Virtual Community: a Sociocognitive Analysis

Neil A. Granitz, Arizona State University
James C. Ward, Arizona State University
[ to cite ]:
Neil A. Granitz and James C. Ward (1996) ,"Virtual Community: a Sociocognitive Analysis", in NA - Advances in Consumer Research Volume 23, eds. Kim P. Corfman and John G. Lynch Jr., Provo, UT : Association for Consumer Research, Pages: 161-166.

Advances in Consumer Research Volume 23, 1996      Pages 161-166


Neil A. Granitz, Arizona State University

James C. Ward, Arizona State University

The internet is a new forum for consumer behavior, a forum that is rapidly expanding, but has been the focus of little published research in marketing outlets. Since 1988, the number of individuals with access to the internet appears to have almost doubled each year, growing to a conservative estimate of 27 million people worldwide in 1994 (Merit 1994). Many consumer and industrial marketers, now have a presence on the internet, or plan to, often on the World Wide Web, whose graphical interface facilitates promotional efforts. The mass media, consumer marketers, and advertising agencies seem to be in the midst of internet discovery and exploitation (Wells 1994, Fawcett 1994, Donaton 1994). Despite the apparent paucity of research about how consumers will behave in computer-mediated communications environments, sweeping speculations are being made about how the information superhighway will change marketing.

Practitioners and academics alike have speculated that as consumer decisions move on-line, the cognitive and social context of decision-making will change in ways that are as yet only partially understood (Benjamin and Wigand 1995). For example, at the individual cognitive level, the increasing availability of extensive, easily retrievable, and easily stored databases relevant to product/service purchases may lower the cost of information search for the typical on-line consumer, perhaps decreasing the proportion of consumers that engage in what some researchers consider a sub-optimal degree of search.

However, the sociocultural, not the cognitive, aspects of internet participation have prompted the most discussion and interest in the media, and have significant implications for consumer behavior on-line. In particular, the idea that virtual communities, novel sociocultural environments, exist in cyberspace, has been widely remarked upon and debated (Barlow 1995, Jones 1995). For students of consumer behavior, the possibility of virtual community raises the issue of whether consumers are turning to the internet for "w.o.m." (word-of-mouth) advice about their purchase decisions.

One of the strongest and most established ideas about the transmission of marketplace information is the importance of interpersonal communication. Extremely consistent research has demonstrated the significance of interpersonal sources in influencing marketplace choice (Price and Feick 1984). Research has shown that interpersonal communication affects preference and choices (Arndt 1967) and that interpersonal communication is often the most important source of information (Katona and Mueller 1955).

Traditionally, many consumers seeking expert, unbiased advice about a purchase have turned to their circle of social ties C family members, co-workers, neighbors, and friends. But in an America increasingly characterized by social disintegration (single parent households, only children, the break-up of the extended family, the increasing rate of career changes and moves), many consumers' circle of ties may be limited and include no one with expertise about their interest. Consumers may be turning to the internet to not just examine ads and order, but to interact with others who share their "consuming passions" in discussion groups such as those in the rec. and alt. areas of the usenet, a part of the internet devoted to news groups whose members communicate with one another by e-mailing messages to a common bulletin board. Many of these groups are devoted to the discussion of products, services, or sources of entertainment.

The reality of virtual community is controversial. Some commentators have documented, in a largely anecdotal sense, their own experience with communities of others on the net (Rheingold 1995, MacKinnon 1995). By some accounts, many have found their internet friends far closer, and more influential in major and minor decisions (including consumer decisions) than friends, co-workers, or even family.

Others have questioned the existence of virtual community by pointing out that participants rarely have any on-going commitment to computer-mediated social relations and often feel no sense of responsibility for the consequences or accuracy of what they say (Kadi 1995). They point out that participants in discussion groups are usually anonymous, physically distant, and not involved in the on-going exchanges of favor for favor that characterize many social relations in the "real" world. Thus, some maintain that the "social ties" that exist are hardly worthy of the name, or are at least of a different character than social ties in face-to-face communities.

