Contrary to the assumptions of normative economic theory, choice preferences are not only driven by our attitudes but also modulated by the experience of previous choices. Brehm’s study (1956) suggested that, after choosing between two similarly attractive options, individuals no longer perceive these options as similar, evaluating the chosen option more positively and devaluating the unchosen option. The devaluation of the rejected option has been repeatedly demonstrated in studies using different versions of “free choice paradigm” (Kitayama et al., 2004; Izuma et al., 2010; Mengarelli et al., 2015; Colosio et al., 2017).

According to the prominent theory of cognitive dissonance (Festinger, 1957), “difficult” choices between similarly appealing options that require the rejection of one of them result in a cognitive discrepancy between dissonant cognitions such as preferences and actions. This discrepancy arises from the need to act in a manner that contradicts one’s preferences and attitudes towards the highly favored option and may subsequently lead to a sense of psychological discomfort, also known as dissonance (Harmon-Jones and Mills, 2019). This discomfort motivates individuals to be consistent with their actions and reduce the dissonance by either devaluing the rejected option or increasing their evaluation for selected one. Thus, the mere act of choosing between similarly preferred options affects individual preferences.

In a typical “free choice paradigm,” participants are asked to rate a set of goods according to their preference (preference task I). Next, they select between two of the items that had similar preference ratings in the first rating task (choice task). Finally, participants are asked to re-rate the original set of goods for the second time (preference task II). According to the theory of cognitive dissonance, after making a difficult choice between two equally preferred items, participants’ preference, guided by the need to resolve conflict, can decrease for the rejected item and increase for the chosen ones. The resulting difference between the assessment of the items in preference task II and preference task I could represent the observable resolution of cognitive dissonance: spreading of alternatives or choice-induced preference changes. Alteration of preference was observed in plenty of studies, either preference devaluation for rejected options (for example, Izuma and Murayama, 2013; Salti et al., 2014; Colosio et al., 2017) or an increase in evaluation for selected ones (for example, Sharot et al., 2010; Izuma and Murayama, 2013).

Importantly, the free-choice paradigm can produce artificial preference changes (see Chen and Risen, 2010; Izuma and Murayama, 2013; Enisman et al., 2021 for a review). Chen and Risen (2010) showed that measured alteration in preference when making a difficult decision may not necessarily be associated with the choice itself; rather, it may be a result of the artifact, while choice and repeated evaluation merely uncover already existing preferences. For example, preference for option 1, measured by rating or ranking, can slightly exceed preference for option 2, although the ratings were equal during preference task I. Therefore, it is likely that in preference task II, preference rating for option 1 will continue to get even higher, producing ostensible changes of preference. To counter this drawback of the “free-choice paradigm,” various control conditions and task modifications have been suggested and investigated (Chen and Risen, 2010; Izuma and Murayama, 2013; Enisman et al., 2021). The use of the brain stimulation approach also may overcome the limitations of the free-choice paradigm. Alterations in preference in making difficult conflictual decisions under region-specific brain stimulation may be attributed to the suppression or enhancement of the neuronal activity in the region responsible for conflict monitoring and resolution, and cannot be attributed to a statistical artifact. Therefore, the substantial effect of well-controlled brain stimulation on the following conflictual decision spreading of alternatives is likely attributable solely to the modulation of neural mechanisms underlying choice-induced preference changes.

Despite the significant progress in studying cognitive dissonance, neurocognitive mechanisms of preference alteration in decision making are still not fully understood. Several studies consistently indicated the involvement of the pMFC (van Veen et al., 2009; Izuma et al., 2010), posterior cingulate cortex (Kitayama et al., 2013; Tompson et al., 2016), dorsolateral prefrontal cortex (Harmon-Jones et al., 2008; Flavia Mengarelli et al., 2015), and nucleus accumbens (Izuma et al., 2010; Kitayama et al., 2013) to post-decisional preference changes. The involvement of other brain regions was not replicated. It is likely these brain regions form a network responsible for detection dissonance and its subsequent resolution (Colosio et al., 2018; Voigt, 2022).

