Back to Vol. 2, No. 4 Table of Contents
[Conversation]
Collaborative U CMC Play E-Commerce Symposium Net Law InfoSpaces


Conduct Control on Usenet

Christine B. Smith

Naval Postgraduate School
Monterey, CA

Margaret L. McLaughlin
Kerry K. Osborne

Annenberg School for Communication
University of Southern California

Table of Contents

Abstract

In this paper we explore the nature of offensive conduct and its treatment on Usenet. Specifically, we examine the frequency, form, and tone of reproaches for misconduct on five newsgroups: rec.arts.tv.soaps; soc.motss; soc.singles; rec.sports.hockey; and comp.sys.ibm.pc.games. Where possible, subsequent accounts offered by offenders are also examined. Results indicate that few individuals respond publicly to their reproachers and that complete "traditional" remedial episodes in Usenet are relatively rare. Discriminant analysis supports a tentative conclusion that different offense types elicit reproaches which vary in form and tone. Furthermore, the tenor and frequency of reproaches for particular offenses vary according to newsgroup, supporting the thesis that norm violations are differentially treated in Usenet "communities." The analyses and discussion include an examination of gender differences in the newsgroups studied.

Introduction

Although initiated primarily for the exchange of research and computing resources, Usenet has become a forum for electronic discussions among those with access to networked computers on topics ranging from hobbies to current events . Usenet promotes the development of new relationships by affording immediate access to thousands of others with similar interests and spheres of expertise (McLaughlin, Osborne, & Smith, 1995). Usenet remains the primary vehicle for exchange of information and resources among computer aficionados, but it also attracts those with more arcane and mundane interests. Netters with diverse sexual tastes tease and titillate (Witmer, 1997). Virtual friends celebrate and console their way through "real-life" events while bemoaning the fates of their favorite soap-opera characters (Baym,1995). Amateur artists distribute and solicit comments on their creations (for example, alt.binaries.pictures.fine-art.graphics; alt.ascii.art). Parents and official agencies try to locate missing children (alt.missing.kids).

Some of the conventions of interactive Usenet participation, i.e., posting and responding to as well as reading newsgroup articles, are dictated by the newsreading software. More sophisticated software enables users to organize newsgroup articles ("posts") by size, sender, date, or "threads," specific discussion topics which are indexed by the word or phrase appearing in the subject heading of the article. Follow-up postings to earlier articles, generated with a "reply" function or its equivalent while reading a particular article, appear with the same header preceded by "Re:" and may be cross-posted to several different newsgroups. Readers wishing to follow a specific thread of discussion may easily do so by scanning the indexed headers and retrieving the texts of only those articles they wish to see.

Other Usenet conventions, such as encrypting sensitive or offensive information or otherwise flagging material with a subject pointer (an acronym, word, or phrase which conveys the nature or content of the post), are newsgroup specific and have evolved as communication rules for posters (Baym, 1995; McLaughlin et al., 1995). Such conventions are usually propagated regularly for the benefit of new readers in informative postings known as FAQs, a well-known and ubiquitous acronym for "frequently asked questions" (Moraes, 1994b). FAQs are usually generated by volunteers, newsgroup regulars who tire of seeing the "same old questions" and as a result begin to post Socratic dialogues of recurring questions and answers. So widespread is the FAQ phenomenon that readers of more esoteric newsgroups have created tongue-in-cheek FAQs which are more on the order of "inside jokes" than they are informative. Nonetheless, they are still posted more or less regularly (See, for example, FAQs for alt.alien.vampire.flonk.flonk.flonk; alt.buddha.short.fat.guy; alt.angst; alt.devilbunnies; and alt.peeves).

In addition to newsgroup-specific norms for communication, overarching Usenet standards, collectively known as "netiquette," guide proper interaction (Von Rospach, 1993; Shea, 1994; Spafford, 1993). For example, one is expected to quote "just enough" of a preceding post to enable readers to follow one's commentary. However, follow-on posters frequently include the entire thread, consisting of not only the original post, but subsequent comments made by other posters as well. The enormous bandwidth-wasting article that results may signify defiance or ignorance of netiquette, but in many cases it merely signifies technological deficiency in using the newsreader editor.

