https://orcid.org/0000-0002-0457-0233
Abstract
Human behavior often aligns with fairness norms, either voluntarily or under external pressure, like sanctions. Prior research has identified distinct neural activation patterns associated with voluntary and sanction-based compliance or non-compliance with fairness norms. However, an investigation gap exists into potential neural connectivity patterns and sex-based differences. To address this, we conducted a study using a monetary allocation game and functional magnetic resonance imaging to examine how neural activity and connectivity differ between sexes across three norm compliance conditions: voluntary, sanction-based, and voluntary post-sanctions. Fifty-five adults (27 females) participated, revealing that punishment influenced decisions, leading to strategic calculations and reduced generosity in voluntary compliance post-sanctions. Moreover, there were sex-based differences in neural activation and connectivity across the different compliance conditions. Specifically, the connectivity between the right dorsolateral prefrontal cortex and right dorsal anterior insular appeared to mediate intuitive preferences, with variations across norm compliance conditions and sexes. These findings imply potential sex-based differences in intuitive motivation for diverse norm compliance conditions. Our insights contribute to a better understanding of the neural pathways involved in fairness norm compliance and clarify sex-based differences, offering implications for future investigations into psychiatric and neurological disorders characterized by atypical socialization and mentalizing.
Introduction
Fairness norms are societal standards that dictate what is considered fair or unfair, influencing various aspects of daily life, including business, politics, and interpersonal interactions (Taing and Chang 2020; Gross and Vostroknutov 2022). Adherence to these norms can be derived by two primary motivations. First, individuals may voluntarily comply with fairness norms when they are perceived as inherently moral, guiding their preferences without external influence or consideration of self-interest (Bicchieri et al. 2022; Gross and Vostroknutov 2022). For example, this may lead to decisions to equitably divide desirable resources, such as money, or to share opportunities with others. Second, compliance with fairness norms can be enforced through external pressures, such as the avoidance of punishment, motivating individuals to adhere to social norms (Nowak and Sigmund 2005; Gächter et al. 2008; Buckholtz and Marois 2012). Examples of this include paying taxes to avoid penalties. Importantly, individual’s behavior regarding fairness is influenced by both of these mechanisms (Chen et al. 2019; Gross and Vostroknutov 2022), with the former termed voluntary fairness norm compliance and the latter termed sanction-induced fairness norm compliance.
Neuroeconomics studies using functional magnetic resonance imaging (fMRI) have identified crucial brain regions involved in compliance with fairness norms (Ni and Li 2021; Wyss and Knoch 2022), including the dorsolateral prefrontal cortex (DLPFC), ventromedial prefrontal cortex (vmPFC), dorsal anterior cingulate cortex (dACC), temporoparietal junction (TPJ), and anterior insula (AI) (Buckholtz and Marois 2012; Feng, Luo, and Krueger, 2015; Ni and Li 2021). The rDLPFC is particularly crucial in norm compliance due to its recognition as a core region for cognitive control and decision making processes (Speitel et al. 2019; Hu et al. 2022; Li et al. 2023), making it the focus of our study. While the rDLPFC mediates both voluntary and sanction-induced compliance with fairness norms, it plays opposing roles in these two types of compliance (Ruff et al. 2013; Chen et al. 2019). However, there is a gap in our understanding of the connectivity of the rDLPFC in this context. Our primary objective is to address this gap and gain insights into the brain network associated with the rDLPFC that underlies compliance with fairness norms.
The second gap in this field lies in its limited exploration of sex-based differences, despite emerging evidence suggesting variations in key neural processes that mediate social decision making. Recent works found that males show greater variability in risk-taking (Lei et al. 2021; Xiao et al. 2022) and social preferences, whereas females tend to demonstrate more consistent and moderate preferences (Thoni and Volk 2021). For example, males tend to display either purely selfish or entirely selfless behavior, while females typically exhibit a preference for equitable sharing (Andreoni and Vesterlund 2001). These sex-based differences are also evident at the neural level (Fumagalli et al. 2010; Lv et al. 2023), particularly within the prefrontal cortex (Chen et al. 2019; Korucuoglu et al. 2020). For instance, female demonstrate reversed activation patterns compared to males in social dilemmas and trust games, particularly in brain regions associated with social decision-making processes (Rodrigo et al. 2014). Therefore, it is anticipated that males and females would show distinct behavior patterns and neural activation in their adherence to fairness norms.
The present study used a behavioral experiment and brain scans with fMRI techniques to examine the neural basis of fairness norm compliance. The behavioral task included games that assessed voluntary and sanction-based compliance with fairness norm, and examined the impact of sanctions on subsequent decision making. Drawing on existing literature and the central tenet of prospect theory that emphasizes the greater impact of losses (such as punishment in this context) compared to gains (Kahneman and Tversky 2013), we hypothesized that (H1) the sanction condition would elicit greater compliance with fairness norms than the voluntary condition (Spitzer et al. 2007). Additionally, considering that sanctions have been shown to diminish cooperation (Abbink et al. 2017) by shifting individuals’ focus towards cost–benefit calculations (Li et al. 2009), we predicted that (H2) the post-sanctions condition would result in reduced compliance with fairness norms compared to the voluntary, pre-sanction condition.