A growing number of studies indicate the critical role of the posterior medial frontal cortex (pMFC) in cognitive dissonance and preference re-evaluation (van Veen et al., 2009; Izuma et al., 2010; Colosio et al., 2017; Voigt et al., 2019; Tandetnik et al., 2021). This part of the brain largely consists of the pre-supplementary motor area (pre-SMA), the dorsal medial frontal cortex (dmPFC), ventral medial frontal cortex (Voigt et al., 2019), and the dorsal anterior cingulate cortex (dACC) (Izuma, 2013; Tandetnik et al., 2021). A number of fMRI studies consistently showed activations in Brodmann areas 10/24/32 in both left and right hemispheres (Izuma, 2013). The pMFC has been associated with monitoring of conflicts, cognitive control, error detection (Carter et al., 1998; Botvinick et al., 1999, 2001; Holroyd Nieuwenhuis et al., 2003; Danielmeier et al., 2011), and reward-based decision making (Williams et al., 2004). Overall, pMFC activity has been linked to performance monitoring and behavior adjustment. Recently, neuroimaging studies have focused on the role of the pMFC in cognitive dissonance and following the difficult choice preference changes (van Veen et al., 2009; Izuma et al., 2010; Jarcho et al., 2011; Kitayama et al., 2013; Voigt et al., 2019). A multichannel electroencephalographic (EEG) study demonstrated that the fronto-central resting state activity predicted the individual strength of preference changes and the magnitude of the dissonance-related neural activity (Colosio et al., 2017). The newest fMRI study showed partitioning of the activity of the medial frontal cortex: vmPFC is associated with expected reward-based decision making, whereas dmPFC is linked with metacognitive aspects of decisions such as deliberation and confidence about the alternatives and choice (Clairis and Pessiglione, 2022). Thus, the activity of the medial frontal cortices at rest affects different aspects of the behavioral effects of cognitive dissonance.

The use of a neuromodulatory approach with the help of repetitive transcranial magnetic stimulation (rTMS) (Izuma et al., 2015) facilitated the unveiling of the causal role of the pMFC in generating and reducing cognitive dissonance in a modified “free-choice paradigm” with a “choice-blindness” procedure. A disruption of pMFC activity, using 1 Hz rTMS right after the choice stage of the free-choice paradigm, significantly reduced the choice-induced preference changes. Although the rTMS approach demonstrated great potential in elucidating the causal relationship between cortical areas, the temporal aspect of the rTMS precludes one from understanding whether choice-induced preference changes take place during preference task II or the choice task. In the past decade, functional neuroimaging studies (e.g., Izuma et al., 2010) have explored the neural underpinning of cognitive dissonance, predominantly during the post-decisional stage of the “free choice paradigm” when subjects rated options again, some time after making difficult choices (Izuma and Murayama, 2013). This is based on the theoretical proposition that cognitive dissonance is experienced after making a difficult decision, which subsequently leads to an increase in preference for the chosen highly attractive option and a decrease in preference for the rejected one. Importantly, the activity of the pMFC was demonstrated already during the making of such a decision (choice task) (Kitayama et al., 2013; Voigt et al., 2019), which supports the hypothesis about the occurrence of preference changes while making a choice. An EEG study with use of “free choice paradigm” demonstrated that difficult decisions during the choice task are associated with stronger evoked elevated activity in the pMFC, reflected in a larger fronto-central error-related negativity (ERN) response, compared to easy decisions (Colosio et al., 2017). A comparison of ERN amplitude between trials featuring difficult and easy choices revealed that the ERN amplitude was higher for difficult ones. Furthermore, the ERN amplitude correlated with the magnitude of choice-induced preference changes. The difference waves (trials in difficult choices versus trials in easy choices) in Cz electrodes position significantly correlated with the extent of spread of alternatives. Thus, a stronger ERN was observed in the Choice task, and the stronger individual preferences were later altered for rejected items in Preference task II (Colosio et al., 2017). Since ERN activity was manifested during choices, the above-mentioned results suggest that the pMFC may be involved in the preference changes at an earlier stage than previously thought. This hypothesis about alteration of preference at an early stage is also supported by the studies of metacognitive aspects of choices (Lee and Daunizeau, 2020; Lee and Holyoak, 2021; Clairis and Pessiglione, 2022).

