Ventromedial prefrontal cortex attenuates the link between physiological heart rate reactions to unfairness and value computations in normative decision making

Abstract

Emotional reactions to norm violations influence how people trade-off selfish interests against social norms, but it remains unknown which brain regions contribute to integrating physiological (e.g. heart rate) responses to norm violations into the decision process. Here, we provide evidence that the ventromedial prefrontal cortex (vmPFC) attenuates the link between value computations in normative choice and physiological heart rate responses to norm violations. Excitatory brain stimulation over the vmPFC reduced aversion to disadvantageous inequality in an Ultimatum game and increased post-decision confidence after acceptance of unfair monetary splits. Physiological heart rate responses were increased in response to splits perceived as unfair. Enhancing vmPFC excitability reduced the sensitivity to unfairness not by directly reducing heart rate reactions to unfairness but by de-correlating them from the decision process. This suggests that vmPFC moderates the association between physiological heart rate responses to norm violations and trade-offs between self-interests versus norm considerations, highlighting its key role for integrating emotional and cognitive processes in decision making.

Introduction

People often have to trade-off their selfish interests against fairness considerations (Guth et al. 1982, Fehr and Camerer 2007). For example, if a person treats us unfairly, we may punish the other for her norm violation even at our own expenses. Experiencing norm violations triggers emotions like anger about the unfair treatment that are accompanied by physiological reactions, for example by increased heart rates (Wu et al. 2019); this physiological heart rate reaction is thought to affect the value we assign to punishing the norm violation (Dunn et al. 2012). Little is known, however, about the neural mechanisms that moderate the influence of emotion-related heart rate responses on decisions involving trade-offs between selfish interests and fairness considerations.

One area that fulfills the criteria required to play a central role in integrating physiological responses into the decision process is, besides other parts of the limbic network (Lindquist and Barrett 2012), the ventromedial prefrontal cortex (vmPFC). Theoretical accounts ascribe the vmPFC important contributions to emotion processing (Rudebeck et al. 2008, Hiser and Koenigs 2018), according to which the vmPFC may moderate the influence of emotion-related physiological responses on decisions in different ways. First, the vmPFC might directly be involved in attenuating physiological responses to unfairness, in line with emotion regulation accounts linking the vmPFC to the voluntary down-regulation of negative emotions (Egner et al. 2008, Etkin et al. 2015, Suzuki and Tanaka 2021). In keeping with this, lesions to vmPFC increased physiological responses to emotionally salient stimuli (Jenkins et al. 2018). From this perspective, vmPFC may attenuate the impact of norm violations on choice behavior by inhibiting the physiological response to perceived unfairness.

Alternatively, the vmPFC might not directly affect emotion-evoked physiological responses but regulate how strongly they influence the decision process. The latter account is consistent with the vmPFC’s putative role for interoceptive processes (Terasawa et al. 2013, Sojka et al. 2025), that is, the ability to perceive, interpret, and integrate physiological processes of the body (Khalsa et al. 2018; Suksasilp and Garfinkel 2022). Interoceptive processes were found to moderate the influence of bodily reactions to unfairness in the Ultimatum game. In particular, individuals with higher interoceptive accuracy, the ability to correctly monitor changes in internal states (Khalsa et al. 2018), showed stronger influences of unfairness-elicited bodily responses on social decisions (Dunn et al. 2012), whereas another study linked interoception (albeit without directly measuring it) to weaker reactions to unfairness (Kirk et al. 2011). While we note that the current study did not directly assess the vmPFC’s involvement in interoception (due to the lack of a task measuring interoceptive processes), the literature on the vmPFC’s role for interoception suggests that vmPFC activity might change normative decision making not by altering the physiological response to unfairness (e.g. increases in heart rates; mediation hypothesis); instead, vmPFC activation might moderate the association between unfairness-elicited heart rate changes and decisions to reject unfair offers (moderation hypothesis).

Existing evidence on the role of the vmPFC for trading-off fairness versus selfish interests provided no conclusive evidence whether vmPFC promotes fairness-related over selfish (i.e. self-payoff maximizing) motives or vice versa. While one line of evidence linked the vmPFC to stronger weighting of selfish interests over fairness norms (Koenigs and Tranel 2007, Gilam et al. 2017), other studies suggest the vmPFC to increase the sensitivity to norm violations (Baumgartner et al. 2011, Feng et al. 2015, Gu et al. 2015, Lo Gerfo et al. 2019). However, the inferences that can be drawn from these findings remain limited because they either did not causally test the contribution of the vmPFC (Baumgartner et al. 2011, Feng et al. 2015) or they lack the necessary spatial resolution to conclusively link changes in social decision making to vmPFC activation (Koenigs and Tranel 2007, Gu et al. 2015, Gilam et al. 2017, Lo Gerfo et al. 2019). Moreover, no previous study assessed physiological responses (e.g. in heart rates) to norm violations, as would be necessary to decide whether the vmPFC affects decision-making via down-regulating physiological responses or via moderating their influence on the decision process. Thus, it remains unknown whether, and if so how, the vmPFC contributes to integrating unfairness-induced physiological responses into trade-offs between selfish interests and fairness considerations.