Next, we tested how different brain regions encode different types of fairness norm compliance. Fairness norm compliance involves the evaluation of conflicting internal (self-interest) and external (social norm) information (Spitzer et al. 2007; Hu et al. 2022). Therefore, the network mediating social norm compliance should encompass the following brain regions. First, the right temporoparietal junction (rTPJ) is involved in mentalizing about the mental states of others (FeldmanHall et al. 2012). Second, the posterior cingulate cortex (PCC) or precuneus responds to “self-consciousness” and encodes participants’ own mental states (Leech and Sharp 2014). Third, the anterior insula (AI), as part of the salience network, mediates the integration of different functional systems in the brain (Wang et al. 2017; Nomi et al. 2018). Fourth, the rDLPFC mediates cognitive control (Kurzban et al. 2013; Yuan et al. 2020; Molnar-Szakacs and Uddin 2022). Lastly, the ventromedial prefrontal cortex (vmPFC) is involved in the evaluation of punishment information and the computation of the value of actions (Castegnetti et al. 2021). We hypothesize that rTPJ, rDLPFC, PCC, AI, and vmPFC mediate fairness norm compliance (H3.1). Furthermore, we propose that rTPJ, rDLPFC and vmPFC may exhibit stronger activation in sanction-induced social norm compliance (H3.2), given the increased mentalizing, computation and inhibition induced by sanctions.
Last, we examined potential sex-based differences in both behavioral and neural patterns. Our prior work found that sex-based difference in behaviors and the involvement of the rDLPFC in fairness norm compliance (Chen et al. 2019). Specifically, enhancing the excitability of rDLPFC in female through anodal transcranial direct current stimulation (tDCS) led to a decrease in their voluntary but an increase in sanction-induced compliance with fairness norm; Interestingly, the effects of anodal tDCS on compliance types were opposite in direction for male, with voluntary transfer increasing and sanction-induced transfer decreasing compared to the sham group. However, there remains a dearth of understanding regarding how fairness norm compliance engages neural mechanisms related to the rDLPFC and influences decision-making in males and females. Given the absence of definitive evidence in the literature, we have refrained from formulating formal hypotheses on this matter. Instead, our aim is to examine the common and distinct neural mechanisms that underlying various norm compliance processes between sexes and different fairness norm compliance conditions.
Material and methods
Participants
Sixty-three right-handed Chinese college students volunteered to participate in this study. Data of eight participants were not included for analysis because of excessive motion during fMRI scanning, leaving 55 participants (27 female and 28 male) with mean age of 20.43 (SD = 1.51, range = 18–24) years. All participants had normal or corrected-to normal vision, and none of them reported any history of neurologic or psychiatric disorders. They gave written informed consent prior to the experiment, which was approved by the Institutional Review Board of Southwest University. They received at least ¥65 for their participation. The final monetary outcome depended on the tokens they earned in the game (see details of the experimental paradigm).
Experimental paradigm and measures
During the scanning, participants engaged in a two-person monetary allocation game, completing both voluntary and punishment conditions as a within-subject factor. Following the procedures outlined in (Chen et al. 2019), each participant acted as the proposer and made decisions in both conditions, interacting with different responders (Fig. 1A). At the start of each trial, both players received an initial endowment of 25 tokens. The proposer was then allocated an additional 100 tokens and tasked with deciding how many tokens to transfer to the responder by adjusting a bar on the screen. The transfer could range from 0 to 100 tokens in increments of 10. Throughout the scanning process, participants were informed that they were interacting with different responders in each trial, and that the feedback in the punishment condition had been predetermined by real individuals (although, in reality, the responder was always a preprogrammed computer). For details on the probability of refusal and the number of tokens subject to punishment, refer to the supplementary materials, Table S1. In the punishment condition, for each token the responder chose to sacrifice in punishing the proposer, the proposer’s earnings were reduced by 5 tokens.
