A Radical Point of View on Dissonance Theory - Perspectives Employing the Original Version of the Theory

Cognitive Dissonance: Reexamining a Pivotal Theory in Psychology - Eddie Harmon-Jones 2019

A Radical Point of View on Dissonance Theory
Perspectives Employing the Original Version of the Theory

Jean-Leon Beauvois and Robert-Vincent Joule

In 1957, Leon Festinger put forth the theory of cognitive dissonance (Festinger, 1957). His book described research and proposed a theory that explained the experimental results presented. It also contained a metatheory that borrowed from the Zeitgeist of the period and incited Festinger to make various generalizations, including one that made a connection between his theory and cognitive-consistency theories. In fact, extracted from the metatheory, the central element of Festinger’s theory boils down to this: A person can experience an unpleasant state of arousal (state of dissonance) that can be quantified by a ratio (the dissonance ratio; see Chapter 2, this volume, for further explanation) and is reduced when this ratio decreases. Cognitions are relevant and taken into consideration only to the degree that they allow for composing this ratio. Further, in 1962, Brehm and Cohen stated that everything nontrivial that this theory offers relates to the dissonance ratio. When we speak below of the “theory of ’57,” we are referring to this central element of Festinger’s presentation.

After functioning on this basis for about 10 years and producing the experimental results and classical paradigms that made its reputation, the theory was revised. These revisions consisted of the introduction of new propositions (assumptions?) that were supposed to explain why the state of dissonance existed. These propositions pulled the theory of dissonance toward a theory of the ego, and in fact, researchers neglected the rate of dissonance ratio and its theoretical implications,1 to deal with a new kind of cognition that Festinger had not foreseen (e.g., I was free to accept or to refuse). In spite of the interest of these revisions and the experimental work that they inspired, we believe that in abandoning the dissonance ratio, these revisions broke with the theory of ’57. The conception that we propose rests, on the contrary, on the idea that we must stick as close as possible to the theory of ’57, which was never disproven and produced the most fascinating products of dissonance theory. Thus, all hypotheses must be derived from the dissonance ratio.

Does this mean that the revisions were ill timed? Of course not! They had a genuine problem as their origin: The effects of dissonance are observed only in certain conditions (e.g., free choice or weighing consequences of the performed behavior). To deal with this problem, we favored a course other than revising the theory.1,2 In fact, we strove to resolve this problem without altering the theory of ’57 but rather by adding to it a single proposition, which was faithful, moreover, to Brehm and Cohen (1962): An act induces a state of dissonance only when there is commitment. Accepting this proposition amounts to making an important theoretical decision. It actually amounts to inserting dissonance theory into a more general theoretical framework, that of the psychology of commitment (Kiesler, 1971). However, this framework does not have affinities with that of ego theories. In fact, the same situation applies to Kiesler’s book as to Festinger’s. It contains a theory packaged within a metatheory. When one takes it out of the metatheory, Kiesler’s theory looks like a theory of external commitment (rather than the person committing to the act, the situation commits the person to the act), and this external commitment is rather incompatible with an ego theory.

This option, which implies the careful distinction between theory and metatheory,3 led us to propose a radical dissonance theory (Beauvois & Joule, 1996), which conserves only the central element of Festinger’s (1957) work, and a theory of external commitment (Joule & Beauvois, 1998), which conserves only the central element of Kiesler’s (1971) book. It goes without saying that returning to the central element of a theory does not mean that one is sticking to a theoretical status quo. It also goes without saying that giving up the metatheories of Festinger and Kiesler can result in new theoretical developments likely to astonish Festinger and Kiesler themselves.

The goal of this chapter is twofold. On the one hand, we want to show how our theoretical developments result in original hypotheses and even in new experimental paradigms. On the other hand, we want to present the particularities of the radical view relative to other versions of dissonance theory. For this, we select from the experimental work on which this radical view is based, and we describe several experimental results that, although compatible with the theory of ’57, seem quite incompatible with the other versions of the theory. Thus, we show that, within the experimental framework of forced compliance: (a) dissonance theory is not a theory of consistency, (b) the reduction of dissonance is not to serve a morally good ego, and (c) dissonance arousal does not necessarily imply personal responsibility for the act. Finally, we show that this radical view allows for the proposal of new experimental paradigms.


Two basic features of the theory of ’57 are important here:

1. The dissonance ratio, the reduction of which corresponds to the dissonance-reduction process, is defined with reference to one and only one cognition. This cognition appears in neither the numerator nor the denominator of the ratio. It represents the participant’s behavior.

 Shortly before Festinger decided to quantify the state of dissonance, structural balance theorists had proposed a measure of structure imbalance (Cartwright & Harary, 1956), which derived from a fundamentally different conception. This measure was the ratio of the unbalanced triads to the total number of triads implied by the structure. It was, therefore, a measure relating to the whole structure under consideration and accorded no particular status to any of the cognitions involved. This type of measure was in perfect conformity with Heider’s premises, which viewed the cognitive universe as a scene contemplated by the perceiver and that satisfied, to a greater or lesser degree, his or her preference for balance (Heider, 1958).

 Festinger chose a different approach. In effect, for him, the evaluation of the total amount of dissonance requires that one define a special element that makes it possible to assign the status of consonant or dissonant to the other cognitions. Therefore, this measure is oriented by a special cognition. We call it the generative cognition (Mills, Chapter 2, this volume, calls it the focal element). Formal and experimental arguments do at least suggest that this cognition is behavioral in nature. Several experiments (see Beauvois & Joule, 1996) allowed us to develop two types of hypotheses: one in which the behavioral cognition was generative, and one in which it was attitudinal. All the results indicate that the generative cognition is provided by the representation of behavior. This is the case of Experiment 1, which we report below.

2. Decreasing the dissonance ratio is compatible with the emergence of inconsistent relations between certain cognitions implicated by the dissonance ratio. The reason for this is that cognitions in the ratio are determined solely by their relation with the generative cognition. So, if the generative cognition is behavioral, the goal of the dissonance-reduction process is not the production of cognitive consistency, but rather the rationalization of the behavior that produced the generative cognition. That is why the dissonance-reduction process may result in greater inconsistency among other cognitions. Therefore, the reduction of the ratio, which for us remains the unconditional objective of the dissonance-reduction process, in no way implies that there is consistency among the cognitions that it contains. These two basic features come into play in the following experiment (Experiment 1).