The goal of the current study was to decide between these competing accounts of vmPFC involvement in normative decision making. To determine how the vmPFC moderates the influence of physiological heart rate responses on normative decision making, participants played an Ultimatum game after either anodal or sham transcranial direct current stimulation (tDCS) targeting the vmPFC. Participants had to decide whether to accept or reject unfair monetary splits of anonymous human proposers. We measured changes in heart rate during the presentation of the offers as indicator of the physiological response to unfairness. Previous studies found that increases in heart rate are associated with rejections of unfair offers, supporting the view that they reflect an emotion-evoked physiological response to inequality (Sarlo et al. 2013, Gilam et al. 2017). Crucially, measuring heart rate responses as physiological reaction to unfairness allowed us to decide between different functional roles of the vmPFC: if the vmPFC down-regulates physiological responses to unfairness, we would expect anodal tDCS over vmPFC to reduce unfairness-related changes in heart rate. Alternatively, if the vmPFC moderates the influence of heart rate responses on the decision process, enhancing vmPFC excitability should change the link between heart rate responses to unfairness and decisions in the Ultimatum game. Moreover, given the vmPFC’s involvement in representing subjective post-decision confidence (Vaccaro and Fleming 2018), we also hypothesized that vmPFC tDCS strengthens post-decision confidence via lowering the sensitivity to fairness violations.

Materials and methods

Participants

Thirty-five healthy volunteers (mean age = 26.5 years, range = 18–35 years, 16 female, 19 male) who were recruited at the Ludwig Maximilians University (LMU) Munich participated in the study. We excluded data from three participants due to technical issues with the heart rate recordings, leaving 32 participants for the analysis. According to a power analysis, 20 participants are sufficient to detect a significant effect (alpha = 5%, power = 95%) assuming an effect size of Cohen’s d = 0.86 observed in a meta-analysis of brain stimulation studies in social decision making (Christian and Soutschek 2022). We used a relatively large power of 95% to minimize the probability of false negatives. This sample size also provides sufficient power according to a previous tDCS study on the Ultimatum game (Gilam et al. 2017). All participants were healthy volunteers without any known psychiatric or neurological disorders, as assessed with a one-item self-report question, and were naive with respect to the aims of the study. Ethical approval was granted by the ethics committee of the psychology department at LMU, and all procedures followed the principles of the Declaration of Helsinki (2013). Participants provided informed written consent prior to participation. Participants showed no contra-indications for tDCS and were informed about its potential side effects (Bikson et al. 2016). As compensation, participants received 10 euros per hour and a monetary bonus depending on their choices (see below).

Stimuli and task design

The study followed a randomized, sham-controlled, single-blind crossover design. All participants took part in two experimental sessions where they received either anodal or sham tDCS over vmPFC (in counterbalanced order) prior to performing the Ultimatum game (Fig. 1a). Participants played the Ultimatum game in the role of the responder and had to decide whether to accept or reject a split of 20 euros offered by an anonymous human proposer. The amount offered ranged from 0 (unfair offer) to 10 euros (fair offer) in steps of 1 euro. If participants accepted an offer, the participant and the proposer obtained the indicated amounts of money; if an offer was rejected, none of them obtained money. Participants therefore had to trade-off their selfish interests (accepting an offer to obtain the bonus) against the costly punishment of norm violations (rejection of unfair offers). It is important to emphasize that choices had real consequences for participants’ and proposers’ payments. In a pre-study, we had collected offers from various proposers, which we then used for the Ultimatum game in the main study. We informed participants that offers were made by real individuals in pre-study and that, at the end of the experiment, one randomly selected decision would be paid out to the participant and a proposer from the pre-study. Moreover, we informed participants that the offers they would receive were stemming from different proposers; this was done to minimize the influence of strategic considerations on accept versus reject decisions that occur in iterative Ultimatum games with real-time proposers.

Figure 1.

(a) Experimental procedure. Participants completed two separate sessions, in which either anodal or sham offline tDCS was applied over the vmPFC for 20 minutes in a counterbalanced within-subject design. After stimulation, participants provided a brief rating of tDCS-induced perceived discomfort and flickering. Following this, participants performed the Ultimatum Game for 20 minutes while their heart rate was continuously recorded. Upon completing the task, participants indicated their perceived unfairness of the offers made by the proposers. (b) Example trial for the Ultimatum Game. Participants were presented with proposals regarding how to split an amount of 20 euros, made by an anonymous proposer (for example, 16 coins for the proposer and 4 coins for the participant). Participants had to decide whether to accept or reject the offer. After each choice, participants rated their subjective confidence to have made the best choice. The next trial started after a short inter-trial interval. (c) tDCS setup: We targeted the vmPFC with a high-definition electrode setup. Current flow simulations with the SimNIBS toolbox (Saturnino et al. 2019) suggest that current density was highest (indicated by warmer colors) in ventral parts of the medial prefrontal cortex.

