Cognitive Dissonance

Abstract

Cognitive dissonance, which MeSH classifies under cognition, is the motivational state that arises when a person holds cognitions that conflict with one another, or holds a belief that conflicts with his or her own behaviour. Much of its evidence comes from three classic laboratory procedures: arguing for a position one rejects, choosing between similar options, and paying a cost to reach a goal. Later revisions disputed what drives the state, while research on arousal, reduction strategies, culture and brain activity examined how it is felt and resolved. A 39-lab replication of the essay procedure found no effect of choice, whereas a meta-analysis of free-choice studies that removed a known artifact found reliable preference change. Three interactive demonstrations cover induced compliance, the spreading of alternatives and the routes to reduction.

Keywords: cognitive dissonance, induced compliance, free-choice paradigm, self-perception, attitude change

The theory of cognitive dissonance was set out in Festinger's (1957) book. More than 60 years later, a single-author review could still describe the idea as an enduring part of social psychology, one that has been revised along the way (Cooper, 2019). Much of the research it inspired asks whether people who argue against their own attitudes come to change them, the question the induced-compliance paradigm was built to answer (Vaidis et al., 2024).

The field is now examining its own foundations. A critical review written by authors who describe themselves as supporters of the theory argued that its central concepts need sharper definition and that no validated instrument exists for measuring the dissonance state itself; the review also noted that procedures for inducing and assessing dissonance have never been standardised, and that measures of how people relieve the state cannot by themselves confirm it was present (Vaidis & Bran, 2019). That paper offered argument rather than new data, but it frames the questions this article returns to.

What Dissonance Is

Motivational state produced by inconsistencies between simultaneously held cognitions or between a cognition and behavior; e.g., smoking enjoyment and believing smoking is harmful are dissonant.

Scope note for the MeSH descriptor Cognitive Dissonance (D003073), U.S. National Library of Medicine.

The definition places dissonance among the forms of cognition but describes it as a state of motivation: something a person is driven to remove, not merely a mismatch noted in passing. An edited volume opens with an introduction to the theory and an overview of the perspectives that have grown up around it (Harmon-Jones & Mills, 2019). Cooper's (2019) review makes a point about the drive itself that is easy to miss: Festinger had no direct evidence that dissonance worked as a drive, later learning-task evidence for drive properties was mixed, and Cooper's own suggestion that the drive is learned is, by his account, speculative.

Cognitive dissonance theory, as a whole, is contested. Its standing turns on two kinds of evidence that point in different directions: a 39-lab test of the induced-compliance procedure did not find its central prediction (Vaidis et al., 2024), and a meta-analysis of free-choice studies that controlled a known artifact did find reliable preference change (Enisman et al., 2021). Both studies are described below.

The action-based model offers one answer to why inconsistency should matter at all. On this view, dissonance is a negative affective state that arises when cognitions bearing on action conflict, because the conflict gets in the way of acting; the model's proponents set this out in a theory paper that reviews experiments in its support (Harmon-Jones et al., 2015). In two experiments, the same laboratory tied the spreading effect to approach motivation: when neurofeedback lowered relative left frontal activity, participants showed less spreading of alternatives, and adopting an action-oriented mindset once the decision was made heightened left frontal activity and spreading alike (Harmon-Jones et al., 2008). The model remains contested in the broader sense that it is one proposal among several in a field that, according to a recent critical review, still lacks agreed concepts and measurement tools (Vaidis & Bran, 2019). Cooper (2019) argues that the action-based model has difficulty accommodating the classic caveats about choice.

Induced Compliance

Induced compliance is the procedure of leading people to say or write something that contradicts their own view and then measuring whether their view shifts. Its founding report was a paper on the cognitive consequences of forced compliance (Festinger & Carlsmith, 1959). Cooper's (2019) retrospective recounts that 1959 experiment, with its two payment levels of $1 and $20, as the start of a tradition whose later work relied on the counterattitudinal essay and stressed the role of decision freedom. Because a $20 payment gave people a ready reason for what they had done, theory predicted it would arouse less dissonance than $1, and the participants offered only $1 did end up shifting their attitudes further than those offered $20 (Cooper, 2019).

