(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Social bonding in groups of humans selectively increases inter-status information exchange and prefrontal neural synchronization [1]

['Jun Ni', 'State Key Laboratory Of Cognitive Neuroscience', 'Learning Beijing Normal University', 'Beijing', 'Idg Mcgovern Institute For Brain Research', 'Beijing Normal University', 'Beijing Key Laboratory Of Brain Imaging', 'Connectomics', 'Jiaxin Yang', 'Yina Ma']

Date: 2024-03

Social groups in various social species are organized with hierarchical structures that shape group dynamics and the nature of within-group interactions. In-group social bonding, exemplified by grooming behaviors among animals and collective rituals and team-building activities in human societies, is recognized as a practical adaptive strategy to foster group harmony and stabilize hierarchical structures in both human and nonhuman animal groups. However, the neurocognitive mechanisms underlying the effects of social bonding on hierarchical groups remain largely unexplored. Here, we conducted simultaneous neural recordings on human participants engaged in-group communications within small hierarchical groups (n = 528, organized into 176 three-person groups) to investigate how social bonding influenced hierarchical interactions and neural synchronizations. We differentiated interpersonal interactions between individuals of different (inter-status) or same (intra-status) social status and observed distinct effects of social bonding on inter-status and intra-status interactions. Specifically, social bonding selectively increased frequent and rapid information exchange and prefrontal neural synchronization for inter-status dyads but not intra-status dyads. Furthermore, social bonding facilitated unidirectional neural alignment from group leader to followers, enabling group leaders to predictively align their prefrontal activity with that of followers. These findings provide insights into how social bonding influences hierarchical dynamics and neural synchronization while highlighting the role of social status in shaping the strength and nature of social bonding experiences in human groups.

Funding: This work was supported by the National Natural Science Foundation of China ( https://www.nsfc.gov.cn/english/site_1/index.html , Projects 32125019 to Y.M.); the STI 2030—Major Projects 2022ZD0211000 to Y.M. ( https://en.most.gov.cn/ ); the Fundamental Research Funds for the Central Universities ( http://en.moe.gov.cn/ , 2233300002 to Y.M.); the Major Project of National Social Science Foundation ( http://www.nopss.gov.cn/ , 19ZDA363 to Y.M.); the start-up funding from the State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University ( https://brain.bnu.edu.cn/English/index.htm , to Y.M.). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

( A ) Experimental setting. During the group interaction, 3 group members’ rDLPFC and rTPJ signals were simultaneously recorded by the same fNIRS system. Shown are snapshots of a control session and a bonding session. ( B ) Illustration of fNIRS probe configurations. Two identical 3 × 2 optode probe sets, with each consisting of 3 emitters (light blue) and 3 detectors (dark blue), were placed on rDLPFC and rTPJ, respectively. The integers in between indicate the recording channels. ( C ) Group members were engaged in turn-taking conversation session. Each turn transition was defined as a discrete pair of utterances from different individuals (depicted by rectangles). The turn-response time was calculated as the time interval of corresponding turn transition. ( D ) Social bonding increased inter-status turn transitions (control: 5.960 ± 2.971, bonding: 8.854 ± 3.759) but not intra-status ones (control: 5.470 ± 3.560, bonding: 6.320 ± 3.499). ( E ) Bonding significantly sped up turn transitions (inter-status: control: 0.997 ± 0.188, bonding: 0.898 ± 0.172; intra-status: control: 0.947 ± 0.284, bonding: 0.909 ± 0.228). ( F/G ) Interpersonal cohesion was positively associated with turn transition frequency, respectively, for inter-status ( F ) and intra-status dyads ( G ). Each solid line represents the least squares fit, with shading showing the 95% CI. ( H ) Bonding selectively increased inter-status (control: 6.310 ± 2.025, bonding: 6.970 ± 1.855) but not intra-status cohesion (control: 6.220 ± 1.870, bonding: 6.240 ± 2.035). Data are plotted as box plots for each condition, with horizontal lines indicating median values, boxes indicating 25% and 75% quartiles, and whiskers indicating the 2.5%–97.5% percentile range. Cross symbols in each box represent the mean values. Data points outside the range are shown separately as circles. *p < 0.05, **p < 0.01, ***p < 0.001. Data used to generate Fig 1D–1H can be found in S1 Data . fNIRS, functional near-infrared spectroscopy; rDLPFC, right dorsolateral prefrontal cortex; rTPJ, right temporal-parental junction.

To address these questions, we applied fNIRS hyper-scanning to 176 three-person groups (the most basic hierarchical group with 1 leader and 2 followers, S1 Table ) and simultaneously recorded neural activities of 3 group members of each group during online within-group interactions. Participants democratically elected a group leader and discussed group strategies for potential intergroup contests after in-group social bonding or no-bonding control manipulation (Figs 1A and S1 ). The in-group social bonding manipulation employed in the current study was adapted from several previously validated procedures [ 15 , 46 – 49 ]. Specifically, we integrated 3 fundamental procedures to manipulate in-group social bonding: (i) shared preference [ 46 ]; (ii) symbolic marker [ 47 ]; and (iii) similarity among group members [ 48 ]. Taking advantage of fNIRS technology that provides noninvasive measures of neural activity with minimal sensitivity to motion artifacts [ 50 ], we measured neural synchronization in the right dorsolateral prefrontal cortex (rDLPFC) and right temporal-parental junction (rTPJ) in inter-status and intra-status dyads ( Fig 1B ). Brain regions of interest in the current study included rDLPFC and rTPJ. Previous studies have shown that the right (but not left) DLPFC was involved in allocating attention and making strategic decision during social interaction [ 51 , 52 ]. The TPJ, particularly in the right hemisphere, is a core region of the mentalizing network and involved in alignment with in-group members regarding consensus decision and group norms [ 53 ]. This ROI choice was also based on earlier studies that have linked neural synchronization in the rDLPFC and rTPJ with social interactive processes [ 32 , 36 , 45 , 54 ]. Moreover, neural synchronization in the rDLPFC activity predicted in-group cooperation during intergroup conflict [ 15 ], while neural synchronization in rTPJ was associated with leader–follower interaction [ 45 ].

