Manipulation check
One-way ANOVA with the seven items assessing emotional states as dependent measures showed that only scores on the anger item (i.e., “How angry do you feel now?”) differ significantly between conditions, F (2, 96) = 5.78, p = .004. Subsequent post hoc test (LSD) showed that participants in the anger condition (M = 3.58, SD = 2.39) reported higher anger than those in the sadness condition (M = 2.19, SD = 2.02), p = .007, as well as those in the neutral condition (M = 2.03, SD = 1.64), p = .002. For the sadness item (i.e., “How sad do you feel now”), although participants tended to report higher sadness in the sadness condition (M = 3.66, SD = 2.42) than in the anger condition (M = 3.27, SD = 1.92) and the neutral condition (M = 3.21, SD = 2.14), these differences were not significant, F (2, 96) = 0.41, p = .665. Therefore, it seems that our manipulation of anger was successful, but the manipulation of sadness was not as powerful as expected.
IAT analyses
A 3 (condition: sadness vs. anger vs. neutral) by 2 (IAT block: Block 4 vs. Block7) ANOVA with repeated measure on the latter factor showed no significant results. As depicted in Fig. 1, the main effect of condition, F (2, 96) = 0.61, p = .548, the main effect of IAT block, F (1, 96) = 0.11, p = .742, and the interaction, F (2, 96) = 0.89, p = .413, were all far from significance. Therefore, DeSteno et al.’s [7] results were not replicated.
Further, in each condition, we used Bayes factor analysis to examine whether the null hypothesis (i.e., RT in Block 4 is not different RT in Block 7) was more likely given the current data. The traditional significance testing can only provide information regarding whether H0 (the null hypothesis) can be rejected. However, rejecting H0 does not necessarily mean there is no difference between conditions. The Bayes factor provides a continuous measure of evidence for H1 (the alternative hypothesis) over H0 [8]. When the Bayes factor is 1, the data favors neither H1 nor H0. As the Bayes factor increases above 1 (toward infinity), the data favors H1 over H0. On the contrary, the data favors H0 over H1 as the Bayes factor decreases below 1 (toward 0). Researchers suggest that a Bayes factor (BF10) that is bigger than 3 would provide substantial support for H1 whereas a Bayes factor (BF10) smaller than 1/3 would provide substantial support for H0. Our results supported the null hypothesis in all conditions (for anger, BF10 = 0.27; for sadness, BF10 = 0.22; for neutral, BF10 = 0.25).
IPD-MD
A 3 (condition: anger vs. sadness vs. neutral) by 3 (pool: individual vs. within-group vs. between-group) ANOVA with repeated measure on the latter factor showed a significant main effect of pool, F (2, 192) = 24.04, p < .001, partial η2 = .20, as shown in Fig. 2. However, both the main effect of condition, F (2, 96) = 0.38, p = .688, and the interaction, F (4, 192) = 0.77, p = .547, were not significant. Post hoc test of the main effect of pool (LSD) showed that participants invested more in the within-group pool (M = 5.45, SD = 3.61) than the individual pool (M = 2.69, SD = 2.89, p < .001) and the between-group pool (M = 1.86, SD = 2.81, p < .001). They also allocated slightly more to the individual pool than the between-group pool, p = .065. These results indicate that participants in all of the three conditions showed ingroup favoritism rather than outgroup derogation.
Explicit attitudes
A 3 (condition: anger vs. sadness vs. neutral) by 2 (group: ingroup evaluation vs. outgroup evaluation) ANOVA with repeated measure on the latter factor showed only a marginally significant main effect of group, F (1, 96) = 3.30, p = .073, partial η2 = .03, as shown in Fig. 3, which suggests that participants evaluated the ingroup (M = 5.85, SD = 1.23) more positively than the outgroup (M = 5.70, SD = 1.19). The main effect of condition, F (1, 96) = 0.71, p = .496, and the interaction, F (2, 96) = 0.96, p = .386, were far from significance. Interestingly, the evaluation of the outgroup was above the midpoint of the scale (i.e., “5”), t (98) = 5.82, p < .001, which showed no sign of outgroup derogation. This pattern held in all of the three emotion conditions (for anger, t (32) = 3.10, p = .004; for sadness, t (31) = 3.95, p < .001; for neutral, t (33) = 3.51, p = .001).