An early extension moved the logic from persuasion to prohibition. In a laboratory experiment with 22 preschool children, a mild threat against playing with a toy led to more derogation of that toy than a severe threat did (Aronson & Carlsmith, 1963). The sample was small, and the stored record of the study is a truncated abstract.

The paradigm's central claim is now in doubt. A multilab constructive replication collected data from 4,898 participants in 39 labs across 19 countries and 14 languages, using the first experiment of Croyle and Cooper as its template and three conditions (Vaidis et al., 2024). Writing a counterattitudinal essay under high choice produced no more attitude change than writing it under low choice, so the core prediction was not supported; the result held up across exclusion rules and variation between labs. Essay writers did differ significantly from a neutral comparison group, which leaves open what produced that difference, but the authors concluded that the paradigm's evidential value is in question. This is the main reason dissonance theory's standing is described here as contested.

The first demonstration lets the reader set the payment and the sense of free choice in a schematic essay task and then switch on the replication result.

Free Choice and Effort

The second classic procedure studies the aftermath of decision making. Brehm (1956) designed an experiment to test the prediction that, after choosing between alternatives, people would come to rate the chosen option as more desirable and the rejected one as less so. The stored abstract of that paper states the predictions but not the results, so it is cited here for the test, not for its outcome. A third tradition concerns effort justification, the tendency to value what has been costly to obtain; its classic study examined how the severity of an initiation affected later liking for a group (Aronson & Mills, 1959).

The free-choice effect has been sought well beyond adult humans. Preschool children choosing among stickers and capuchin monkeys choosing among coloured M&M's both devalued the option they had passed over, while no such shift appeared in a control condition (Egan et al., 2007). Those results bear on the origins of the effect and are taken up again under culture and development.

A methodological problem then reshaped the paradigm. Chen and Risen (2010) combined a mathematical proof with experiments to show that spreading of alternatives, the widening gap between ratings of chosen and rejected items, can appear even when no attitude has changed. The authors reported experiments indicating that the conditions behind the proof appear to hold, argued that the problem extends to every variant of the paradigm, and suggested that conclusions drawn from it be reassessed. A later critical review, which combined simulation with a small meta-analysis, agreed that the artifact can produce spreading without dissonance; it concluded that choice-induced preference change exists but is small in magnitude, estimated from four studies that addressed the critique, and that earlier findings need to be established again (Izuma & Murayama, 2013).

The strongest current evidence comes from designs that remove the artifact. A meta-analysis of 43 such studies with a combined N = 2,191 found that choice changed preferences, d = 0.40, 95% CI [0.32, 0.49], and found no evidence of publication bias; the analysis included Bayesian methods (Enisman et al., 2021). That estimate concerns choice behavior in the free-choice paradigm and says nothing directly about the induced-compliance effect. Figure 1 sets the naive spread beside the spread that remains once noise alone is accounted for.

Dissonance research also left the laboratory early: Festinger and colleagues covertly joined a group that expected the world to end, an observational field study framed as a test of how believers respond when a prophecy fails (Festinger et al., 1956). A meta-analysis of selective exposure found a moderate preference for information that agrees with what people already believe (d = 0.36) (Hart et al., 2009). That preference was weaker when a person's view had already been validated, and it reversed into a preference for disagreeing information when such information served a goal (Hart et al., 2009).

The second demonstration simulates participants who rate, choose and rate again, and shows how much spread rating noise produces by itself.

Revising the Theory

Self-perception theory offered a rival account. Bem (1967), writing from a radical behaviourist position and reporting supporting experiments, argued that people judge their own attitudes much as they judge other people's, from behaviour, so that the classic findings need no drive toward consistency. A later paper took an integrative view in which dissonance and self-perception each have their own proper domain of application (Fazio et al., 1977). Self-perception theory remains contested as a full account: experiments showing that dissonance is felt as discomfort, one of which was designed to exclude a self-perception explanation, count against a purely inferential reading (Elliot & Devine, 1994), as does the finding that a placebo pill could absorb the effect (Zanna & Cooper, 1974).