At the behavioral level, we tested whether social bonding facilitated inter-status interaction, intra-status interaction, or both, and if so, whether such bonding effect on intra- and/or inter-status interaction was linked to the facilitation of in-group cohesion and leader’s influence. Moreover, in-group social bonding may increase not only in-group cohesion, but also hostility towards out-group members [ 15 , 29 , 30 ], resulting in increased intergroup discrimination. We thus also examined how in-group social bonding influenced intergroup discrimination, especially for individuals with different social status (i.e., group leader and followers). At the neural level, we employed functional near-infrared spectroscopy (fNIRS) hyper-scanning to measure neural synchronization among group members. Recent neuroscience research has suggested inter-brain neural synchronization (INS) as a reliable indicator of social interactions [ 31 – 33 ]. Neural synchronization emerges in a variety of social encounters, including interactions between peers [ 34 ], romantic partners [ 35 ], parent and child [ 36 , 37 ], teacher and student [ 38 ]. The degree of neural synchronization was predictive of interaction quality [ 39 ]. Of particular interest, neural synchronization among group members has been suggested as a candidate mechanism mediating within-group interaction [ 15 , 40 ], and in-group bonding increased within-group neural synchronization [ 15 , 41 ]. However, previous studies merely focused on the egalitarian group, suffering from not being able to differentiate inter- and intra-status interactions and leaving the social bonding effect on hierarchical interactions an open question. Here, we aimed to reveal whether and how social bonding influenced neural synchronization within a hierarchical group and, in particular, the inter-status and intra-status neural synchronization. Recent research has documented stronger neural synchronization during social interactions between individuals with different social roles than those with the same roles (e.g., teacher–student versus student–student, [ 42 , 43 ]; leader–follower versus follower–follower, [ 44 , 45 ]). Therefore, it could be expected that in-group social bonding would have differential effects on inter-status and intra-status neural synchronizations.

Specifically, we asked how social bonding facilitated interpersonal interactions within hierarchical groups and examined here at both the behavioral and neural levels. The hierarchical structure places individuals at different positions in the group (i.e., individuals with different social statuses), such as group leader and followers, varying in levels of resources, social influence, or competence [ 10 , 23 ]. Sensitivity to status information and recognizing one’s relative social status in the group are essential for successful social interaction, and interpersonal interaction within a hierarchical group is shaped by different status relationships [ 3 ]. Interpersonal interactions within a simply structured hierarchical group are thus classified into 2 fundamental, status-related types: (i) interactions between individuals of different social status, e.g., leader–follower interaction (henceforth inter-status interaction); and (ii) interactions between individuals of the same social status, e.g., follower–follower interaction, the most common representative of intra-status interactions. We further asked whether and if so, how social bonding differentially influenced these 2 types of interpersonal interactions within a hierarchical group. Previous studies have demonstrated distinct functions of inter-status and intra-status interactions within hierarchical groups [ 24 , 25 ]. Inter-status interaction facilitates the exchange of asymmetric information between the group leader and followers [ 26 ], as well as leader–follower coordination and alignment [ 27 ]. On the other hand, intra-status interaction fosters reciprocal relationships and peer support among individuals with similar social status (e.g., followers) [ 1 , 28 ]. Therefore, we expected that social bonding would exert different effects on inter-status and intra-status interpersonal interactions within a hierarchical group. Specifically, we examined whether social bonding would exhibit stronger or weaker effects or even opposite effects on inter- versus intra-status interactions.

Most social groups, from the basic family unit to professional organizations, and societal institutions, are hierarchically structured [ 1 , 2 ]. The hierarchical structure and different status relationships are one of the most fundamental features of social groups and shape interpersonal interactions among group members [ 3 , 4 ] to facilitate group stability [ 5 ] and maximize group productivity [ 6 ]. However, hierarchical structure comes with challenges and costs for social groups [ 7 ]. In hierarchical groups, high-ranking individuals may bully subordinates and usurp a disproportionate share of resources, social influence, and reproductive opportunities [ 8 , 9 ], which may amplify intragroup inequality and competitions [ 10 ], undermine the authority and legitimacy of group leaders [ 7 ]. Small groups overcome these problems through in-group social bonding [ 11 , 12 ], an adaptive means of forming, strengthening, and maintaining interpersonal connections with in-group members [ 13 – 15 ]. Social bonding exercises, such as grooming behaviors in nonhuman primate, collective rituals, traditions, and team-building activities in human society [ 16 – 19 ], have been widely adopted to facilitate leader influence [ 20 ], increase group cohesion [ 21 ], and reinforce the hierarchical structure [ 22 ]. Yet, despite the significance and impact of in-group social bonding, the neurocognitive mechanisms underlying the effects of social bonding on hierarchical groups remain largely unknown.

( A / B ) Leaders (vs. followers) showed stronger rDLPFC-rTPJ functional connectivity at channel-pairwise level (28 rDLPFC-rTPJ channel pairs survived FDR correction for 49 channel pairs, A ) and the grand mean level (i.e., averaged coherence value of 49 channel pairs between the rDLPFC and rTPJ, B , displayed as box plots, with horizontal lines indicating median values, boxes indicating 25% and 75% quartiles and whiskers indicating the 2.5%–97.5% percentile range. Cross symbols in each box represent the mean values. Data points outside the range are shown separately as circles). ( C ) The functional connectivity between channel 9 in the rDLPFC and the rTPJ (averaged across 7 channel pairs) was associated with leader-to-follower neural alignment at channel 9. The solid line represents the least squares fit, with shading showing the 95% CI., **p < 0.01. Data used to generate Fig 5A–5C can be found in S1 Data . FDR, false discovery rate; rDLPFC, right dorsolateral prefrontal cortex; rTPJ, right temporal-parental junction.