Netiquette and other standards of communicative practices have developed so that newsgroup participants can digest an immense quantity of information as quickly, efficiently, and economically as possible. For example, a feature of most newsreading software is the ability to ignore unwanted posts, either those from designated authors or with designated subject headings, consigning them to a "kill" file (Moraes, 1994a). This reduces substantially the number of articles one is exposed to in a given newsgroup, an important feature for time- and/or money-conscious readers who pay for their online time directly or indirectly with associated costs for transport and storage of Usenet messages.

Technological proficiency, demonstrated knowledge of the FAQ, and conformity to newsgroup practices are among the conditions of acceptance in many well-established newsgroups. Failure to observe net and newsgroup standards is quite common, however, and may range from "relatively innocuous errors in the use of newsreading software to actions characterized as 'net terrorism'" (McLaughlin et al., 1995, p.95). Whether such behavior is caused by error or choice, it often provokes comment from other users. Reprimands are rampant, ranging from private (e-mailed) admonitions to public censure across several newsgroups to crusading efforts to divest offenders of their net access.

While we can effectively gauge the nature and extent of corrective episodes conducted publicly in newsgroup discussions, it is important to note that an indeterminate number of such episodes may occur privately, either wholly or in part, via e-mail. For example, when a poster is publicly excoriated for offensive behavior, he or she may choose to respond to detractors privately and individually rather than encourage further communal discussion with a public response. Presumably, those who take offense at the behavior of others may exercise the same restraint and voice their concerns directly to offenders in e-mail messages.

Nevertheless, conduct-correcting episodes which occur in the public arena are numerous enough to warrant scrutiny. In the study of recurring discursive practices in Usenet, we gain a better understanding of emergent standards of behavior and communal responses to violations of those standards.

Tsk, Tsk! Taking Transgressors to Task on Usenet

McLaughlin et al. (1995) analyzed postings from five popular newsgroups and, using the concept of a reproach as defined in the literature on accounts and explanations (Cody & McLaughlin, 1990a; Cody & Braaten, 1992; McLaughlin, Cody, & Read, 1992; Schonbach, 1990), generated a taxonomy of reproachable conduct on Usenet. Offending posts, those which presented behavior sufficiently in violation of normative expectations to prompt comment and spark remedial discussion and debate, were analyzed for type of offense(s). Seven preliminary categories resulted:

Incorrect/novice use of technology : e.g., editing and formatting errors, multiple postings or signatures, failing to use follow-on option

Bandwidth Piggery : e.g., excessively long article or signature, quoted material longer than comment, indiscriminate cross-posting, asking a frequently answered question

Violation of Usenet Conventions : e.g., incorrect or missing subject headers, failing to encrypt offensive material, posting to an inappropriate newsgroup or otherwise demonstrating lack of regular reading

Violation of Newsgroup Conventions : e.g., failing to use spoiler warnings, lack of familiarity with and failure to use appropriate subject headers or abbreviations, failing to conform to group spirit or style and group traditions regarding appropriate topics

Ethical Violations : e.g., posting private email or personal information about others without permission, misattributions or misquoting of sources, harassment of individual posters

Inappropriate Language : e.g., flaming (personal attacks, ridicule), hostile or coarse language, linguistic affectations which distract or detract from message content

Factual Errors : e.g., spelling and grammatical errors, mistakes with respect to names, dates, places, and events, errors in summarizing others' posts

(adapted from McLaughlin et al., 1995)

Some offenses are likely seen as more egregious than others, and it is equally possible that what one newsgroup condemns, another condones. For example, while flaming (directing a particularly vindictive or hostile post at another user) is perfectly acceptable on some newsgroups (e.g., alt.fan.warlords, alt.flame, and alt.irc), vicious verbal attacks invite censure on many socially-oriented newsgroups. In this chapter, we explore further the nature of offensive conduct and its treatment on Usenet. Specifically, we examine the frequency, form, and tone of reproaches and, when available, offenders' subsequent accounts.