Other early responses refined the theory or challenged its premises. A chapter gave a current perspective on the theory (Aronson, 1969), and a critique in the American Psychologist asked whether the effects reflect private reasoning or a public display (Tedeschi et al., 1971). Cooper's review recounts Aronson's argument that dissonance requires one of the conflicting cognitions to concern the self, rather than arising from any inconsistent pair (Cooper, 2019).

The new look was presented as a fresh look at dissonance theory (Cooper & Fazio, 1984), and Cooper's (2019) later review describes it as a revision in which responsibility for an aversive consequence drives the state. The new look has itself been challenged. In three experiments, the first of which crossed high or low choice with a pleasant- or unpleasant-tasting beverage, dissonance occurred even when the counterattitudinal act produced no aversive consequences, though the abstract reports no statistics (Harmon-Jones et al., 1996).

Self-affirmation theory was set out in a chapter on self-affirmation and how people maintain the integrity of the self (Steele, 1988). On cross-cultural evidence, its predictions depend on culture. In a free-choice study, Canadian participants reduced dissonance and Japanese participants did not, a contrast the authors related to independent and interdependent ways of construing the self (Heine & Lehman, 1997). Four later experiments found dissonance among both Eastern and Western participants but in different forms, with interdependent and independent affirmations working differently by culture (Hoshino-Browne et al., 2005). A later review of self-affirmation research treats writing about core values as a buffer against threats to self-integrity, with benefits that in some cases have lasted months or years (Cohen & Sherman, 2014).

Stone and Cooper (2001) proposed a self-standards model of dissonance. It is one of several later formulations, alongside ideas such as vicarious dissonance, and Cooper's (2019) review notes that the theory has needed many moderating conditions; how these revisions fit together remains contested.

The most recent reformulation comes from outside social psychology. Kaaronen (2018) proposed treating dissonance as a form of prediction error within predictive coding, with inconsistencies nested at several levels, and suggested that responses to it can guard against overfitting, although in today's world some responses may yield models of the world that fit too loosely and lean toward confirming evidence. The paper is a conceptual argument with no data, and it offers its evolutionary rationale as a potential one.

Arousal and Discomfort

If dissonance is a motivational state, it should be felt. Zanna and Cooper (1974) tested this with misattribution in a 2 x 3 experiment: participants who wrote a counterattitudinal essay under different levels of choice had taken a placebo pill described as producing tension, relaxation or no effects. Where the pill offered an explanation for feeling tense, the usual attitude change virtually disappeared, and a pill described as relaxing enhanced the effect. The abstract does not state the sample size.

Physiological recording followed. Across two induced-compliance experiments, the first reproduced the usual attitude change, and the second, which added physiological recording, found that writing counterattitudinal essays under high choice raised nonspecific skin conductance responses, a marker of arousal (Croyle & Cooper, 1983). The second experiment produced no attitude change, possibly because participants attributed their arousal to the recording device itself.

Elliot and Devine (1994) then measured the felt state directly. In two induced-compliance experiments, a self-report affect measure showed that dissonance was experienced as psychological discomfort, and the second experiment was arranged to rule out a self-perception explanation. The finding located dissonance among the emotions people can report, not only in a physiological signal.

Whether discomfort is what produces the effects is less clear. A synthesis that applied meta-analytic structural equation modelling across five research paradigms found no support for discomfort as the mediator of dissonance effects, and guilt emerged as a predictor instead (Kenworthy et al., 2011). The authors argued that a theory integrating guilt and dissonance is stronger than either alone, and the evidence available here is the abstract alone.

Reducing Dissonance

Attitude change is not the only way out. A series of four studies found that people can reduce dissonance through trivialization, deciding that the inconsistent behaviour or issue was unimportant, and that this route was taken when their prior attitude had been made salient (Simon et al., 1995). Trivialization resembles everyday rationalization: the inconsistency is not resolved but made to matter less.

