First, it was explored whether social comparison orientation moderated any of the expected effects of target under evaluation and evaluation. To do this, hierarchical moderated regression analyses were conducted to test the extent to which orientation toward social comparison interacted with the dummy coded experimental variables. Social comparison orientation did not yield significant main effects, neither did it qualify any of the reported effects in the current study. Thus, individual differences in orientation toward social comparison was not included in further analyses.

Next, the dependent variables were submitted to a two-way (Target under Evaluation X Type of Evaluation) analyses of variance (ANOVAs). Means and standard deviations for all measures are presented in Table 6. First, the ANOVA on social desirability revealed a significant main effect for target under evaluation and a significant main effect for type of evaluation. Participants judged statements about the partner less undesirable (M = 3.60) than statements about the self (M = 3.07), F(1, 162) = 26.78, p < .001. As expected, noncomparative evaluations were seen as less undesirable (M = 3.59) than comparative evaluations (M = 3.05), F(1, 162) = 29.10, p < .001. These main effects were qualified by the predicted Target x Evaluation interaction, F(1, 162) = 5.64, p < .02. As can be seen in Table 6, the interaction arose because participants evaluated comparative statements as less desirable than absolute statements, and more so when the statements focused on superior qualities of the self.

The ANOVA on feelings of sympathy indicated that partner-evaluations evoked more sympathy (M = 3.47) than self-evaluations (M = 2.67), F(1, 162) = 24.32, p < .001. Type of evaluation by the discussant also significantly affected participants' sympathy for the discussant, F(1, 162) = 15.51, p < .001, with participants who had read noncomparative evaluations felt more sympathy for the discussant (M = 3.38) than participants who had read comparative evaluations (M = 2.73). There was a significant interaction between target and evaluation, F(1, 162) = 10.74, p < .001. As can be seen in Table 6, participants expressed the least sympathy for the discussant who provided statements that emphasized the superiority of the self. Duncan's multiple range tests between the remaining three condition revealed no significant differences, indicating that someone who describes the partner in comparative or in noncomparative terms, and someone who engages in noncomparative self-evaluations can expect the same amount of sympathy for his or her evaluations.

As expected, a reverse pattern of results was found for feelings of annoyance. A main effect of target of evaluation was observed, F(1, 162) = 17.47, p < .001, indicating that partner-evaluations evoked less annoyance (M = 1.89) than self-evaluations (M = 2.56). Also, the main effect of type of evaluation was significant, F(1, 162) = 10.34, p < .01, revealing that participants who read noncomparative evaluations felt less annoyed about the statements (M = 1.98) than those who read comparative evaluations (M = 2.51).

Although the Duncan's multiple range tests on the levels of sympathy and annoyance revealed an identical pattern of results (see Table 6), the analysis of variance failed to support the interaction effect between target and evaluation, F(1, 162) = 2.34, p = .13. Notwithstanding the absence of a significant interaction effect on the annoyance ratings, the overall pattern of results suggests that especially self-enhancing evaluations of the self as a partner elicited the least positive feelings for the discussant, whereas the remaining types of evaluations yielded similar levels of sympathy and annoyance.

Further, an ANOVA on the level of presumed relationship satisfaction yielded a significant interaction between target and evaluation, F(1, 60) = 3.82, p < .06. As Table 6 reveals, this interaction largely derives from participants' tendency to rate the relationship of someone evaluating the self (as a partner) in comparative terms as less satisfying than the relationship of a person who provides self-evaluations in a noncomparative manner, whereas the reverse is true when someone evaluates the partner. Thus, the effects of social comparison on the perceived quality of the relationship are believed to be most beneficial when someone perceives his or her intimate partner as better the others, whereas self-enhancing social comparison is to a lesser extent associated relationship satisfaction. No other effects approached statistical significance.

As in Studies 4 and 5, it was examined whether the obtained effects on social desirability were mediated by participants' sympathy for the discussant or by their feelings of annoyance about the discussant's statements, and vice versa. Participants social desirability ratings were highly correlated with their levels of sympathy and annoyance, rs = .67, and -.71, respectively, both ps < .001. Accordingly, four 2 (Target under Evaluation) by 2 (Type of Evaluation) ANCOVAs were conducted to determine whether the effects remained significant when the variance explained by the sympathy and annoyance ratings was partialled out.

The first analysis covaried participants' scores on the sympathy index out of the scores on the social desirability index.¹¹ When the effects of sympathy were statistically removed, the main effects of target under evaluation and type of evaluation still emerged, F(1, 161) = 7.40, p < .01, and F(1, 161) = 13.37, p < .001, respectively. However, the interactive effect was no longer reliable, F < 1, ns. The second ANCOVA covaried participants' scores on the social desirability index out of their scores on the sympathy index (see footnote 5). In contrast with the main effect of evaluation, the main effects of target under evaluation remained, F(1, 161) = 5.39, p < .03, as did the interaction, F(1, 161) = 5.31, p < .03. As was the case in Study 5, these findings also suggest it were feelings of sympathy that mediated the interactive effect on social desirability.

We next examined whether feelings of annoyance did account for the effects on social desirability by conducting an analysis after adding annoyance scores as a covariate (see footnote 5). Reliable effects for target under evaluation and type of evaluation still emerged, F(1, 161) = 9.36, p < .01, and F(1, 161) = 17.49, p < .001, respectively. In this analysis, the interactive effect was marginally significant, F < 3.25, p = .07. In a fourth ANCOVA, participants' scores on the social desirability index were covaried with their scores on the annoyance index (see footnote 5). In contrast to the aforementioned ANOVA on the annoyance scores, no significant effect effects emerged, Fs < 1, ns. Hence, perceptions of social appropriateness and desirability appeared to mediate the effects of target and evaluation on feelings of annoyance.