Cyberspace is increasingly seen as a medium for social interaction, despite ongoing skepticism that the technology is inimical to meaningful social interactions. The study reported here set out to estimate the frequency and correlates of friendship development via the Internet of a national random sample of 601 Internet users. The results show that a substantial minority of users (14 percent) had established friendships via the Internet. Those with greater Internet experience or skills appeared more likely to make friends via the Internet. There appeared to be only weak or in most cases non-existent statistical relationships with demographic variables, measures of traditional forms of interaction, and measures of personality attributes. Those making friends via the Internet reported a significant number of such friendships - 70 percent reported four or more Internet friends. Those with longer experience of the Internet reported more friendships. Of those who developed friendships through the Internet, three out of five reported meeting one or more of their Internet friends. Overall, our survey suggests that the Internet is indeed a medium where friendships can be developed; that Internet friendships frequently lead to face-to-face meetings; and that success at friendship creation appears more related to Internet experience and skills than social and personality characteristics.
We seek insight on the nature and extent of friendship formation among on-line users. In light of some early research on computer-mediated communication (CMC), one might have predicted the uses of cyberspace for social communication to be infrequent and constricted, and the subsequent relationship formation minimal. For example, Baron (1984), based partly on some of Kiesler and associates’ work (1984), wrote that CMC "is ill-suited for such ‘social’ uses of language." (1984: 136) Others have held that for technical reasons – the medium is too restrictive – that not much in the way of lively social relationships would develop. (c.f. Rice, 1987; Rice and Love, 1987)
The growing social dimension in cyberspace
However, since that time, a plethora of works on cyberspace have described various aspects of the explosion of social communications, including such ostensibly inappropriate places as task-oriented groups and computer-supported cooperative work (CSCW) sites. An intriguing aspect of many of these studies has been the finding of how extensively "social" and informal behavior intrudes itself in the "business-like" sphere of task-oriented groups and educational spheres. This occurs despite routine system administration warnings that the computer resource is only to be used for official business and that all activities are monitored. Numerous case studies of CMC have shown that "the social" is an important glue that binds together the task-oriented aspects of CMC, and in some cases even supplants them. (Rice, 1987) This work has been complemented by research on the functioning of psychological help-groups, IRCs, MUDs, MOOs, and even on-line dating services, all of which are essentially social and "affect" (as opposed to task) oriented. The recent emergence of web-based chat-rooms which offer "community" would seem to provide additional evidence refuting the "non-social" nature of CMC. Baym summarizes a decade of research as demonstrating "the ways in which people have appropriated the commercial and non-commercial networks demonstrate that CMC not only lends itself to social uses but is, in fact, a site for an unusual amount of social creativity." (Baym, 1995: 160).
Cyberspace friendship: theoretical viewpoints
But does social activity and creativity translate into meaningful friendships and relationships? Two schools of thought on cyberspace suggest that the answer is, for practical purposes and in most cases, no.
The first school of thought holds that CMC technology is too inherently antithetical to the nature of human life for meaningful relationships to form. (Stoll, 1995) To illustrate, Beninger and Baudrillard argue CMC is basically alienating and unfulfilling. To type is not to be human, to be in cyberspace is not to be real. Rather all is pretense and alienation, a poor substitute for the real thing. Ipso facto, cyberspace cannot be a source of meaningful friendships. (Beninger, 1987; Baudrillard, 1983; Numes, 1995)
A second school of thought maintains the technology is basically too limited to provide a useful basis for relationship formation. Hence, CMC inherently leads to "experimentation" (that is mis-representing oneself to others who cannot directly learn what the truth is) about one’s identity and qualities. The atmosphere would be dominated by trickery, lechery, manipulation, emotional swindles. As a result, so much posturing, "gender-switching" and faking of identities would take place that it would be extremely difficult for any real relationships to be created let alone maintained. (Turkle, 1996)
Both these schools of thought, often informed by the expositions of postmodernist theorists, have been everted by Wynn and Katz (1997). Wynn and Katz essentially argue on phenomenological and empirical grounds that doubtless such dissembling does exist in cyberspace, as it does everywhere else from street-corner prostitution and telephone pranks, to carnival celebrations and "lonely hearts" newspaper ads. They hold, however, that despite the existence of anti-, non-, and pseudo-social behavior among some denizens of cyberspace, what counts in this regard is the results for most of the people most of the time. They show that cyberspace is not phenomenologically special or new, that group norms and rules are enforced, that penalties exist for violations of ethical codes, and that lots of people are sincere and integrative in their attempts at cyberspace communication. (Witness, they say, the rise of personal home pages.) What does matter, ultimately, is the results and the real consequences of behaviors in cyberspace. Within the subject under discussion, the import of their work can be seen as a theoretical bolstering for a position that cyberspace-originated friendships can be sincere, meaningful, and durable.