The observation that the bonding effect was selectively exhibited on the inter-status dyads, especially in a leader-to-follower manner, led us to further examine the bonding effects respectively in leaders and followers. The functional connectivity between rDLPFC and rTPJ has been shown to play a key role in perspective taking, mental inference, and information integrating [ 66 , 67 ]. As leader-to-follower neural alignment may indicate situations in which group leaders predict followers’ mental states or perspectives [ 64 ], as well as when followers strategically attend to and track the group leader [ 68 ], we compared rDLPFC-rTPJ functional connectivity between leaders than followers. Furthermore, we tested whether rDLPFC-rTPJ functional connectivity could account for the leader-to-follower neural alignment. To this end, we applied cross-correlation analysis to assess the functional connectivity of rDLPFC and rTPJ in leaders and followers (Methods). Results showed that, leaders (versus followers) showed stronger rDLPFC-rTPJ connectivity ( Fig 5A for channel-pairwise rDLPFC-rTPJ connectivity, 28 rDLPFC-rTPJ channel pairs survived FDR correction for 49 channel pairs, S7 Table ; Fig 5B for the grand mean rDLPFC-rTPJ connectivity, two-way mixed-model ANOVA, F 1, 174 = 12.006, p = 6.679 × 10 −4 , η 2 = 0.065; LMM: F 1, 348 = 12.438, p = 4.77 × 10 −4 ). We next correlated the strength of rDLPFC-rTPJ connectivity (averaged connectivity between channel 9 in the rDLPFC and each channel in the rTPJ) in leaders and followers respectively with the leader-to-follower neural alignment at channel 9 in the rDLPFC (Methods). Results endorsed a positive relationship between rDLPFC-rTPJ functional connectivity in leaders (but not followers) and the leader-to-follower neural alignment (leader: r 176 = 0.177, p = 0.019, Fig 5C , for average of lagged inter-status neural alignment, S9 Fig for correlations to neural alignment on each time lag; follower: r 176 = 0.004, p = 0.955), suggesting that leaders with stronger rDLPFC-rTPJ connectivity predicted their followers’ neural activity to a greater degree.

Moreover, such leader-to-follower neural alignment facilitated intergroup discrimination. Specifically, we correlated leader-to-follower neural alignment with intergroup discrimination and showed that stronger leader-to-follower neural alignment (when leaders’ neural activity preceded that of followers +1 to +6 s) predicted larger intergroup discrimination ( Fig 4D , FDR corrected for 21 time lags, S6 Table ). Moreover, this relationship was especially true in the bonding but not control condition (averaged neural alignment of +1 to +6 time lag, bonding: r 89 = 0.281, p = 0.008, control: r 87 = 0.081, p = 0.455, Figs 4E and S7 for each time lag separately). Control analyses were conducted for the intra-status INS for −10 to 10 s (in 1-s increment). Neither the bonding effect ( S8A Fig ) nor the correlation with intergroup discrimination ( S8B Fig ) was significant on intra-status neural alignment at any time lags. Taken together, the bonding-elevated inter-status INS selectively emerged when leaders’ neural activity preceded that of followers, indicating that leaders predicted or anticipated followers’ mental states and followers tracked the leader’s mental states to achieve the leader-to-follower neural alignment under in-group social bonding condition. Such leader-to-follower neural alignment may further result in stronger intergroup discrimination under in-group social bonding.

While the time-lagged neural alignment in the rTPJ was significant in both the leader-to-follower and follower-to-leader directions (with peak centered at 0 second, S6A Fig ), regardless of the bonding/control conditions ( S6B Fig ), the time-lagged neural alignment in the rDLPFC was significant mainly in the leader-to-follower direction and modulated by social bonding. We conducted independent t tests on time-lagged neural alignment between bonding and control conditions on each time lag and revealed a right-skewed bell-shaped curve for the bonding effect ( Fig 4B ). To be specific, bonding (relative to control condition) significantly facilitated inter-status neural alignment on +1 to +6 time lags (with peak centered at +5 s, Fig 4B and S4 Table ), indicating that leaders’ neural activity preceded that of followers for 1 to 6 s (survived multiple corrections for 21 time lags). Separate analyses for bonding and control conditions confirmed significant increase (contrast to zero) of inter-status neural alignment only occurred in the bonding condition ( Fig 4C and S5 Table ); however, neither the leader-to-follower nor follower-to-leader neural alignment was significant in the control condition ( Fig 4C ).

( A ) Illustration of time lag analysis. The neural activity of the leader is shifted forwards (backwards) in relative to that of followers in the positive (negative) time lags, indicating leader-to-follower (follower-to-leader) alignment. ( B ) Bonding (vs. control) facilitated leader-to-follower neural alignment on +1 to +6 time lags (peaked at +5 s), survived FDR multiple correction. The dashed line indicates the corrected significance threshold. The significant time lags (survived multiple correction) are highlighted with the horizontal line on the x-axis. ( C ) Significant increases of inter-status neural alignment on +1 to +6 time lags was only found in the bonding condition. Shaded areas represent SE. ( D / E ) Inter-status neural alignment was positively correlated with intergroup discrimination on +1 to +6 time lags in the bonding (but not control) conditions (correlation coefficients on each time lag of −10 to +10 lags, D ; correlations for the averaged inter-status neural alignment, E ). Correlations were performed by Pearson’s correlation coefficient analysis. Each solid line represents the least squares fit, with shading showing the 95% CI. Data used to generate Fig 4B–4E can be found in S1 Data . FDR, false discovery rate; rDLPFC, right dorsolateral prefrontal cortex; SE, standard error.