 Festinger and Carlsmith (1959) realized that the condition that produced the greatest attitude change also produced less argumentation about the position being defended. Rabbie, Brehm, and Cohen (1959) observed incidentally that in a counterattitudinal role-playing situation, participants who had produced the greatest number of arguments in favor of the position being defended were also the ones whose attitudes changed the least. These observations suggested an inverse relationship between the elaborateness of argumentation and attitude change. In fact, these findings were never formalized by dissonance theorists, who, on the contrary, mentioned them as mere curiosities. Yet they clearly fit quite well into the strict (let us say “radical”) view of the theory of ’57: The counterattitudinal arguments that participants produce furnish cognitions that are consistent with the counterattitudinal behavior precisely because it consists of defending the argued viewpoint. There is nothing shocking about this statement to a dissonance theorist, who “naturally” recognizes that arguments consistent with the initial private attitude imply defending it by writing an attitudinal essay. As such, every argument that psychologically implies the viewpoint being defended must be regarded as a consonant cognition and, as such, is a dissonance-reducing cognition. It is thus easy to see why, in a counterattitudinal role-playing situation, the more arguments participants find supporting the attitude that they defend and against their initial attitude, the less dissonance they experience, and consequently, the less they change their mind to reduce dissonance. We would, therefore, expect an inverse relation between argumentation and attitude change. Beauvois, Ghiglione, and Joule (1976) obtained results supporting this hypothesis. For participants who were free to choose, as opposed to those with no choice, the more time allotted to supporting the counterattitudinal position, the less the participants’ attitudes changed. Joule and Levèque (1993) tested this hypothesis more recently with a 2 × 3 factorial design, with participants in a classical counterattitudinal roleplaying situation (individual training).

 In the Joule and Levèque (1993) experiment, 120 participants (literature students at the University of Provence) had to write a persuasive essay favoring the counterattitudinal position that “leisure activities are a waste of time for students.” All participants had volunteered to participate in what was said to be a study on persuasion. Half were told that they were free to accept or refuse to write the essay (free-choice instructions), whereas the other half heard no such statement (Independent Variable 1). A third of the participants were given 20 min to write the advocacy essay before assessing their attitude. Another third were given 5 min before assessing their attitude, and the final third expressed their attitudes immediately, before beginning to write (Independent Variable 2). The postexperimental attitude was measured on a 21-point scale, where the participants rated their degree of agreement with the position they were asked to defend (Dependent Variable 1). Of course, the arguments produced by the participants in the 20-min condition significantly outnumbered those generated in the 5-min condition (Dependent Variable 2). The results are given in Table 3.1.

TABLE 3.1. Effects of Counterattitudinal Advocacy on Attitude


Free-choice condition

No-choice condition

0 min

5 min

20 min

0 min

5 min

20 min















Note. For attitude, the higher the figure, the more closely the measured attitude conforms to the counterattitudinal act. In a control group (n = 20), the mean attitude was 2.60. Judges who were unaware of conditions were asked to assess the arguments furnished by the participants. The arguments given in the 5-min conditions were found also in the 20-min conditions, along with new arguments deemed acceptable by the judges. From “Quelques Limites des Reinterpretations Commodes des Effets de Dissonance [Some Limitations of Convenient Reinterpretations of Dissonance Affects],” by J.-L. Beauvois, R. Ghiglione, and R.-V. Joule, 1976, Bulletin de Psychologie, 29, p. 760. Copyright 1976 by Bulletin de Psychologie. Reprinted with permission.

As for the postexperimental attitude, the expected interaction between the two independent variables was significant and did indeed exhibit the predicted pattern: In the free-choice condition, the more time the participants had to find arguments, the further away their postexperimental attitudes were from the position being defended (i.e., less attitude change). The opposite pattern was obtained in the no-choice condition.

As in the Beauvois et al. (1976) study, the results observed in the free-choice condition support the hypothesis that producing more arguments reduces the dissonance generated by the behavior executed during the counterattitudinal role-playing. Note that these results are incompatible with self-perception theory: No self-perception view could possibly be used to derive the idea that participants are less likely to adhere to the position they are defending as the number of arguments they produce increases.

These results fit fully with the theory of ’57, with the arguments found being cognitions that implicate psychologically the counterattitudinal behavior and that therefore increase the denominator of the dissonance ratio (i.e., add consonant cognitions). They support the view that the behavioral cognition functions as the generative cognition in the evaluation of the total amount of dissonance. It is self-evident that taking the private attitude as the generative cognition would force us to the opposite expectation, with participants experiencing more dissonance, the more time they had to find arguments opposed to their private attitude. Results do not support this expectation. Moreover, results go against the widely accepted idea (see the classical presentations of dissonance theory: Feldman, 1966; and especially, Zajonc, 1968) that dissonance theory is a theory of consistency. It is indeed difficult to see how there could be a consistency effect in the fact that participants change their attitude even less when they come up with numerous arguments against it. Brehm and Cohen (1962) even judged dissonance theory to be ambiguous in this respect, claiming that another hypothesis opposing the one we have just set forth was just as compatible with the theory and, needless to say, more compatible with the consistency axiom (“the higher the quality of the participant’s arguments, the more dissonance there should be, assuming an initial disagreement with the advocated position,” p. 34). But as we have already seen, the inverse relationship between argument quantity and attitude change is not ambiguous from the standpoint of dissonance theory.


The principal reformulations of dissonance theory (Aronson, 1968, 1969; Cooper & Fazio, 1984; Wicklund & Brehm, 1976) have gradually turned the original theory into a cognitive-defense theory, describing the mechanisms used by responsible participants concerned about their own morality. These reformulations date back, as we noted previously, to the discovery in the sixties of factors that Kiesler (1971) considered to be the conditions for commitment (e.g., free choice, irrevocability, and consequences). We would now like to demonstrate that even if one agrees that commitment is necessary, it does not have to be interpreted in terms of the morally good self. Let us begin by showing that we can derive hypotheses from dissonance theory that are incompatible with this interpretation. Contrary to what the versions based on the morally good self and centered around the idea of lying would lead us to expect, dissonance can be increased by a perfectly moral act: telling the truth.