Empirical work to date has focused on specialized case histories
As further evidence in support of friendship formation, at least on an initial basis, there is a host of case studies of use groups or special interest bulletin boards that describe a variety of friendships springing up among members, often leading to efforts to meet face-to-face, either as a group from the general list, or between two members who are drawn to each other. However, there is little in the empirical literature on what happens to these friendships over time, that is, whether they continue or are ephemeral.
Although by no means representative, the media have also reported many stories dealing with relationships which have formed on-line then migrated to face-to-face, often with disastrous results (thus the media attention). In our on-going efforts to interview Internet users, many have reported making friends on-line. In some cases they travel to meet these new friends, crossing continents or even oceans. And in some cases they break-off long-standing relationships or even marriages to be with their new friend. But is on-line friendship formation, short of the more dramatic instances cited above, something rare, and thus comes to our attention because they are exotic, surprising, or unusual? Or is on-line friendship formation rather commonplace, and thus likely to be encountered in the course of ordinary reporting and research on cyberspace behavior?
The case studies which have been reported make no pretense to being representative, but instead are quite limited in their scope and time-span coverage. Thus we remain largely in the dark in terms of what most of the denizens of the Internet are doing most of the time in terms of relationships. Despite this important substantive gap in the literature concerning the frequency and growth of on-line relationships, a few studies do stand out as making contributions to this area.
A notable example in this regard is the study by Parks and Floyd (1996) who surveyed via email usenet denizens to probe how often personal relationships form in newsgroups, who has them and how close or developed they become, and if friendships started in this venue, whether they migrate to other settings. They contacted 528 individuals, 176 of whom responded (response rate, 33 percent). They conclude that women are more likely to have formed personal relationships through newsgroups (72% versus 55% of men), but no other demographic variables, including age, or personality factors, seem to make much of a difference. The primary association with number of friendships is the temporal length and intensity of newsgroup involvement. They conclude that "developing personal relationships on-line is more a function of simple experience . . .As people get used to and involved with their favorite newsgroups over time, they appear to start developing personal relationships with one another." (p. 87). They also observed that in the majority of cases, relationships that begin on-line migrate to other settings, such as face-to-face or phone chats.
Research aims
Despite the attention given to the blossoming of particular cases of friendships in cyberspace, we have little information about the statistical frequency with which they occur among the general user population. Also, the absence of knowledge about potential demographic and attitudinal correlates of friendship formation, thus theory-building and evaluation is curtailed.
Hence our purpose is to look at the frequency and correlates of friendship development on the Internet. We examine issues comparable to those of Parks and Floyd. However they focused only on those who posted messages to newsgroups, and relied on email for responses. These respondents, by their self-demonstrated expertise, activism, and special interest, are a special sub-population of Internet and email users. By contrast, we sought to sample a nationwide survey of all Internet users. We believe this is a useful complement since we are dealing with a representative sample of all users. This in turn allows us to understand the situation of more typical users, and hence be able to discuss with more persuasiveness the likely societal impact of the Internet and email in friendship formation. As a result, broader generalizations can be advanced about the possible impact of expanded on-line participation of Americans.
We sought to analyze the extent of friendship formation through the Internet by carrying out a national phone survey. Among our aims were to quantify the amount of friendship formation and to analyze how this related to various demographic variables (gender, age education, income and race). We also wanted to quantify the extent friendship formation led to people meeting in person.
Data for our study was taken from an October 1995 national random telephone sample, surveyed by a commercial firm under our direction. The survey yielded about 2,500 respondents of which 8 percent reported being Internet users, 8 percent reported being former Internet users, 69 percent reported being aware of the Internet but not being users, and 15 percent reported not being aware of the Internet. The sample of Internet users was augmented by a national random telephone sample of about 400 Internet users. Both surveys asked about demographic and occupational characteristics, social attributes and personality traits.