The experimental design. (A) In this experimental setup, each participant acted as the proposer and made decisions in both conditions, interacting with different responders. At the start of each trial, both players received an initial endowment of 25 tokens. The proposer was then allocated an additional 100 tokens and tasked with deciding how many tokens to transfer to the responder by adjusting a bar on the screen. In the voluntary condition, the proposer's transfer of X tokens serves as the proposal, while in the punishment condition, the responder can accept or reject the proposal and penalize the proposer by reducing their payoff. If the proposal is accepted, the transfer is executed; otherwise, the responder can invest Y tokens from their initial endowment to penalize the proposer by deducting 5 times the invested amount. The block sequence during the scanning involves three consecutive blocks: Voluntary (V), punishment (P), and voluntary after punishment (VP). (B) Each block consisted of 25 trials of the same game type, with each game being explained on the screen before commencing the block. In both the voluntary and punishment conditions, participants were presented with an axis marked 0–100, and the initial bar positions were randomly determined. Participants indicated their preference by adjusting a highlighted bar’s position using corresponding left (decrease 10 tokens) and right (increase 10 tokens) buttons and confirming their choices by another one. They were instructed to make their selections as swiftly as possible within a 6-second response window. Only in the punishment condition were participants informed of the ostensible player’s response (“accept” or “reject and punish”). Subsequently, the screen displayed the payoff of this trial, followed by a jittered intertrial interval before commencing a new trial.
Participants completed three blocks of task in the following order: voluntary (V), punishment (P) and voluntary after punishment (VP, a repeated voluntary condition post-sanctions). Each block consisted of 25 trials of the same game type, with each game being explained on the screen before commencing the block (Fig. 1B). In both the voluntary and punishment conditions, participants were presented with an axis marked 0–100, and the initial bar positions were randomly determined. Participants indicated their preference by adjusting a highlighted bar’s position using corresponding left (decrease 10 tokens) and right (increase 10 tokens) buttons and confirming their choices by another one. They were instructed to make their selections as swiftly as possible within a 6-second response window. Only in the punishment condition were participants informed of the ostensible player’s response (“accept” or “reject and punish”). Subsequently, the screen displayed the payoff of this trial, followed by a jittered intertrial interval before commencing a new trial. At the conclusion, one trial was randomly chosen, and 10% of the participants’ payoff was converted into real money and added to their payment, with the accumulation of tokens determining their final payoff.
This experimental paradigm effectively captures the voluntary fairness compliance (V condition), sanction-based compliance (P condition), and the learning effect of sanctions/punishment (VP condition). In the V condition, token transfer is based on the pure social norm compliance, while in the P condition, the threat of punishment may prompt participants to give more to avoid sanctions. The difference between the V condition and VP condition reflects the impact of punishment on participants.
Exit questionnaire
Following the fMRI scanning, participants completed an exit questionnaire to assess their experiences in the scanner and their perceptions of fairness, serving as a manipulation check. Participants were asked to rate their concentration and seriousness on a scale from 0 (low concentration/careless) to 100 (high concentration/carefully). Fairness beliefs were measured using a series of seven questions, such as “How fair do you think it is to give the responder a transfer of 0/10/20/30/40/50/60?” with responses ranging from 1(very unfair) to 7(very fair). A benchmark of perceived fairness was established when a response in the sequence was ≥4 (neutral). In addition, we captured anger expectations related to how the responder is expected to feel after their offers; how many punishment tokens a responder should invest at different transfer levels and how they felt when they were punished.
fMRI acquisition and Preprocessing
Functional magnetic resonance images were collected using a 3 T Siemens Magnetom Prisma magnetic resonance scanner at the SWU Brain Imaging Center equipped with a 64-channel head coil. The images were acquired using a gradient echo planar imaging (EPI) sequence with interleaved multi-band. The imaging parameters were as follows: repetition time (TR) of 2000 ms, echo time (TE) of 30 ms, flip angle of 90°, field of view (FOV) of 192 mm2, matrix size of 112 by 112, slice thickness of 3 mm, and MB-factor of 2. The structural image acquisition, used for registration, was scanned using a T1-weighted MPRAGE sequence with the following imaging parameters: TR of 2530 ms, TE of 2.98 ms, flip angle of 7°, FOV of 256 mm2, and slice thickness of 1 mm.
Preprocessing and statistical analyses were performed using SPM 12 (Wellcome Department of Imaging Neuroscience, London, United Kingdom; www.fil.ion.ucl.ac.uk/spm). To correct for head motion, all functional images were realigned to the first image and subsequently unwarped to remove residual movement-related variance because of susceptibility-by-movement interaction. Co-registration of the structural images with the functional data was followed by segmentation of the co-registered images to obtain the normalization parameters. These parameters were then used to normalize the images to a standard space (Montreal Neurological Institute template, MNI) and resampled to a spatial resolution of 2 mm × 2 mm × 2 mm. Last, spatial smoothing was performed using a Gaussian kernel of 8 mm full-width-at-half-maximum.
fMRI data analyses
General linear modeling (GLM) was employed for the first level analysis of functional scans from each participant. All regressors were modeled using a canonical hemodynamic response (HRF) with time dispersion derivatives, allowing for a variation of plus or minus a second in the peak response (Engelmann et al. 2019). Three regressors of interest, which correspond to the three task conditions (V, P, and VP) were modeled for the decision period for the entire period of reaction time on each trial, from the onset of the decision screen until participants pressed the confirm button, accurately capturing the pure duration of decision-making (Engelmann et al. 2019). Last, other regressors of no interest were included in the model. Fixation periods were not modeled and thus acted as an implicit baseline.