On each trial, we first presented participants with an offer (e.g. “4 euro for you, 16 euro for other”) on a screen. Participants had to indicate their choice to accept or reject the offer by pressing the left or right arrow key (key-response assignment was counterbalanced across participants) within 4.5 s. We instructed participants that there were no correct or incorrect responses in the task and that they should simply choose the option they preferred most. Afterwards, participants had to rate their confidence in having made the best choice (which should depend on the strength of their preference for the chosen over the unchosen option) on a continuous Likert scale ranging from 0 (not confident at all) to 100 (very confident) within 6 s, followed by an inter-trial interval of 2 s (Fig. 1b).

Procedure

All participants took part in two experimental sessions in which they received either anodal or sham tDCS over vmPFC (in a counterbalanced order). Offline tDCS was applied for 20 minutes prior to the Ultimatum Game (see below for details). Following tDCS, participants performed 96 trials of the Ultimatum Game (∼20 minutes) in randomized order while we continuously recorded their heart rate. At the end of the experiment, participants were asked to rate as how unfair they perceived each offer (i.e. offers from 0–10 euros as used in the decision task) on a continuous scale from 0 (not unfair at all) to 100 (very unfair). These ratings were collected after the Ultimatum game to ensure that they did not influence participants’ choices in the Ultimatum game. More specifically, assessing the perceived unfairness of offers before the decision task might lead to consistency effects where participants adjust their choices to avoid inconsistencies with their previous ratings (Cialdini and Trost 1998). Alternatively, measuring perceived unfairness after each choice in the Ultimatum game would potentially have increased the total duration of the experimental tasks beyond the duration of tDCS-induced after-effects. At the end of the experiment, one choice was randomly selected for the bonus payment.

tDCS protocol

We used a 16-channel tDCS stimulator (neuroConn, Ilmenau, Germany) to apply tDCS with a current strength of 1.5 mA (peak-to-peak) for 20 minutes in line with the safety recommendations for tDCS (Bikson et al. 2016, Woods et al. 2016). This protocol is thought to result in tDCS-induced aftereffects for at least 30 minutes (Kuo et al. 2013, Polanía et al. 2018). We placed one round high-definition anode (diameter = 2 cm) over the vmPFC (position Fpz according to the international EEG 10/20 system), surrounded by three cathodes over F3, F4, and under the chin. The electrode under the chin allowed us to induce a current flow deeper into the brain to stimulate large parts of vmPFC (Junghofer et al. 2017, Kroker et al. 2022, 2025; Fig. 1c). To verify that tDCS targeted the vmPFC, we performed a region of interest (ROI) analysis of the electric fields generated by tDCS using the SimNIBS toolbox (Saturnino et al. 2019). The vmPFC ROI (MNI coordinates: x = −3, y = 54, y = −12) was defined based on a previous study reporting the vmPFC to correlate with offer fairness in the Ultimatum game (Baumgartner et al. 2011). The electric field in the vmPFC ROI was 0.13 V/m, whereas the field strength in an ROI for the right dlPFC (x = 45, y = 24, z = 21) from the same study was only 0.06 V/m. This suggests our tDCS setup to show stronger effects on vmPFC than on dlPFC.

For anodal tDCS, the stimulation lasted a total of 20 minutes (ramp-up: 15 seconds, ramp-down: 15 seconds). Note that we employed anodal rather than cathodal tDCS because cathodal tDCS was reported to often result in paradoxical excitatory effects (Jacobson et al. 2012). In the sham condition, the current was ramped up for 15 seconds and directly afterwards ramped down for 15 seconds. In the following 20 minutes, the sham protocol followed a repeated pattern of stimulation-free intervals of 28 seconds, ramp-up phases of 1 second, directly followed by ramp-down phases of 1 second. This was done to improve participants’ blinding to the current tDCS condition. At the end of the stimulation, participants had to rate the perceived discomfort and flickering during the stimulation on rating scales from 0 (not at all) to 7 (very much) (time limit = 4 seconds). We observed no significant differences in perceived discomfort, t(31) = 1.80, P = 0.08, or flickering, t(31) = 1.83, P = 0.08, between anodal and sham tDCS. Nevertheless, given these non-significant trend-level effects, we entered discomfort and flickering ratings as control variables to all statistical analyses to rule out that any influences of tDCS on choice behavior might be related to the perceived discomfort of the stimulation.

ECG protocol

Continuous electrocardiography (ECG) data were recorded with two bipolar electrodes below the left pectoral muscle and the left clavicle, as well as one ground electrode below the right clavicle. ECG data was recorded with a Brain-Vision QuickAmp amplifier, using a sampling rate of 500 Hz.

ECG analysis

Preprocessing of ECG data was performed using the FieldTrip toolbox (Oostenveld et al. 2011) and MATLAB 2021b (MathWorks Inc. Natick, MA, USA). ECG data were filtered using a 40-Hz low-pass filter and a 1-Hz high-pass filter. We extracted heart rate responses from the time window of interest (presentation of offer for 4 seconds). We used the Pan–Tompkins algorithm in the BioSigKit toolbox (Sedghamiz 2018) for detecting R-peaks in the time window for each trial. Based on these, we calculated the latency between R-peaks within the time window of interest (Kaiser et al. 2021). We converted the mean latencies between peaks for each trial to beats per minute. Because the pattern of results was robust to using heart period instead of heart rate (as reported below in the Results section), we retained heart rate to facilitate comparison with Dunn et al. (2012). Based on the identified R-peaks, we excluded outlier trials for which the heart rate deviated more than three standard deviations from a participant’s mean (2.9% of all trials). Note that the study was not designed to measure heart-rate variability, which requires a minimum time window of at least 30 or 60 s with stable physiological conditions (Burma et al. 2021, Bernardes et al. 2022). This is not provided by the current experimental design where the decision period lasted only 4.5 s and physiological demands strongly differed between trials as a function of the unfairness of a given offer.