But first, let us analyze this situation used in Festinger and Carlsmith’s (1959) classic experiment. Participants were induced to perform two consecutive behaviors likely to generate dissonance. The first behavior consisted of accomplishing a particularly boring and mindless task. Thus, the cognition, “This task is boring” would imply the opposite psychologically of the cognition “I will do this work,” just as the cognition “I’m thirsty” implied the opposite psychologically of “I do not drink.” The second behavior consisted of telling a peer that the task was interesting (counterattitudinal advocacy). Joule and Girandola (1995) demonstrated that this situation would be regarded as one of double compliance (Joule, 1991), making it necessary to consider the relationship between the two behavioral cognitions. This relationship would be obviously consonant if we confined ourselves to dual relationships involving two cognitions only, as required by the 1957 definition of psychological implication and dissonance relations. Dissonance theory treats relationships two at a time, and psychological implication indeed only looks at relationships between two cognitions. Saying that a task is interesting goes quite well (is consonant) with having carried out that task. This is true regardless of one’s attitude toward the task, and the attitude constitutes a third cognition. In short, the actual accomplishment of the task provided Festinger and Carlsmith’s participants who said the task was interesting with a consonant cognition. They would thus experience less dissonance than participants who said the same thing but had not been required to carry out the task.

What happens when participants are led to tell the truth, namely, to say that the task is uninteresting? This time, the two acts are inconsistent with each other. The requested statement, which involves telling the truth, should thus increase the dissonance induced by task execution. In short, once they have executed the tedious task, participants having “lied” should experience less dissonance than participants having told the truth. These conjectures have been confirmed by the experiment conducted by Joule and Girandola (1995), described next (Experiment 2).

The participants, all volunteers, were 80 female literature students at the University of Provence, assigned to the four cells of a 2 × 2 design. They were asked to take part in unpaid research ostensibly concerning the effects of concentration on performance. Half had to accomplish a tedious task (turning knobs on a board for 13 min). The other half simply had the task described to them and were clearly told that they would not have to perform it (Independent Variable 1). Then all participants described the task to a peer, either positively (counter attitudinal role playing) or negatively (attitudinal role playing; Independent Variable 2), using arguments supplied by the experimenter (e.g., “It was very enjoyable” or “I had a lot of fun” vs. “It was tedious” or “I got bored”). Finally, the participants rated their attitude toward the task on an 11-point scale (dependent variable). This scale was strictly identical to the one used by Festinger and Carlsmith (1959; the scale that produced the most significant findings). The results are given in Table 3.2.

TABLE 3.2. Attitude Toward a Tedious Task


Positive presentation

Negative presentation

With Task



Without Task



Note. Entries are participant ratings of the task on a 11-point scale ranging from −5 (not at all interesting) to 5 (very interesting). In Control Situation 1 of simple compliance (task execution only), the participants’ mean rating of the task was −0.2. In Control Situation 2 of simple task assessment, the mean rating by the participants who neither executed nor presented the task was −2.1. From “Tâche Fastidieuse et Jeu de Rôle dans le Paradigme de la Double Soumission [Tedious Task and Role-Playing in the Double Compliance Paradigm],” by R.-V. Joule and F. Girandola, 1995, Revue Internationale de Psychologic Sociale, 8, p. 109. Copyright 1995 by Presses Universitaires de Grenoble. Reprinted with permission.

Festinger and Carlsmith’s (1959) main finding was replicated: Participants who had to accomplish and positively present the task had a better attitude toward it than Control Group 2 participants, who only had to rate the task (−0.5 vs. −2.1). But what these results showed above all was that participants who performed the tedious task found it more interesting after telling a peer it was boring than after telling a peer it was interesting (1.5 vs. −0.5). In addition, participants who performed the task and negatively described it found it more interesting than Control Condition 1 participants, who only had to perform the task (1.5 vs. −0.2). This last result was absolutely incompatible with a self-perception view of dissonance phenomena.

Thus, the immorality of the “lie” was not behind the dissonance experienced by the Festinger and Carlsmith (1959) participants, nor was the fact that they had tricked their peer. Joule and Girandola’s (1995) results showed that participants would have felt even more dissonance, had the researchers asked them to tell the truth. Thus, counterattitudinal advocacy does not evoke dissonance because it is immoral but, from a theoretical perspective, because there are some cognitions that would have implied doing the opposite. From this same theoretical perspective, it is possible that the same would apply to proattitudinal advocacy. We have just seen, for example, that having previously performed the task sufficed for the participant to experience more dissonance for telling the truth than for telling a lie. This double-forced compliance effect has been replicated and extended in other research (Joule & Azdia, 2003).

This conclusion prompts us to think about the role played by the consequences of the act. Indeed, it is probably because they shared this somewhat moral interpretation of dissonance that certain researchers (in particular, Cooper & Fazio, 1984) insisted so strongly on the importance of what they called the act’s aversive consequences (here, tricking a peer), to the point of making it the core of their “new look.” Granted, we are not questioning the necessity of the commitment, and even less so the importance of the act’s consequences as a commitment factor. Nevertheless, having shown in this situation that telling the truth to a peer (and in doing so, suggesting that he or she not agree to perform the tedious task) leads to an increase in dissonance compels us to reconsider the theoretical role sometimes ascribed to these consequences. As we have just seen (see also Harmon-Jones, Brehm, Greenberg, Simon, & Nelson, 1996), they need not be morally aversive at all for dissonance to be generated. Note once again that we have adhered strictly to the stipulations of the theory of ’57.


Since Brehm and Cohen (1962), dissonance theorists accepted the idea that the simple presence of relations of inconsistency among cognitions is not a sufficient condition for the arousal of dissonance. In fact, the 1960s were to prove a rich source of experiments showing that the primary reward effect of dissonance in forced-compliance situations is observed only if the participants are allowed to choose whether to perform the requested act. When participants do not have this choice, the dissonance effect is replaced by a reinforcement effect (Holms & Strickland, 1970; Linder, Cooper, & Jones, 1967). At the same time that the critical importance of free choice was being demonstrated, researchers revealed the significance of other cognitions relating to the public or anonymous nature of the problematic behavior (Carlsmith, Collins, & Helmreich, 1966), the irrevocable or reversible nature of the act (Helmreich & Collins, 1968), and, above all, the role of cognitions concerning its consequences (Calder, Ross, & Insko, 1973; Cooper & Worchel, 1970). Like free choice, such cognitions seem to function as more or less necessary conditions for dissonance arousal. In our view, the idea of commitment, in the form proposed by Kiesler (1971), provides the best conceptual synthesis of this rich research tradition and allows us to hypothesize that the induction of a state of dissonance in a forced-compliance situation requires the participant’s commitment to the problematic behavior. In fact, Kiesler treated all the cognitions that have been studied as preconditions for dissonance effects (e.g., free choice, public nature of the behavior, irrevocability, and consequences) as variables affecting a person’s level of commitment. Since 1971, a number of dissonance theorists have placed at least two of the conditions of commitment (free choice and consequences) at the heart of their formulations. To the extent that they thought they had discovered the psychological reason for the arousal of dissonance in these conditions of commitment, they were led to revise Festinger’s original theory, turning it into a theory of personal responsibility (Wicklund & Brehm, 1976) or a theory of the cognitive management of the consequences of behavior (Cooper & Fazio, 1984).