Our main interest was examining sex-based differences in decision making across different conditions. We specifically targeted group comparisons within the main conditions (V, P, and VP) and the differences between conditions (P-V and VP-V). All statistical tests were carried out at the voxel-level threshold of P < 0.001 and the cluster level with a corrected P < 0.05, using family-wise error (FWE) correction methods.
Regions of interest (ROIs) and generalized psychophysiological interactions analysis (gPPI)
Our analyses are based on the region of interest (ROI) known as the right dorsolateral prefrontal cortex (rDLPFC). We derived the ROI mask by intersecting the neurosynth term "dlpfc" with the AAL right middle frontal cortex mask, using neurosynth.org and the WFU Pickatlas. To search for the various circuits of the rDLPFC under different conditions, we employed the generalized form of context-dependent psychophysiological interactions (gPPI) to establish couplings with the rDLPFC at a within-subject level. This was achieved using the gPPI toolbox (McLaren et al. 2012) and the same GLM as mentioned above. Subsequently, we conducted between-subject level tests to identify potential differences. The activation clusters were determined based on a voxel-level threshold of P < 0.001 and a cluster level threshold of corrected P < 0.05, using FWE methods.
Furthermore, to characterize the activation and functional connectivity patterns, we used rfxplot (Gläscher 2009) to extract the regression coefficients (percent signal change) for the canonical HRF regressors from 6-mm spheres around the peak voxel of individual participants, which demonstrated the BOLD response or functional connectivity in various conditions.
Results
Manipulation checks
The majority of participants were able to concentrate on the experiment (mean ± SD: 77.91 ± 18.65) and expressed a high level of seriousness (mean ± SD: 83.95 ± 15.68). They believed that higher transfers were associated with greater fairness, lower anger, and fewer tokens used for punishment (Table S2). Importantly, there were no significant effect of sex (Fs < 2.89, ps > 0.095) on fairness beliefs. When experiencing punishment, participants reported feeling anger (65.45%), understanding (20%) or ambivalence (14.55%), with no sex differences (Fisher’s exact test was used, P = 0.625 < 0.05), details in Table S3.
Behavioral results
The average transfers by sex and task condition are shown in Table 1. As shown in Fig. 2A, fairness perception was significantly correlated with fairness behavior (i.e. amount of transfer) in V (Spearman’s r = 0.52, P = 0.005), P (Spearman ‘s r = 0.31, P = 0.104), and VP (Spearman ‘s r = 0.43, P = 0.024) conditions for male respectively, but not in female (|r|s < 0.14, ps > 0.494). However, the differences in correlations between sexes were not statistically significant (Table S4). Multivariate analysis of variance for fairness perceptions and transfers of different condition as outcome demonstrated no significant main effect of sex (Fs(1,53) < 1.04, ps > 0.313). There was a significant main effect of task condition (F(2, 108) =64.68, P < 0.001, ηp2 = 0.545), with transfers in the P condition being higher than in the V condition (P < 0.001), while transfers in the VP condition were lower than in the V condition (P < 0.001) at alpha =0 0.025. Refer to the results in Fig. 2B for more details.
Behavioral results by task and sex. (A) We tested the correlation between fair beliefs and fair behavior. The X-axis represents the average transfers of each participant made by participants in the V, P, and VP conditions, while the Y-axis represents the participant’s benchmark of perceived fairness. (B) The average transfers were displayed for each condition by broken down by sex. (C) The response times (RT) of each condition were also displayed by categorizing by sex. It is important to note that confidence bounds around regression lines reflect a 95% confidence interval, and error bars represent standard deviation. Additionally, significance levels are indicated by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001).
The amount of transfers (mean ± SD) by sex and task condition.
| Sex | N | Fairness Perception | V condition | P condition | VP condition |
|---|---|---|---|---|---|
| Female | 27 | 34.81 ± 13.97 | 29.90 ± 17.34 | 44.18 ± 9.67 | 23.11 ± 16.08 |
| Male | 28 | 30.36 ± 18.15 | 28.46 ± 15.22 | 43.96 ± 5.80 | 24.01 ± 15.57 |
| Sex | N | Fairness Perception | V condition | P condition | VP condition |
|---|---|---|---|---|---|
| Female | 27 | 34.81 ± 13.97 | 29.90 ± 17.34 | 44.18 ± 9.67 | 23.11 ± 16.08 |
| Male | 28 | 30.36 ± 18.15 | 28.46 ± 15.22 | 43.96 ± 5.80 | 24.01 ± 15.57 |
The amount of transfers (mean ± SD) by sex and task condition.