Statistical analysis

We analyzed choices, confidence ratings, and heart rate responses in the Ultimatum game with general(ized) linear mixed models (GLMMs) using the lme4 package in R (Bates et al. 2014). The alpha threshold was set to 5% (two-tailed) for all analyses. To test the hypothesis that vmPFC tDCS changes the acceptance of unfair offers, we regressed binary choices to accept or reject an offer (0 = reject, 1 = accept) on fixed-effect predictors for tDCS (0 = sham, 1 = anodal), Offer (z-transformed amount of money offered to participant), and the interaction term. All predictors were also modelled as random slopes in addition to participant-specific random intercepts. As covariates of no interest, we controlled for tDCS-induced discomfort and flickering in all statistical analyses.

In further GLMMs, heart rate responses (which were approximately normally distributed) were regressed on predictors for tDCS, Offer, Choice (0 = reject, 1 = accept), and all interaction terms. Because the perceived unfairness of offers and post-decision confidence ratings were skewed, we analyzed them with ordinal hierarchical mixed regression analyses as implemented in the ordinal package. Lastly, to test whether vmPFC tDCS changes the influence of heart rate responses on choices, we conducted a further GLMM where binary choices were regressed on predictors for tDCS, trial-specific Heart rate, Offer, and all interaction terms. This allowed testing the hypothesis that vmPFC moderates the influence of the physiological response to unfairness on decisions in the Ultimatum game.

To further assess whether vmPFC tDCS affects preferences for unfair outcomes, we also fitted an inequality aversion model to the binary choice data with a hierarchical Bayesian approach (Soutschek et al. 2024). We used a simplified version of the Fehr–Schmidt model (Fehr and Schmidt 1999) that allowed us to quantify how individuals trade-off selfish interests against their aversion to disadvantageous inequality (note that inequality was always disadvantageous rather than advantageous in our task as the offered amount for the participant was never larger than the amount for the proposer). In more detail, we assumed that the utility of accepting versus rejecting an offer was given by the following equation:

$${\text{Utility}}_{\text{accept}}=\text{offer}–\text{inequality}\times \left({\beta }_{\text{inequality}}+\text{tDCS}\times {\beta }_{\text{inequality}\_\text{anodal}}\right)$$ (1)

Here, β_inequality is a free parameter quantifying the aversion against disadvantageous inequality, the inequality term represents the difference between the amounts for the receiver and the proposer for a given offer, tDCS is a dummy-coded variable that is 0 for sham and 1 for anodal tDCS, and β_{inequality_anodal} captures the influence of anodal tDCS on inequality aversion. A preference for accepting or rejecting an unfair offer is indicated by Utility_accept > 0 and Utility_accept < 0, respectively. In a further model, we tested whether the utility of unfair offers is also affected by heart rate responses by adding trial-by-trial changes in heart rate (HR) to Equation 1:

$${\text{Utility}}_{\text{accept}}=\text{offer}–\text{inequality}\times \left({\beta }_{\text{inequality}}+\text{tDCS}\times {\beta }_{\text{inequality}\_\text{anodal}}+\text{HR}\times \left({\beta }_{\text{HR}}+\text{tDCS}\times {\beta }_{\text{HR}\_\text{anodal}}\right)\right)$$ (1′)

We used a standard softmax function to translate utilities into binary choices:

$$\mathrm{P}\left(\text{accept}\right)={\left(1+\hspace{0.17em}\text{exp}\hspace{0.17em}\left(-{\text{Utility}}_{\text{accept}}{\times }_{ }\left({\beta }_{\text{consistency}}+\text{tDCS}\times {\beta }_{\text{consistency}\_\text{anodal}}\right)\right)\right)}^{-1}$$ (2)

where β_consistency is a choice consistency (“inverse temperature”) parameter indicating how consistently participants chose the higher utility option, whereas β_{consistency_anodal} quantifies the influence of anodal tDCS on choice consistency. For all model parameters, we estimated the parameter values under sham as well as the influences of anodal tDCS. We estimated parameters with a hierarchical Bayesian approach using the dbern module in JAGS. We used non-informative flat priors for the group parameters under sham and normally distributed priors for the tDCS effects. We assumed that individual parameters are normally distributed around the group-level parameters. We computed four chains, each with 30 000 samples (burn-in = 5000). The $\widehat{R}$ values for all parameters were ≤ 1.01, indicating model convergence. For statistical inference, we assessed whether 95% highest density intervals (HDI_95%) included zero.

