions of the impactof technology on social structure often focus on theimpact of alternative infrastructures on communica-tion activity without simultaneously considering theconsequences of those changes for size (Butler 1999).The empirical results presented above indicate thattheories of online social structure that consider eitherfactor alone are likely to be, at best, incomplete. Fordevelopers of online communities, these results implythat identification of a "critical mass" is inherently tiedto the features of the communication activity, and thatanalyses of technology or policy impact on a group'scommunication activity must also take into accountimpacts on future resource availability. The concept ofan onhne social structure's critical mass of members isonly meaningful when coupled with assumptions about



the type and volume of communication activity. Con-versely, technological or procedural design choicesthat affect the type and volume of communication ac-tivity must be evaluated in terms of their impact onthe attraction and retention of members (and hence theavailability of resources in the future).

Limitations and Future ResearchThe resource-based model of sustainable social struc-tures is not a theory of technological impacts, butrather a theory of social structure that allows us to con-sider the consequences of technology.

To create a general framework, the current studyfocuses on developing the constructs of resource avail-ability, benefit provision, and social structure sustain-ability with minimal models of resources and com-munication. This approach furthers our understandingof the general problem of social structure sustainabilityin several ways. First, by articulating the resource-based model in terms of elementary constructs that areboth simplified and grounded in prior empirical andtheoretical studies, this work can serve as the foun-dation for computational and analytical models thatconsider other aspects of the dynamics of online andtraditional social structures. The resource-based frame-work describes a process of resource accumulation andbenefit generation. The empirical analysis considersevidence for the nature of specific relationships thatcomprise the process. However, it is beyond the scopeof this study to systematically examine the full impli-cations of that process model to understand the long-term evolution of social structures, as might be done infuture computational or analytical modeling studies.

Developing the resource-based theory with minimalmodels of resources and communication also allows itto be used as a common framework for studies of bothonline and traditional social structures in a variety ofcontexts. Although the individual operational con-structs (size, communication volume, etc.) and thelinks between them have been considered in priorstudies of traditional social structures, the resource-based theory of sustainability has not been directlytested in those contexts. Studies of sets of constructsand processes in traditional groups, associations, or or-ganizations would strengthen claims regarding the

generality of the resource-based theory of sustainablesocial structures. Additional work might also be doneto develop the resource-based model of sustainable so-cial structure as a more detailed description of the fun-damental problem of organizing. In this vein, futurework might benefit from consideration of prior studiesof organizational demographics, group composition,and communication processes and structure (e.g.,Berthod et al. 1996, Bonito and Hollingshead 1996) assources for elaborating the constructs and processesunderlying the model.

Likewise, the empirical work presented here isbased on a sample of online social structures that makeuse of one type of technology (centralized e-mail serv-ers), have minimal internal structures, and exist in apublic network (the Internet). Although this samplerepresents a common type of online social structure,future work that considers other types of technologyand different contexts would also increase the validityof the general model as a description of one of thefundamental problems of social organization. Whileknowledge of the fundamental problems is valuable,practitioners also benefit from systematic examina-tions of the tools and designs they can use. In the caseof social structure development, future work mightconsider whether structures based on pull technolo-gies, which require that users explicitly request eachmessage (e.g., WWW conferencing), or hybrid infra-structures, which combine traditional and computer-mediated communication infrastructures, differ sig-nificantly in terms of their sustainability from thestructures considered here which use a push technol-ogy (e-mail), presenting messages to members withoutthe need for an explicit request. Assessment of the im-pact of internal structures, such as moderation andmember screening on a structure's ability to attract andretain members would also be useful when developingthe social and managerial infrastructure to support on-line communication. Finally, consideration of socialstructures operating within other, larger contexts, suchas within an organization or a well-defined commu-nity, would provide additional insight into the factorsthat underlie the relationships between size, commu-nication activity, and sustainability (Rice 1987). In allof these cases, the resource-based model of sustainablesocial structures provides a theoretical framework for



examining the impacts of new technologies on the vi-ability of long-term social structures.