| Sex | N | Fairness Perception | V condition | P condition | VP condition |
|---|---|---|---|---|---|
| Female | 27 | 34.81 ± 13.97 | 29.90 ± 17.34 | 44.18 ± 9.67 | 23.11 ± 16.08 |
| Male | 28 | 30.36 ± 18.15 | 28.46 ± 15.22 | 43.96 ± 5.80 | 24.01 ± 15.57 |
| Sex | N | Fairness Perception | V condition | P condition | VP condition |
|---|---|---|---|---|---|
| Female | 27 | 34.81 ± 13.97 | 29.90 ± 17.34 | 44.18 ± 9.67 | 23.11 ± 16.08 |
| Male | 28 | 30.36 ± 18.15 | 28.46 ± 15.22 | 43.96 ± 5.80 | 24.01 ± 15.57 |
The response times (RT) of participants in the decision phase were analyzed using repeated measures ANOVA with the factor of sex. We found a significant effect of task condition (F(2,106) = 11.65, P < 0.001, ηp2 = 0.180), indicating that the RT of decisions were slower in the V condition (mean RT = 2.88 s) than that in the P condition (mean RT = 2.68 s; P < 0.001, 95% CI = [0.094, 0.299]) and VP condition (mean RT = 2.64 s; P < 0.001, 95% CI = [0.133, 0.338]). Both male and female group showed a similar main effect of task condition, with females only showing a significant difference between the V condition and the VP condition (see Fig. 2C). There were no significant main effects or interaction effects for sex and task condition (Fs < 1.74, ps > 0.181), and a marginally significant main effect of sex in the RT difference between the P condition and V condition (F(1,53) = 3.38, P = 0.072).
Neuroimaging results
Whole brain results
The results of whole brain analysis are presented in Table S5, and it was found that using individual fairness perception as a covariate did not change the results. The study first examined the differences in brain activation during fairness norm compliance in different conditions (Table S5.1). The contrast of P versus V conditions revealed increased activation in regions including left middle frontal cortex (MNI [−48, 22 44]), left inferior occipital cortex (MNI [−26, −100, −12]) and left postcentral cortex (MNI [−26, −100, −12]). Compared to V condition, VP condition had stronger activity in bilateral inferior frontal cortex (MNI [−5, 20, 24; 62, 18, 16]), bilateral precuneus (MNI [2, −34, 28]), left TPJ (MNI [−46, −50, 26]), right inferior occipital cortex (MNI [22, −100, −4]) and left middle occipital (MNI [−30, −62, 32]).
Secondly, we analyzed the fMRI data to identify sex-based neural differences among conditions (details in Table S5.2). We found that males had stronger activity in the rTPJ (MNI [42, −66, 32]), bilateral PCC (MNI [10, −56, 32; −16, −64, 24]), right ventromedial prefrontal cortex (vmPFC, MNI [10, 48, −4]), middle temporal cortex (bilateral, MNI [54, −34, 0; −56, 50, 0]) and middle occipital cortex (MNI [34, −88, −6]) compared to females in the V condition. Also, compared with females, males had stronger activity in left postcentral gyrus (MNI [−44, −40, 64]) in the P condition. There were no sex-based neural differences in the VP condition.
Next, we examined neural differences among conditions. The precuneus cortex (MNI [10, −70, 30]) showed a significant interaction effect between sexes and condition (P vs V). There was no significant interaction effect between sexes and condition in other contrasts (e.g. VP vs V or VP vs P). We also examined which regions showed greater activity due to punishment in females vs males. Compared to the V condition, in the P condition, the middle temporal cortex (MNI [−60, −10, −6]), PCC (MNI [−10, −60, 24]), and inferior occipital cortex (MNI [−32, −98, −10]) had greater activity in females, whilst the postcentral cortex (MNI [−38, −32, 46]) had greater activity in males. We next focused on how the punishment experience affected voluntary fairness norm compliance. Females had stronger activity in the PCC (bilateral, MNI [8, −38, 24; −14, −62, 26]), TPJ (bilateral, MNI [−42, −52, 24; 46, −44, 18]), rDLPFC (MNI [44, 24, 18]), left vmPFC (MNI [−6, 64, 12]), middle temporal cortex (MNI [−60, −30, −2]) and inferior occipital cortex (MNI [−34, −92, −10]) when deciding in the VP condition compared with the V condition, while males had stronger activity in the Lingual gyrus (MNI [−22, −98, −16]) in the VP compared to the V condition.
Last, we examined in our ROIs whether the relationship between mean transfers and activity in different conditions. We also found that the signals of rDLPFC and rTPJ increased with the decrease in the transfer in the P condition (r = −0.39, P = 0.039 and r = −0.45, P = 0.015, respectively) in the male group (Fig. 3A and B), but not in females (|r|s < 0.354, ps > 0.073).