Results

vmPFC tDCS strengthens self-interests over fairness concerns

First, we aimed to disentangle between conflicting accounts on the role of the vmPFC for normative choice (Koenigs and Tranel 2007, Baumgartner et al. 2011, Gilam et al. 2017, Lo Gerfo et al. 2019) by testing how anodal vmPFC tDCS influences the acceptance of unfair offers. As to be expected, participants were more willing to accept unfair splits the more money was offered to them, β = 5.57, z = 10.40, P < .001. Anodal versus sham tDCS over vmPFC increased acceptance rates, β = 1.13, z = 2.45, P = .01, independently of the fairness of the offer, tDCS × Offer: β = −0.47, z = 0.92, P = .36 (Fig. 2a). Thus, vmPFC tDCS strengthened the weight assigned to economic self-interests relative to fairness considerations in the Ultimatum game. We observed no correlations between the individual strength of the tDCS effect on choices and stimulation-induced discomfort, r = 0.05, P = 0.78, or flickering ratings, r = 0.11, P = .56. This is at variance with the assumption that the tDCS effect on choices might be driven by the sensory perception of the stimulation (note that we controlled for these sensory effects in all statistical models). To account for the possibility that the strength of the stimulation effect might decay with increasing time-on-task in an offline protocol, we added a predictor for trial number (z-transformed) to the analysis. The positive sign of the tDCS × Trial interaction: β = 0.46, z = 2.10, P = .04, suggests that—if anything—stimulation effects on acceptance rates become stronger rather than weaker over the course of an experimental session.

Figure 2.

(a) Anodal versus sham tDCS over vmPFC increased acceptance rates independently of the magnitude of the monetary offer. (b) Higher offers were perceived as less unfair than lower offers, but there was no evidence that vmPFC tDCS changed the self-reported unfairness of offers. In (a) and (b), we showed regression lines fitted to the mean acceptance rates and unfairness ratings, respectively, with errors bars representing standard error of the mean. (c, d) Consistent with the observed higher acceptance rates under anodal tDCS, a computational model of inequality aversion suggests that increasing vmPFC excitability (c) significantly reduced aversion to disadvantageous inequality while leaving (d) choice consistency unaltered. Boxplots indicate inter-quartile ranges; black dots represent individual data points. Asterisks indicate significant effects (*P < .05).

Contrary to the tDCS effects on decision-making, vmPFC tDCS did not affect the perceived unfairness of offers (measured after the decision task but still under the influence of tDCS): while lower monetary offers were considered as increasingly unfair, ordinal regression analysis: β = −3.91, z = 12.75, P < .001, there were no significant tDCS effects on perceived unfairness, tDCS: β = 0.03, z = 0.08, P = .93, tDCS × Offer: β = −0.25, z = 0.85, P = .40 (Fig. 2b). Bayes factors computed via the Akaike information criteria (Jarosz and Wiley 2014) indicated that the null hypotheses (no tDCS effect) were more likely than the alternative hypotheses (tDCS affects perceived unfairness) for both the main effect of tDCS, BF01 = 2.70, and the tDCS × offer interaction, BF01 = 1.89. Note that the GLMM on acceptance decisions provided no evidence that stimulation effects were weaker at the end of an experimental session, so this null effect cannot be explained by the timing of the unfairness ratings. There was thus no evidence that enhancing vmPFC excitability alters the normative evaluation of unfair offers per se; instead, vmPFC stimulation changes how fairness is traded-off against selfish interests.

To corroborate these findings, we fitted a computational model of inequality aversion to the choice data that quantified the aversion against disadvantageous unequal splits. Participants showed disadvantageous inequality aversion under sham, β_inequality: mean HDI = 0.41, 95% HDI = [0.02, 0.98], which was significantly reduced under vmPFC tDCS, mean HDI = −0.27, 95% HDI = [−0.50, −0.06] (Fig. 2c). In contrast, there were no stimulation effects on choice consistency (inverse temperature), β_consistency: mean HDI = 0.11, 95%HDI = [−0.06, 0.32] (Fig. 2d). This is relevant because previous studies related vmPFC activation to value representations and choice consistency (Bartra et al. 2013, Bowren et al. 2018), and strengthening such value representations through anodal tDCS should have resulted in increased choice consistency. Taken together, our findings suggest a role of vmPFC for strengthening selfish interests over fairness concerns in normative choice.