The typical remedial episode consists of four components: the failure event, a reproach, an account, and an evaluation of that account (i.e., honoring it or rejecting it). Reproaches may be explicit or implicit, depending on the situational dynamics or communication style of the reproacher. Reproaches may range in harshness or severity from a simple "What happened?" to a tirade laced with profanity and attacks on the offender's self-esteem (Schonbach, 1990). Two of the more common forms of accounts offered are excuses and justifications (Scott & Lyman, 1968). Other forms include apologies, concessions, and denials (Schlenker, 1980; Schonbach, 1990). Apologies and excuses are generally perceived as more mitigating, i.e., more likely to result in an account being honored, whereas justifications and denials are usually deemed more aggravating and could escalate conflict (Cody & McLaughlin, 1990b). Concessions, depending on the extent of apology included and/or restitution offered, may be mitigating or aggravating (Schonbach, 1990).

It is reasonable to wonder if severe reproach forms result in similarly aggravated accounts or, as may often be the case with new or infrequent posters, a tendency to eschew response. Little research has been done on the relationship between forms of reproach and subsequent accounts; however, a review of the literature on account episodes in face-to-face settings (Cody & Braaten, 1992) suggests at least one possible relationship in this medium: in the case of particularly aggravating reproaches the form and tone of a reproach influence the form and tone of a corresponding account. For example, a reproach which is rudely phrased and targeted at the offender's self-esteem may be more likely to result in a defensive rather than apologetic account. Also, humorous or sarcastic reproaches may elicit similar responses in an unusually free-wheeling environment which prizes wit more than it does behavioral conformity or social niceties (Coates, 1993). And, as suggested above, the form and tone of reproaches may be a factor in whether an offender responds publicly by offering an account to the newsgroup at large.

Somewhat more is known about the influence of gender on remedial episodes. Women, for example, tend to offer fuller apologies than do men and engage in more relational repair work with accounts (Gonzales, Pederson, Manning, & Wetter, 1990; Schonbach, 1990). Pasting this hypothesis to the medium of Usenet is problematic, however. Not only do men have greater access to technology (Balka, 1993; Truong, 1993), their dominant posting behaviors, even on newsgroups specifically oriented to women's discussion (Kramarae & Taylor, 1992), may affect both the level of women's participation and women's posting behaviors. Gender study in electronic media is therefore a task of increasing complexity, but because researchers have begun to take note of an inequity of female participation on the internet we make a tentative survey of gender differences as they pertain to conduct-correcting episodes on Usenet.

Method

As a preliminary exploration of reproach characteristics and account sequences in Usenet, we return to our earlier data, a 3.09 MB corpus of conduct-correcting episodes which constitutes roughly 15 percent of all postings to five popular newsgroups -- comp.sys.ibm.pc.games, rec.sport.hockey, soc.motss (members of the same sex), soc.singles, and rec.arts.tv.soaps -- over a three week period (see McLaughlin et al., 1995, for a detailed description of sampling procedures). Ten "offenders" were randomly selected from each of the five newsgroups. All posted messages which followed-up an offender's article with corrective comments were saved for analysis. Each corrective message was considered a "reproach episode" for the purposes of defining cases. The original offending article was also analyzed in all but one episode, and in this case, a sufficient amount of the offending post was quoted by reproachers to warrant including the case. If the offender posted a response to a reproach during the data collection period, that message was retained as part of the episode data. Finally, as McLaughlin et al.'s data suggest, reproachers themselves are occasionally taken to task for bullying newcomers, flaming, or nit-picking. If a follow-on poster (excepting the offender) reproached another reproacher, a new case was defined and coded as a reproach episode. Gender of reproachers and offenders was coded where information was available and reasonably unambiguous. The number of messages posted by each individual during the data collection period was also recorded. A detailed account of the coding procedures is available in Smith, McLaughlin, and Osborne, 1997.