A narrative review of the reduction literature concluded that the process by which people reduce dissonance has received too little study, and it recommended designs that offer participants several modes of reduction at once (McGrath, 2017). That review is available here only in abstract.

Hypocrisy induction turns reduction toward behaviour change. Sexually active young adults placed in a hypocrisy condition after advocating condom use for AIDS prevention bought more condoms than those in control conditions (Stone et al., 1994). The same approach has been applied to health behavior during the pandemic: in an experiment with a follow-up a week later, a dissonance condition led to more reported compliance with COVID-19 safety measures and more vaccination seeking than three non-dissonance control conditions (Pearce & Cooper, 2021). The behaviour in that study was self-reported, and the abstract does not give the sample size.

The third demonstration lets the reader vary the importance and salience of an attitude and the availability of an affirming value, and shows which route to reduction the reviewed findings would lead one to expect.

Culture and Development

The free-choice effect does not appear in the same form everywhere. Canadian participants in a free-choice study reduced dissonance while Japanese participants did not, although the abstract gives no sample size (Heine & Lehman, 1997). Four experiments then showed that Japanese participants did justify their choices when cues to self-relevant others were present, whether through questionnaire primes or schematic faces, but showed no dissonance without such social cues (Kitayama et al., 2004). A further series comparing European Canadians, Asian Canadians and Japanese participants found dissonance in both cultural groups, triggered and relieved in different ways, with bicultural participants differing from monocultural ones; sample sizes are not reported in the abstract (Hoshino-Browne et al., 2005).

Developmental and comparative evidence asks how early and how widely the effect appears. Choice-induced devaluation of a rejected option was found in preschool children and in capuchin monkeys, with no such effect in a control condition (Egan et al., 2007). This evidence bears on child development, but it comes from the free-choice paradigm and predates the artifact critique described above, and it does not show that children or monkeys experience the discomfort adults report.

Neural Basis

Brain imaging has asked where the conflict registers and what predicts its resolution. In an fMRI experiment that used a Solomon four-group design, activity in the dorsal anterior cingulate cortex and anterior insula predicted later attitude change, and these effects were absent in the control group (van Veen et al., 2009).

A free-choice fMRI study with 20 adults rating 160 food items found that rejected items were devalued after choice beyond a control for the artifact (P = 0.036, one-tailed), while the rise for chosen items was not significant (P = 0.41) (Izuma et al., 2010). Its analyses concerned regions including the dorsal anterior cingulate cortex, the dorsolateral prefrontal cortex and the anterior striatum. The sample was small, some region-of-interest thresholds were uncorrected, anterior insula activity appeared only at a lenient threshold, and only food items were used.

Activity during the decision itself has also been linked to what follows. An fMRI study found that brain activity during the decision phase related to later decision-induced attitude change, and the authors suggested that rationalization resembles reappraisal, a form of emotional regulation (Jarcho et al., 2011). That mechanism was suggested, not demonstrated, and the abstract reports no sample.

Electroencephalography adds timing. In a free-choice EEG study of 45 adults, 42 of whom were analysed, difficult choices were followed by a frontocentral negativity resembling the error-related negativity seen in performance monitoring (Colosio et al., 2017). The authors noted that the signal's identity as an error-related negativity still needs confirmation, that the correlation between the evoked response and preference change was only a trend at the frontocentral electrode FCz (p = 0.093), and that a reminder cue in the design may have inflated the effect.

Causal evidence is rarer. In one experiment, lowering relative left frontal activity through neurofeedback left participants with less spreading of alternatives; in a second, a postdecisional action-oriented mindset heightened both left frontal activity and spreading (Harmon-Jones et al., 2008), the result that anchors the action-based model described earlier.

Discussion

Two findings stand on firm ground. Choosing between options changes preferences even when the artifact that once clouded the free-choice paradigm is designed out, with an effect of d = 0.40 across 43 studies and no sign of publication bias (Enisman et al., 2021). And people who act against their attitudes in induced-compliance experiments report psychological discomfort, a result obtained in two experiments with a self-report measure (Elliot & Devine, 1994).