It is clear that for these researchers commitment to one’s behavior is a key element of the revised theory. Was it really necessary to change the basic assumptions of the original theory to introduce commitment? Before addressing this question, we clarify the meaning of the most important factor of commitment: free choice.

Indeed, the traditional way to manipulate free choice (i.e., “You are entirely free to do or not to do what I ask you. It’s up to you”) authorizes two interpretations with very different theoretical implications. Free choice can mean that the participant agrees (or, in a few cases, refuses) to execute the particular act requested of him or her, such as writing an essay in favor of police intervention, refraining from eating or drinking, stopping smoking temporarily, or saying that a task is interesting. No one would deny that this is the traditional interpretation. In fact, this is the interpretation that allows for understanding the effects of dissonance in terms of self-perception. Yet manipulating free choice can be interpreted to mean something else, namely, that the participant agrees (or, in a few cases, refuses) to comply with the obedience relationship proposed by an experimenter in a research framework. It is obvious that this is not the same choice as above. Replying “No, I don’t want to do that particular thing, but you can ask me to do some other thing” (first interpretation) is not the same as replying “I have no reason to comply with your demands” (second interpretation). In the first case, accepting means agreeing to perform a specific act (and, thus, to be held responsible for that act), as several post-Festinger theorists assume. In the second case, which is closer to our idea of what a forced compliance contract is (Beauvois & Joule, 1996, pp. 146—154), accepting means being willing to comply with the experimenter, in which case the experimenter can be held responsible for whatever happens. The following description of Experiment 3 (Beauvois, Bungert, & Mariette, 1995) shows that the commitment necessary to induce a state of dissonance corresponds to the second interpretation of free choice (for a more recent discussion of commitment, see Joule, Girandola, & Bernard, 2007).

Three independent variables were manipulated in a 2 × 2 × 2 design. In every case, participants were asked in the end to write a counterattitudinal essay, for which they did or did not choose the topic (Independent Variable 1: commitment to act). Before writing the essay, they either chose to or were assigned to (Independent Variable 2: compliance commitment) an experimental situation in which they had to perform problematic behaviors. The third independent variable dealt with the problematic behaviors announced to participants before the manipulation of the second independent variable. For the first half of the participants, the stated problematic behavior was writing counterattitudinal essays, like that which they would genuinely have to write (paradigmatic-sequence condition). For the second half of the participants, these problematic behaviors consisted of doing tedious tasks (which they would never be asked to perform in reality: nonparadigmatic-sequence condition). Further details with regard to this third independent variable are described in the following paragraphs.

In the paradigmatic-sequence condition, the participants were either given the choice or not (Independent Variable 2) to take part in a counterattitudinal role-playing situation. Before they made their decision or, in the no-choice condition, before they were told to do the behavior, Beauvois et al. (1995) presented the participants with the topics (all counterattitudinal) that they might be asked to write about (i.e., drivers under 16 years old no longer allowed to drive when accompanied by an adult, shorter holidays, or scholarships to be limited to students achieving an average grade of 14/20, approximately a “B” in the American system, in the preceding year). In the commitment-to-compliance condition, the experimenter said that the participants were entirely free to agree or to refuse to take part in the research and asked them to make their own decision. However, they were not told which of these essays they would ultimately have to produce but that this would be decided later. In contrast, in the other condition (no commitment to compliance), the experimenter stated that the research had been requested by the government and made no mention of the possibility of refusing to take part. Independent Variable 1, commitment to the counterattitudinal act, was then manipulated. In the commitment condition, the experimenter reformulated the three topics and asked the participant which one he or she would like to write about. In the no-commitment condition, the experimenter told the participant to write about “today’s topic,” which were topics that matched those chosen by the participants in the commitment to the counterattitudinal act condition.

In the no-paradigmatic-sequence condition (Independent Variable 3), the participants were either given the choice or not, to take part in a tedious task experiment. Before manipulating the commitment to compliance (statement of participant’s free choice to participate in the experiment), the experimenter gave examples of problematic behaviors that were very different from the ones the participants would actually have to accomplish (e.g., glue a piece of confetti on every occurrence of the letter a in a long text, copy three pages of the telephone book, or take a lengthy test involving crossing out symbols). The commitment-to-compliance manipulation was the same as the paradigmatic-sequence condition. In the commitment-to-compliance condition, the experimenter told the participants that they were entirely free to agree or to refuse to take part in the research. Then, once commitment to compliance was manipulated, the experimenter claimed she had made a mistake and told the participants that the task would be something completely different from what had just been said. Without repeating the statement that participants were free to participate, she went on to the counter-attitudinal essays paradigm and manipulated the chosen or nonchosen topic variable. In summary, all participants wrote a counterattitudinal essay for which they had chosen or not chosen the topic (commitment to act vs. no commitment to act). Before this, they either could choose to accept or were assigned to a situation (commitment to compliance vs. no commitment to compliance), with or without the knowledge of what problematic act this situation really involved (paradigmatic sequence vs. nonparadigmatic sequence). Finally, after the completion of the essays, a postexperimental questionnaire was used to assess three attitudes (vacations, driving, and scholarships).

Let us consider first the paradigmatic-sequence condition. As shown in Table 3.3, participants who changed attitudes the most, and thereby reduced dissonance the most, were the ones who had been given the choice to enter or not, into the situation (commitment to compliance), but who were required to write about the topic of the day. This result has important theoretical implications.

