Results

Hypothesis 1, predicting differences in participation for people depending on their familiarity with computer-mediated communication concepts, was analyzed through the t-test, examining the difference of means between participants and non-participants. Of the people who successfully completed the survey, 41 did not participate in the computer-mediated conference beyond the initial one-time sign-up requirement (non-participants) while 73 signed on more than once (participants). As noted previously, there were several items in the survey that assessed their prior knowledge of computers in general, attitude towards computers, as well as more specific computer-mediated communication terminology. (T-tests between the participants and the non-participants were conducted for each of these items with mixed results.) The t-tests for the difference of means between these two groups for general computer use and the perceived ability to explain the advanced computer terminology were not significant (for general computer usage, participant M = 6.31, n = 74, non-participant M = 6.0, n = 41, t = -1.19, n.s.; for ability to explain terminology, participant M = 7.4, n = 74, non-participant M = 8.1, n = 41, t = -1.02, n.s.). However, significant differences between participants and non-participants were found for attitude towards computer use (participant M = 21.8, n = 73, non-participant M = 20.0, n = 41, t = -1.94, p = .056) and for familiarity with computer-mediated communication terminology (participant M = 6.7, n = 69, non-participant M = 5.6, n = 38, t = -3.17, p < .01). Thus, it seems that while specific familiarity and attitude towards computer-mediated communication affected participation, reported computer use and familiarity with general computer terminology was not related with participation in the conference beyond the initial requirement. Hypothesis 1, then, is supported only for domain-specific prior knowledge.

Hypothesis 2 predicted differences in the volume of participation for the pseudonymous and real-name conferences. The measure of participation for each type of conference (pseudonymous and real-name) consisted of two types: the total number of responses (number of times the participants responded to items) divided by the number of participants and the total number of sentences (number of sentences of all the responses for all the participants) divided by the number of participants. Participation was measured by the average number of responses and the average number of sentences per participant. T-tests for the difference of means between the pseudonymous and the real-name conferences were conducted for these two measures. Overall, volume in the pseudonymous conference was higher than in the real-name conference (for number of responses, pseudonymous M = 6.9, n = 37, real-name M = 4.2, n = 37; for number of sentences, pseudonymous M = 27.0, n = 36, real-name M = 18.4, n = 37), but only the total number of responses measure was significant, and only at the p < .10 level (for number of responses, t = -1.88, df = 62, p = .065; for number of sentences, t = -1.30, df = 66, n.s.). Thus partial support for hypothesis 2 was obtained.

As noted previously, choice of gender was coded into two categories, same gender and choice of gender that did not reveal the gender of the participant (gender masking). In testing hypothesis 3, cross-tabulations of gender by gender choice were conducted. As can be seen in Table 1, we obtain patterns that are consistent with our expectations. Males have a higher tendency to choose pseudonyms that retain the same gender while females have a higher tendency to choose pseudonyms that mask their gender. Pearson chi-square tests are highly significant (c2 = 14.15, df = 1, p < .001). Thus, hypothesis 3 is strongly supported.

In analyzing hypotheses 4 through 7, the four relational categories (reference to others, self-reference, supporting references, and emotional text) were used as various indicators of social interdependence. Thus, four separate corollary tests, utilizing each of the indicators as dependent measures, were conducted for each of the hypotheses 4 through 7.

Women were hypothesized to have a greater tendency to exhibit social interdependence (hypothesis 4). We performed a series of t-tests for the difference in means of each gender for the four relational categories. Table 2 summarizes the mean scores of the testing variables for each gender. Significant differences between gender were observed for references to other responses ( t = -2.15, df = 69, p < .05), self-references ( t = -3.26, df = 70, p < .005), and supporting references ( t = -2.39, df = 71, p < .05), but not for emotional text ( t = -.37, df = 69, n.s.). As noted previously, these measures reflect the percent of their occurrence per sentence. Examination of the means show that the relationships are in the predicted direction, i.e., that there is a higher percentage of each type of response for the females than for males. Thus, we find support for hypothesis 4 in three of the four corollary tests.

Hypotheses 5 and 6 predicted different relational patterns for the two types of conferences (pseudonymous and real-name) for each gender. Thus, analyses of hypotheses 5 and 6 consisted of separating the file according to gender and performing a series of t-tests, comparing the means of the two types of conferences for the four relational content measures, for each gender. Tables 3 and 4 display the means and the significance levels of the t-test for males and females respectively. Overall, significant results are obtained for males in two of the relational content measures, self-reference and supporting response. Thus, for males, we obtain significant results for half of the corollaries, providing partial support for hypothesis 5. However, for females (hypothesis 6), none of the corollaries are supported, as we fail to obtain significant difference in means of the relational content measures between the two conferences. Thus, our results show that for males, being pseudonymous in a computer-mediated communication context led to changes in relational interaction in some instances. However, we find no evidence of a change in relational behavior for females between the pseudonymous and the real-name context.

Four separate 2 (gender) x 2 (conference type) analysis of variance (ANOVA) tests using each of the four content analysis variables (reference to other responses, self-reference, supporting reference, and emotional text) as dependent variables were conducted to assess the presence of interaction effects predicted by hypotheses 7. Tables 5a and 5d in the Appendix contain the complete ANOVA results. As can be seen in the four tables, only one measure, self-reference, yielded mildly significant (p < .1) interaction effects. Examination of the means (obtained from Tables 3 and 4) shows that the relationship is in the predicted direction, i.e., more self-reference in the pseudonymous conference than in the real-name conference for males and lower self-reference in the pseudonymous conference than in the real-name conference for females. Thus, we obtain partial support for hypothesis 7, as only one of its corollaries provide us with significant results.

In sum, some support is obtained for all but one of our hypotheses, hypothesis 6. We found evidence of increased participation for subjects who reported greater familiarity with, and more positive attitudes towards, CMC. Subjects in the pseudonymous conference in general participated more than people in the real-name conference. Another interesting finding is that males did not mask their gender while females did, as evidenced by the majority of men choosing pseudonyms that are male, as compared to the huge majority of the women who chose either male or neutral pseudonyms. Also, increasing patterns of social interdependence are found for women compared to men, and men in the pseudonymous conference than in the real-name conference. However, not much difference is observed between women in the two conferences.