Students of consumer behavior have so far been largely absent from the debate about virtual community. However, the social and technical environment may have wide-ranging influence on the interpersonal aspects of on-line consumer behavior. Marketers have long known the critical role of interpersonal ties in a variety of phenomena studied under headings such as word-of-mouth, reference group influence, group decision-making, and opinion leadership/diffusion of innovations. If interactive computer-mediated discussion groups represent a new technical and perhaps cultural context for social relations, the study of community on the net, and its influence on the content, patterns, and structure of interchange, is vitally important for consumer researchers to understand.

Face-to-face communities include, minimally, social actors, social ties, and communications among the actors, channelled by the pattern of ties among them. Each of these, and their interrelations, should be studied in the context of consumption-related communications in computer-mediated groups. The possible differences that may exist between face-to-face and cyberspace groups are illustrated by the differences between w.o.m. (word-of-mouth) and w.o.l. (word-on-line). In an internet discussion forum, anyone, even a complete stranger, can break into the discussion of the community at any time. Such behavior is not considered unusual, forward, or impolite as it might be in face to face groups. In fact, many "netizens" would consider saying something as better than merely "lurking." Furthermore, a questioner often addresses the group as a whole, or whoever is on-line, not just one target individual. Everyone is empowered to reply, although a significant question is how many do, and in particular who really does much of the question answering, explaining, and advice giving in usenet groups. The replies to a question create a conversational "thread," a record of the discussion, perhaps spanning several days, about a topic. Here again is a significant difference of w.o.l. from w.o.m. W.O.M. is said and expires, except perhaps in the memory of the listener. W.-O.-L. becomes part of a public record, that may influence hundreds or thousands more "listeners" than the original target of the comment. W.O.L. and other sociocultural aspects of internet discussion groups seem sufficiently novel phenomena to deserve further study.

The nature of communication is one aspect of consumer behavior in discussion groups, but it is created by individuals and channelled by their social ties. Thus, another important issue to understand in the context of computer-mediated discussion groups is the character of the participants and how their individual differences (e.g., in expertise about the topic) interact with social ties and structure.

Ward and Reingen (1990) have emphasized the importance of studying how social and cognitive processes interact on group decision-making. An attempt to relate individual differences, social structure, and communication flows, would seem to offer a useful approach for the study of virtual community on the internet. If virtual communities exist in computer-mediated discussion groups, they represent a novel context for social relations and the relation of individual differences to the content and structure of these relations. Social actors on the usenet create roles for themselves in a decontextualized environment. This freedom from the usual indicants of role and status (e.g., wealth, appearance) may increase the salience of other social resources, such as expertise about the focal topic, as an indicant of status. A sociocognitive approach suggests a focus on the relation of individual status indicants on the net, such as expertise, to social structure, and the flow and content of communication.

In the present study, we explored the nature of virtual community on a usenet discussion group and its relation to the flow of consumption-relevant communication. The usenet refers to a set of approximately 7500 discussion groups distributed world-wide across the internet. Under a variety of headings, users can discuss issues ranging from medical ethics to Melrose place. The usenet is only a small part of the internet, but it is representative of the rapidly growing parts of the net that provide forums for individuals to discuss topics of interest. These forums exist on not only the usenet, but on commercial services such as American Online, local bulletin board systems (e.g., the San Francisco bay area's Well), and on other parts of the net. These forums have been the focus of discussions about virtual community, and are relevant to marketers because of their popularity, and their frequent devotion to products and services.

We approached our study with several specific research questions, influenced by an interest in the relation between social and individual processes in human communities. First, we expected to find at least some evidence of community in a usenet group, as shown by ongoing communications defining a social structure consisting of a pattern of ties among the participants. Second, we expected cognitive and social structures to be related. Since the purpose of a usenet group devoted to a product or service is information exchange about the focal topic, knowledge should translate into social power. Celsi et. al. (1993) noted this when they found that within-group status increased as one gained experience and formed interpersonal relationships. Given the uneven distribution of knowledge likely about even a low-tech product or service, we expected more expert participants to be at the center of the group's social structure. Similar to the findings of Schouten and McAlexander (1995) regarding novice Harley-Davidson bikers and "experienced" Harley-Davidson bikers, we expected novices to direct questions towards experts and to accept their comments about a variety of issues. In accordance with Raven (1965), we expected expert power to confer the status to arbitrate opinion on the net. We also expected to find that sociocognitive structures would channel the content of communication. That is, we expected experts to be the actors most likely to correct and critique other users' opinions, and novices to be most likely to ask questions. Finally, although we anticipated finding virtual community on the usenet, we expected that only a small minority of participants in even a special interest group would be involved in extensive interaction with one another. Our speculation about this last issue was based upon exploratory observation of usenet groups.


