Abstract
Memories serve as a “database” of the self and people often produce distorted memories that support their self-concepts. One, surprisingly untested, possibility is that cognitive dissonance may be one mechanism by which people may misremember their past. We tested this hypothesis using an induced-compliance paradigm: participants either chose or were forced to write a counterattitudinal essay supporting a tuition increase and were afforded the opportunity to reduce dissonance via attitude shift or denial of responsibility. They then reported their memories for the experimental instructions and their initial attitudes (assessed two days prior to the laboratory session). Participants who chose to write the essay exhibited the predicted attitude-shift effect, and were more likely to misremember their initial attitudes and the experimental instruction than those who were forced to write the essay. Overall, our results provide evidence that cognitive dissonance may yield memory distortion, filling a significant gap in the motivated cognition and memory literatures.
We would like to thank Nicholas Bonomo and Marita Salwierz for their help in data collection.
We would like to thank Nicholas Bonomo and Marita Salwierz for their help in data collection.
Notes
1 A full discussion of attitude formation is beyond the scope of this paper. However, interested readers may wish to consult Bodenhausen and Gawronski (Citation2013) and Cunningham and Zelazo (Citation2007) for perspectives on how and when attitudes are formed.
2 Additionally, the denial-of-responsibility effect size estimate we used to determine our sample size was derived from Gosling et al. (Citation2006). They obtained large effects (as large as d = 3.32) with rather small samples (Ns = 51 and 64), suggesting that perhaps sampling error contributed to their estimates (see Ferguson & Heene, Citation2012). The true effect size may be smaller than we were able to detect with our sample.
3 To ensure that any differences on these items were the product of dissonance-induced memory distortion and not differences in the inherent “memorability” of the High vs. Low choice instructions, we conducted a pilot test using an independent sample of 60 undergraduates. These participants (n = 30 in each group) received the same Choice manipulation, but were asked to write an essay opposing a tuition increase (i.e., not a counterattitudinal essay). Participants then completed the same memory items. There was no difference in the number of memory errors between the two groups for either the Instruction item (High choice: 6.67%; Low choice: 10%), χ2(1, N = 60) = .22, p = .64, or the Statement item (High choice: 6.67%; Low choice 13.3%), χ2(1, N = 60) = .74, p = .39; the two instructions do not differ in their inherent memorability.
4 In hierarchical loglinear analysis, all lower-order effects subsumed by a significant interaction term are retained in the final model. Thus, our final model contained the Choice × Instruction association, as well as the “main effects” of Choice and Instruction.