In this study, we applied transcranial direct current stimulation (tDCS) over the pMFC to probe the critical role of the pMFC in choice-induced preference changes and its contribution to cognitive dissonance during decision-making. The tDCS is a non-invasive neuromodulation technique that temporarily enhances (more often, anodal stimulation) or reduces (more often, cathodal stimulation) cortical excitability. This effect is achieved through applying a constant weak electrical current through an electrode placed on the surface of the scalp. Importantly, tDCS may result in facilitation of, or interference with the targeted brain region activity underlying changes of behavior (Nitsche and Paulus, 2001; Nitsche et al., 2008; Brunoni et al., 2012). This technique has been recently employed to explore the role of the medial frontal cortex in the modulation of error processing and performance monitoring (i.e., the modulation of the ERN and feedback-related negativity) in both clinical (Reinhart et al., 2015) and healthy populations (Bellaïche et al., 2013; Reinhart and Woodman, 2014). Here, we have conducted two sham-controlled experiments with delivering cathodal tDCS of the pMFC (Experiment 1) and anodal tDCS of the pMFC (Experiment 2). Using tDCS, we do not anticipate any effect of brain stimulation on ostensible preference changes due to statistical artifact found by Chen and Risen (2010) for option 1 and option 2. Therefore, any significant differences in preference changes across stimulation conditions could be predominantly attributed to the causal role of the pMFC in evoked by difficult choices spread of alternatives.

By applying tDCS at the preliminary decision stage of the “free choice paradigm,” we expected to exert control on the cortical excitability of the pMFC, and thus observe either a reduction (after cathodal stimulation) or an increase (after anodal stimulation) of the choice-induced preference changes compared to the non-stimulated (sham tDCS) condition, particularly after making difficult choices. The previous study by Colosio et al. (2017) demonstrated more explicit and accurately interpreted alteration of preference after hard choices specifically for the options which were rejected. Therefore, to specify hypothesis and test the stimulation effect, in this study, we were mainly interested in the alteration of preference for rejected option under tDCS, expecting a decrease in the devaluation of the attitude towards declined options in difficult decisions.

The comparison of preferences for food items in A and B sets in pre-study proved that the sets had similar mean ratings (Set A = 4.70 ± 0.87; Set B = 4.69 ± 0.88). The independent t-test showed no significant difference between preferences for food items in sets A and B: t₍₂₂₂₎ = 0.06, p = 0.94. The Shapiro–Wilk test for normality ensured that set A (W = 0.991, p = 0.215) and set B (W = 0.990, p = 0.121) were sampled from normal distribution.

In the current study, we used cathodal and anodal tDCS of the pMFC right before the choice task of the “free choice paradigm” to investigate the neural mechanism of the cognitive dissonance and subsequent choice-induced preference changes.

Regardless of the stimulation, we replicated a general behavioral effect of cognitive dissonance in both experiments: the preferences for items rejected in self-difficult (conflictual) choices significantly decreased after making the choice, compared to self-easy (non-conflictual) and computer (self-irresponsible) choices. This effect was observed regardless of the type of non-invasive tDCS: preference re-evaluation was detected in both Experiment 1 with inhibitory (cathodal) tDCS and Experiment 2 with excitatory (anodal) tDCS.