TABLE 3.3. Commitment to Compliance and Free Choice of the Problematic Act in a Forced-Compliance Situation

Commitment to compliance

Paradigmatic sequence

Nonparadigmatic sequence


No choice


No choice






No choice





Note. The higher the number, the more closely the participant’s attitude conforms to the position defended in the selected essay (dissonance effect). From “Forced Compliance: Commitment to Compliance and Commitment to Activity,” by J.-L. Beauvois, M. Bungert, and P. Mariette, 1995, European Journal of Social Psychology, 25, p. 24. Copyright 1995 by John Wiley and Sons. Reprinted with permission.

If the traditional free choice were indeed the choice to execute or not to execute the specific behavior requested, then the choice of one particular act among three in Experiment 3 would indeed be the best approximation of this type of free choice. Yet being able to choose one of three acts did not generate dissonance in this case. On the contrary, it seemed to have reduced it when the participant was committed to compliance, and this is what we expected.4 Imagine a participant who has just agreed to comply with the experimenter, knowing that he or she will have to execute a counterattitudinal behavior but not knowing which. Once this commitment to compliance is obtained, the participant is given the choice between various obviously counterattitudinal behaviors. In such a decision-making situation, being able to choose must reduce the dissonance ratio, compared with a classical forced-compliance situation, in which the participant is proposed one and only one act. Indeed, insofar as the chosen alternative is the least discomforting for the participant (the lesser of three evils), it must reduce the total amount of dissonance by creating consistent cognitions (the higher cost of the nonselected alternatives psychologically implies choosing the selected one). The fact that having chosen one out of three acts in Experiment 3 reduced dissonance is thus perfectly compatible with the theory of ’57.

The most important point in this experiment is that commitment to compliance must be considered as the factor that aroused the dissonance in the commitment-to-compliance-nonchosen-issue situation. These results were confirmed in the nonparadigmatic-sequence condition. We discovered that participants who committed to compliance by agreeing to perform boring tasks modified their attitude in favor of the essay topic when it was imposed. Here again, the choice of one of three counterattitudinal acts did not induce dissonance per se. However, in the commitment-to-compliance condition, being able to choose a topic does not seem to have reduced the dissonance (as in the paradigmatic condition). This is probably because in the nonparadigmatic condition, the participants had been told nothing about the two nonchosen topics when they had to pick one of three, so they had not already implicitly agreed to write about them (during the commitment-to-compliance manipulation). In effect, avoiding two topics one has never heard about and therefore never agreed to write about does not provide any consonant cognitions, unlike the case in which participants could choose to write about the least problematic of three topics.

In opposition to the traditional understanding of free choice, it is indeed commitment to compliance, not commitment to a particular counterattitudinal act, that is the condition needed to induce a state of dissonance. The results of Experiment 3 point out the limitations of a view of the dissonance-reduction process that reduces it to the management of responsibility. In our minds, revisions based on responsibility, the anticipation of aversive consequences, and other similar concepts stem from a faulty interpretation of what free choice really is in classic forced-compliance experiments. Of course, it is quite understandable that participants who experience the feelings involved in having chosen to perform the particular counterattitudinal act just performed have some problems about their own values and have trouble accepting that the act has aversive consequences. But these feelings have nothing to do with the state of dissonance. Note first of all, against this view, that dissonance is induced in a number of situations void of moral implications. Such is the case in situations of abstinence from smoking, for example, or in situations involving eating an unappetizing dish. Note also, and still opposing this view, that the traditional free-choice effects are not incompatible with very slight although real differences in the feeling of freedom between participants who had free choice and those given no choice: Either they all globally experience a strong feeling of constraint (Steiner, 1980), or on the contrary, they all experience a strong feeling of freedom (Beauvois, Michel, Py, Rainis, & Somat, 1996). In the studies mentioned by Steiner (1980), as well as in those described by Beauvois et al. (1996), dissonance effects are observed only in participants assigned to a free choice condition, whether they experience a strong feeling of freedom (as in Beauvois et al.) or, on the contrary, a strong feeling of obligation (as reported by Steiner).

Furthermore, for us, the key element for dissonance arousal is not the feeling of freedom, but rather whether the participant is said to have choice. In any case, the results of Experiment 3, described here, and those of another experiment described by Beauvois et al. (1995), more clearly pinpoint the limits of personal responsibility in a “morally good self” and the ideological confusion this view implies: Are participants morally responsible when they accept a condition of obedience and perform an imposed, unexpected act, which they do (no refusals observed) simply because they accept their state of compliance with the experimenter? Even if we obviously have to answer no to this question, the results show that participants nevertheless experience cognitive dissonance right from the very moment they are told they are free to comply or not to comply.


In our minds, no theory other than dissonance theory can make sense out of all of the effects presented here. This also applies to the theories devised by Festinger’s critics (e.g., self-perception and impression management) and even to the revised versions proposed by his followers. Yet if dissonance theory indeed remains the only theory that can make sense out of these effects, then it is not just one of many theories of dissonance. Does the radical theory really diverge that far from the theory of ’57? The answer to this question is no to the extent that the radical theory conserves the key element of this theory of ’57. In the remainder of this section, we discuss what is necessary for it to do so.

Strictly calculating the dissonance ratio and accepting its implications. The use of this ratio has several repercussions: Insofar as the state of dissonance is calculated from relationships (both dissonant and consonant) between cognitions, the calculation requires making the important theoretical distinction between the state of dissonance and the presence (vs. absence) of dissonant relationships between cognitions. The fact that Festinger used the word dissonance to refer to both of these instances may have led to the assumption that a state of dissonance exists whenever dissonant relationships between cognitions exist.

Radically calculating the total amount of dissonance means recognizing that all cognitions do not have the same status. To calculate, one of the cognitions must be designated as the generative cognition. This cognition allows us to say that the other cognitions (the ones we put in the numerator or denominator of the dissonance ratio) are consistent (whenever there is psychological implication) or inconsistent (whenever there is implication of the obverse of the generating cognition). The other cognitions enter into play only as a result of potential psychological implications that link them to the generating cognition or to its obverse.