Next, we aimed to probe the directionality of the inter-status neural synchronization. We specifically asked whether social bonding influenced the leader-to-follower (i.e., neural activity of leaders preceded that of followers) or follower-to-leader (i.e., neural activity of followers preceded that of leaders) neural alignment, or both. The leader-to-follower neural alignment reflects situations in which the group leader leads and followers follow, while the follower-to-leader neural alignment may indicate instances where followers take the lead and group leader follows. We thus conducted time-lag analysis that has been employed in previous studies to reveal the directional influence between leader and follower’s neural activity [ 63 – 65 ]. The time course of leader’s neural activity was shifted relative to that of the followers from −10 to 10 s (in 1-s increment). Positive time lags indicated leader-to-follower neural alignment and negative time lags reflected follower-to-leader neural alignment ( Fig 4A ). On each time lag, the coherence values of inter-/intra-status dyads were recomputed for both resting-state and interaction stage. The coherence value increase (i.e., lagged INS during interaction minus that during resting) were used to indicate lagged neural alignment and submitted into subsequent analysis (Methods). It should be noted that the time lag analyses were conducted for the channels that showed increased neural synchronization between leader and followers during the interaction stage, i.e., channel 3 in the rTPJ and channel 9 in the rDLPFC.

Next, we aimed to reveal whether and how the inter-status or intra-status INS within a group was linked to behaviors towards in-group and out-group members. We found that stronger inter-status (but not intra-status) INS in the rDLPFC was predictive of stronger intergroup discrimination (i.e., donations to in-group versus out-group members; inter-status: r 176 = 0.216, p = 0.004; intra-status: r 176 = −0.096, p = 0.206, Fig 3H ). Further modulation analysis compared the Fisher-transformed correlation coefficients and confirmed selective prediction of inter- (versus intra-) status INS on intergroup discrimination (z = 2.94, p = 0.003). These results suggested that leader and followers synchronized their rDLPFC activity in a way that predicted how they differently treated in-group and out-group members.

We next conducted 2 sets of validation analyses to exclude the possibility that the observed INS was partially reflected participants sharing the same environment or performing the same task. First, within the bonding and control conditions, we generated 176 within-condition three-person pseudo-groups by randomly grouping a real leader and 2 real followers from different original groups in the same bonding or control condition as 1 three-person pseudo-group (Methods, S5A Fig ). We recalculated the inter- and intra-status INS for pseudo groups and repeated these procedures for 1,000 times. We then conducted nonparametric permutation tests on the observed effects of the real interacting groups against the 1,000 permutation samples. This analysis confirmed that both the main effect of hierarchy in the rTPJ (real group: 0.041, permutation: 95% CI: −0.020, 0.031, 99% CI: −0.029, 0.039, p = 0.003, Fig 3F ) and the interactive effect of hierarchy × bonding in the rDLPFC (real group: 0.043, permutation: 95% CI: −0.020, 0.031, 99% CI: −0.028, 0.039, p = 0.003, Fig 3G ) in the real groups exceeded the upper limit of 99% CI of the permutation distributions. Second, similar analysis conducted on the cross-condition permutation samples (i.e., 1 leader and 2 followers randomly from the bonding or control condition were organized into a pseudo-group, S5B Fig ) again confirmed the observed effects in the real interacting groups ( S5C and S5D Fig ). Taken together, the 2 validation analyses confirmed that the observed bonding and/or hierarchy effects on INS in the real interactive groups were not due to same experimental environment or performing the same task. In addition, we further eliminated potential influence of global physiological noises by (i) using a wavelet-based denoising method [ 61 ]; and (ii) controlling the globally co-varying signals in the hierarchy × bonding ANCOVAs (i.e., including the global mean of INS across all channels as a covariant, [ 62 ]). These 2 complementary analyses well replicated the aforementioned patterns ( S3 Table ).

Interestingly, we observed a significant hierarchy × bonding interaction on INS at channel 9 in the rDLPFC (F 1, 174 = 9.577, p = 0.002, η 2 = 0.052, survived 14-channel-wise FDR-correction; LMM: F 1, 348 = 8.912, p = 0.003, Fig 3D and S2 Table ). Specifically, social bonding selectively increased the inter-status INS (independent-sample t tests, t 174 = 2.357, p = 0.020, Cohen’s d = 0.355, 95% CI: 0.003, 0.035) but decreased intra-status INS (t 174 = −1.998, p = 0.047, Cohen’s d = −0.301, 95% CI: −0.047, 0.000). Taking another perspective to interpret the interaction, we observed stronger inter-status (versus intra-status) INS in the bonding condition (paired-sample t tests, t 88 = 2.521, p = 0.013, Cohen’s d = 0.267, 95% CI: 0.005, 0.043), which was comparable even in an opposite trend in the control condition (t 86 = −1.875, p = 0.064, Cohen’s d = −0.201, 95% CI: −0.039, 0.001). Interestingly, we found that inter-status but not intra-status INS in the rDLPFC was predictive of how the group interaction was perceived. Independent rater perceived the groups with stronger inter-status INS in rDLPFC more cohesive (inter-status INS: r 175 = 0.177, p = 0.019, Fig 3E ; intra-status INS: r 175 = 0.085, p = 0.265).