We analyzed our sample of 601 Internet users from the perspective of when they first started using the Internet - 308 reported starting in 1995, 132 in 1994, 79 in 1993, and 82 started pre-1993. We found that recent cohorts are increasingly female, younger, less well-educated, more likely to be working part-time or a be a student, and less well-off. We also examined the marital status and racial/ethnic mix of Internet users, neither of which appeared to change across Internet cohorts. Appendix 1 gives the details.
In the context of friendship formation via the Internet our survey asked four key questions of Internet users:
Friendship formation
We first asked Internet users whether they made any acquaintance exclusively through the Internet whom they now considered to be their friend. Of our 601 Internet users, about one out of seven (82 respondents) or 14 percent reported knowing people in this way.
Demographic variables: We examined statistically how propensity to form friendships though the Internet related to demographic variables. We only found two weak statistical correlations between propensity to form friends via the Internet - those with moderately low educational achievements and those with middle incomes being more likely to form friendships through the Internet.
Gender status (chisq = 0.01, sig = 0.94 with 1 df), marital status (chisq = 6.1, sig = 0.11 with 3 df), age (chisq = 5.8, sig = 0.33 with 5 df), employment status (chisq = 1.9, sig = 0.59 with 3 df) and race (chisq = 2.4, sig = 0.66 with 4 df) did not relate statistically to propensity to form friendships through the Internet.
Our survey showed that those less well academically qualified reported
being more likely to make friends via the Internet. Seventeen percent of
those who had not received a college degree reported making friends through
the Internet, as compared to 11 percent of those with a college degree
only and 10 percent of those with some post-graduate education.
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
Regarding household incomes, those
with medium household incomes ($35-74,000) reported being more likely to
make friends via the Internet. Sixteen percent of those whose household
incomes lay in the range $35-74,000 reported making friends through the
Internet, as compared to 9 percent of those with household incomes up to
$34,000 and 9 percent of those with household incomes $75,000 and above.
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
Our survey data indicates that the older cohorts of Internet users reported being more likely to form friendships through the Internet. Twenty-one percent of the pre-1993 cohort reported forming friendships via the Internet, while 15 percent of the 1993 cohort, 17 percent of the 1993 cohort, and 10 percent of the 1993 cohort did so.
|
|
||||||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
The experience levels were all highly
correlated with the other more general measures of Internet usage, and
all of them correlated with propensity to make friends via the Internet.
For each Internet feature, heavy users had a greater propensity to make
Internet friends. The strongest statistical relationship was with experience
of usenet and user groups where 18 percent of the 601 Internet users reported
being heavy users and among these heavy users 28 percent reported having
made friends via the Internet, as compared to the 40 percent of the 601
Internet users who reported not having heard of usenet and newsgroups,
and among these non-aware users only 7 percent reported having made friends
via the Internet.
|
|
|||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Other strong statistical relationships where with MUDs/MOOs and listservs. Although few Internet users had experience with MUDs/MOOs, those that had were highly likely to make friends through the Internet. Of the 601 Internet users only 5 percent (29 users) reported being light or heavy users of MUDs/MOOs, but of these 31 percent (9 users) reported having made friends via the Internet. Of the 13 percent of Internet users who reported being light or heavy listserv users, 23 percent reported having made friends via the Internet, as compared to the 71 percent of Internet users who reported not having heard of listservs, and among these non-aware users 11 percent reported having made friends via the Internet.
Measures of traditional social
activity: Our survey also explored respondent’s social connectivity
by asking them how often they contacted other people by letter or phone,
how many religious, leisure and community organizations they belonged to,
and how frequently they met friends and family. The specific questions
are given in Table 7. Interestingly, none of these measures correlated
with propensity to make friends via the Internet. This was not because
we lacked data - most contingency tables contained about 200 data points.
Thus in our sample, propensity to make Internet friends does not appear
to correlate with traditional sociability characteristics.