The results of this study suggest that the use ofcomputer-mediated communication infrastructures isnot likely to fundamentally change the problems un-derlying the development of sustainable social struc-tures. Size and communication activity have both posi-tive and negative effects on the sustainability of onlinesocial structures. Larger listservs are better able to at-tract members, but they are also less able to keep them.Likewise, listservs with more communication activityare more able to attract members, but less able to retainthem. Thus, while new communication technologiesmay serve to reduce spatial and temporal constraintson communication (Rice 1980, Sproull and Kiesler1990), the growth of online social structures does notappear to be limited only by the availability of inter-ested participants. Rather, as with traditional socialstructures, developing and maintaining sustainableonline social structures requires that the fundamentalproblem of balancing the positive and negative im-pacts of size and communication activity be solved inorder to maintain a resource pool for the future whileproviding benefits for the members in the present. Thissuggests that rather than focusing on computer-mediated communication technologies as revolution-ary forces that fundamentally change the problem ofsocial organization, researchers and practitioners wouldbe better served by theories that characterize them asadditional tools in the organizers' repertoire for deal-ing with certain fundamental problems. While themodel presented in this work provides an initial step,there is much work that remains to be done in the de-velopment of a practical understanding of the chal-lenges of organizing and the true opportunities pro-vided by new technologies in the realm of developingand maintaining sustainable social structures.

AcknowledgmentsThe author would like to acknowledge the valuable comments andhelp provided by Kathleen Carley, Robert Kraut, and RichardMoreland, the valuable guidance provided by the Associate Editor,and the suggestions of the three anonymous reviewers.

ReferencesAbbot, A. 1988. The System of Professions: An Essay on the Division of

Expert Labor. University of Chicago Press, Chicago, IL.Arrow, H., J. E. McGrath, J. L. Berdahl. 2000. Small Groups as Complex

Systems: Formation, Coordination, Development, and Adaptation.Sage Publications, Thousand Oaks, CA.

Baron, R. M., D. A. Kenny. 1986. The moderator-mediator variabledistinction in social psychology research: Conceptual, strategic,and statistical considerations. /. Personality and Social Psych.51(6) 1173-1182.

Baumgartel, H., R. Sobol. 1959. Background and organizational fac-tors in absenteeism. Personnel Psych. 12 431-443.

Baym, N. 1993. Interpreting soap operas and creating community:Inside a computer-mediate fan culture. /. Folklore Res. 30 143-176. [Also appears in S. Kiesler (Eds.), Culture of the Internet]

Berthod, M., F. Sudweeks, S. Newton, R. Coyne. 1996. "It makessense": Using an autoassociative neural network to explore typ-icality in computer mediated discussions. S. Rafaeli, F.Sudweeks, M. McLaughlin, eds. Network and Netplay: VirtualGroups on the Internet. AAAI/MIT Press, Cambridge, MA. 191-220.

Bonito, J. A., A. B. Hollingshead, 1996. Participation in small groups.B. R. Burelson, ed. Communication Yearbook 20. Sage Publica-tions, Thousand Oaks. CA.

Bonner, H. 1959. Group dynamics: Principles and applications. Ronald,New York.

Bossard, J. H. 1945. Law of family interaction. Amer. ]. Soc. 50 292-294.

Brinthaupt, T. M., R. L. Moreland, J. M. Levine. 1991. Sources of op-timism among prospective group members. Personality and So-cial Psych. Bull. 17(1) 36-43.

Butler, B. S. 1999. The dynamics of cyberspace: Examining and mod-eling online social structure. Phd Dissertation, Carnegie MellonUniversity, Pittsburgh, PA.

Cartwright, D. 1968. The nature of group cohesiveness. D.Cartwright, A. Zander, eds. Group Dynamics: Research and The-ory, 3rd Edition. Harper & Row, New York.

Cleland, S. 1955. Influence of Plant Size on Industrial Relations. Prince-ton University Press, Princeton, NJ.

Connolly, T. 1997. Electronic brainstorming: Science meets technol-ogy in the group meeting room. S. Kiesler, Ed. Cidture of theInternet. Lawrence Erlbaum, Mahwah, NJ.