Other parts of the picture are weaker or open. The induced-compliance effect, long treated as the theory's signature, was not supported in a 39-lab replication with 4,898 participants, which found no difference between high and low choice, although essay writers differed from a neutral-essay group (Vaidis et al., 2024). What the dissonance state is, and how it should be measured, remain unsettled questions; a critical review found no validated instrument for the state and no standard procedures for inducing or assessing it (Vaidis & Bran, 2019). Whether discomfort is the engine of the effects is also unresolved, since a cross-paradigm synthesis found no support for its mediating role and pointed to guilt instead (Kenworthy et al., 2011).

These threads converge on a single problem. The paradigms that established dissonance infer the state from the behaviour that is supposed to relieve it, so a failure to find attitude change cannot distinguish an absent state from one relieved some other way, such as trivialization. Designs that measure the state directly, offer several reduction routes at once and include no-choice or neutral comparisons would separate these possibilities; the work reviewed here shows that such designs are being called for, not that they have been run at scale.

Current Directions

Much of the research front since 2016 has been reassessment. The multilab replication of the induced-compliance paradigm, with 4,898 participants in 39 labs, has made the paradigm's evidential value an open question (Vaidis et al., 2024). Calls for clearer concepts, standardised induction and validated measures of the state come from researchers who remain committed to the theory (Vaidis & Bran, 2019), while the free-choice literature has been consolidated by meta-analysis of artifact-free designs (Enisman et al., 2021).

Methods and theory are also widening. EEG work has looked for a fast performance-monitoring signal after difficult choices, though its identity is not yet confirmed (Colosio et al., 2017), and a conceptual proposal has recast dissonance in predictive-processing terms without yet testing it (Kaaronen, 2018). Applied work continues, for example on self-reported pandemic safety behaviour (Pearce & Cooper, 2021), while a review of management research found that the field relies on an outdated version of the theory and overlooks its recent refinements (Hinojosa et al., 2017). Reviews of reduction strategies (McGrath, 2017) and of the theory's history over more than 60 years (Cooper, 2019) set the agenda for this work.

Common Misconceptions

Dissonance always ends in attitude change.

Attitude change is only one route. Across four studies, people could reduce dissonance by trivializing the issue instead, particularly when their earlier attitude had been made salient (Simon et al., 1995).

The classic $1 and $20 effect is settled fact.

The induced-compliance paradigm that grew from that experiment failed its central test in a 39-lab replication with 4,898 participants, which found no difference between high-choice and low-choice conditions (Vaidis et al., 2024).

A spread between chosen and rejected items proves that choosing changed preferences.

Spreading of alternatives can arise with no change in attitude at all, in every variant of the free-choice paradigm (Chen & Risen, 2010). Only designs that correct for the artifact, for example with a comparison in which the second rating does not follow the choice, can attribute the spread to the choice.

Dissonance works the same way in every culture.

Japanese participants in a series of four experiments showed choice justification only when primed with cues to self-relevant others, and no dissonance without them (Kitayama et al., 2004).

Glossary

Action-based model.

A proposal that dissonance arises because conflicting cognitions relevant to action interfere with acting; it is one contested formulation among several.

Cognitive dissonance.

An unpleasant motivational state produced when a person holds cognitions that conflict with each other or with his or her behaviour; a 2019 critical review found no reliable instrument for measuring the state.

Counterattitudinal advocacy.

Arguing, usually in a written essay, for a position one does not hold; the standard task in induced-compliance research.

Effort justification.

The tendency to value an outcome more because obtaining it was costly or unpleasant.

Free-choice paradigm.

A procedure in which people rate options, choose between two, and rate them again; any shift in ratings needs a correction for the artifact, such as a no-choice baseline, before it can be credited to the choice.

Hypocrisy induction.

Making people aware that they do not practise a behaviour they have publicly advocated, with the aim of changing what they do.

Induced compliance.

Leading people to act against their own attitudes, often under varied levels of choice or reward, to see whether the attitude moves toward the act.

Misattribution.

Crediting one's own arousal or tension to a source other than its actual cause, such as a pill or a recording device.