|
|
|
|
|
|
|
(i) by letter? (ii) by phone? (iii) by fax? (iv) by leaving a message by voice mail or answering machine? |
No. of ph calls No. of faxes No. of messages |
|
(v) religious? (vi) leisure? (vii) community? |
No. of orgs No. of orgs |
|
(ix) In your social life you are frequently away from home? (x) Of the ten people living closest to your home, how many do you know personally? |
SA/A/N/D/SD* No. of people |
|
|
|
|
|
|
|
|
|
|
(i) You like using the phone ("extroversion") (ii) You feel you have more to do that you comfortably handle ("feelings of stress") (iii) You’ve had problems in the past because you missed and important phone call ("need to be in touch") (iv) There are often times when you urgently need to get through to another person ("need to be in touch") (v) You sometimes like to do things that are a little dangerous ("willingness to take risks") (vi) Generally, you’re the first person among your friends to go out and try something new (willingness to try new thing") |
SA/A/N/D/SD SA/A/N/D/SD SA/A/N/D/SD SA/A/N/D/SD SA/A/N/D/SD |
|
|
|
|
("willingness to take risks") |
|
|
|
|
Number of friendships formed
For the 81 users who reported establishing friendships via the Internet, a substantial number of friendships were reported. Thirty percent of the group (24 respondents) reported having established friendships with 1-3 people, 40 percent (32 respondents) with 4-10 people, 22 percent (18 respondents) with 11-30 people, and 9 percent (7 respondents) with 31 or more people.
Demographic variables: We examined statistically how the number of friendships formed through the Internet related to a number of demographic variables. We found only one statistically significant results (sig levels < 0.05) - those who had never married reported making fewer friends. There were three further marginally statistically significant results (0.05< sig level < 0.10) - those in full-time employment, the more highly educated and older people reported making more friends. There was no gender or household income effect. We had too few respondents in the non-white racial categories to test whether there was a race/ethnic effect.
Among our sample of Internet users who had established friendships, those who had never married appear to establish fewer friendships - 24 percent of this group reported 6 or more friendships, while 51 percent of the married group and 62 percent of the divorced/separated group reported having 6 or more friendships.
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
||||||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
In our sample 59 percent of the pre-1993 cohort and 73 percent of the 1993 cohort reported establishing 6 or more friendships through the Internet, as compared to only 27 percent of the 1994 cohort and 39 percent of the 1995 cohort.
|
|
||||||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
Measures of traditional social connectivity and personality measures: We had too few data to carry out any meaningful statistical tests.
Internet friendships leading to meetings
According to our survey, a majority of people who reported making friends through the Internet met one or more of them. Of the 81 respondents who reported making friends via the Internet, 60 percent of them reported meeting one or more of these friends.
Those reporting higher number of Internet friends were more likely
to have met at least one of them. Only 29 percent of those with 1-4 Internet
friends reported meeting at least one of their Internet friends, whereas
74 percent of those with 5-12 Internet friends and 79 percent of those
with more than 12 Internet friends reported meeting at least on of their
Internet friends.
|
|
|||||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|
|
|
||
|
|
|||||
Eighty-nine percent of those with a college degree but without postgraduate
education met friends made via the Internet, while only 49 percent of those
with less than a college degree and 55 percent of those with postgraduate
education met friends made via the Internet.
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
Measures of traditional social connectivity and personality measures: Again, we had too few data to carry out any meaningful statistical tests.
Number of Internet friends met
For the 49 users who reported meeting Internet friends, a substantial number of Internet friends were met. Thirty-seven percent of the group (18 respondents) reported meeting up with 1-3 Internet friends, 29 percent (14 respondents) with 4-10 Internet friends, 22 percent (11 respondents) with 11-30 Internet friends, and 12 percent (6 respondents) with 31 or more Internet friends.
Excluding one outlier who reported meeting 50 of his/her 500 Internet friends, the proportions of friends met to friends known was almost 65 percent. This proportion appeared to be independent of the number of friends known.
Demographic variables: We examined statistically how the number of Internet friends met related to our set of demographic variables. We found only one statistically significant result - older respondents reported meeting more Internet friends. We found no gender, work status, marital status, educational, household income or cohort effect.
Only 15 percent of teens or twenties
reported meeting 6 or more Internet friends. Older respondents reported
meeting more Internet friends. Fifty-three percent of those in their thirties
or forties and 83 percent of those 50 and over reported meeting 6 or more
Internet friends.
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
Usage and expertise levels: There were no statistically significant results.
Measures of usage of specific Internet features: Again, there were no statistically significant results.
Measures of traditional social connectivity and personality measures: Once again, we had too few data to carry out any meaningful statistical tests.
Statistical models
Using logit models (Agresti, 1990: 92) we sought to identify the key variables for predicting propensity to make Internet friends and to quantify the relative importance of each key variable, keeping the other key variables fixed. Of all the demographic variables, general measures of Internet usage and experience levels of Internet features we considered, the key variable appeared to be experience of usenets/user groups. The odds ratio of forming Internet friendships was estimated to be 1.45 times higher for users than non-users. Appendix 2 provides the details.