We considered two alternatives that might explain the impact of vmPFC tDCS on inequality aversion: First, the vmPFC might directly be involved in down-regulating physiological responses (e.g. heart rate) to unfairness (mediation hypothesis). According to the second possibility, the vmPFC does not affect heart rate responses to unfairness per se but attenuates the influence of heart rate responses to unfairness on the decision process (moderation hypothesis). These possibilities make dissociable predictions for the impact of tDCS on heart rate responses to unfairness as indicator of the reaction of the autonomous system to perceived unfairness: The first alternative predicts that enhanced vmPFC excitability decreases inequality aversion via reducing changes in heart rate in response to unfair offers (mediation hypothesis). We first aimed to validate that changes in heart rate are indeed sensitive to unfairness: while the heart rate did not correlate with the objective amounts offered to participants or with the decision made, all t < 1.20, all P > .20, a separate GLMM suggested that heart rate changes were related to the self-reported perceived unfairness of offers, β = 0.74, t(912) = 2.30, P = .02. This suggests that heart rates reflect the subjective unfairness of offers rather than monetary self-interests (which are related to the objective magnitude of an offer). Moreover, this physiological heart rate response to unfairness tended to be stronger for rejected compared with accepted offers, Unfairness × Choice: β = −0.90, t(88) = 1.96, P = .06. Post-hoc tests suggest that perceived unfairness correlated with heart rate responses only for rejected, β = 0.60, t(134) = 2.78, P = .006 (Fig. 3a), not for accepted offers, beta = −0.06, t(25) = 0.17, P = .87 (Fig. 3b). This is consistent with the interpretation that an increase in heart rate reflects the physiological response to offer perceived as unfair, which biases choices toward offer rejection. However, there was no evidence that vmPFC tDCS changed the relationship between perceived unfairness and heart rate, tDCS × Unfairness: β = 0.12, t(81) = 0.21, P = .84, BF01 = 1.89; tDCS × Unfairness × Choice: β = −0.42, t(313) = 0.70, P = .48, BF01 = 10.28. The Bayes factors indicated stronger evidence for the null (no tDCS effects) over the alternative (tDCS affects heart rate) hypothesis. Together, these findings do not support the first alternative according to which the vmPFC is involved in down-regulating heart rate responses to unfairness.

Figure 3.

(a, b) Stimulation effects on heart rate changes during decision making. Under sham, heart rates significantly increased as a function of perceived unfairness for (a) rejected offers, but not for (b) accepted offers. We observed no significant effects of vmPFC tDCS on heart rates. Instead, vmPFC tDCS attenuated the influence of heart rates on decision making: (c) Under sham, higher hear rates during low, unfair offers increased the likelihood of rejecting the offer. (d) This influence of heart rates on decision making was reduced under anodal vmPFC tDCS. (a–d) depict regression lines fitted to the mean heart rate responses (a, b) and acceptance rates (c, d) as function of perceived unfairness and offer magnitude, respectively. Errors bars represent standard error of the mean.

As robustness check, we assessed whether our results are robust to analyzing heart periods (interval between two heartbeats) instead of heart rates (heartbeats per minute). The perceived unfairness of offers was correlated with shorter heart periods, Unfairness, β = −0.01, t(88) = 2.64, P = .01, and this relationship was stronger for rejected than for accepted offers, Choice × Unfairness: β = 0.01, t(67) = 2.58, P = .01. There was no evidence for significant tDCS effects on heart periods, all P > .41. Thus, the analysis of heart periods replicates the heart rate findings.

According to the second alternative, vmPFC does not reduce the perceived unfairness of offers or the heart rate response to unfairness but moderates how the physiological heart rate reaction to unfairness influences choice behavior (moderation hypothesis). To assess whether the impact of vmPFC tDCS on choices was moderated by heart rate changes, we added trial-by-trial measures of heart rate as a further predictor to the model regressing binary choice on tDCS and Offer. Consistent with the findings reported above, increases in heart rates were associated with an enhanced likelihood of rejecting unfair offers, Offer × Heart rate, β = 1.88, z = 7.07, P < .001. Importantly, this association between heart rate and choice behavior was reduced under vmPFC tDCS, tDCS × Offer × Heart rate, β = −1.27, z = 4.52, P < .001 (we note that this interaction was robust to replacing heart rates with heart periods: tDCS × Offer × Heart period, β = 0.98, z = 3.75, P < .001). Separate GLMMs for sham and anodal tDCS suggest that faster heart rates were linked to higher rejection rates for low, unfair offers under sham, β= 1.86, z = 6.53, P < .001 (Fig. 3c), whereas this effect was weaker under anodal tDCS, β = 0.55, z = 2.53, P = .01 (Fig. 3d). This suggests the vmPFC to play a central role in attenuating the relationship between physiological heart rate responses to unfairness and decisions to reject the offer. To corroborate these findings, we tested whether vmPFC stimulation moderates the association between heart rate responses to unfairness and decision making also in our computational model of inequality aversion by adding heart rate as covariate to the computational model described above. In fact, under anodal tDCS the utility of accepting unfair offers was increased for higher heart rates, mean HDI = 0.04, 95% HDI = [0.01, 0.09]. Taken together, this provides converging evidence that the vmPFC is involved in attenuating the link between physiological heart rate responses to unfairness and decisions in the Ultimatum game.