To examine whether social bonding differently influenced the inter- and intra-status INS, we submitted INS in each channel of the rDLPFC and rTPJ to 2 (hierarchy: inter- versus intra-status dyads) × 2 (bonding: bonding versus control) mixed-model ANOVAs and LMMs (hierarchy and bonding as fixed effects, group as a random effect). Significant effects were identified after false discovery rate (FDR)-corrected for multiple comparisons for 14 channels. First, the analysis revealed stronger INS in the rTPJ for the inter-status than intra-status dyads (main effect of hierarchy at channel 3, F 1, 174 = 10.207, p = 0.002, η 2 = 0.055, survived 14-channel-wise FDR-correction; LMM: F 1, 348 = 9.684, p = 0.002, Fig 3B and S2 Table ). Moreover, stronger inter-status INS in the rTPJ was associated with stronger social influence of the group leader (r 176 = 0.188, p = 0.012, Fig 3C ), suggesting that followers perceived their leaders more influential when their rTPJ activity synchronized with that of leader to a greater degree.

( A ) Illustration of inter-status neural synchronization calculation. The concentration changes in oxygenated hemoglobin (oxy-Hb) were simultaneously collected in each channel from each member of the three-person group. The cross-correlations between oxy-Hb time series of leader–follower pairs were generated through WTC analysis, and the 2 pairs were then averaged to indicate INS of inter-status dyads. Comparison of coherence values between group interaction and resting-state identified INS specific to group interaction and the frequency band of interest. (B ) Stronger inter-status (vs. intra-status) INS at channel 3 in the rTPJ. Data are plotted as box plots for each condition, with horizontal lines indicating median values, boxes indicating 25% and 75% quartiles and whiskers indicating the 2.5%–97.5% percentile range. Cross symbols in each box represent the mean values. Data points outside the range are shown separately as circles. ( C ) Positive association between inter-status INS at channel 3 in the rTPJ and group leader’s social influence (Pearson’s correlation analysis). The solid line represents the least squares fit, with shading showing the 95% CI. ( D ) Bonding increased inter-status INS at channel 9 in the rDLPFC (control: 0.001 ± 0.053, bonding: 0.019 ± 0.055) but decreased that of intra-status dyads (control: 0.019 ± 0.076, bonding: −0.004 ± 0.079). ( E ) Positive association between inter-status INS at channel 9 in the rDLPFC and perceived group cohesion. ( F/G ) INS validation by nonparametric permutation tests. We compared the hierarchy main effect in the rTPJ and the interaction effect in the rDLPFC of real group against within-condition permutation distributions (n = 1,000). The observed effects of hierarchy in the rTPJ ( F ) and of hierarchy × bonding interaction in the rDLPFC ( G ) exceeded the upper limits of 99% CI of the permutation distributions. ( H ) The inter-status (but not intra-status) INS at channel 9 in the rDLPFC was associated with intergroup discrimination. *p < 0.05, **p < 0.01. Data used to generate Fig 3B–3H can be found in S1 Data . INS, inter-brain neural synchronization; rDLPFC, right dorsolateral prefrontal cortex; rTPJ, right temporal-parental junction; WTC, wavelet transform coherence.

We applied fNIRS to each hierarchical group and simultaneously recorded all group members’ neural activity, captured by the dynamic hemodynamic signals, from the rDLPFC (7 channels, Fig 1B ) and the right temporoparietal junction (rTPJ, 7 channels, Fig 1B ), during resting-state and interaction stages. Consistent with previous studies [ 15 , 34 – 37 ], we operationalized the INS in terms of wavelet transform coherence (WTC). The WTC value indicates the cross-correlation between 2 fNIRS time series of concentration changes in oxygenated hemoglobin (oxy-Hb) in dyads of individuals as a function of frequency and time. Within each three-person group, we calculated the coherence values from the leader–follower dyads to index the inter-status INS, and the coherence value from the follower–follower dyads to index the intra-status INS (Methods, Fig 3A ). We were interested in the INS specific to group interaction, thus focused on the INS increases during group interaction relative to the resting-state. We compared coherence values between the resting-state and group interaction to identify the frequency band of interest (FOI, Methods, Fig 3A ). Moreover, the INS specific to group interaction was indicated by the FOI-averaged coherence differences (Group interaction—Resting) and then submitted into the following analyses.

We concluded our behavioral analysis by examining how in-group social bonding influenced followers’ perception of the leader (i.e., leaders’ social influence and social attraction). We found that followers in groups under social bonding (versus control) perceived their leaders as more influential (t 174 = 2.313, p = 0.022, Cohen’s d = 0.348, 95% CI: 0.103, 1.301, Fig 2C ) and more attractive (t 174 = 2.944, p = 0.004, Cohen’s d = 0.444, 95% CI: 0.265, 1.339, S4A Fig ). Moreover, the perceived social influence and social attraction of leaders were positively associated with in-group cohesion, especially evident with inter-status cohesion (social influence: r 176 = 0.765, p = 4.183 × 10 −35 , Fig 2D ; social attraction: r 176 = 0.702, p = 1.743 × 10 −27 , S4B Fig ; weaker but also with intra-status cohesion, social influence: r 176 = 0.435, p = 1.548 × 10 −9 , S4C Fig ; attraction: r 176 = 0.287, p = 1.107 × 10 −4 , S4D Fig ; slope test: social influence: z = 5.04, p = 1.164 × 10 −7 ; attraction: z = 5.36, p = 2.081 × 10 −8 ), suggesting that followers perceived their leaders as more influential and attractive in more cohesive groups, especially when they coordinated better with the leaders. Importantly, we established a potential mediation path that the effects of social bonding on perceived social influence (Indirect effect = 0.520, SE = 0.238, 95% bootstrap CI: 0.073, 1.001, Sobel test, Z = 2.224, p = 0.025, Fig 2E ) and attraction (Indirect effect = 0.426, SE = 0.196, 95% bootstrap CI: 0.057, 0.0832, Sobel test, Z = 2.224, p = 0.026, S4E Fig ) in leaders were fully mediated by inter-status cohesion (Methods).