For the number of friends made, the key variable was the year of starting to use the Internet. No other variables added any predictive ability. For users starting in 1993 and before, the odds of having made six or more Internet friends are 1.99 times higher than for those starting in 1994 and 1995. We had many fewer data points for this model than for the model relating to propensity to make friends, and hence our observations are much more tentative. Again, Appendix 2 provides the details.
Developing friendships in Cyberspace: The results of our survey appear to show that a substantial minority of Internet users have established friendships via the Internet. Propensity to make friends via the Internet is most strongly correlated with both general measures of Internet usage and experience, and experience of specific Internet features, suggesting that it takes time to become comfortable with the dynamics of friendship creation in cyberspace.
There was also some indication that those with lower educational achievements and those with middle incomes were more likely to form friendships via the Internet. Unlike some other studies we found no gender effect. Somewhat surprisingly, we found no statistical relationships between propensity to make friends and a wide range of measures of traditional forms of social connectedness and measures of personality attributes. This points to the Internet de-emphasizing the importance of sociability and personality differences.
Numbers of Cyberspace friends: Those who reported making friends via the Internet appear to have established a substantial number of Internet friendships - 70 percent reported having 4 or more Internet friends. This may mean that for those with the social skills and the confidence to make friends via the Internet that the Internet is a good medium to establish and sustain friendships.
We found that earlier cohorts had more Internet friends, perhaps again attesting to the need to learn how to use the Internet to develop friendships. However, we found no statistical relationships between the number of cyberspace friends and experience levels with specific Internet features.
We also found that people who had never married made fewer friends. This intriguing result raises questions about the motivation for friendship creation via the Internet, a subject we did not explore in our survey. Are married people more comfortable making friendships via the Internet? Are those who have never married more interested in friendship creation via face-to-face communication? The answers to these questions and other motivational questions must await further research.
We had insufficient data to explore possible relationships between numbers of cyberspace friends and traditional social connectedness and personality measures.
Cyberspace friendships leading to meetings: A large proportion (three out of five) of those who reported establishing friendships via the Internet met one or more of their Internet friends in person. This suggest that for these people, at least, the Internet is an important media for establishing face-to-face friendships. Those reporting higher numbers of Internet friends were more likely to have met at least one.
We found that those with college degrees but with no postgraduate education met more Internet friends than those with lesser or greater educational achievements. This perhaps suggests that this college-degree group may be more highly motivated to meet Internet friends. Why this should be requires further research.
Self-identified "above average" or "excellent" users were more likely than "novice" or "average" users to meet friends made via the Internet, perhaps emphasizing the importance of Internet skills in friendship formation.
Numbers of Cyberspace friends met: Of those who reported meeting Cyberspace friends, we again found that that they met a substantial number - three out five reported meeting 4 or more friends. This further supports the view that for these people the Internet is an important media for establishing face-to-face friendships. The proportion of Internet friends met to Internet friends known was nearly 65 percent.
Younger people appeared to meet fewer friends. We can speculate (without any direct evidence) that this might be due to financial barriers inhibiting travel or that younger people often find it easier to make friends face-to-face among the local population than do middle-aged people.
Skill at using the Internet seems to be most important determinant of friendship formation on the medium. Those who indicate that they are highly sociable do not appear to be at any advantage; equally, those with low sociability levels do not appear to be at a disadvantage. Similarly those whom one might expect would be more successful at making friends - the extroverted, those willing to take risks, those willing to try new things - do not seem to be at an advantage either. Some demographic groups appear likely to be more successful, but it is difficult to conclusively interpret the results given the limitations of the data.
The Internet, it is often claimed, is not a transparent medium for transferring social cues. This claim has given rise to the wide circulation of cartoon showing one dog, sitting at a computer, telling another dog sitting on the floor, "On the Internet, no one can tell you are a dog." And it is also often claimed that physical-social cues are usually necessary to establish friendships. It may be that those who are skilled in computer communication are also more skilled in transmitting the social cues necessary for friendship formation. Currently many users, especially recent arrivals, find the Internet cumbersome and difficult to use. (Kraut, et al. 1996) This may in turn inhibit friendship formation. Hence unless there is substantial improvements in what the technology is and how it can be used, it may retard friendship formation over that medium. On the other hand, people have been busy. The growing array of social interaction in cyberspace is truly impressive. Casual perusal uncovers an astonishing variety of chat rooms, "members forums" and self-described "communities" (Cf. Waters & Barrus, 1997; Rockwell, 1997). And we hardly need mention the plethora of dating services, including international "lonely hearts" web sites. All this serves as testimony to the indomitable nature of the human spirit, and to the desire for people to be social and have social relationships, even in the relatively desiccated environment of cyberspace.