The Rec.Food.Drink.Coffee group (a group devoted to the discussion of coffee consumption) was chosen for study because it met several criteria. First, the group focused on consumption. Second, the group had been in existence for some time ( several years), so patterns of social structure had a chance to emerge. Third, the group was currently active, with well over one hundred members contributing. Finally, the group concerned a commonly purchased product, coffee. We felt that study of a group focused on more esoteric or technical issues might be prone to exaggerate the role of expertise or social ties in channelling communication.

We examined the complete record of discussion for two separate two week periods in early February and early March. In total, 204 people contributed to the usenet discussion group in the periods studied. The choice of two separate periods not too widely separated in time kept the data to a manageable size and provided a more longitudinal perspective while not sacrificing the continuity of the group. The sample had to be cleansed of discussions that began prior to the start of data collection, because without a complete record, we found it hard to identify the participants and often the topic of the string. However, whether a string was answered or not answered - if it began in one of the periods, it was included in the analysis. This lead to a total number of 24 strings studied in period 1 and 31 strings studied in period 2. Usenet members may post messages directly to one another's E-mail addresses. These communications could not be captured for our study.

Transcript Analysis

Usenet discussions are not "real-time." Users send mail to one another which is posted in the discussion groups' file. Participants may address the group as a whole, perhaps by asking a question, or may specifically "talk" to another participant by incorporating his or her name, or often a portion of some previous message being commented upon. Participants seemed careful to identify the conversation they were contributing to, a habit that facilitated analysis of the data. The order and hierarchical structure of the report, along with dates, names, and addresses on the messages themselves, helped organize the conversation for analysis. Two types of analysis was performed on the data - content analysis and social network analysis.

(i) Content Analysis

The content analysis had several objectives. The first was to determine the type of consumption related comment. We classified comments into the following categories: questions (how do I make espresso?), product recommendations (buy the Acme espresso machine), advice about how to do something related to coffee (this is how you make good espresso), explanation of how something works (here's how the Acme espresso machine works), comments (adding to a conversation C you should also consider this when buying an espresso machine), disagreements (you are wrong about how the Acme espresso machine works), announcements (Acme has a new machine available), and what we classified as assorted replies. The total length in number of words of individual contributions by category was also estimated and recorded.

The second objective of the content analysis was to measure the number of different aspects of coffee consumption that a particular individual mentioned in his or her total volume of comments. The number of different attributes in a person's cognitive structure for a product is one measure of his or her expertise about the product. After a thorough preliminary study of the transcripts, we developed a list of 34 different coffee-related concepts that we scored all the transcripts for. We then computed the total number of comments in each category for each individual under study, and labelled the variable "aspects".



(ii) Social Network Analysis

The social network analysis began by creating a participant by participant interaction matrix. Each message was coded for its originator and its target (the message it referred to). Messages with no particular reference were coded as being directed to the group as a whole. We then counted the number of times each individual contributed messages concerning each other individual's messages. To fit the requirements for network analysis, the data matrix was dichotomized and symmetrized, and then analyses of actor centrality and clique membership were performed using the Ucinet social network analysis software.


The data resulting from the content analysis and the individual level data from the network analysis (individual centrality scores) were entered into a person (204) by variable data file and further analyzed by spss. For relevant analyses, experts were distinguished from novices by a score of four or more on the aspects of coffee noted (complexity) measure. This quantitative measure of "expert" was qualitatively checked by one of the researchers who is an "expert" on coffee and who was familiar with virtually all the on-line experts' arguments. The resultant list of experts matched his subjective judgement of who was expert very well.


Little data has been reported on the content of what is said on usenet discussion groups, particularly as it relates to consumption. Table 1 shows the results of a content analysis.

The column aggregating both expert and novice communications shows that over 20 percent of the total words were devoted to product recommendations, and another 20 percent were devoted to discussions of how to use a product (e.g., how to use an espresso machine). The other categories (e.g., explanations) include passages indirectly concerning consumption (e.g., why world coffee prices increased), and other passages more relevant to purchase decisions (e.g., why inexpensive coffee grinders do not work well). Overall, the majority of talk in the group was highly relevant to product or service purchase decisions.

