Subjects (n=114) were students enrolled in a large upper-level lecture class at a midwestern university. The students voluntarily filled out questionnaires that asked about prior computer experience as well as basic background information prior to their participation in the conference. Of the 114 people who completed the questionnaires (53 males and 61 females), approximately seventy-five subjects participated beyond the initial sign-up. Content analysis was conducted only for those people who participated beyond the initial requirement.
Two parallel conferences, one requiring the participants to use their real names and one having the participants use pseudonymous names, were run. The electronic conference that the subjects participated in was devised as an optional part of a regular course in the social sciences. The conference was introduced as a potential source for information concerning the course as well as a venue for discussions about topics of interest for the course participants. The subjects were provided with computer accounts and were required to sign on at least once and electronically register their names, real or pseudonymous, to the conference. Participation beyond that was entirely voluntary. The subjects were randomly assigned to either the real- name conference or the pseudonymous conference. The existence of the two conferences was not announced to the participants initially. The conference ran for approximately two months.
In each conference, participants could respond to Òitems,Ó or topics of interest that other participants created. Generally, there were more items in the pseudonymous conference than in the real-name conference. Since participants interacted in different conferences, the topics of discussion that participants responded to in one conference did not necessarily appear in the other. Thus, in order to keep the two conferences as similar as possible, conference organizers attempted to introduce similar topics in both conferences. The participant's response always appeared next to the participant's name or pseudonym, making it possible for other participants to recognize the source of each response. The entire transcripts of both conferences were content analyzed for relational characteristics.
Bales's (1950) interaction process analysis (IPA) has been the most commonly utilized measure in research on the interpersonal aspects of CMC (Walther, 1992; e.g., Rice & Love, 1987; Hiltz, Johnson, & Turoff, 1986). This method categorizes each response in the interaction as either socioemotional or task-oriented. However, this task-social dichotomy has been criticized for its inability to account for the multidimensional relational qualities and for its rigid categorization of each response into only one category (McGrath, 1984; Walther, 1992). Attempts at addressing these deficiencies such as Rice & Love's (1987) inclusion of an additional category and Walther & Burgoon's (1992) use of self-report data on multidimensional relational aspects have provided somewhat more robust measures. However, we feel that a more appropriate measure of the multidimensional relational qualities should examine the various aspects of the content of the interaction. Thus, the present study analyzed the entire transcript of two computer-mediated conferences and coded for the presence of the following four relational categories.
Each response was coded for the occurrences of each of the relational categories and all the responses for each subject was summed. The average intercoder reliability for the four categories was .81 (Holsti, 1969). Although this is not an exhaustive list of possible human relational dimensions, it is inclusive of the central multidimensional aspects of relational computer-mediated communication.
For each participant, volume of response was also measured, by counting the total number of responses, as well the total number of sentences within each response.
The total number of occurrences of the different categories for each subject is influenced by the total volume produced by that subject. That is, the higher the number of responses by a subject, the higher the total count of relational category occurrences. In order to avoid confounding differences in discourse patterns with verboseness, it was necessary to convert the raw count of category occurrences into proportions. The total count in each relational category for each subject was thus summed and divided by the total number of sentences for that subject, producing a measure of the percentage of each relational category per sentence. These proportions provide a measure of the relative use of the different relational categories unaffected by total volume.
The survey, which was introduced separately as a voluntary part of the course, was conducted prior to the onset of the computer-mediated conference. No link between the conference and the survey was made. Questions in the survey included items concerning the course in general as well as those concerning people's prior attitude and knowledge of computers. Specifically, questions assessed people's general usage of computers (an 8 item multiple choice question asking how often subjects use a computers), attitude towards various computer usage (four items: computers helps me make better use of my time; computers are difficult to use; computer conferencing is difficult to use; electronic mail is difficult to use), familiarity with four computer-mediated communication terms (FTP , Compuserve /Prodigy /America-On-line, computer conferencing, electronic mail), and perceived ability to explain three advanced computer technology concepts (computer bulletin boards, multimedia, internet).
Initially, three categories were utilized in assessing the gender of each pseudonym: same gender (i.e., the gender of pseudonym was same as the gender of the participant), neutral gender, and cross (different) gender. Then, the neutral gender and different gender categories were collapsed in order to analyze the extent of same gender choice (same gender) versus choice of gender that did not reveal the gender of the participant (gender masking).