If our survey results can be generalized, one conclusion seems inescapable: hundreds of thousands of new face-to-face friendships have been established via the Internet. Even though they may be relatively small in percentage terms, when extrapolated to millions of on-line users, it appears that huge numbers of new friendships have been formed. What we could not explore in this survey was the nature and quality of these friendships, neither could we probe their outcomes and implications. These important issues must await another investigation.
Agresti, A. 1990. Categorical Data Analysis, John Wiley & Sons, New York.
Baron, N. S. 1984. Computer-mediated communication as a force in language change. Visible Language, 18 (2): 118-141.
Baudrillard, Jean. 1983. Simulations. Translated by Paul Foss, Paul Patton and Philip Beitchman. New York: Semiotext(e).
Baym, Nancy K. 1995. The emergence of community in computer-mediated communication, Pp. 138-163 in Steven G. Jones, Cybersociety: Computer-mediated communication and community. Sage, Thousand Oaks, CA.
Beninger, James R. 1987. Personalization of mass media and the growth of pseudo-community. Communication Research . 14: 352-71.
Kraut, Robert, Scherlis, W. Mukhopadhyay, T., Manning, J. and Kiesler, S. 1996. The HomeNet field trial of residential internet. Communications of the ACM, 38, 12: 55-63.
Kiesler, Sara, Siegel, H. & McGuire, T.W. 1984. Social psychological aspects of computer-mediated communication. American Psychologist. 39 (10): 1123-1134.
Numes, Mark. 1995. Jean Baudrillard in cyberspace: Internet, virtuality, and postmodernity. Style. 29 (2): 31-27.
Parks, Malcolm R. & Kory Floyd 1996. Making friends in cyberspace. Journal of Communication, 46: 80-97.
Rheingold, Howard, 1993. The virtual community: Homesteading on the electronic frontier. Reading, MA: Addison Wesley.
Rice, Ronald E. 1987. Computer-mediated communication and organizational innovation. Journal of Communication, 37, 65-94.
Rice, Ronald E. and G. Love. 1987. Electronic emotion: Socioemotional content in a computer mediated communication network. Communication Research, 14, 85-108.
Rockwell, R. 1997. An infrastructure for social software. IEEE Spectrum, 34, 3: 26-31.
Turkle, Sherry. 1996. Virtuality and its discontents: Searching for community in cyberspace. The American Prospect. 24, Winter: 50-57.
Stoll, Clifford. 1995. Silicone snake oil. New York: Doubleday.
Parks, Malcolm R. & Kory Floyd 1996. Making friends in cyberspace. Journal of Communication, 46: 80-97.
Water, R. C., and Barrus, J. W. 1997. The rise of shared virtual environments. IEEE Spectrum. 34, 3: 20-25.
Wynn, Eleanor, and James Katz, 1997. Hyperbole over cyberspace. The Information Society, in press.
We analyzed our sample of 601 Internet users from the perspective of when they first started using the Internet - 308 reported starting in 1995, 132 in 1994, 79 in 1993, and 82 started pre-1993. We found that recent cohorts are increasingly female, younger, less well-educated, more likely to be working part-time or a be a student, and less well-off. We also examined the marital status and racial/ethnic mix of Internet users, neither of which appeared to change across Internet cohorts.
Our survey indicates that recent cohorts are increasingly female - 28 percent of the pre-1993 cohort reported being female, 39 percent of 1993 cohort and 34 percent of 1994 joiners were female, while 45 percent of the 1995 cohort reported being female.
|
|
||||||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
|
|
||||||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
|
|
||||
| Percent who started using the Internet (num resp) |
|
|
|
|
| HS
or some college
/technical school only |
|
|
|
|
| College degree highest level achieved |
|
|
|
|
| Some post-grad work up to Ph D |
|
|
|
|
| Num of resp |
|
|
|
|
|
|
||||
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
We also examined the racial/ethnic mix of Internet users. This, too, did not appear to change over Internet cohorts, although for most racial/ethnic groups we did not have enough data to carry out a statistical test. Eight-five percent of Internet users reported being white, 4 percent black, 4 percent Asian, and 3 percent Hispanic.