Reproaches

Reproaches were first analyzed for nature of offense(s) using McLaughlin et al.'s taxonomy of reproachable conduct on Usenet. As posters are frequently rebuked for more than one offense, a given case may present one or more of the following categories: technological ineptitude, bandwidth waste, violation of Usenet norms, violation of newsgroup norms, ethical transgressions, language faults, and factual errors. Reproach messages were then rated on the magnitude of each of eight characteristics (see below), which were then reformulated into the following four bi-dimensional properties:

AFFECT:
  • friendly: the reproacher treats the offender in a cordial or congenial manner
  • hostile: obvious belligerence toward the poster; may or may not entail the use of profanity

INTENT:
  • helpful: the reproacher offers correct information, aid, or politely informs offender of the applicable "netiquette" rule with the obvious aim of preventing repeat offenses
  • sarcastic: implicit or explicit derision or ridicule; may be condescending and the reproach implicit; intent is to demean

STYLE:
  • witty: an obvious attempt to be witty or funny without sarcasm; reproach may be implicit
  • factual: a simple statement of offending behavior

REPROACH ORIENTATION:
  • directed at behavior: reproach directed at behavior rather than the person
  • directed at person: reproach directed at person more than behavior
Accounts were rated on identical attribute scales. Additionally, accounts were rated (1=not at all characteristic; 5=extremely characteristic) on the following properties offered by Cody and McLaughlin and their colleagues (See, for example, Braaten, Cody, & Bell, 1990; Cody & McLaughlin, 1988; McLaughlin, Cody, & O'Hair, 1983): apologetic, conciliatory, excusatory, justificatory, and denying.

Results

Only a fraction of offenders responded publicly to their reproachers during the data collection period (n=13, 25%), and of these, only four accounts were publicly evaluated in newsgroup discussion. Given the small number of complete "traditional" corrective sequences (i.e., failure event, reproach, account, evaluation (Cody & Braaten, 1992), data analysis was largely confined to reproach characteristics and descriptive statistics for the overall sample and each of the five newsgroups.

The overall sample yielded 83 reproach episodes in which 52 posters were chastised by 70 other posters for a total of 128 offenses (poster N=116; 6 reproachers were in turn reproached for their offensive reprimands, thus increasing the number of offenders). One person rebuked another for four different infractions, and several people were reproached for two or three offenses. All of McLaughlin et al.'s (1995) reproachable conduct categories were represented; however, not all categories were present in every newsgroup.

Over 75 percent of poster transgressions fell into three of the seven offense types: violation of newsgroup norms, which includes failure to demonstrate knowledge of the FAQ or undermining the communal spirit of the newsgroup (N=40; 31%); inappropriate language, which is associated with flaming (N=36; 28%); and errors, which may involve spelling and grammar as well as facts (N=24; 19%). Participants in the newsgroup rec.sport.hockey were the most active reproachers, accounting for roughly 40 percent of both total offenses and total reproach episodes. Failure to conform to newsgroup norms and flaming were the source of most of the reproach episodes.

Discriminant Analysis

A discriminant function analysis was conducted in order to determine if reproaches differed in tenor among the three most common offenses. Because several reproach episodes involved recriminations against offenders for discursive improprieties along with violation of group norms, a fourth classification reflecting these dual offense incidents (language + groupnorms) was included in the analysis. To test the hypothesis that different offense types elicit reproaches significantly different in tone and form, a stepwise discriminant method was used in which the dependent measures were affect, intent, style, and reproach orientation. Nine of the 83 reproach episodes did not involve any of the three major offense types and were consequently dropped from the analysis. Complete details of all of the statistical analysis are available here.

Posters who were reproached for language and errors were likely to differ with respect to the style, intent, and orientation of the reproaches they received (e.g., posters' errors elicited factual, helpful, behavior-oriented reproaches vs. the more sarcastic and person-oriented reproaches directed against "flaming"). Reproaches directed to language violations differed from those targeted at violations of group norms; here, affect, intent, and reproach orientation were all likely to differ. Comparison of posters who were reproached for either group norm infractions or language improprieties to those who were reproached simultaneously for both differed in orientation and intent (e.g., somewhat more obviously directed at the person as opposed to the behavior, and more obviously sarcastic as opposed to helpful).