Constant, D., L. Sproull, S. Kiesler. 1996. The kindness of strangers:The usefulness of electronic weak ties for technical advice. Or-gan. Sci. 7 119-135.

Daft, R. L., A. Y. Lewin. 1993. Where are the theories for the "New"organizational forms? An editorial essay. Organ. Sci. 4(4) i-vi.

Davidow, W., M. Malone. 1992. The Virtual Corporation. Harper, NewYork.

Davies, S. 1988. Choosing between concentration indices: the Iso-concentration curve. Economica 46 67-75.

Diehel, M., W. Strobe. 1987. Productivity loss in brainstorminggroups: Towards the solution of a riddle. /. Personality and SocialPsych. 53 497-509.

Feld, S. 1982. Structural determinants of similarity among associates.Amer. Soc. Rev. 47 797-801.

Fine, G. A., R. Stoecker. 1985. Can the circle be unbroken? Smallgroups and social movements. E. Lawler, Ed. Advances in GroupProcesses 1. JAI Press, Greenwich, CT 1-28.


BUTLERMembership Size, Communication Activity, and Sustainabitity

Finholt, T., L. Sproull. 1990. Electronic groups at work. Organ. Sci. 141-64.

Forsyth, Danelson R. 1990. Group Dynamics 2nd. Ed. Brooks/Cole,Pacific Grove, CA.

Fulk, J., A. J. Flanagin, M. E. Kalman, P. R. Monge, T. Ryan. 1996.Connective and communal public goods in interactive com-munication systems. Commun. Theory 6(1) 60-87.

Furlong, M. S. 1995. An electronic community for older adults: TheSeniorNet network. /. Commun. 39(3) 145-153.

Galbraith, J. R. 1973. Designing Complex Organizations. Addison-Wesley, Reading, MA.

Greene, W. H. 1993. Econometric Anatysis, 2"'' Ed. Macmillan Pub-lishing, New York.

Hagel, J. Ill, A. G. Armstrong, 1997. net.gain: Expanding marketsthrough virtual communities. Harvard Business School Press,Boston, MA.

Hare, A. P. 1976. Handbook of Small Group Research 2nd Ed. Free Press,New York.

Haveman, H. A. 1993. Organizational size and change: Diversifica-tion in the savings and loan industry after deregulation. Admin.Sci. Quart. 38(1) 20-47.

Hiltz, S. R. 1985. Online Communities: A Case Study of the Office of theFuture. Ablex Publishing, Norwood, NJ.

Hirschman, A. 0.1964. The paternity of an index. Amer. Econom. Rev.54(5)761.

Hof, R. D., S. Browder, P. Elstrom. 1997. Internet communities. Bus.Week (May 5).

Homans, G. C. 1950. The Human Group. Harcourt, Brace, and World,New York.

Indik, B. P. 1965. Organization size and member participation: Someempirical tests of alternative explanations. Human Relations 18339-350.

James, J. 1953. The distribution of free-forming small group size.Amer. Soc. Rev. 18 569-570.

Kaufer, D. S., K. M. Carley. 1993. Communication at a Distance: TheInfluence of Print on Sociocuttural Organization and Change. Law-rence Erlbaum, Hillsdale, NJ.

King, S. 1994. Analysis of electronic support groups for recoveringaddicts. Interpersonal Comput. Tech. 2(3) 47-56.

Kraut, R. E., P. Attewell. 1993. Electronic mail and organizationalknowledge. Working paper, Carnegie Mellon University, Pitts-burgh, PA., W. Scherlis, T. Mukhopadhyay, J. Manning, S. Kiesler. 1996.The HomeNet field trial of residential internet services. Com-mun. ACM 39(12) 55-63.

Krech, D., R. S. Crutchfield. 1948. Theory and Problems of Social Psy-chology. McGraw Hill, New York.

Lipnack, J., J. Stamps. 1997. Virtual Teams: Reaching Across Space,Time, and Organizations with Technology. John Wiley and Sons,New York.