In this Appendix we report on our use of statistical models to determine the key variables and to quantify the relative importance of each key variable, keeping the other key variables fixed, in the prediction of:
log[p/(1-p)] = k + b1 + b2 + + bn
where p = the probability of making friends via the Internet
(or probability of making 6 or more friends);
k = a constant;
b1 , b2 . . . bn are n categorical independent variables, each with
two or more levels.
Our approach has been to calibrate models with separate sets of variables - demographic variables, general measures of Internet usage, and experience levels with Internet features, and then take the statistically most important variables from these models to calibrate a general model encompassing all three groups of variables.
We have taken a simplified approach to defining the levels of the categorical variables. Where a variable appears to be independent of or linearly related to the probabilities being considered, we dichotomized the variable. Where the relationship could, based on our earlier analysis be more complex, we introduced an appropriate number of levels. For example, we found earlier that those with medium household incomes ($35-74,000) reported being more likely to make friends via the Internet than those with lower and higher household incomes. To accommodate this behavior, we established three levels - up to $34,000, $35-74,000, and $75,000 and above.
We have adopted this approach for two reasons - dichotomized variables are simpler to interpret, and minimizing the number of levels for each independent variable helps achieve a good ratio of number of observations to numbers of populations. Finally, to estimate the model parameters we used the CATMOD procedure in SAS (Software Release 6.07.02).
Propensity to make friends: demographic variables
The results of our
model below indicate that there are two (marginally) significant variables
- highest educational level achieved (transition from up to college to
college degree) and income level (transition from household income $35-74k
to $75k and above). Moreover, the model cannot be rejected on the grounds
of poor goodness of fit.
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
Those with college degrees are less likely than those whose education has not reached the achievement of a college degree to make friends via the Internet. The estimated odds of making a friend are estimated to be 0.66 lower for those with college degrees. Similarly, those with household income above $75,000 are less likely than those with household incomes in the range $35-74,000 to make friends. The estimated odds of making a friend are estimated to be 0.70 lower for those with household incomes above $75,000.
Propensity to make friends: general measures of Internet usage
For this model the
most significant variable is self-identified skill level. Those reporting
"above average" or "excellent" skill levels are more likely to make friends
- the odds of above average/excellent users making a friend are estimated
to be 1.42 times higher than for novice/average users.
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||
Propensity to make friends: experience levels with Internet features
Despite the fact that the experience levels of all the Internet features correlate to some degree with propensity to make friends, the dominant variable is usenet/user group experience, with none of the other variables adding any additional statistical explanatory power. This model again has an acceptable likelihood ratio (p = 0.24).
Users of usenets/user groups are much more likely than non-users to form Internet friendships. The odds of forming Internet friendships are estimated to be 1.48 times higher for users than non-users.
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
Propensity to make friends: general model
For the general model, we used the following independent variables - educational achievement (3 levels), household income (3 levels), self-assessed skill level (2 levels), frequency of use (2 levels), usenet/user groups usage (2 levels), and MUDs/MOOs usage (2 levels). For this model, the key variable again appears to be usenet/user group experience. As before, users of usenets/user groups are much more likely than non-users to form Internet friendships. The odds of forming Internet friendships are estimated to be 1.45 times higher for users than non-users. This model has a marginally acceptable likelihood ratio (p = 0.10).
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|||
|
|
|
|
|
|
||
|
|
|
|
|
|||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
||||||
Number of friends made: demographic variables
In order to run the model we converted the "number of friends" made into a dichotomized variable - small number of friends (up to and including five) and large number of friends (six and more). For the demographic model, we reduced the number of variables considered to marital status, work status, age and educational level, to ensure a reasonable observations to population ratio.
The results of our model below indicate that none of these variables is significant in predicting number of friendships made.
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
|
|
|
|||
|
|
||||||
Number of friends made: general measures of Internet usage
For this model the
most significant variable is the year of starting using the Internet. Those
starting in 1993 and earlier are likely to have more friends - for users
starting in 1993 and before the odds of having made six or more Internet
friends are estimated to be 1.99 times higher than for those starting 1994
and 1995. The likelihood ratio for this model is high but just about acceptable.
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
||
|
|
||||||
Number of friends made: experience levels with Internet features
None of the Internet features measures appeared to relate to number of Internet friends made so we did not calibrate this model.
Number of friends made: general model
There appeared to be
only one significant variable - year of starting - so we did not calibrate
the general model.