The sex-based difference in the activation and functional connectivity patterns in different fairness compliance conditions. Sanction-induced fairness norm compliance was associated with the neural activity of the rDLPFC (A) and rTPJ (B). The connectivity between rDLPFC and rdAI was correlated with the average of transfer (C). The trend of rDLPFC-rdAI connectivity in three conditions were displayed for females (D) and males (E), respectively. The dot plots in (D) and (E) represent the individual participant's mean strength of connectivity in each condition and are connected to illustrate the trend.
gPPI results
We applied gPPI to examine whether there are sex-based differences in connectivity in the different conditions. Results are shown in Table S6. Compared to females, we found stronger rDLPFC connectivity with right dorsal anterior insular (rdAI, MNI [44, 10, 2]), precentral (MNI [−46, 6, 48]) and postcentral cortex (MNI [28, −28, 52]) during voluntary fairness compliance in males. We observed a stronger connectivity between the rDLPFC and the inferior frontal cortex (MNI [−48, 36, 10]) in males compared to females in the P condition. In the VP condition, there was stronger rDLPFC connectivity with vmPFC (bilateral, MNI [−6, 52, −2]), ACC (MNI [−2, 44, 4]) and left inferior frontal cortex (MNI [−54, 30, −4]) in males vs females.
Next, we examined neural connectivity differences between conditions. Compared to the V condition, females had stronger connectivity between rDLPFC and middle occipital cortex (MNI [32, −72, −2]) in the P condition. Experiencing the punishment (VP-V), females presented stronger connectivity, compared to males, in the rDLPFC to right precuneus (MNI [22, −60, 26]), rTPJ (MNI [30, −58, 40]), rdAI (MNI [46, 16, 0]), precentral cortex (MNI [42, 6, 36]) and inferior parietal cortex (MNI [−30, −54, 42]). Males presented stronger connectivity between the rDLPFC and the left middle posterior insula (mid_PI, MNI [−38, −8, 4]).
Last, we considered the brain-behavior correlations which the relationship between mean transfers and rDLPFC functional connectivity is different in the different conditions. Given the diverse functions of the insular cortex, which is divided into three distinct subdivisions (Deen et al. 2011; Chang et al. 2013; Glasser et al. 2016; Krueger et al. 2020; Zhao et al. 2023), we sought to determine the specific involvement of each subdivision in our results. We used the three masks which represent a tripartite functional parcellation of the right insular cortex into the dorsal anterior insular (dAI), the ventral anterior insular (vAI), and the posterior insular (PI) based on the templates by Chang (Chang et al. 2013) in NeuroVault.org (Gorgolewski et al. 2015). We found that the rDLPFC-rdAI coupling increased with the decrease in voluntary norm compliance (r = −0.40, P = 0.041) in females (Fig. 3C), but not in males (r = −0.11, P = 0.584). The connectivity strength of rDLPFC-rdAI was on the rise in females (Fig. 3D), while the connectivity strength only decreased in the P condition in males (Fig. 3E).
Discussion
In this study, a monetary allocation game was used to examine how individuals make decisions about fairness norm compliance in different situations. We also used fMRI to measure brain activity and to compare differences between men and women. The results of the study confirmed previous findings (Chen et al. 2019) that participants were less likely to share money in a voluntary situation, but more likely to share money when there was a punishment involved. The fMRI results also showed differences in brain activity between the different conditions. Interestingly, men showed only a slight change in brain activity across the conditions, while women showed more significant differences. Additionally, the connectivity between the rDLPFC and rdAI regions of the brain, which is related to fairness norm compliance, suggests that this process involves integrative processing rather than intuitive behavior. Overall, the study highlights the important role of sex differences in adherence to fairness norms.
Behavioral and neural signatures of different fairness norm compliance
In the voluntary fairness norm compliance condition, the self-interested strategy reflects the goal of money-seeking and dictates zero transfer. Behavioral results of our study suggested that people may act based on pre-existing fairness norms (Chen et al. 2019). Specifically, participants’ transfers in the voluntary condition were about 29.17, mirroring their perception of fairness.
In contrast to previous research (Chen et al. 2019), our study found that the sanction condition led to greater compliance with fairness norms compared to the voluntary condition, supporting our hypothesis (H1). Threats of punishment or sanctions are effective in encouraging people to conform to social norms, as they are seen as a way to ensure that rule breakers are held accountable (Mathew and Boyd 2011). This suggests that individuals are sensitive to the possibility of punishment and may act in accordance with fairness norms to avoid it (Gianotti et al. 2018). Our neuroimaging data also supports these findings. Specifically, we observed significant differences in brain activity between the voluntary and punishment conditions, particularly in the left DLPFC and inferior parietal cortex. While our results did not fully support H3.2, they did align with previous research by Spitzer et al. (2007), which found increased activation in the bilateral DLPFC when comparing punishment to a control condition. In our study, we also observed activation in the parietal cortex when comparing punishment to voluntary conditions, suggesting that participants may engage in strategic decision-making based on expected outcomes and rational choices (Li et al. 2009). In other words, participants may be more focused on calculating potential gains and losses rather than solely being motivated by social factors, leading to behaviors aimed at maximizing their earnings.