Posting Patterns of Reproachers and Offenders

Reproachers were somewhat more prolific than offenders, averaging 22.5 (sd 30.2) messages to offenders' 17.9 (sd 24.7), but this difference was not significant: t(118)= -.91 (p=.18). One-time posters accounted for 25 percent of offenders, while only 7 percent of reproachers posted once during the data collection period. Furthermore, those who responded publicly to their reproachers appeared to have a much higher rate of posting than those who failed to respond publicly during the data collection period (M=27.15, sd 36.6 and M=14.87, sd 18.9, respectively). However, given the lower power associated with so few responses and the broad range of posting frequency by offenders, it is not surprising that the difference failed to reach significance, F(1,50)=2.47 (p=.12). With the exception of rec.sport.hockey, reproachers posted more often than offenders in each newsgroup. The newsgroup soc.singles was by far the most prolific in our sample; both reproachers and offenders had the highest mean number of messages (67.8, sd 55.56 and 35.6, sd 45.58, respectively). In sum, newsgroups differed significantly in frequency of posting by both offenders, F (4,47)=2.76 (p=.038) and reproachers, F(4,65)=8.05 (p=.000).

Reproach Characteristics: Variations among Newsgroups

Reproach message characteristics were examined according to their effects on offenders' responses (a very small sample, N=13) and for characteristic trends and differences across groups. Newsgroups differed significantly on affect, intent, and reproach orientation, but not on style of reproach. Results indicated pc gamers were friendliest, while motss and hockey vied for the title of "most hostile." Hostile reproaches were somewhat more likely to prompt public responses from offenders, but not significantly so. Reproaches aimed at those who publicly responded did not differ significantly in affect orientation from those who failed to respond.

Although groups differed similarly with intent to help or demean offenders--pc gamers were considerably more helpful toward offenders while singles more frequently injected their reproaches with sarcasm--degree of helpfulness versus sarcasm was not a significant factor in whether an offender offered a public account.

Newsgroups did not differ significantly on style; witty and factual reproaches were equally prevalent across groups, although pc gamers tended to use a more factual approach. Witty reproaches prompted fewer offender rejoinders; however, this tendency was not statistically significant. Not surprisingly, perhaps, posters from the three newsgroups which may be construed as more social-- soc.motss, soc.singles, and rec.arts.tv.soaps--displayed somewhat more humor and wit. When humor/ wit is rated as a unidimensional attribute, significant differences emerge among newsgroups in terms of reproach humor and/or wittiness F(4,78)=2.56 (p=.04).

Reproach orientation focused on the extent to which reprimands targeted offenders' behavior or constituted personal attacks on the offenders themselves. Hockey enthusiasts resorted to more ad hominem reproaches than did other newsgroups, while pc gamers were decidedly behavior-oriented. The newsgroups in our sample differed significantly on the tendency to condemn the person F(4,78)=3.35 (p=.0138). Furthermore, of all the reproach characteristics rated in our sample, person-directedness, as a unidimensional attribute of reproach messages, was the only property which had a significant effect on offender responses, F(2,80)=5.16 (p=.007). Reproaches that prompted response were more person-directed than the reproaches that went unanswered.

For example, the following post violated soc.singles norms regarding personal ads:

>Hi, I'm a 23 year old graduate student and would like to
>communicate with any females on this news net.
>---------------(Posted for a non-net friend)----------------

A witty, but person-directed reprimand resulted:

Well, Howdy! Finally, a request for female that doesn't
specify species--you wouldn't believe how many people
on this net want a woman, which of course means a person.
*giggle*

My name is Susa, and I'm a five-year-old Lemur in the
Philly Zoo. My measurements are 12-12-12, which is
considered quite sexy for a lemur *giggle* we all fail the
pencil test *giggle*
My hobbies include running around, climbing trees, and
picking lice; I hope you have a nice thick head of hair!

I only write to stupid people who post personals on
soc.singles; the other ones are too smart for me--we
lemurs may be very_cuddly *giggle* but we tend to be
on the low end of the smarts scale. I know that with that
post, you'll be really_dumb for a human, and perfect for
me! *giggle*

P. Smith

To which the hapless "friend of a non-net person" responded:

In reference to my posting a few hours ago...I have just
discovered that this is the wrong news group! Thanks to
so many people, among others, so if
you'll all quit sending me more messages, I move on.
OK? But those who seem to have nothing better to do
feel free to do whatever you want!