Malone, T., J. Yates, R. Benjamin. 1987. Electronic markets and elec-tronic hierarchies: Effects of information technology on marketstructure and corporate strategies. Commun. ACM 30(6) 484—497.

Markus, M. L. 1990. Toward a "Critical Mass" theory of interactivemedia. J. Fulk, G. Steinfield, eds. Organizations and Comjnuni-cation Technology. Sage, Newbury Park, GA. 194-218., T. Gonnolly. 1990. Why CSGW applications fail: Problems inthe adoption of interdependent work tools. Proc. Conf. ComputerSupported Work Association for Computer Machinery, NewYork, 371-380.

Marwell, G., P. Oliver. 1993. The Critical Mass in Collective Action: AMicro-Social Theory. Gambridge University Press, New York.

McGlelland, D. G. 1985. How motives, skills, and values determinewhat people do. Amer. Psych. 40 812-825.

McGormick, N., J. McGormick. 1992. Gomputer friends and foes:Gontent of undergraduates' electronic mail. Comput. in HumanBehavior 8 379^05.

McPherson, M. 1983. The size of voluntary organizations. SocialForces 61(4) 1045-1064.. 1990. Evolution in communities of voluntary organizations.J. V. Singh, Ed. Organizational Evolution New Directions. Sage,Newbury Park, GA. 224-245., Thomas Rotolo. 1996. Testing a dynamic model of social com-parison: Diversity and change in voluntary groups. Amer. Soc.Rev. 61 179-202.

Meyer, G. R. 1989. The social organization of the computer under-ground. Master's thesis. Northern Illinois University, Dekalb,IL.

Milgram, S., L. Bickman, L. Berkowitz. 1969. Note on the drawingpower of crowds of different sizes. /. Personality and Social Psych.13 79-82.

Moreland, R. L., J.M. Levine. 1982. Socialization in small groups:Temporal changes in individual-group interactions. L. Berkow-itz, ed. Advances in Experimental Social Psychology. AcademicPress, New York. 15 137-192., , M. L. Wingert. 1996. Greating the ideal group: Gompo-sition effects at work. E. H. Witte, J. H. Davis, eds. Understand-ing Group Behavior, Vol. 2. Small Group Processes and Interper-sonal Relations. Lawrence Erlbaum Associates, Mahwah, NJ.11-35.

Nunamaker, J. F., A. R. Dennis, J. S. Valacich, D. R. Vogel, J. F.George. 1991. Electronic meeting systems to support groupwork. Commun. ACM 34 40-61.

Ogan, C. 1993. Listserver communication during the Gulf War: Whatkind of medium is the Bulletin Board? /. Broadcasting and Elec-tronic Media 37(2) 177-196.

Olson, M. 1965. The Logic of Collective Action: Public Goods and theTheory of Groups. Harvard University Press, Gambridge, MA.

Osborn, A. F. 1957. Applied Imagination. Scribner, New York.Ostrom, E. 1990. Governing the commons: The evolution of institu-

tions for collective action. Gambridge University Press, NewYork.

Petty, R. E., S. G. Harkins, K. D. Williams, B. Latan6.1977. Theeffectsof group size on cognitive effort and evaluation. Personality andSocial Psych. Bull. 3 579-582.

Porter, L. W., E. E. Lawler III. 1965. Properties of organization struc-ture in relation to job attitudes of job behavior. Psych. Bull. 6423-51.

INFORMATION SYSTEMS RESEARCHVol. 12, No. 4, Decetnber 2001 361


Rafaeli, S., R. J. LaRose. 1993. Electronic bulletin boards and 'publicgoods' explanations of collaborative mass media. Commun. Res.20(2) 177-197.

Rheingold, H. 1993. The Virtual Community. Addison-Wesley, Read-ing, MA.