Given the observed relationship between heart rate changes and perceived unfairness, one might ask whether the stimulation effect on the influence of heart rate responses and normative decision making is driven by the impact of tDCS on the link between decisions and perceived unfairness rather than the heart rate. To test this possibility, we conducted a further GLMM that regressed choices in the Ultimatum game on predictors for heart rate and perceived unfairness as well as their interactions with tDCS and Offer. We replicated the significant tDCS × Offer × Heart rate interaction, beta = −0.79, z = 2.59, P = .010, suggesting that vmPFC tDCS modulated the link between heart rate and choice even when only the unique variance in heart rates (i.e. the variance that is not shared with perceived unfairness) is considered. However, we also found a significant tDCS × Offer × Perceived unfairness interaction, beta = −1.40, z = 3.34, P = .001. This indicates that vmPFC stimulation changes not only the association of choices with heart rates but also with perceived unfairness.

vmPFC tDCS increases decision confidence after self-benefitting choices

Given that anodal vmPFC tDCS reduced the influence of unfairness-related heart rate changes on decision making, we next asked whether the vmPFC affects not only choices themselves but also the post-decision confidence in having made the best choice (using ordinal hierarchical regressions as confidence data were strongly left-skewed). Post-decision confidence was lower after accepted compared with rejected offers, main effect of Choice: β = −0.63, z = 2.32, P = .02, and under anodal compared with sham stimulation, tDCS: β = −0.61, z = 2.22, P = .03. Moreover, we observed that vmPFC stimulation significantly reduced the link between the choice made (accept versus reject) and post-decision confidence, tDCS × Choice: β = 0.70, z = 3.54, P < .001 (Fig. 4): While under sham confidence was lower after rejected relative to accepted offers, β = −0.67, z = 2.44, P = .01, there were no significant differences in confidence after acceptance versus rejection decisions under anodal tDCS, beta = 0.10, z = 0.43, P = .67. Taken together, enhancing vmPFC excitability increased both the willingness to accept unfair offers and the subjective post-decision confidence that accepting versus rejecting an offer was the best choice.

Figure 4.

Influence of vmPFC tDCS on post-decision confidence. Under sham, confidence was higher after rejected compared with accepted offers, and this difference between accepted and rejected offers was significantly reduced under anodal tDCS. Boxplots show participant-specific mean confidence ratings, asterisks indicate significant effects (*P < .05).

Discussion

The vmPFC was ascribed a central role in integrating emotions into the decision process, but it remains unknown whether vmPFC influences decisions via directly down-regulating emotion-related physiological responses, for example in heart rates, or via attenuating the link between physiological responses and decision making. Here, we provide evidence that the vmPFC is involved in attenuating the relationship between unfairness-related heart rate responses and decisions involving trade-offs between selfish interests and fairness considerations. First, we found the self-reported perceived unfairness of rejected offers, but not the monetary self-interest associated with offers, to be linked to increased heart rate responses, consistent with the idea that changes in heart rate reflect the physiological response of the autonomous nervous system to (perceived) unfairness (Dunn et al. 2012). There was no evidence for an influence of vmPFC tDCS on the perceived unfairness of offers or on changes in heart rate, which does not support the view that vmPFC directly affects normative evaluations of unfairness or inhibits physiological heart rate reactions related to unfairness. Instead, our findings suggest that enhancing vmPFC excitability attenuates the link between unfairness-related heart rate changes and rejection decisions in the Ultimatum game. Higher heart rates biased the rejection of unfair offers only under sham, and this link between changes in heart rate and normative decisions was diminished under anodal tDCS. Interestingly, vmPFC stimulation weakened not only how normative choices are predicted by unfairness-related heart rate changes but also by the perceived unfairness of offers, suggesting that the vmPFC might play a general role for attenuating the influence of affective valuations on decision making. While previous neuroimaging and lesion research yielded inconclusive results regarding the vmPFC’s role for trading-offs between fairness and self-interest (Koenigs and Tranel 2007, Baumgartner et al. 2011, Feng et al. 2015, Gu et al. 2015, Gilam et al. 2017, Lo Gerfo et al. 2019), the current high-definition stimulation protocol provides evidence that vmPFC strengthens the focus on self-interest relative to fairness considerations by lowering the sensitivity to fairness violations relative to selfish interests. These results are consistent with findings from other domains of decision making according to which vmPFC reduces the influence of action costs like risk, losses, or waiting time on decision making (Ciaramelli et al. 2021, Kroker et al. 2022, Cecchi et al. 2024), suggesting that vmPFC may generally strengthen reward maximization over cost avoidance.

The vmPFC may be able to moderate the link between heart rate responses to unfairness and decision making due to its central role in a distributed neural decision network. The vmPFC receives input from several regions involved in value computation and emotion processing, including insula, dlPFC, amygdala, and striatum (Lindquist and Barrett 2012, Ghaziri et al. 2017, Jackson et al. 2020). The insula was related to emotion processing (Gogolla 2017), including violations of social norms (Sanfey et al. 2003, Harle et al. 2012). In addition, insula activation and connectivity between insula and vmPFC play a role for the interoceptive awareness of bodily signals (Craig 2002, Critchley and Harrison 2013, Kirk et al. 2014, Jarrahi et al. 2015, Sagliano et al. 2019). Interoceptive awareness of unfairness-related bodily responses was shown to reduce the influence of unfairness on social decisions (Kirk et al. 2011), indicating that interoceptive awareness of physiological responses may allow decoupling emotional responses from the decision process. Based on the vmPFC’s putative role in interoception (Hurliman et al. 2005, Sojka et al. 2025), it is possible that the weaker association between heart rate changes and normative decision making under anodal compared with sham vmPFC tDCS in the current study could be related to stimulation effects on interoception. However, we note that we did not directly measure interoceptive processes in the current study, such that the current results do not allow drawing any inferences regarding the vmPFC’s contribution to interoception. We note that fairness considerations may influence social decisions not only via unfairness-triggered physiological responses; for example, the right dlPFC was hypothesized to encode abstract representations of social norms (Knoch et al. 2006), and disrupting dlPFC activation reduces the value assigned to punishing norm violations in vmPFC (Baumgartner et al. 2011). Moreover, other studies observed vmPFC stimulation to down-regulate emotion-related physiological reactions in tasks where participants were explicitly instructed to direct attention to their emotions (Hilz et al. 2006, Szeska et al. 2022). In situations requiring no intentional down-regulation of emotions such as in value-based decision making, vmPFC may attenuate the influence of emotion-triggered physiological responses to value computations rather than by directly inhibiting these bodily signals.