Several people wrote in support of offenders. Four offenders were defended once; two offenders were supported by two or more people. Six reproachers were in turn reproached by other posters, usually for flaming or nit- picking. In one episode, a poster requested strategic information about a popular computer game and was accused (rather subtlely) of software piracy:

>It should be in the book that comes with the game!
(Unless you have a pirated copy.)

In defense of the original offender, another poster recalled that the manual was missing from his legitimately purchased copy of the same game:

[snip...] So, , next time don't be too quick to shout
PIRATE! There may be a perfectly reasonable
explanation.

The hypothesis suggested by Cody and Braaten (1992) that form and tone of reproach influence form and tone of corresponding accounts was in part supported. Specifically, as suggested by Coates (1993), humorous or witty reproaches were strongly correlated with humorous or witty accounts (r =.86; p < .001). Form of account, i.e., apology, concession, excuse, justification, and denial, was not significantly correlated with reproach characteristics.

Gender Democracy In a Male Domain?

In keeping with popular "net" wisdom that males greatly outnumber females in this medium, the posters in our sample were predominantly male (78% of posters whose gender was clearly established; N=109). Furthermore, seventy percent of the reproaches were targeted at males, 32.7 percent at females; and 4.1 percent were targeted at individuals whose gender was indiscernible from information available. Males accounted for two thirds of the 52 individuals reproached (N=33), females 31 percent (N=16). The remaining 6 percent were those of unknown gender (N=3). Females comprised an even smaller proportion of those issuing reprimands (13%); however, female reproachers were more prolific than female offenders (mean number of messages=38.22 and 16.69). In fact, female reproachers outposted all others, but not significantly so. Same-sex reproaches were most common (71%); females reproached females 7 times while males reproached each other 52 times. Cross-gender reproaches occurred 20 percent of the time; men reproached women more frequently than women reproached men. A breakdown of the distribution of posters by gender across newsgroups is available in Smith, McLaughlin, and Osborne (1997). One might expect women to be more temperate in the tone of their reproaches. Such was not the case. Female reproachers were no more friendly and helpful, or less hostile, than were males. Neither was wit the province of any particular gender. When rated unidimensionally, however, sarcasm F(2,67)=3.42 (p=.038) and person- directedness F(2,67)=5.32 (p=.007) were significantly more characteristic of reproaches from males and those of unknown gender.

Recall that only 13 people offered accounts for their conduct. Of these, ten were male and three were female. The between group differences for type of account--apology, conciliation, excuse, justification, and denial--were not significant. Although women were less apologetic than men, they tended to offer more mitigating accounts (e.g., they conceded error, provided excuses, or attempted to justify their behavior while men more often denied wrongdoing). Men and women did not differ significantly on message characteristics of their accounts; however, male offenders tended to be somewhat more sarcastic F (1,11) = 4.09 (p =.068), while women exhibited more humor and wit.

Discussion

One of the unique aspects of remedial episodes in an asynchronous communication context like Usenet is the ease with which offenders can "duck out" of accounting for their behavior. Usenet participants frequently fail to respond to reproaches, at least publicly. Given the lack of reliable lurking statistics and the ease with which offenders can ignore their reproachers it is difficult to assess why a particular reproach goes unanswered. Our data suggest that offenders targeted by reproachers are often such infrequent posters that the likelihood of their responding to reproaches is greatly reduced, and in fact, they might not see the reproach at all. Many communicators in this medium have a propensity to display somewhat less inhibition than communicators in face-to-face interactions (McCormick & McCormick, 1992; Turner, 1990; Wilson, 1993), and the resulting lack of conversational niceties may shock or dismay the net neophyte.

Alternatively, as we discussed earlier in this chapter and as is evident in the offender's response to the "Susa the lemur" post, many of the discursive episodes originating in Usenet are continued via more private channels, such as e- mail. Still, those regular posters wishing to maintain a "presence" within the newsgroup community (MacKinnon, 1995; Rheingold, 1993) may be more likely to respond to reproaches with some accounting for their putative offenses. A positive relationship is indicated between active participation (as determined by frequency of posts) and the likelihood of publicly responding to a reproach. One distinct possibility meriting further study is potential interactive effects between degree of participation and reproach characteristics. Also, it may be the case that "net-newbies" and frequent posters are distinct populations with regard to factors influencing likelihood of response.