( A ) In-group social bonding increased intergroup discrimination to a greater degree in leaders (control: 40.314 ± 39.547, bonding: 67.649 ± 38.752) than followers (control: 47.971 ± 28.122, bonding: 57.261 ± 26.918). ( B ) Bonding increased out-group hate in leaders (control: 19.858 ± 22.151, bonding: 27.814 ± 29.810) but not in followers (control: 28.557 ± 20.210, bonding: 26.810 ± 18.040). ( C ) Under social bonding, followers perceived greater social influence of the leader (control: 6.270 ± 2.061, bonding: 6.970 ± 1.963). Data are plotted as box plots for each condition, with horizontal lines indicating median values, boxes indicating 25% and 75% quartiles and whiskers indicating the 2.5%–97.5% percentile range. Cross symbols in each box represent the mean values. Data points outside the range are shown separately as circles. ( D ) Leader’s social influence was positively associated with inter-status cohesion (Pearson’s correlation analysis). Each solid line represents the least squares fit, with shading showing the 95% CI. ( E ) Bonding increased perceived social influence of the leader through enhancing inter-status cohesion. *p < 0.05, ***p < 0.001. Data used to generate Fig 2A–2E can be found in S1 Data .

Next, we examined whether social bonding influenced behaviors toward in- and out-group members differently (or not) in individuals of different social statuses (i.e., group leader and followers). Participants completed 2 economic games related to intergroup discrimination: (i) an intergroup dictator game (IDG) where participants donated to in-group and out-group members [ 15 , 58 ]; (ii) an intergroup prisoner’s dilemma-maximizing differences game (IPD-MDG) where participants self-sacrificed separately to benefit in-group members (“ingroup love”) and to derogate out-group members (“outgroup hate”) [ 59 , 60 ]. We found that groups in the bonding (versus control) condition donated more to in-group members than to out-group members in the IDG (F 1, 174 = 26.406, p = 7.375 × 10 −7 , η 2 = 0.132, Fig 2A ), and such bonding-facilitated intergroup discrimination was stronger in group leaders than followers (hierarchy × bonding: F 1, 174 = 6.109, p = 0.014, η 2 = 0.034; leader: t 174 = 4.631, p = 7.087 × 10 −6 , Cohen’s d = 0.698, 95% CI: 15.685, 38.983; follower: t 174 = 2.239, p = 0.026, Cohen’s d = 0.338, 95% CI: 1.101, 17.479; LMM: F 1, 348 = 6.256, p = 0.013, Fig 2A ). In the IPD-MDG, participants showed stronger in-group love (paired t test: ingroup love (Mean ± SD): 34.332 ± 16.650, outgroup hate: 26.319 ± 15.764, t 175 = 3.966, p = 1.065 × 10 −4 , Cohen’s d = 0.299, 95% CI: 4.025, 12.001). Interestingly, the interactive effect of bonding and hierarchy was observed in out-group hate (F 1, 172 = 4.470, p = 0.036, η 2 = 0.025; leader: t 172 = 1.995, p = 0.048, Cohen’s d = 0.302, 95% CI: 0.084, 15.830; follower: t 174 = −0.605, p = 0.546, Cohen’s d = −0.091, 95% CI: −7.443, 3.950; LMM: F 1, 346 = 3.898, p = 0.049, Fig 2B ) but not in in-group love (F 1, 172 = 0.445, p = 0.506, η 2 = 0.003). Taken together, in-group social bonding increased intergroup discrimination and “hate” towards outgroup, especially in group leaders.

At the end of the experiment, we asked participants to report subjective evaluations on inter- and intra-status cohesion. First, we found that the frequency of group communication predicted group cohesion. In groups with more frequent communications, their group members reported a higher level of group cohesion (r 172 = 0.206, p = 0.006). Interestingly, more inter-status turn transitions selectively predicted inter-status cohesion (r 173 = 0.165, p = 0.029, Fig 1F ), but not intra-status cohesion (r 173 = 0.052, p = 0.496). Similarly, more intra-status turn transitions predicted a higher level of intra-status cohesion (r 172 = 0.162, p = 0.033, Fig 1G , but not inter-status cohesion, r 172 = 0.146, p = 0.055). Second, social bonding selectively facilitated inter-status cohesion (t 174 = 2.261, p = 0.025, Cohen’s d = 0.340, 95% CI: 0.082, 1.238) rather than intra-status cohesion (t 174 = 0.040, p = 0.968, Cohen’s d = 0.010, 95% CI: −0.562, 0.602), confirmed by a significant interaction between hierarchy (inter- versus intra-status) and bonding (bonding versus control) on in-group cohesion rating (F 1, 174 = 4.914, p = 0.028, η 2 = 0.027, Fig 1H ). In addition, within-group interactions under social bonding were also perceived as more frequent and cohesive by third-party observers (Methods, S3 Fig ).

The number of turn-transition and turn-response time was then calculated separately for inter-status (a discrete pair of utterances between a group leader and a follower) and intra-status (a discrete pair of utterances between 2 followers) communications. We compared these measurements between social bonding and control conditions using hierarchy (leader versus follower) × bonding (bonding versus control) ANCOVAs (controlling for the total length of utterances) and corresponding linear mixed models (LMMs, with hierarchy and bonding as fixed effects and group as a random effect). This analysis revealed that social bonding increased the frequency of inter-status (compared to intra-status), communications to a great extent (increased the number of leader–follower turn transitions, bonding × hierarchy: F 1, 172 = 9.951, p = 0.002, η 2 = 0.055, inter-status: t 174 = 5.658, p = 6.182 × 10 −8 , Cohen’s d = 0.853, 95% CI: 1.885, 3.904, intra-status: t 173 = 1.587, p = 0.114, Cohen’s d = 0.240, 95% CI: −0.206, 1.900; LMM: F 1, 347 = 7.673, p = 0.006, Fig 1D ) and shortened the turn response time (bonding main effect: F 1, 166 = 9.793, p = 0.002, η 2 = 0.056, especially for inter-status turns: t 172 = −3.406, p = 0.001, Cohen’s d = −0.516, 95% CI: −0.154, −0.040; LMM: F 1, 340 = 12.924, p = 3.72 × 10 −4 , Fig 1E ). These results together suggested that social bonding was efficient in increasing group communication, especially promoted more frequent and responsive inter-status interactions and strengthened inter-status social connections.