Besides value computations, the vmPFC was also related to representing confidence judgements (De Martino et al. 2013, Vaccaro and Fleming 2018, Shapiro and Grafton 2020). Consistent with this view, our data suggest that vmPFC tDCS increased decision confidence for accepted versus rejected offers. Although it was suggested that separate neural signals in vmPFC encode value and confidence (Lebreton et al. 2015), the confidence in having made the best choice scales with the difference in value between choice options (Bobadilla-Suarez et al. 2020, Shapiro and Grafton 2020). We speculate that, by attenuating the association between heart rate responses and decision making, vmPFC tDCS may have increased not only the value of unfair offers (as suggested by our computational model of inequality aversion) but also the subjective post-decision confidence after accepting unfair offers as a result of a stronger preference for accepting versus rejecting unfair offers. Alternatively, it is possible that the stimulation effects on confidence reports stem from changes in post-decision evaluation processes rather than from changes in the decision process itself. Given that the focus of our study was on the vmPFC’s role for normative decision making, we emphasize that the confidence results should be considered as exploratory. In any case, the current study provides support for the hypothesized link between vmPFC activation and decision confidence.

It is worth mentioning some limitations of the current study. First, while the spatial precision of our high-definition tDCS setup that minimizes stimulation effects on regions such as dlPFC is a strength compared to previous vmPFC tDCS studies, we note that without neuroimaging data we cannot fully rule out tDCS effects on other areas like dlPFC, lateral orbitofrontal cortex, or subgenual anterior cingulate cortex. The stronger electric fields in the vmPFC compared to other areas just implies that the observed behavioral effects are more likely to stem from stimulation of the vmPFC compared to other areas, and we emphasize that our stimulation setup has a higher spatial specificity than previous setups targeting the vmPFC (Gilam et al. 2017, Lo Gerfo et al. 2019). Moreover, previous studies reported that stimulating the dlPFC leads to a stronger weighting of fairness over selfish motives (Knoch et al. 2006, 2008), contrary to behavioral pattern observed here. However, the current design does not allow us to conclusively determine which specific medial prefrontal subregions contributed to the observed effects. Therefore, any claims regarding the spatial specificity of our results should be interpreted with caution. We moreover note that we did not include an active control condition (for example, cathodal tDCS or another brain region) because a further stimulation condition might have led to stronger task repetition effects. Although we statistically controlled for any confounding influences of the sensational experience of tDCS on choice behavior, we cannot fully rule out also site-unspecific stimulation effects at least partially contributed to the observed result pattern. Lastly, a further limitation is that we cannot distinguish between different motives in the current version of the Ultimatum game. Besides inequality aversion, rejection versus acceptance decisions can be motived by norm enforcement or punishment motives (Branas-Garza et al. 2014). Thus, it remains possible that the observed heart rate changes in the Ultimatum game are evoked by stronger punishment intentions rather than inequality aversion per se.

Taken together, the current findings deepen our understanding of the vmPFC’s contribution to computing trade-offs between social norms and self-interests. The vmPFC is causally involved in attenuating the link between physiological heart rate responses to fairness violations and the decision process, thereby reducing aversion to disadvantageous inequality and strengthening subjective confidence when accepting unfair offers. This advances our knowledge of the brain areas causally underlying the interaction between physiological reactions to emotions and cognition in social and potentially also non-social domains of decision making.

Author contributions

Patricia Christian (Conceptualization [equal], Formal analysis [equal], Investigation [lead], Visualization [equal], Writing—original draft [equal]), Jakob Kaiser (Conceptualization [equal], Methodology [equal], Writing—review & editing [equal]), Georgia E. Kapetaniou (Investigation [equal], Writing—review & editing [equal]), Alexander Soutschek (Conceptualization [equal], Formal analysis [equal], Supervision [lead], Visualization [equal], Writing—original draft [equal]), and Simone Schütz-Bosbach (Conceptualization [equal], Writing—review & editing [equal])

Conflicts of interest

None declared.

Funding

A.S. received support from an Emmy Noether fellowship from the German Research Foundation (SO 1636/2-2) and an Exploration grant from the Boehringer Ingelheim Foundation.

Data availability

The raw data underlying the findings are available on Open Science Framework (OSF; https://osf.io/d6m54/overview?view_only=36364c95eab747f6b4e6292d6cab3be6).