Radically calculating the dissonance ratio also means only considering those relationships involving the generative cognition. Indeed, relationships are included in this calculation only to the extent that they link the generative cognition (or its obverse) to other cognitions. This implies that certain relationships, and in particular those between the cognitions in the numerator or denominator of the dissonance ratio, are not part of the calculation of the total amount of dissonance. For instance, anyone would agree that Festinger and Carlsmith (1959) were correct in ignoring the potential relationship between personal attitude and reward, that is, between two cognitions included in the dissonance ratio because of the relationships (inconsistency for the former and consistency for the latter) they had with the generative cognition (the counterattitudinal behavior). Note that in the present case, the relationship between these two cognitions is irrelevant. There may exist cases however, in which there is a relevant relationship between two cognitions in the dissonance ratio. If so, should it also be ignored in the calculation of total dissonance? The answer is yes. We have even seen that total dissonance could be decreased by the generation of new inconsistencies between the cognitions in the dissonance ratio. In Experiment 1, for instance, the counterattitudinal advocacy that led to the production of cognitions that were inconsistent with the participant’s personal attitude was accompanied by less overall dissonance, because the total amount of dissonance was reduced by those cognitions.

Finally, radically calculating the total dissonance calculation implies considering only those relationships that link two cognitions (two-term relationships). Coming back to the role of the reward in the Festinger and Carlsmith (1959) experiment we can agree once again—because the findings support their reasoning—that Festinger and Carlsmith were quite right to call the relationship between the reward and the counterattitudinal behavior consistent. Yet it would suffice to bring a third cognition into the picture, personal attitude, for the relationship between the reward and the behavior to change in nature, because the participants might think that they were being bribed by the experimenter. In this case, the reward would become outright immoral, even aversive, and would generate dissonance. This type of reasoning is invalid, because the findings clearly show that rewards reduce dissonance. So why, then, would this reasoning become valid when we looked at other possible cases of relationships among three cognitions, in particular, when the consequences of the act were at stake? Indeed, the very idea of aversive consequences (Cooper & Fazio, 1984), presumably responsible for dissonance, relies on the consideration of three-term relationships (the relationship between one’s act and its consequences being modified by one’s personal attitude).

Granting a particular status to commitment cognitions. Discovery of commitment factors had a strong impact on the evolution of dissonance theory. Was it really necessary to change it? First of all, if we limit ourselves to two-term relationships, which are the only ones defined in the theory, then commitment cognitions (e.g., “I was told I was free to accept or refuse,” “What I do will have such and such a consequence,” and “I will not be able to go back on my word”) do not really fall within the scope of the theory of ’57. Such cognitions are obviously relevant, because they condition the dissonance-reduction process. They are definitely consistent with the act. Indeed, it makes no sense to contend that knowing one is free to do something (psychologically) implies that one does not do it. Commitment cognitions thus pose a real theoretical problem: Although relevant, they are not inconsistent with the act, even though their presence is necessary to induce a state of dissonance. This is the reason why theorists quickly veered away from the theoretical constraints of the 1957 version of dissonance theory, especially those involving the calculation of the dissonance ratio. They began to reason in a very flexible fashion by intuitively ascertaining a state of dissonance that participants experience as they engage in a sort of reasoning based on three, if not four, cognitions (“I say x, but I am against x, and yet I was free not to say x” or “I say x, but I am against x, and it is even worse because my act is going to have such and such a consequence”).

We think there is an alternative, one that is in keeping with the experimental practices and data. This alternative fits into one proposition: Commitment to an act is a necessary condition (but insufficient, because the act must also be discomforting, i.e., counterattitudinal or countermotivational) for the induction of a state of dissonance. This proposition has three implications. The first goes without saying: The theory of ’57 can remain unchanged. The second is that it forces us to carefully examine this mandatory dissonance-inducing commitment. Reflection about this problem should give rise to a second branch of a more complete theory (see Experiment 3 above). The third is that dissonance theory is a local theory of the psychology of commitment; more precisely, it is a theory of the effects of commitment to a problematic act. From an experimental point of view, one can imagine that there is no arousal below a certain threshold of commitment and that experimental operationalizations (declaration of free choice or the salience of consequences) ordinarily allow us to attain this threshold. This proposition allows us to distinguish two types of cognitions: cognitions generating arousal (commitment cognitions) and cognitions composing the total amount of dissonance (specific cognitions). This perspective is justified first because commitment cognitions are not consistent or inconsistent cognitions and so cannot be located in the dissonance ratio (e.g., “I was free to accept or to refuse”). It is justified also by the fact that some cognitions facilitate commitment, therefore arousal, even though they reduce the total amount of dissonance. This is the case of internal explanations. These explanations increase commitment because they reinforce the link between the individual and the behavioral act. But because they psychologically imply the act, they are consistent cognitions, which reduce the total amount of dissonance (see Beauvois et al., 1996, for experimental support).

Thus, on a formal level we can distinguish commitment cognitions, which define the committing character of the generative cognition, from specific cognitions. Both types of cognitions are represented in the dissonance ratio on either the right or the left side of the equal sign in the formula:

Dgb(F, C,…) = A/A + R + …

where D is the total amount of dissonance induced by the generative cognition g relative to the behavior b and F (free choice) and C (consequences) are commitment cognitions without which there would be no dissonance arousal. Their status comes from the theory of commitment. A (attitude) and R (reward) are specific cognitions that appear in the numerator and denominator of the dissonance ratio. These cognitions allow for the quantification of the state of dissonance.


By focusing on the generative cognition as a behavior-related cognition and on the rationalization of behavior, the radical theory has made it possible to explore two new paradigms (for a review, see Beauvois & Joule, 1996): double forced compliance (see Experiment 2 above), and act rationalization.

In the double forced compliance paradigm, we are interested in the dissonance-reduction process after the execution of two behaviors, at least one of which is discomforting. Let us consider a participant who has produced two behaviors, B1 and B2. Naturally, if this participant is to experience a state of dissonance, it is necessary and sufficient for one of these behaviors to contradict the participant’s attitudes or motivations A. Let us suppose that the discrepant behavior is B. If we reason on the basis of behavior B1 alone, then we will consider the participant to be in a classic forced-compliance situation. However, if we also consider behavior B2, we must take account not only the relations between this new behavior and the attitudes or motivations A of the participant (inconsistent, consistent, or neutral relations) but also the relations that exist between behavior B2 and behavior B1 (inconsistent, consistent, or neutral relations).

First of all, let us consider only the relevant relations that may exist among B1, B2, and A. If we assume the inconsistent relation between B1 and A to be constant, then logic indicates the presence of four possible cases, as shown in Figure 3.1.

