From the Illusion of Choice to Actual Control: Reconsidering the Induced-Compliance Paradigm of Cognitive Dissonance

Reanalysis

Given potential limitations of the experimental procedure to manipulate choice perceptions, ² we explored whether self-reported perceived choice influenced dissonance reduction in Vaidis et al.’s (2024) data set. We aggregated the two manipulation-check items on choice perception (r = .41) and probed their role in behavior-consistent attitudinal change (while controlling for prior attitudes). An analysis of covariance yielded a significant interaction between perceived choice and dissonance induction (i.e., the contrast between writing a self-chosen or counterattitudinal essay; see Table 1 and Fig. 1). Simple slope analysis indicated that the effect of dissonance induction was nonsignificant at low choice perceptions but significant at higher levels.

These results persisted in robustness checks (without controlling for prior attitudes and in separate analyses of the choice perception items; see R script at https://osf.io/hb58p/?view_only=d2902b3f2e884b7383d2ab4214c72e93).

OPEN IN VIEWER

Table 1.

Interaction Effects Between Self-Reported Perceived Choice and Condition Contrasts on Attitudes in Favor of Tuition Fees (Dissonance Reduction), With Bonferroni-Corrected Post Hoc Simple Comparisons

				95% CI
		b	SE	Lower	Upper	β	p
(Intercept)		12.88	0.19	12.51	13.25	.00	< .001
Prior attitude		0.16	0.09	–0.02	0.33	.05	.076
HC-NE – HC-CE		–1.19	0.46	–2.09	–0.29	–.20	.010
LC-CE – HC-CE		0.03	0.43	–0.82	0.88	.00	.947
Perceived choice		0.04	0.05	–0.05	0.13	.03	.417
HC-NE – HC-CE × Perceived Choice		–0.37	0.12	–0.61	–0.14	–.26	.002
LC-CE – HC-CE × Perceived Choice		0.05	0.10	–0.15	0.25	.04	.614
Simple comparisons at levels of  perceived choice
–1 SD	HC-NE – HC-CE	0.41	0.77	1.43	2.26	.06	1.00
	LC-CE – HC-CE	–0.19	0.57	–1.56	1.18	–.03	1.00
M	HC-NE – HC-CE	–1.19	0.46	–2.29	–0.08	–.20	.031
	LC-CE – HC-CE	0.03	0.43	–1.01	1.06	.00	1.00
+1 SD	HC-NE – HC-CE	–2.78	0.58	–4.19	–1.38	–.46	< .001
	LC-CE – HC-CE	0.25	0.65	–1.30	1.80	.04	1.00
Simple effects of perceived choice  by condition
	HC-NE	–0.23	0.09	–0.42	–0.04	–.16	.015a
	LC-CE	0.20	0.07	0.05	0.34	.14	.007
	HC-CE	0.15	0.07	0.01	0.28	.10	.040

Note: The overall interaction between condition (HC-NE vs. HC-CE vs. LC-CE) and perceived choice was significant, F(2, 1098) = 7.07, p < .001, η²_partial = 0.01. CI = confidence interval; HC-NE = neutral (self-chosen) essay with high-choice condition; HC-CE = dissonant essay with high-choice condition; LC-CE = dissonant essay with low-choice condition.

^aWhen excluding participants who defied the essay instructions (e.g., HC-NE participants who wrote essays to counteract increased tuition), as in Vaidis et al. (2024)’s secondary analysis, the simple effect of perceived choice in the HC-NE condition turned nonsignificant (β = -.13, p = .086), whereas the other effects remained unchanged.

OPEN IN VIEWER

Fig. 1.

Simple slopes of self-reported perceived choice on attitudes in favor of tuition fees (dissonance reduction) by condition, with 95% confidence intervals.

Discussion

The interaction between dissonance induction and perceived choice indicates that behavior-consistent attitudinal change occurred only among participants who genuinely felt a sense of choice over the counterattitudinal behavior. This finding contradicts the conclusions of Vaidis et al. (2024), who questioned the role of choice in dissonance. They observed dissonance reduction both after making participants’ choice salient (experimental condition) and without such choice reminders (control condition), whereas our reanalysis indicates actual choice perceptions are indeed a crucial precondition of dissonance reduction. Specifically, we observed no significant dissonance reduction in participants who reported low levels of choice perceptions.

Our findings are consistent with the theoretical underpinnings of the induced-compliance paradigm. Therefore, instead of discarding the framework, we suggest that the original methodology for manipulating choice perceptions is insufficient. The shortcomings of this experimental method might stem from recent societal and scientific practices: If participants are desensitized to obligatory reminders of voluntary study participation, such disclosures will hardly manipulate feelings of choice. However, this reflects a post hoc response to Vaidis et al.’s findings, and our reanalysis is limited by relying on two manipulation-check items.

The present perspective highlights the need for dissonance research to refine its methods and theoretical frameworks in multiple ways. First, a promising avenue may be to borrow from models and methods from personal-control research (Hamann et al., 2023; Pauer, Rutjens, & van Harreveld, 2024; Skinner, 1996). This could involve manipulating not only behavioral control (perceived choice) but also control over behavioral outcomes, compensatory control, and generalized control. Second, researchers can draw on behavioral tasks (e.g., Berger & Wyss, 2021; Lange & Dewitte, 2021) designed to induce conflict between behavioral engagement, its outcome, and a given attitude object. These tasks readily lend themselves to manipulating choice perceptions and dissonant behavior under well-controlled conditions. Third, various additional methods beyond traditional laboratory experiments are essential for advancing the generalizability and reliability of the field (Diener et al., 2022; Grosz et al., 2020; Yarkoni, 2022), such as by investigating naturally occurring instances of dissonance and the role of choice perceptions in daily life (Pauer, Rutjens, Hofmann, & van Harreveld, 2024). Finally, historical data could yield meta-analytic insight into the trends in effect sizes of the induced-compliance paradigm over time (Muthukrishna et al., 2021). Overall, the implications of the failed replication open up fruitful opportunities for advancing cognitive-dissonance research.