Additionally, we noticed a significant decrease in voluntary compliance after sanctions compared to voluntary compliance before sanctions (support H2), and most participants reported feeling angry when they were punished (see Table S3). One possible explanation for this is that the decisions made after being punished are influenced by the emotions experienced post-punishment (Bechara 2004; Chen et al. 2019). The reluctance to transfer resources in voluntary compliance post-sanctions may be a strategy for participants to maintain their cost–benefit calculations and restore emotional balance (Zheng et al. 2017). However, the average transfer amount was still relatively high at 23, which is well above what would be expected based on standard economic theory (Kahneman et al. 1986). This suggests that the fairness norm still plays a role in participants' decision-making. Furthermore, the contrast between voluntary compliance and voluntary compliance post-sanctions revealed different brain responses in the bilateral DLPFC, TPJ, and bilateral occipital cortex, which are areas associated with self-control and theory of mind processes involved in prosocial behavior (Speer and Boksem 2020). These findings provide partial support for H3.1.
Sex-based differences in behavioral and neural patterns within different fairness norm compliance
The study found that males' transfer of resources in the voluntary condition aligned with their perception of fairness, while there was no significant correlation between transfers and fairness beliefs in females. One possible explanation for this gender difference is that males may have been more genuine in expressing their feelings and allocation in the post-study questionnaire, while females may have been influenced by the experiment's intentions and provided biased self-reports (Trelohan 2021). Although the behavioral patterns of males and females were similar across different conditions, further insights can be gained from neural imaging results (Cascio et al. 2015).
Firstly, our findings revealed that males showed stronger neural in brains regions associated with social cognition and self -related processes (Van Overwalle 2009; Scholz et al. 2017) when behaving in a voluntary fair manner, which supported our previous hypotheses (H3.1). This suggests that males may require more neural processing to behave fairly compared to females. The increased connectivity between the rDLPFC and the rdAI in males suggests that they may integrate more information to balance values and comply with fairness norms voluntarily, based on the fact that rdAI responds to integrate information from different functional systems in the brain and coordinate brain network dynamics (Molnar-Szakacss and Uddin, 2022; Nomi et al. 2018; Wang et al. 2017), and the rDLPFC is a key node in the selecting and integrating task-relevant information for goal-directive behavior (Kurzban et al. 2013; Yuan et al. 2020). On the other hand, females exhibit a negative representation of rDLPFC-rdAI connectivity, indicating that they may have an easier path towards reaching voluntary fairness decisions (Zakim and Mitchell 2013), possibly due to cultural expectations and gender stereotypes (Soutschek et al. 2017). This suggests that fairness norms may be more internalized and followed voluntarily by females (Gross and Vostroknutov 2022).
Secondly, the results under punishment conditions provide us with intriguing insights. As in prior research, the sequence-specific design led to a decrease in response time for button presses (Thomas and Nelson 2001; Lum et al. 2023). However, compared to the voluntary condition, males exhibited a greater decrease in response time than female. At neural level, inferior parietal cortex had stronger activation than female, and the role for rDLPFC and rTPJ was less demanding when male gave more transfer in punishment condition. The rTPJ links to theory of mind and is positively correlated with donation behavior (Gianotti et al. 2018). These results suggest that males may turn to strategic calculations rather than social decisions in punishment condition. Moreover, the relationship between rDLPFC and transfer is consistent with our previous research (Chen et al. 2019), which found that the sanction-induced transfer has decreased under anodal tDCS compared to sham. Regardless of whether it is the intuitive prosociality model or the reflective model, the rDLPFC appears to integrate information and inhibit intuitive motives during conflict (Yamagishi et al. 2016; Zinchenko et al. 2021). Previous research have overlooked sex differences, and our previous study (Chen et al. 2019) and current study results suggest that there may be sex-based difference in intuitive motivation to some extent (Strang et al. 2015). Males giving more money directly in punishment condition and female presenting intuitive prosociality in voluntary may be an internalized and intuitive norm-abiding behavior. The intricacies of evaluating the dynamics of social interactions are inherently more challenging than non-social interaction and introduce the potential for assessment errors (Yamagishi et al. 2016). Clearly, further studies are needed to be conducted to investigate the role of the rDLPFC in norm compliance directly.
Lastly, compared to female, males showed stronger connectivity between rDLPFC and vmPFC (ACC), left inferior frontal cortex in voluntary post-sanctions condition. The ACC mediate conflict monitoring and strategic adaptation of decision making. The ACC has been postulated as the cerebral region where diverse information or decision variables are integrated and update. Moreover, the vmPFC is involved in value computations. This implies that males may rely more on contextual and adjusted value computations to adapt to specific circumstances.