In Experiment 1, we observed that cathodal (inhibitory) tDCS of the pMFC particularly diminished choice-induced preference changes on declined options in difficult choices compared to the sham stimulation. This result supported the causal role of the pMFC in preference changes while making a difficult choice: suppressing the activity of the pMFC by cathodal tDCS prior to the choice reduced the reevaluation of the preference for rejected options. However, this result was demonstrated only in direct comparison using t-test, and effect size of the stimulation was comparatively small (Cohen’s d = 0.28). Further investigation of the general influence of cathodal tDCS on preference re-evaluation using linear mixed effects models (LME) did not show the significant effect of the tDCS.

In Experiment 2, we found no significant effect of anodal tDCS of the pMFC on the preference changes: neither in a focused analysis of the rejected items in self-difficult trials in comparison with sham condition nor analyzing data using linear mixed effect model.

One of the main possible reasons for not finding the strong effect of stimulation in two experiments as we expected is high probability of getting a false negative result. Having a limited knowledge of the neuromodulatory effects of tDCS on the activity of the pMFC in the cognitive dissonance studies poses difficulties to a prior calculation of the appropriate sample size in order to obtain reliable result. The posterior calculation of the statistical power, based on the observed effect size in Experiment 1, did not reach 80%, which makes a false negative outcome highly likely. Descriptive statistics (Supplementary Tables S1, S2) for choice-induced preference changes under stimulation and without it showed substantial heterogeneity and variability. We invited participants without neurological and psychiatric diseases and the use of any medication asked them not to drink coffee and alcohol on the day of the experiment and excluded those who experienced extreme fatigue and discomfort during the experiment. More attention should undoubtedly be paid to controlling the participants’ states in tDCS-experiments, due to the severe variability in the stimulation effect. These results can be used in subsequent tDCS-studies for prior calculation of the required sample size based on the statistical power and enhance the experimental design.

Another explanation of the current outcome in Experiment 2 is asymmetry in inhibitory-excitatory effects of non-invasive tDCS. This is supported by the results of a number of previous studies that have demonstrated heterogeneity of anodal and cathodal stimulation (Fregni et al., 2005; Karim et al., 2010; Mengarelli et al., 2015; for a meta-analysis see Jacobson et al., 2012). Some studies have specifically reported that there was no significant behavioral modulatory effect of anodal tDCS (Karim et al., 2010; Fagerlund et al., 2015; Conley et al., 2016). Further tDCS studies of conflict monitoring and resolution are needed to reconcile the asymmetry in stimulation effects and, in particular, to differentiate between the influence of anodal stimulation. Subsequent tDCS studies of choice-induced preference changes should pay specific attention to searching for the optimal target of brain stimulation. For example, evidence suggests that the more anterior subregions of the pMFC (FPz site) did not result in any modulatory effect on the ERN (Bellaïche et al., 2013). Systematic calculating of the electric field across many studies or the use of the high-definition tDCS could also be beneficial in reconciling the tDCS results across studies.

One further debatable point is the potential compensation of the effect of tDCS of the pMFC by the activity of unaffected brain areas, such as the dorsolateral prefrontal cortex (DLPFC). Recent investigations provide further evidence that a whole brain network is involved in the process of preference changes (Colosio et al., 2018; Voigt, 2022). For example, neuroimaging studies indicated an important role for the DLPFC in cognitive dissonance (Harmon-Jones and Harmon-Jones, 2008; Harmon-Jones et al., 2011; Mengarelli et al., 2015). Mengarelli et al. (2015) down-regulated the DLPFC by a 15 min, 1 mA cathodal tDCS. Offline stimulation of the left DLPFC significantly reduced the post-decision preference changes, and hence suggested that the left DLPFC plays an important role in the behavioral effects of cognitive dissonance. The role of the DLPFC in choice-induced preference changes is still under discussion, but it is thought to contribute to more general cognitive control mechanisms, regardless of whether conflicts is present (Harmon-Jones et al., 2011; Izuma et al., 2015). Interestingly, Ridderinkhof et al. (2004) proposed the existence of a functional pMFC-DLPFC network which supervises performance monitoring and executions. Further studies should focus on the development of possible controls for electromagnetic stimulation which can elucidate the interaction between the pMFC and DLPFC in choice-induced preference changes.