Rice, R. E. 1982. Communication networking in computer-conferencing systems: A longitudinal study of group roles andsystem structure. M. Burgoon, ed. Communication Yearkbook,Vol. 6. Sage, Beverly Hills, CA. 925-944.. 1987. New patterns of social structure in an information soci-ety. J. Schement, L. Lievrouw, eds. Competing Visions, ComplexRealities: Social Aspects of the Information Society. Ablex, Nor-wood, NJ. 107-120.. 1990. Computer-mediated communication system networkdata: Theoretical concerns and empirical examples, lnternat. J.Man-Machine Stud. 32 627-^7., G. Love. 1987. Electronic emotion: Socioemotional content in acomputer-mediated communication network. Commun. Res.14(1) 85-108.

Roberts, T. L. 1998. Are newsgroups virtual communities? Proc.CHI'98, Los Angeles, CA, April 18-23.

Robey, D., M. Boudreau. 1999. Accounting for the contradictory or-ganizational consequences of information technology: Theo-retical directions and methodological implications. Inform. Sys-tems Res. 10(2) 167-185.

Rubenstein, C. M., P. Shaver. 1980. Loneliness in two northeasterncities. J. Hartog, R. Audy, eds. The Anatomy of Loneliness. Inter-national Universities Press, New York.

Scott, W. R. 1992. Organizations: Rational, Natural and Open Systems,3"* Ed. Prentice Hall, Englewood Cliffs, NJ.

Shaw, M. B. 1981. Group Dynamics: The Psychology of Small Group Be-havior, y^ ed. McGraw-Hill, New York.

Simon, H. A., B. Bonini. 1958. On the size distribution of businessfirms. Amer. Econom. Rev. 607-617.

Slater, P. E. 1958. Contrasting correlates of group size. Sociometry 21129-139.

Sproull, L., S. Kiesler. 1990. Connections: New Ways of Working in theNetworked Organization. MIT Press, Boston, MA.

Stogdill, R. M. 1959. Individual Behavior and Group Achievement. Ox-ford, New York.

Thorn, B. K., T. Connolly. 1990. Discretionary databases. J. Fulk, C.Steinfield, eds. Organizations and Communication Technology.

Sage, Newbury Park, CA.Thraser, Fredric M. 1927. The Gang. University of Chicago, Chicago,

IL.Turoff, M. 1991. Computer-mediated communication requirements

for group support. /. Organ. Computing 1 85-113.Walther, J. B. 1994. Anticipated ongoing interaction vs. channel ef-

fects on relational communication in computer-mediated inter-action. Human Commun. Res. 20(4) 473-501.. 1996. Computer-mediated communication: Impersonal, inter-personal, and hyperpersonal interaction. Commun. Res. 23(1) 3-43.

Wellman, B. 1990. The place of kinfolk in community networks. Mar-riage and Family Rev. 15 195-228.

. 1995. Lay referral networks: Using conventional medicine andalternative therapies for low back pain. Soc. Health Care 12 213-238., J. Salaff, D. Dimitrova, L. Garton, M. Gulia, C.Haythornthwaite. 1996. Computer Networks as Social Networks:

Collaborative Work, Telework, and Virtual Community. Annual Re-

view of Sociology 22 213-238.

, S. Whortley. 1989. Brothers' keepers: Situating kinship rela-tions in broader networked of social support. Sociological Per-spectives 32 273-306., . 1990. Different strokes from different folks: Communityties and social support. Amer. J. Soc. 96 558-588.

Whittaker, S. 1996. Talking to strangers: An evaluation of factorsaffecting electronic collaboration. Proc. CSCW'96 Boston, MA,November 16-20.

Winter, D. G. 1973. The Power Motive. Free Press, New York.Wittenbaum, G. M., G. Stasser. 1996. Management of information in

small groups. J. L. Nye, A. M. Brower, eds. What's Social aboutSocial Cognition?: Research on Socially Shared Cognition in Small

Groups. Sage, Thousand Oaks, CA 261-282.

Izak Benbasat, Senior Editor. This paper was received on January 4, 2000, and was with the author 9 months for 2 revisions.


Date post:	23-May-2017
Category:	Documents
Upload:	cristina-toma
View:	244 times
Download:	3 times

Butler (2001) - Resource-Based Model of Online Social Structures

Documents