Sex-based differences in neural patterns across different fairness norm compliance
The neural findings revealed slight changes in neural patterns in three conditions for males, while females exhibited more significant neural differences in subsequent conditions. Both whole-brain analysis and gPPI results indicated notable variations in neural activity patterns among females across conditions. In comparison to voluntary actions, females showed stronger activation in the left middle temporal cortex and left PCC during punishment. The PCC is known to encode one's own mental states (Leech and Sharp 2014), while the middle temporal cortex is associated with social perception (Liu et al. 2023). This suggests that females adjust their strategy to consider the balance between self-interest and the interests of others.
After being threatened with punishment, females show a more complex decision-making process in the voluntary post-sanctions compared to the voluntary pre-sanctions. They exhibit stronger activation in the brain's social cognition system and self-related processes, as well as a connection between the rDLPFC and inferior parietal cortex, TPJ, and right insular (Li et al. 2009). This suggests that females integrate more information in the voluntary post-sanctions instead of relying on intuitive prosociality. These findings partially support the idea proposed by Fehr and Rockenbach (2003) that the presence of punishment may shift the motivation for cooperative behavior from being spontaneous to becoming obligatory in order to avoid sanctions. This process essentially involves displacing internal motivation with external incentives. Therefore, when punishment is removed and internal motivation has been displaced, and external incentives are no longer present, compliance with fairness norms significantly decreases. Another possible explanation is that this may restore emotional equilibrium. This highlights the need for more in-depth investigation in future studies.
Our research revealed interesting differences in the connectivity of the rDLPFC-rdAI between males and females across three task conditions. Females showed an increasing trend in connectivity, while males showed a V-shape trend. The rdAI is responsible for identifying homeostatically relevant stimuli, while the rDLPFC integrates information and inhibits intuitive motives during conflict. This suggests that the rDLPFC-rdAI connection may signify a process of integrative processing rather than intuitive behavior. It appears that fairness norm compliance could serve as a more accessible decision-making template in various situations. Females seem to exhibit more intuitive adherence to fairness norms, but their decision-making process becomes more complex in response to contextual changes. On the other hand, males may adhere to fairness norms as an advantageous strategy for optimal decision-making in punishment conditions, but more consideration seems to be required in voluntary conditions. These findings also shed light on the role of the rDLPFC as discussed in our previous work. Interventions targeting the rDLPFC may influence this connection, leading to alterations in the integration of internal and external information, thereby affecting fairness norm compliance between sexes.
The findings of the current study provide strong support for the existence of sex differences in compliance with fairness norms. However, it is important to acknowledge potential limitations. Firstly, we did not specifically examine the impact of the social context setting, and it is possible that interactions with real versus virtual humans could yield different results. Nonetheless, previous research using a similar setting and participants has shown that participants trust the experimental instructions, suggesting that the context setting may not have a significant impact on the current findings. Secondly, the imaging results only provided indirect evidence for these differences and did not allow us to definitively determine the origins of sex-based differences in fairness norm compliance, despite offering detailed insights into this variability. Future studies could investigate the effects of hormones or genes to further clarify these differences.
In summary, our study shows that there are differences between men and women in how they adhere to fairness norms. Women tend to act more intuitively in voluntary situations, while men are more strategic, especially when facing potential punishment. Men also tend to give more money directly to avoid punishment, reflecting a more instinctive adherence to norms. Our fMRI analysis confirmed the role of the rDLPFC in fairness norm compliance and highlighted the importance of the rDLPFC-rdAI neural pathway in this process. These findings could be crucial for understanding and investigating psychiatric and neurological disorders that affect social and mentalizing abilities. We urge further research to explore the role of the rDLPFC-rdAI neural pathway in such disorders.
Author contributions
Wanting Chen conceptualization, data curation, formal analysis, investigation, visualization, writing—Original draft, writing—Review & editing Zhibing Xiao conceptualization, data curation, formal analysis Ofir Turel conceptualization, writing—Original draft, writing—Review & editing Shuyue Zhang conceptualization, writing—Review & editing Qinghua He conceptualization, investigation, methodology, project administration, supervision, writing—Original draft, writing—Review & editing
Funding
This work was supported by research grants from the National Natural Science Foundation of China (31972906); Project of Sichuan Key Laboratory of Psychology and Behavior of Discipline Inspection and Supervision in 2023 (SJX230101); the Innovation Research 2035 Pilot Plan of Southwest University (SWUPilotPlan006); the High-end Foreign Expert Introduction Program (G2022168001L) and the Open Research Fund of the State Key Laboratory of Cognitive Neuroscience and Learning (CNLZD2102).
Conflict of interest statement: All authors declare no conflict of interest.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