The influence of social and cognitive factors on the content of communication is evident when we contrast the distribution of type of comment by expert vs. novice. Novices asked virtually all the questions. Experts rarely violated their social role as experts by asking a question. Instead, the ten experts (selected by the complexity of their cognitive structure for coffee) contributed over one third of the words devoted to product recommendations, almost half the comments about others' remarks, and about two-thirds of the remarks that explicitly disagreed with others. These data strongly suggest that the more expert participants in the group played the role of arbiters of group opinion. They, more than anyone else, decided if some remark about coffee was "wrong", and these evaluations were rarely challenged by less expert participants. Overall, experts contributed about one third of the group's content. Thus, we observe an individual cognitive difference (expertise) related to a social role (arbiter of group opinion) channelling communication (the direction of questions) and its content (critical of others or not). The reluctance of experts to ask questions of the group seems at first somewhat surprising, but when we consider that in an on-line environment, the only way to establish and reinforce one's role is through communication, we might be less surprised about the expert's refusal to step out of role in their postings.

We next created a correlation matrix of the communication scores of each actor, the expertise scores of each, and the social network centrality scores of each. The matrix of pearson correlations is shown in Table 2. All significance tests are two-tailed, based on a null hypothesis of zero correlation. The correlations show the strong relation of social structure to the flow of communication in the usenet group. Higher network centrality was significantly correlated with total length of remarks (r=.46), virtually every specific category of remark, and especially disagreement (r=.40). The one exception is asking questions, which shows no significant correlation with network centrality.

Overall, the clique analysis identified 93 cliques of 3 or more actors in the usenet group. Of the 204 participants, 98 were in a clique and the rest were in dyads or isolates. Thus, about half of the participants were connected to a "group." All the expert participants were in one or more cliques. The participant with the highest expertise (aspects) score (20) was also a participant in the most number of cliques (19), or about 20 percent of the total. This actor also had the highest centrality score of any of the experts.

Table 2 also shows the strong positive relationship between expertise and social network centrality (r=.63). Experts in this usenet group, perhaps because of the social value of their knowledge, enjoy privileged network positions with many incoming links. The correlation of disagreement with centrality (r=.40), and the expertise measure (r=.73), provides some support to the validity of the expertise measure. Despite the high correlation between network centrality and expertise, the latter was much more strongly correlated (p<.05) to disagreement, perhaps because evaluation is usually a social function of experts.



We performed t-tests of expert vs. novice mean difference scores across the communication content and network centrality measures. The tests show a pattern of significant communication differences between experts and novices. Experts had higher mean network centrality (15.9 vs. 3.1, t=3.5, p=.01) than novices, and of course, higher aspect (# of coffee attributes employed) scores (7.2 vs. .5, t=9.0, p=.01). Novices asked significantly more questions than experts (11.3 for novices vs. 1.0 for experts, t=3.6, p<.01), but experts had higher mean scores for disagreements (69.5 vs. 1.7, t=2.1, p=.07), comments (63.2 vs. 3.7, t=2.2, p=.05), product recommendations (123.8 vs. 11.7, t=1.9, p=.09), and explanations (73.6 vs. 15.0, t=1.9, p=.06). Experts also had a marginally higher mean for how-to advice (99.4 vs. 12.1, t=1.8, p=.10). The small sample size of experts (n=10), and high variances resulted in significance levels less than one might expect given the magnitude of the mean differences. These expert versus novice differences were re-tested defining expert participants mentioning three "aspects" of coffee instead of four. The results were the same for 7 out of 8 tests. A chi-square test (p<.05) also confirmed that content categories vary across the expert and novice distinctions. Overall, experts were clearly influential sources of W.O.L. in the group examined.


We began our project with the objective of exploring how individual characteristics, in particular expertise, related to the social structure of a usenet discussion group, and how the individual and social dimensions of the group related to the pattern of consumption-relevant communication within the group. Our sample of interactions strongly confirmed our expectations. Social and cognitive structure were closely related on the usenet. More central network actors seemed to be more expert about coffee, as measured by the complexity of their knowledge structures about coffee. These more central, expert actors appeared to function as informational resources and arbiters of opinion on the net. However, our suggestion about the apparent influence of experts is an inference drawn from the structure of the discourse, not a conclusion based on any measure of actual influence.