Although many neuroimaging studies demonstrated that the pMFC plays a central role in conflict monitoring, cognitive control and conflict resolution, little is certain about the chronometry of neuronal mechanisms of choice-induced preference changes. One of the first studies to show fMRI signatures of cognitive dissonance at the post-decisional stages in a “free choice paradigm” demonstrated that more conflicted decisions were associated with the larger pMFC activity during preference task II, compared to less conflicted decisions (Izuma et al., 2010). Of note, the majority of previous literature studying cognitive dissonance and choice-induced preference changes in the ‘free choice paradigm,” focused on the neural activity after decision during preference task II (Izuma and Murayama, 2013). For example, TMS of the posterior medial frontal cortex (pMFC) decreased preference changes only if applied at the later stages of the paradigm – right before preference task II (Izuma et al., 2015). However, relation between the pMFC and post-decisional preference changes is not always supported by experimental finding. For example, Kitayama et al. (2013) also showed elevated activity of the pMFC during making difficult conflictual choices (compared to easy ones), but found no correlation between the activity of the pMFC and post-decisional attitude changes. The neuroimaging study of Jarcho et al. (2011) examined the decisional phase of the decision-based cognitive dissonance paradigm and observed increased activity of the pMFC regions during the decision but not after it. Voigt et al. (2019) demonstrated, using fMRI and eye tracking, that activity of the DLPFC and pMFC, as well as and fixation duration during the making of hard decisions, predicted the magnitude of subsequent preference changes. Our study supports this evidence. Importantly, the duration of the tDCS after-effect is still a matter of debate: some studies have reported that a 20 min, 1.5 mA stimulation could generate a modulatory effect for several hours (Nitsche and Paulus, 2001; Nitsche et al., 2003; Reinhart and Woodman, 2014). In that case, in our study, cathodal tDCS could inhibit cortical activity during both the choice task and preference task II. Thus, new protocols should be developed to differentiate neural activity of the DLPFC and pMFC in the mechanisms of choice-induced preference changes.

Another interesting question is about the metacognitive aspects of preference changes in decision making. Some studies showed that at least partially preference changes can be attributed not to the fact of making difficult choices or the necessity of rearranging preferences to resolve conflicts, but rather to the internal refinement of the choice based on the certainty of pre-choice value judgment and confidence about the options in the decision (Lee and Daunizeau, 2020; Clairis and Pessiglione, 2022). Lee and colleagues provided a computational model for the online metacognitive control of decisions (Lee and Daunizeau, 2021; Lee et al., 2023). The fMRI study showed that value-based decision making and metacognitive evaluation of the option can be separated even at the neuronal levels (Clairis and Pessiglione, 2022). Therefore, it is important in future studies of the preference changing in making difficult decisions to disentangle the effects of the choice when comparing the expected values of the options and the subjective metacognitive process regarding this choice.

Generally, we traced the neuromodulatory (inhibitory) effect of cathodal tDCS on choice-induced preference changes. This effect was consistent with the proposed role and temporal dynamics of the pMFC: inhibiting the pMFC through cathodal tDCS, a key region in conflict detection and behavioral adjustments, prior to the making of a difficult decision, decreases the preference changes. This effect, however, was rather small, manifested only in direct comparisons with placebo stimulation and showed an asymmetry to the anodal (excitatory) tDCS, which did not demonstrate an increase in preference changes.

The original contributions presented in the study are publicly available. This data can be found here: https://osf.io/abpqj.

The studies involving humans were approved by Institutional Review Board of the National Research University Higher School of Economics. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

ER: design of the study, data collection for experiment 2, statistical data analysis for experiment 1 and experiment 2, writing all sections of the manuscript, and figures and tables creation. MC: conceptualization and design of the study, stimuli development, and data collection for experiment 1. AS and VK: supervising the study and revision the manuscript. All authors contributed to the article and approved the submitted version.