New look.

A revision of dissonance theory that tied the state to responsibility for an aversive consequence; later experiments that found dissonance without aversive consequences challenged it.

Self-affirmation.

Shoring up a sense of self-worth by affirming a valued part of the self, which can offer a route out of dissonance other than attitude change; its role differs across cultures.

Self-perception theory.

The rival view that people infer their attitudes from observing their own behaviour, without any drive toward consistency; contested as a complete account of dissonance findings.

Self-standards model.

A revision of dissonance theory built around the standards people hold for themselves; how it relates to other revisions remains contested.

Spreading of alternatives.

The growing gap after a choice between ratings of the chosen and the rejected option; part of it can arise without any change in preference.

Trivialization.

Reducing dissonance by judging the inconsistent behaviour or issue to be unimportant, instead of changing the attitude.

Key Researchers

Elliot Aronson

(University of California, Santa Cruz (Professor Emeritus)). Social psychologist known for the severity-of-initiation and forbidden-toy experiments and, later, for using hypocrisy to change behaviour. Wikipedia · Wikidata · Faculty page

Daryl J. Bem

(Cornell University (Professor Emeritus)). Social psychologist who proposed self-perception theory, which reads dissonance findings as inferences people draw from watching their own behaviour. Wikipedia · Wikidata · Faculty page

Joel Cooper

(Princeton University). Social psychologist who recast dissonance around responsibility for unwanted consequences and later proposed the self-standards model. Faculty page

Leon Festinger

(The New School for Social Research; 1919–1989). Social psychologist who proposed the theory of cognitive dissonance and, with Carlsmith, ran the first induced-compliance experiment. Wikipedia · Wikidata · Faculty page

Eddie Harmon-Jones

(University of New South Wales). Social psychologist and neuroscientist who developed the action-based model of dissonance and linked it to frontal brain activity. ORCID · Wikipedia · Faculty page · Google Scholar

Keise Izuma

(Kochi University of Technology). Social neuroscientist who imaged preference change after choice and set out the methodological problems of the free-choice paradigm. Faculty page

Claude M. Steele

(Stanford University (Emeritus)). Social psychologist who proposed self-affirmation theory, in which affirming an important value can stand in for changing an attitude. Wikipedia · Wikidata · Faculty page

Frequently Asked Questions

What is cognitive dissonance?

On one current proposal, cognitive dissonance is a negative affective state aroused not by every conflict between cognitions but specifically when cognitions relevant to action conflict and get in the way of acting (Harmon-Jones et al., 2015).

Did the classic induced compliance effect replicate?

A multilab replication with 4,898 participants in 39 labs found no difference in attitude change between high choice and low choice conditions, so the central prediction was not supported, although essay writers did differ from a neutral group (Vaidis et al., 2024).

Does making a choice really change preferences?

A meta-analysis of 43 studies that used designs free of the known statistical artifact found that choosing changed preferences, with an effect of d = 0.40 and no evidence of publication bias (Enisman et al., 2021).

Is dissonance physiologically arousing?

In the second of two experiments, which added physiological recording, writing counterattitudinal essays under high choice raised nonspecific skin conductance responses but produced no attitude change; the first had reproduced the usual attitude change (Croyle & Cooper, 1983).

Do people always change their attitudes to reduce dissonance?

No. Across four studies, people could reduce dissonance by deciding the issue was unimportant, a route called trivialization, especially when their prior attitude had been made salient (Simon et al., 1995).

Do children or animals show dissonance effects?

Preschool children and capuchin monkeys devalued options they had rejected in a choice, and no such effect appeared in a control condition (Egan et al., 2007).

Does culture change how dissonance works?

Experiments with European Canadian, Asian Canadian and Japanese participants found dissonance in both Eastern and Western groups, but in different forms, and the abstract does not report sample sizes (Hoshino-Browne et al., 2005).

Can dissonance be used to change health behaviour?

Sexually active young adults who were placed in a hypocrisy condition after advocating condom use bought more condoms than participants in control conditions (Stone et al., 1994).

References