We performed conversation analysis [ 55 ] on the transcripts of within-group communication for each group. The within-group communications were operationalized on a turn-taking basis ( Fig 1C ). Therefore, we focused on the number of utterances, the number of turn transition, and turn response time. Turn transition refers to the exchange of utterances among different group members. The number of turn transitions is suggested to reflect the frequency of mutual understanding and engagement [ 56 ], with more turn transitions indicating more interactive and engaging communications between group members. Turn response time is measured by the time interval between turns, with faster response times reflecting stronger social connection and more efficient and interactive communication [ 57 ].

Discussion

Social bonding has been recognized as a potent strategy to enhance in-group interaction and cohesion across species [16–21] and to reinforce hierarchical structures [22]. However, to date, the neurocognitive mechanisms underlying the effects of social bonding on hierarchical interactions remain largely elusive. We applied a multi-brain hyper-scanning approach to real-time within-group communication and differentiated interpersonal interactions within a hierarchical group into 2 status-related types: inter-status and intra-status interactions. By doing so, we provide evidence that in-group social bonding exerts distinct effects on individuals of different statuses and on dyads of different status-related relationships. Specifically, social bonding selectively facilitates communication frequency and responsiveness, as well as synchronized and leader-proceeding neural alignment for inter-status dyads but not for dyads with the same social status. These findings distinguish between inter-status and intra-status interactions and shed lights on the distinct neurocognitive mechanisms through which social bonding shapes group dynamics in hierarchical groups.

Hierarchical groups are characterized by complex relationships among group members [3], where their behaviors and neural systems dynamically coordinate and interact [4]. However, most neuroscience studies on leadership and social hierarchy have examined the brains of leaders or followers in isolation [69,70], focusing on leading behaviors or the distinct roles leaders and followers respectively played in a group [71]. This approach has overlooked the interactive nature of hierarchical relationships, limiting our understanding of how different individuals dynamically interact within a hierarchical group. Recently, the emergence of second-person, interactive neuroscience has underscored the significance of studying how the brains of socially interacting individuals entrain to support social interaction and relationships [31–33]. Through the lens of interactive neuroscience, we were able to characterize distinct behavioral and neural profiles of 2 types of interpersonal interactions based on the status relationship of interactive dyads. Within a hierarchical group, group members of different, rather than same, status engaged in more frequent and rapid information exchange, and frequent inter-status interaction was closely related to in-group cohesion for the aggregated group. At the neural level, inter-status interaction was featured with stronger neural synchronization in the rTPJ, a key brain region involved in mentalizing and taking perspectives of others, especially dissimilar others [72]. This finding suggests an important role of perspective-taking and social mentalizing in inter-status interaction, while interactions among individuals with the same status may require less mentalizing efforts and gain social support through “shared mind.” Together, the distinct behavioral and neural profiles for inter-status and intra-status interactions highlight the importance of differentiating status-related interactions and relationships in future work to understand leadership and group dynamics in hierarchical groups.

By differentiating between inter-status and intra-status interactions in hierarchical groups, we are able to arbitrate between 2 possible mechanisms: whether social bonding enhances group dynamics in hierarchical groups regardless of the interaction types (i.e., enhancing both inter-status and intra-status interactions in a similar way), or if it is modulated by the interaction type, selectively promoting either inter- or intra-status interaction. Our results, which revealed distinct behavioral and neural effects of social bonding on inter- and intra-status dyads, provide evidence in support of the latter. Specifically, social bonding mainly facilitates inter-status (rather than intra-status) communication and cohesion, and selectively enhances neural alignment between the rDLPFC activity of leaders and followers. Synthesizing and extending findings from previous studies [13,15], we suggest that the mechanisms by which social bonding operates within a group depends on the type of group structure. Unlike non-hierarchical groups in which social bonding fosters interaction and cohesion among group members indiscriminately, social bonding in hierarchical groups primarily serves to foster and reinforce connections among individuals with different social statuses. It selectively enhances mutual understanding and information exchange between leaders and followers rather than among peers, highlighting the importance of status-related interactions in shaping the effects of social bonding within a hierarchical group.

We further investigated how social bonding facilitated neural entrainment between individuals of high and low status, specifically addressing whether this entrainment occurred in a high-to-low status or low-to-high status direction or both (i.e., a bidirectional manner). By employing time-lagged analysis, we revealed the temporal dynamics of the inter-status synchronization. The prefrontal activity of the group leader preceded that of followers by 1 to 6 s, indicating a unidirectional neural alignment from group leader to followers after social bonding. Previous studies have associated such sequential neural alignment with anticipatory processing, reflecting active engagement and predictive processing of others’ behaviors and intentions during social interactions [63–65]. For example, in dyadic communication, the listener’s prefrontal activity often precedes that of the speaker, and such anticipatory neural response facilitated mutual understanding and successful communication [65]. Therefore, our findings of leader-to-follower neural alignment suggested a potential underlying process through which social bonding influenced inter-status interaction: the group leader actively engaged in anticipating and predicting the mental states of followers, enabling more frequent communication and coordination with them. Moreover, the effects of social bonding on synchronized and leader-proceeding neural entrainment were only observed in the rDLPFC rather than the rTPJ. Previous work has evidenced a crucial role of the prefrontal cortex, particularly the DLFPC, in predicting forthcoming sensory inputs (e.g., short sound, [73]), social cues (e.g., eye gaze, [63]), and social dominance and future social interactions [74], suggesting DLPFC as a hub region for monitoring errors and updating predictions of upcoming inputs to prepare for appropriate responses [73]. Therefore, the impact of social bonding on rDLPFC synchronization and leader-to-follower alignment may provide a neurocognitive account for increased leader initiation and more frequent, efficient leader-follower information exchange. Interestingly, we showed that group leaders with stronger DLPFC-TPJ functional connectivity exhibited a greater degree of neural alignment with followers. This link suggested that the exchange and integration of information between DLPFC and TPJ may support predictive neural alignment from leader to follower. Taken together, in-group social bonding may enable the group leader to actively adopt followers’ perspective, consider their potential behaviors and intentions, and align predictively with them.