FIGURE 3.1. Double Compliance Situations: Relevant Relationships


Note: — = inconsistent relationship; + = consistent relationship.

Let us now consider both the relevant and irrelevant relations among B1, B2, and A. If we once again assume the relation of inconsistency between B1 and A to be constant, then logic this time presents us with 9 possible cases, namely, the 4 cases presented in Figure 3.1 and the 5 cases presented in Figure 3.2.

FIGURE 3.2. Double Compliance Situations: Relevant and Irrelevant Relationships


Note: — = inconsistent relationship; + = consistent relationship.

Of these theoretically possible cases, two have been studied in particular detail: Cases I and 2. These are easier to imagine than the others, possibly because they correspond to balanced triads. Festinger and Carlsmith’s (1959) situation implemented Case 1 (B1: perform a boring task; B2: lie). Joule and Girandola’s (1995) “truth situation” (Experiment 2) implemented Case 2. In a study by Joule (1991), the first behavior (B1), produced by smokers, consisted of refraining from smoking for an evening (B1). The second behavior consisted of writing an essay either against (B2: Case 1) or in favor (B2: Case 2) of smoking. As expected on the bases of radical considerations, the participants who had to write an essay against tobacco after being induced to refrain from smoking found it most difficult to go without tobacco and felt the greatest need for tobacco.

In the act rationalization paradigm, we are interested in the conditions that are likely to lead a participant who has just carried out a discomforting act to rationalize that act by carrying out another discomforting act. The dissonance state is one of tension, which as such must be reduced. One can imagine that the participant will adopt the most convenient route or the first one available. But whatever the case may be, the arousal can only be reduced by two principal kinds of processes. These processes serve to make the act less problematic. Some processes affect the commitment to the act and reduce this commitment, and other processes reduce the total amount of dissonance by changing one or several existing cognitions or by producing new ones.

In ordinary circumstances, the behavior cannot be denied, and the easiest route available is the change of the private attitude that would have implied the contrary. This is most likely what happens in daily life, and it is natural that it was the first studied and remains the most traditional route of study. But if the participant is given the chance to trivialize his behavior and therefore make it less or even not at all problematic (to the point of no longer having commitment: a great distance between the participant and his act), it is probable that he or she will resort to this route (Simon, Greenberg, & Brehm, 1995). This route is probably less spontaneous than the preceding one (except in the case of a very polarized or salient initial attitude) because in daily fife people are reluctant to say to themselves or others that they engage in trivial behaviors. But the participant can also reduce this tension by any other route that becomes available.

We have more specifically studied act rationalization. In fact, alongside the classic forms of rationalization (attitude change and trivialization), which we might term cognitive rationalization, we have been persuaded of the possibility of another form of rationalization. We have given the name act rationalization to this new type of rationalization because the problematic behavior that underlies the generative cognition is rationalized (and thereby rendered less problematic) by the production of a new, more problematic behavior rather than by cognitive realignment, which has been the classic object of investigation by dissonance theorists.

The hypothesis of the alternative nature of the cognitive path and behavioral path of dissonance reduction has received substantial support (Beauvois, Joule, & Brunetti, 1993; Joule, 1996): Act rationalization is hindered when cognitive rationalization is promoted, and in contrast, it is promoted when cognitive rationalization is blocked. For example (Beauvois et al., 1993, Experiment 1), smokers who had just accepted a first period of abstinence (18 hr without smoking) were proposed a second, much longer abstinence period (6 days without smoking), which was the final request. Before making this final request, the experimenter asked half of the participants to write down the reasons that led them to accept the first abstinence period; nothing of this sort was asked of the other half. Thus, we gave the first group, but not the second, the chance to cognitively rationalize their first abstinence. Fewer of the participants who had the chance to rationalize the first abstinence (26.1%) accepted the final request than of those who did not have this rationalization opportunity (82.6%).

Since the discovery of the conditions necessary for inducing a state of dissonance, dissonance theory has greatly evolved, to the point where previously rival theories are now considered as ways of thinking about dissonance reduction. We do not want to imply here that this evolution has been infertile or that it has not produced substantial data. It remains the case, however, that this evolution has led to the neglect of essential points of Festinger’s theory and that this theory can still teach us a lot. And who knows, what it still has to teach us may contain many new surprises.


Aronson, E. (1968). Dissonance theory: Progress and problems. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, & P. H. Tannenbaum (Eds.), Theories of cognitive consistency: A sourcebook. Chicago, IL: Rand McNally.

Aronson, E. (1969). The theory of cognitive dissonance: A current perspective. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 4, pp. 1—34). New York, NY: Academic Press. http://dx.doi.org/10.1016/S0065-2601(08)60075-1

Beauvois, J.-L., Bungert, M., & Mariette, P. (1995). Forced compliance: Commitment to compliance and commitment to activity. European Journal of Social Psychology, 25, 17—26. http://dx.doi.org/10.1002/ejsp.2420250103

Beauvois, J.-L., Ghiglione, R., & Joule, R.-V. (1976). Quelques limites des reinterpretations commodes des effets de dissonance [Some limitations of convenient reinterpretations of dissonance affects]. Bulletin de Psychologie, 29, 758—765.

Beauvois, J.-L., & Joule, R.-V. (1996). A radical dissonance theory. London, England: Taylor & Francis.

Beauvois, J.-L., Joule, R.-V., & Brunetti, F. (1993). Cognitive rationalization and act rationalization in an escalation of commitment. Basic and Applied Social Psychology, 14, 1—17. http://dx.doi.org/10.1207/s15324834basp1401_1

Beauvois, J.-L., Michel, S., Py, J., Rainis, N., & Somat, A. (1996). Activation d’explications internes et externes du comportement problematique dans une situation de soumission forcee [Activation of internal and external explanations of problematic behavior in a forced compliance situation]. In J.-L. Beauvois, R.-V. Joule, & J.-M. Monteil (Eds.), Perspectives cognitives et conduites sociales: 5. Contextes et contextes sociaux [Cognitive Perspectives and social conduct: 5. Contexts and social contexts]. Neuchatel, Switzerland: Delachaux et Niestle.

