Dissonance within the social self: exploring the effects of norm conflict between ingroups

ABSTRACT

Most people are members of different groups. While the norms of these groups can align, they can also be contradicting, leading to dissonance within the social self. This is different from dissonance at the individual level because it is based solely on individuals’ membership in ingroups with conflicting norms. Building on the Social Identity Approach and Cognitive Dissonance Theory, we assume that norm conflict between ingroups increases psychological discomfort, particularly for people who are highly identified with both ingroups. Norm conflict was manipulated by showing participants the results of bogus surveys, indicating that their ingroups agree or disagree (in Experiment 1 on the topic of self-driving cars; in Experiment 2 on how to act in a moral dilemma). We tested several strategies to cope with norm conflict between ingroups, namely, lowering the credibility of the norm conflict information and/or norm prototypicality (Exp. 1 & 2), disidentification (Exp. 1), as well as compartmentalization and the restorative function of agentic groups (Exp. 2). In line with our assumptions, results indicated that norm conflict between ingroups led to increased psychological discomfort in Experiment 1 (N = 870) and Experiment 2 (N = 812). This was not moderated by the two ingroup identification levels. The results of Experiment 1 indicated that lowering perceived prototypicality of the norms was a coping strategy for high identifiers. No effects emerged for other coping strategies. Both experiments show evidence of dissonance within the social self and highlight the necessity for further examinations of its consequences and boundary conditions.

KEYWORDS:

• Social identity
• multiple ingroups
• dissonance
• norms
• norm conflict

Acknowledgments

Data collection of Experiment 1 was funded by PsychLab, a service of the Leibniz Institute for Psychology (ZPID). The authors thank the action editor and reviewers for their invaluable contributions to this manuscript, as well as Lisa Koziel and Janine Stollberg for providing data used for pre-testing. The introduction and methods of this manuscript are based on the first author's unpublished master thesis ”Social Identity and Norm Conflict between Ingroups”.

Data Availability statement

The data and code that support the findings of these studies are openly available in PsychArchives at https://doi.org/10.23668/psycharchives.8415 (Exp. 1), https://doi.org/10.23668/psycharchives.12868 (Exp. 2), and https://doi.org/10.23668/psycharchives.12894 (Code).

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/23743603.2023.2214965

Notes

1. While participants rated their attitude and attitude strength regarding all nine topics, each participant was assigned to only three norm manipulations. Hence, only a subsample of n = 40 was presented norms towards self-driving cars. The credibility measure included only two items as one item was dropped due to its low item-scale correlation.

2. The response scale was labelled −4 to 4.

3. The panel provider respondi automatically oversamples approximately 10% to make up for exclusions. In line with our preregistration, we excluded 5 participants who disagreed with data processing, one who indicated not answering seriously at all, and 20 speeders (time < median/2). Whilst we did not pre-register this, we initially excluded one response recorded as a technical test trial by the panel provider respondi. Participants who were not in our target group (aged 20–30, city-dweller, neutral attitude) or failed an instructional manipulation check item were filtered out during data collection.

4. Of note, data quality screenings showed that there were limitations to the data quality. 32 participants gave the same answer (e.g. “1” or “5”) on more than 80% of the scale items and 54 participants wrote nothing related to the study or “I don’t know” when asked to summarize the bogus survey.

5. For the first 23 participants the labelling of the manipulation check read 0–8 rather than 1–9. Excluding these cases did not change the results: t(839.36) = − 40.59, p < .001, d = −2.78.

6. Welch-Test for unequal variances.

7. When excluding outliers at ±3 studentized residuals: BE = 0.01, SE = 0.04, t(859) = 0.19, p = .846, partial R² = .000. In all other analyses we similarly checked results when excluding outliers at ±3 studentized residuals (if there were any) and likewise did not find this to change the reported results.

8. Simple slopes showed no significant effect of norm conflict on prototypicality for participants with high identification with city-dwellers and low identification with the generation, BE = 0.14, SE(HC3) = 0.12, t(862) = 1.15, p = .253, or for participants with low identification with city-dwellers and high identification with the generation, BE = 0.16, SE(HC3) = 0.10, t(862) = 1.65, p = .099.

9. The predictors were centered for this model.

10. The response scale was labelled −4 to 4.

11. This material has been used by Stollberg et al. (2015). Their pre-test (N = 40) shows that the group in the agentic picture is perceived as significantly more agentic than in the non-agentic picture for the city context (group demonstrating M = 5.65, SD = 0.96; group shopping M = 1.86, SD = 0.90, t(39) = 14.26, p < .001, d = 2.25).

12. Again, and in line with our pre-registration, we excluded 5 participants who disagreed with data processing, one who did not fill out the questionnaire seriously at all, 64 who failed at least one instructional manipulation check, and 28 speeders (time < median/2). Whilst we did not pre-register this, we initially excluded two responses recorded as a technical test trial by the panel provider clickworker.

13. Of note, data quality screenings showed better results for Experiment 2 than for Experiment 1. No participant gave the same answer on more than 80% of the scale items and only 8 participants wrote nothing related to the study or “I don’t know” when asked to summarize the bogus survey.

14. As in Experiment 1, we checked all results when excluding outliers at ± 3 studentized residuals (if there were any) and did not find this to change the reported results.

15. The sample is smaller for this test, as answers on the card sorting task were not mandatory.

16. Simple slopes showed a significant effect of norm conflict on credibility for participants with high identification with Europeans and low identification with students, BE = − 0.62, SE(HC3) = 0.15, t(610) = − 4.00, p < .001, but no significant effect for participants with low identification with Europeans and high identification with students, BE = − 0.28, SE(HC3) = 0.15, t(610) = − 1.94, p = .052.

17. The three models all included a main effect of the dummy testing norms in favor (vs. against) using the lever (on psychological discomfort: BE = − 0.41, SE (HC3) = 0.17, t(806) = − 2.38, p = .018; on prototypicality: BE = 0.53, SE (HC3) = 0.22, t(806) = 2.46 p = .014; on credibility: BE = 0.86, SE (HC3) = 0.17, t(806) = 5.14 p < .001). Further, there were main effects of attitude on psychological discomfort: BE = 0.21, SE (HC3) = 0.05, t(806) = 3.93, p <.001, and on credibility, BE = - 0.18, SE (HC3) = 0.06, t(806) = - 3.19 p = .002. The predictors were centered for these three models.

18. The predictors were centered for this model.

Additional information

Funding

Data collection of Experiment 1 was funded by PsychLab, a service of the Leibniz Institute for Psychology (ZPID).