Sociocognitive structure channelled communication and its content. Less socially central participants were more likely than more central participants to ask questions of the group. The more central actors were less likely to do so, perhaps because asking a question of the group would have undermined their role as experts who knew all the answers, and thus had little need of asking questions. Central actors were much more involved than most participants in answering questions. These answers included product recommendations (XYZ is a good firm to mail order coffee from), how-to advice (e.g., how to modify a particular brand of coffee grinder so it works better), and explanations (e.g., why a fine grind is needed for Turkish coffee). The asymmetrical structure of discourse in these groups suggests that central actors were exercising expert power (French and Raven 1959), and had real influence on other participants. Whether advice was followed is an issue that could be followed up in future research.

One implication of these findings for marketers is that a relatively small number of participants on usenet groups may strongly influence the opinions of hundreds of other participants, and perhaps thousands of "lurkers" who peruse the group discussion without contributing.

More central actors enjoyed "prestige" mainly because of their expertise. In our sample of strings, the central actors made little of such indicators of status as occupation, income, or place of residence that may influence whose opinion is respected in face-to-face groups. Also, obviously, the appearance (attractiveness, height, dress) of these actors could not influence other's reliance upon them. Although more research is necessary, these observations suggest that opinion leaders in computer-mediated groups like the usenet may rely on differently weighted bases of social power than those found in face-to-face groups.

Earlier we raised the issue of the nature and extent of "virtual community" in usenet groups and similar forums on the internet. Our data provide some insight into this debate. We found cliques within the usenet that often centered around expert participants. Thus, interaction had social structure, perhaps one sign of community. Although the data entered into our network analysis reflected only the presence or absence of a tie, our initial tabulation of interactions among actors found that most of the participants in cliques were interacting many times over a period of several weeks. Thus, social ties influenced communication, another sign of community. Finally, we noted that only a small number of the participants had any degree of social centrality. A slight majority were not in cliques, and only contributed a few lines of comment. Thus, virtual communities that influence consumption decisions appear to exist on the internet, but the percent of users involved to any extent in such groups may be small.


Marketers, more than academics, have taken the initiative in calling for research on internet consumer behavior. However, available wisdom appears sketchy, and is more often based upon opinion or personal experience than systematic research. The explosion of computer-mediated consumption behavior is an opportunity for consumer researchers to do timely, much needed studies that could be of considerable theoretical interest because of the unique character of interaction on the net.

The consumption phenomena on the internet that seem high priorities for exploration include W.O.L. (word on-line), referral networks, opinion leadership, diffusion of innovations, information search (and retention), involvement with interactive ads, the character of discourse, attitudes toward marketing, characteristics of users, expert-novice differences, the character of community, and the extent to which participants rely on one another for consumer advice, to name only a few of the most obvious potential issues.

The computer-mediated discussion groups have enduring characteristics as social and cultural environments that justify re-exploring issues previously studied in more conventional environments. Discussion groups such as the one we investigated communicate to one another in a way different in many respects from face-to-face discussion. First, the communication is written, not spoken. Contributors to the discourse have the opportunity to carefully ponder and craft their remarks. Furthermore, they do so with a complete record of the comments they are responding to at their disposal. They need not pause, interrupt themselves, or limit the length of their remarks to give another speaker his or her turn, as would be expected in conversational discourse. Finally, contributors are no doubt conscious that their remarks become part of a public record, that hundreds, if not thousands may access or respond to.

Additionally, the participants interact in a social environment that is different in more than just a physical sense. They are separated from the social and cultural contexts that influence the character of face-to-face exchanges. This raises fascinating issues about the interaction between the individual and the group in the creation of identities and roles. In most social contexts, the individual's choice of identity and role is highly constrained by the circumstances of their background, appearance, status, neighborhood and workplace. On the internet, the individual is free to choose the groups he or she wishes to identify with, and create a role as "expert" divorced of prior social disqualifications from such roles. Confronted with this opportunity, do consumers merely re-enact their daily roles, or create new and different ones?.