Brehm, J. W., & Cohen, A. R. (1962). Explorations in cognitive dissonance. Hoboken, NJ: Wiley. http://dx.doi.org/10.1037/11622-000

Calder, B. J., Ross, M., & Insko, C. A. (1973). Attitude change and attitude attribution: Effects of incentive, choice, and consequences. Journal of Personality and Social Psychology, 25, 84—99. http://dx.doi.org/10.1037/h0034264

Carlsmith, J. M., Collins, B. E., & Helmreich, R. L. (1966). Studies in forced compliance: I. The effect of pressure for compliance on attitude change produced by face-to-face role playing and anonymous essay writing. Journal of Personality and Social Psychology, 4, 1—13. http://dx.doi.org/10.1037/h0023507

Cartwright, D., & Harary, F. (1956). Structural balance: A generalization of Heider’s theory. Psychological Review, 63, 277—293. http://dx.doi.org/10.1037/h0046049

Cooper, J., & Fazio, R. H. (1984). A new look at dissonance theory. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 17, pp. 229—266). New York, NY: Academic Press.

Cooper, J., & Worchel, S. (1970). Role of undesired consequences in arousing cognitive dissonance. Journal of Personality and Social Psychology, 16, 199—206. http://dx.doi.org/10.1037/h0029830

Feldman, S. (1966). Motivational aspects of attitudinal elements and their place in cognitive interaction. In S. Feldman (Ed.), Cognitive consistency (pp. 75—108). New York, NY: Academic Press. http://dx.doi.org/10.1016/B978-1-4832-2828-0.50008-8

Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.

Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 58, 203—210. http://dx.doi.org/10.1037/h0041593

Harmon-Jones, E., Brehm, J. W., Greenberg, J., Simon, L., & Nelson, D. E. (1996). Evidence that the production of aversive consequences is not necessary to create cognitive dissonance. Journal of Personality and Social Psychology, 70, 5—16. http://dx.doi.org/10.1037/0022-3514.70.1.5

Heider, F. (1958). The psychology of interpersonal relations. Hoboken, NJ: Wiley. http://dx.doi.org/10.1037/10628-000

Helmreich, R., & Collins, B. E. (1968). Studies in forced compliance: Commitment and magnitude of inducement to comply as determinants of opinion change. Journal of Personality and Social Psychology, 10, 75—81. http://dx.doi.org/10.1037/h0026282

Holms, J. G., & Strickland, L. H. (1970). Choice freedom and confirmation of incentive expectancy as determinants of attitude change. Journal of Personality and Social Psychology, 14, 39—45. http://dx.doi.org/10.1037/h0028617

Joule, R.-V. (1991). Practicing and arguing for abstinence from smoking: A test of the double forced compliance paradigm. European Journal of Social Psychology, 21, 119—129. http://dx.doi.org/10.1002/ejsp.2420210203

Joule, R.-V. (1996). Une nouvelle voie de reduction de la dissonance: La rationalisation en acte [A new model for dissonance reduction: Act rationalization]. In J.-L. Beauvois, R.-V. Joule, & J.-M. Monteil (Eds.), Perspectives cognitives et conduites sociales: 5. Contextes et contextes sociaux [Cognitive Perspectives and social conduct: 5. Contexts and social contexts; pp. 293—307]. Neuchatel, Switzerland: Delachaux et Niestle.

Joule, R.-V., & Azdia, T. (2003). Cognitive dissonance, double forced compliance, and commitment. European Journal of Social Psychology, 33, 565—571. http://dx.doi.org/10.1002/ejsp.165

Joule, R.-V., & Beauvois, J. L. (1998). La soumission librement consentie [Freely granted submission]. Paris, France: Presses Universitaires.

Joule, R.-V., & Girandola, F. (1995). Tâche fastidieuse et jeu de rôle dans le paradigme de la double soumission [Tedious task and role-playing in the double compliance paradigm]. Revue Internationale de Psychologie Sociale, 8, 101—116.

Joule, R.-V., Girandola, F., & Bernard, F. (2007). How can people be induced to willingly change their behavior? The path from persuasive communication to binding communication. Social and Personality Psychology Compass, 1, 493—505. http://dx.doi.org/10.1111/j.1751-9004.2007.00018.x

Joule, R.-V., & Levèque, L. (1993). Le changement d’attitude comme fonction du temps d’argumentation [Attitude change as a function of argumentation length]. Unpublished manuscript, Provence University, Aix-en-Provence, France.

Kiesler, C. A. (1971). The psychology of commitment: Experiments linking behavior to belief. New York, NY: Academic Press.

Linder, D. E., Cooper, J., & Jones, E. E. (1967). Decision freedom as a determinant of the role of incentive magnitude in attitude change. Journal of Personality and Social Psychology, 6, 245—254. http://dx.doi.org/10.1037/h0021220

Rabbie, J. M., Brehm, J. W., & Cohen, A. R. (1959). Verbalization and reactions to cognitive dissonance. Journal of Personality, 27, 407—417. http://dx.doi.org/10.1111/j.1467-6494.1959.tb02363.x

Simon, L., Greenberg, J., & Brehm, J. (1995). Trivialization: The forgotten mode of dissonance reduction. Journal of Personality and Social Psychology, 68, 247—260. http://dx.doi.org/10.1037/0022-3514.68.2.247

Steiner, I. D. (1980). Attribution of choice. In M. Fishbein (Ed.), Progress in social psychology (pp. 1—47). Mahwah, NJ: Lawrence Erlbaum.

Wicklund, R. A., & Brehm, J. W. (1976). Perspectives on cognitive dissonance. New York, NY: Wiley.

Zajonc, R. B. (1968). Cognitive theories in social psychology. In G. Lindzey & E. Aronson (Eds.), Handbook of social psychology (Vol. 1, pp. 320—411). Reading, MA: Addison-Wesley.


This chapter is reprinted from the first edition of this book. The editor added some citations to relevant studies that have occurred since the original publication.

1For example, in establishing a dissonance ratio and putting the relevant, considered perceptions in the numerator or denominator, one can consider only the psychological implications of two perceptions at a time.

2Further, this course was more or less suggested by Brehm and Cohen (1962), who were the first to introduce the idea of commitment. Similar to Kiesler (1971), we think that the concept of commitment was a throwaway construct for dissonance theorists.

3Experimental practice is a good criterion. One needs a theory to explain experimental practice and experimental results.

4The interaction between these two independent variables was statistically significant in the paradigmatic-sequence condition.