The relationship between reproach characteristics and the nature of corresponding accounts may be one of the more promising areas of research in this medium. We saw a distinct tendency for humorous and/or witty reproaches to inhibit likelihood of response on the one hand, but on the other hand, when an account was offered, it was likely to match the reproach in humor and wit. A methodological concern is ensuring an adequate number of complete account sequences for comparative analysis. Our data collection period was three weeks, and this yielded a huge data corpus, but as our sample is a relatively small proportion of the total data corpus, our results should be viewed as preliminary and interpreted with caution.

The discriminant analysis supported a tentative conclusion that different offense types tend to elicit reproaches which vary in form and tone. Also, there is sufficient evidence to suggest that reproaches for multiple offenses, particularly those associated with each other like language improprieties and group norm violations, reflect even less tolerance of offenders' behavior than do reproaches aimed at a single offense. However, these observations are tentative precisely because flaming and other language improprieties were negatively associated with normative behavior in most of the newsgroups we studied. Future studies should strive to examine newsgroups from the standpoint of their similarities and differences in overall descriptive variables such as mission (e.g., social vs. informative), style (e.g., freewheeling or straightforward), and spirit (e.g., degree of communality; tolerance). Further study is recommended in the area of functionally describing and distinguishing newsgroups.

The tenor and frequency of reproaches varied according to type of offense and individual newsgroup, supporting the thesis that norm violations are differentially treated in Usenet. Newsgroup participants appear to be less tolerant of those who violate group standards and practices, particularly when it comes to potentially undermining the communal "spirit" of the group (McLaughlin et al., 1995), and, with the exception of errors, somewhat more forgiving of other offenses. The high incidence of reproaches stemming from poster errors perhaps underscores the original mission of Usenet. Despite the fact that much of the traffic on Usenet is demonstrably social in spirit, Usenet is still a medium for exchange of information, and even when expressing an opinion, one needs to adhere to Grice's (1975) "quality" maxim of truthfulness.

Flaming or similarly antisocial behavior was not well tolerated in any of the newsgroups we studied, but this, too, should be interpreted with caution. Incidences of flaming were not equal across the five groups. One possible approach for future analyses would be to analyze the same number of flaming incidents across newsgroups for quantitative and qualitative differences in corresponding remedial episodes.

The gender analysis in this study strengthens a growing concern regarding unequal representation of women on the net. While female participation in Usenet is probably increasing along with the overall growth in usage, our study does not indicate female participation in Usenet is catching up to that of males. It has been noted that women have less access to computing technology (Balka, 1993; Truong, 1993). Kramarae and Taylor (1992) have voiced a concern that men's dominating posting behaviors, such as sending more messages, introducing more new topics, and disagreeing with others more frequently, may inhibit women's participation. Our findings did nothing to dispel Kramarae and Taylor's concern; however, some encouraging trends may be seen in our sample. For example, we found that female reproachers were slightly more prolific than any other group of posters, perhaps indicating a somewhat stronger voice among women who participate frequently. Despite such emerging trends, the biological divide in virtual spaces still deserves ample study from multidisciplinary perspectives.

References

About the Authors

Christine Biship Smith is a PhD Candidate at the Annenberg School for Communication, University of Southern California, and is a research associate with the Department of Systems Management at the Naval Postgraduate School, Monterey, CA. She is currently on leave of absence and residing in Virginia while she completes her dissertation on the growth and scope of the web presence provider industry.
Address: 8600 Ordinary Way, Annandale, VA 22003.

Margaret L. McLaughlin is Professor of Communication, Annenberg School for Communication, University of Southern California, Los Angeles, and a faculty member of the Integrated Media Systems Center. She is Co-Editor of the Journal of Computer-Mediated Communication, and Co-Editor of the forthcoming Network and Netplay: Virtual Groups on the Internet (AAAI/MIT Press). She is Principal Investigator for the USC Interactive Art Museum project
Address: Annenberg School for Communication, University of Southern California, Los Angeles, CA 90089, USA.