The social and technical environment of the internet may be uniquely facilitative for the pursuit of consumer research. For example, Ward and Reingen (1990) have emphasized the importance of studying how social and cognitive processes interact in group decision-making. Unfortunately, access to a precise sequential history of interactions, and their content, is difficult in most field settings except computer-mediated communication environments such as usenet groups. The precise record of discussion groups could facilitate analysis of communications focusing on more than their content. Discourse analysis is the study of the structure of conversation C the structure of who speaks, in what way, and how much. Consumer researchers could employ discourse analysis to better understand such issues as opinion leadership.

Furthermore, researchers need not be limited to passive analysis of the products of interaction on the usenet. Participants could easily be interviewed via E-mail. The possibility exists of asking participants to come to the researchers' own internet site, where they might be asked to interact with software there. Either approach would allow the collection of questionnaire measures of influence, opinion leadership, social structure, etc. The internet presents some difficult sampling issues, and some real sampling opportunities. At present, internet users are not "representative." They are likely to be male, higher in socioeconomic status, and more computer-literate than average. Furthermore, identification of group participants is often difficult since some contribute only periodically and others are perpetual "lurkers" who monitor but do not contribute to the conversation. Longitudinal data collection seems a necessity. The sampling opportunity represented by the internet is the almost unprecedented access offered to narrowly defined groups of consumers from around the world. A sample of consumers with even the most rare or bizarre interest is now but a mouse click away.


Barlow, John Perry (1995), "Is There a There in Cyberspace?" Utne Reader, No. 68 (March-April), 53-59.

Benjamin, Robert and Rolf Wigand (1995), "Electronic Markets and Virtual Value Chains on the Information Superhighway," Sloan Management Review, (Winter), 62-72.

Celsi, Richard L., Randall L. Rose, and Thomas W. Leigh (1993), "An Exploration of High Risk Leisure Consumption through Skydiving," Journal of Consumer Research, 20 (June), 1-23.

Donaton, Scott (1994), "OK to put Ads on Internet, but Mind your Netiquette," Advertising Age, Vol.65 No.18 (April), 3.

Fawcett, Adrienne Ward (1994) "Interactive Awareness Growing," Advertising Age, Vol. 65 No. 42, 30.

French, John R.P. jr. and Bertram H. Raven (1959), "The Bases of Social Power," in Studies in Social Power. ed. Dorwin Cartwright, Ann Arbor: University Michigan Press, 150-167.

Jones, Steven G. (1995), "Understanding Community in the Information Age," in Cybersociety: Computer-Mediated Communication and Community, ed. Steven G. Jones, Thousand Oaks, California: Sage Publications, 10-35.

Kadi, M.(1995), "Welcome to Cyberia," Utne Reader, Vol. 68, 57-59.

Katona, George and Eva Mueller (1955), "A Study of Purchase Decisions, " in The Dynamics of Consumer Reaction, ed. Lincoln H. Clark, New York: New York University Press, 30-87.

MacEvoy, Bruce and Linton Freeman (1987), UCINET: A Microcomputer Package for Network Analysis. Irvine, Ca: University of California, School of Social Sciences, Mathematical Science Group.

MacKinnon, Richard C. (1995), "Searching for the Leviathan in Usenet," in Cybersociety: Computer-Mediated Communication and Community, ed. Steven G. Jones, Thousand Oaks, California: Sage Publications, 112-137.

Merit Network, Inc. (1994), "Internet Growth Statistics." URL:

Price, Linda and Lawrence Feick (1984), "The Role of Interpersonal Sources in External Search: An Informational Perspective," in Advances in Consumer Research lX , ed. Thomas C. Kinnear, Ann Arbor Mi: Association for Consumer Research, 250-253.

Raven, Bertram H. (1965), "Social Influence and power, " in Current Studies in Social Psychology, ed. I.D. Steiner and M. Fishbein, New York: Holt Rinehart and Winston, Inc. 371-378.

Rheingold, Howard (1995), "The Virtual Community," Utne Reader, No. 68 (March-April), 60-65.

Schouten, John W. and James H. McAlexander (1995), "Subcultures of Consumption: An Ethnography of the New Bikers," The Journal of Consumer Research, 22 (June), 43-61.

Ward, James C. and Peter H. Reingen (1990), "Sociocognitive Analysis of Group Decision Making among Consumers," The Journal of Consumer Research, 17 (December), 245-262.

Wells, Melanie (1994), "Desperately Seeking the Superhighway," Advertising Age, Vol. 65 No. 43, 14.