The ratings of social desirability, sympathy, identification, and the two forms of attributions were submitted to two-way (Target under Evaluation X Direction of Comparison) analyses of variance (ANOVAs). Means and standard deviations in all conditions are displayed in Table 5.

The ANOVA on the social desirability ratings yielded a significant main effect of direction of comparison such that public evaluations in a enhancing manner were seen as less undesirable (M = 3.00) than deprecating evaluation (M = 2.11), F(1, 60) = 40.21, p < .001. Furthermore, the ANOVA revealed a marginal tendency to consider statements about the partner as less undesirable (M = 2.68) than statements about the self (M = 2.43), F(1, 60) = 3.04, p < .09. However, as in Study 4, these main effects were qualified again by the predicted target under evaluation x direction of comparison interaction, F(1, 60) = 9.93, p < .01. As is evident in Table 5, the interaction arose because participants evaluated enhancing statements as less undesirable than deprecating statements, but particularly so when the target of evaluation was the partner.

Further, ANOVA on the sympathy ratings yielded neither a significant main effect of direction of comparison, nor a main effect of target under evaluation did not yield significant effects on sympathy, F < 1.53, ns. However, a significant interaction effect was found, F(1, 60) = 19.49, p < .001. Inspection of Table 5 reveals that self-enhancing statements evoked less sympathy than partner-enhancement, whereas the reverse was true for self- and partner-deprecation.

Subsequently, ANOVA on the identification levels revealed a marginally significant main effect of direction of comparison, F(1, 60) = 3.87, p < .06, indicating that participants identified more with an enhancing than with a deprecating discussant (Ms = 2.48 vs. 1.77). Target under evaluation did not yield significant effects on identification, F < 1, ns. However, a marginally significant interactive effect was found on identification, F(1, 60) = 3.87, p = .05. As can be seen from the data in Table 5, the identification levels showed a similar, though somewhat weaker, pattern of results as was observed with participants sympathy ratings. That is, participants identified less with a self-enhancing discussant than with a partner-enhancing discussant, whereas the reverse was true for self- and partner-deprecating discussants.

The ANOVA's performed on both attribution ratings yielded again a significant main effect of direction of comparison on self-enhancement attributions, F(1, 60) = 34.37, p < .001, and a marginally significant effect on the compliance attribution, F(1, 60) = 2.96, p < .10. Condition means showed that relative to the deprecating evaluation condition, enhancing evaluations were attributed more strongly to self-enhancement (Ms = 2.33 vs. 3.23) and to compliance (Ms = 2.31 vs. 2.60). Finally, the predicted interaction of target under evaluation and direction of comparison was marginally significant for the self-enhancement attributions, F(1, 60) = 3.82, p < .06. As can be seen from Table 5, self-enhancement was more strongly attributed to a motive to enhance than partner-enhancement, whereas the reverse was true in the deprecation conditions. No other main effects or interaction effects approached statistical significance.

An ANOVA of the number of qualities revealed a main effect of direction of comparison, F(1, 25) = 11.04, p < .01, whereas no significant effect was found of target under evaluation, F < 1, ns. Thus, regardless of whether the participants evaluated themself or their intimate partner, those in the enhancing conditions generated a greater number of qualities (M = 7.93) than those in the deprecating conditions (M = 4.33). This main effect was modified by a significant interaction between direction of comparison and target under evaluation, F(1, 25) = 6.50, p < .02. As Table 5 shows, Duncan's multiple range tests indicated that participants generated more extensive descriptions when they described their partner in an enhancing manner than when they provided self-enhancing descriptions. In contrast with the expectation, there were no significant differences between the number of qualities in the self- and partner-deprecating conditions, and neither did the number of qualities in the self-enhancing condition differ from both deprecating conditions.

As noted earlier, participants's ratings of social desirability could have been influenced by their levels of sympathy or identification, and vice versa. That is, if the effects of target under evaluation and type of evaluation on the social desirability ratings are mediated by participants' feelings of sympathy for the discussant or identification with the discussant, the effects on social desirability should remain reliable even when the effects of their ratings of sympathy and identification are statistically removed from the ratings of social desirability, and vice versa. Both ratings were related with social desirability, r(64) = .33, p < .01, and r(64) = .70, p < .001, respectively.

To explore whether the effects on social desirability were mediated by sympathy or identification, 2 (Target under Evaluation) by 2 (Direction of Comparison) ANCOVAs were performed after adding sympathy or identification as a covariate.⁹ First, controlling for sympathy did eliminate the interactive effect, F(1, 59) = 1.18, ns, whereas the main effect of direction of comparison on social desirability remained significant, F(1, 59) = 59.74, p < .001. The absence of a significant interaction, the marginally significant main effect of direction of comparison on sympathy notwithstanding, suggests that the effects of the experimental manipulations on the social desirability ratings were mediated by feelings of sympathy for the discussant. Secondly, the ANCOVA on social desirability was repeated after adding identification as a covariate (see footnote 3) revealed a somewhat different pattern. Both the main effect of direction of comparison on social desirability remained, F(1, 59) = 16.01, p < .001, as did the two-way interaction, F(1, 59) = 5.62, p < .03.

Also, ANCOVAs on the sympathy and identification ratings were conducted treating desirability as a covariate (see footnote 3) to investigate the reverse possibility, namely that the effects on sympathy and identification are mediated by participants' perceptions of social desirability of the verbal statements. The interactive effect of direction of comparison and target under evaluation remained statistical significant after partialling out the desirability variable, F(1, 59) = 9.41, p < .01. In contrast, an ANCOVA on participants' identification ratings with desirability as a covariate failed to find any effect, Fs < 1, ns. In short, these findings suggest that the interactive effect between direction of comparison and target under evaluation on social desirability was mediated by feelings of sympathy.