In-group social bonding prioritizes the allocation of cognitive resources, emotional attachments, and neural entrainment to inter-status interactions within hierarchical groups, and fails to yield comparable effects on intra-status interactions. This discrepancy cannot be attributed to a ceiling effect resulting from potentially preexisting close bonding among individuals of the same-status due to shared similarities [75], which may limit the extent to which external bonding manipulation can further enhance their interactions. This interpretation is supported by the absence of differences in communication frequency and responsiveness, and neural synchronization between intra-status and inter-status dyads in the control condition without external bonding manipulation. Moreover, the bonding effects on neural synchronization exhibited a distinct pattern, with social bonding decreasing neural synchronization in the rDLPFC for intra-status dyads while increasing it for inter-status dyads. Given the critical role of the rDLPFC in top-down regulation of social attention [76], we propose that the observed decreases in neural synchronization in the rDLPFC of intra-status dyads may reflect a disengagement of attention from fellow members, potentially accompanied by a reallocation of attention towards the group leader [77]. Consistently, followers under social bonding engaged in more frequent and responsive communications with group leaders (compared to other fellow members, as shown in Fig 1E) and perceived leaders as more influential and socially attractive. Such an upward attention shift may contribute to the maintenance of structural stability by facilitating follower’s understanding of the intentions and/or preferences of the group leader while minimizing potential competition and violations [74,77,78]. Together, these findings put forward the hypothesis that the effects of social bonding on modulating neural couplings and redirecting attentional engagement from intra-status to inter-status interaction may serve to reinforce the hierarchical structure.

Extending previous findings of bonding effects on egalitarian group, our work reveals the critical roles of social status in shaping the strength and nature of the social bonding experience in hierarchical groups, which operates at both the individual and dyadic levels. At the individual level, social bonding facilitates the initiation and engagement of high-status individuals in group communications, while increases the responsiveness of low-status individuals to group leader and their perceptions of high-status individuals. At the dyadic level, social bonding exerts distinct effects on the inter-status and intra-status dynamics, potentially through the mechanisms of top-down predictive alignment and bottom-up attentional shift. Specifically, social bonding increases a leader’s forward-prediction of follower’s neural activity, while suppresses follower’s neural and attentional entrainment with same-status fellows. These plausible neurocognitive pathways helped us to synthesize social bonding effects, providing neural accounts for the effect of social bonding on group dynamics within hierarchical social contexts. The establishment of social bonds between leader and followers may serve to alleviate inter-status inequality and competition, foster inter-status coalitions, and maintain social hierarchy [71].

Our findings may be limited to the current experimental settings and could raise a number of exciting research questions for future studies. First, nonverbal communication, such as gestures, facial expressions, and eye contact, plays a crucial role in real-life social interactions. However, since our participants were restricted to online communication through typing, the effects of social bonding on nonverbal hierarchical interaction remain unexplored. Second, the current study was conducted within simple three-person hierarchical groups with leaders who represented symbolic, perhaps prestige-style leaders, democratically elected by group members. These leaders made more contributions in intergroup economic games and established positive connections with followers, but lacked the authority to sanction fellow members or allocate resources. These settings deviated from those commonly encountered in complex real-life scenarios. Therefore, caution should be exercised when attempting to generalize our findings and future work is encouraged to explore the generality and specificity of social bonding effects across diverse leadership styles. Third, our study examined the three-person group, which is the minimal unit of a hierarchical group with only 2 levels of social hierarchy. While group dynamics and leadership in such a small-scale society are arguably representative, it lacks the complexity of group structure institutions. It will be interesting for future research to test whether the bonding effect would be weakened or amplified by the despotic power of nonhuman animal groups or the complexity of large-scale social networks.

Finally, it should be noted that the current dataset was obtained from participants of a specific cultural background, i.e., East Asian Chinese individuals. This raises the question of whether the observed effects in the current sample can be generalized to individuals from other cultures. Individuals from East Asian cultures place emphasis on group cohesion, interpersonal connection, and social hierarchy [79–81]. In comparison to Western cultures, followers in East Asian cultures tend to display higher levels of obedience and commitment towards their group leader while also encouraging more supportive leadership [82]. Therefore, one may expect cultural differences in how leaders and followers interact, especially after in-group social bonding, in hierarchical groups. To explore this possibility further, we conducted a preliminary examination by assessing individual differences in cultural values within our sample. Previous studies have suggested that cultural group differences in the neural activity underlying social cognition may be mediated by cultural values, such as interdependence of self-construal [83]. In the current study, we employed the Self-Construal Scale [84] to assess individual variations in cultural value of interdependence. We found that individual differences in independence did not influence behavioral and neural indices related to inter- or intra-status interactions nor did they affect observed bonding effects on behavioral and neural indices. These results suggested that our findings may be insensitive to culture-specific values. However, we acknowledge that the lack of modulation by cultural values could potentially be attributed to minimal variability in culture values within a single cultural context. It is important for future cross-cultural research to directly test whether and how our current findings can be generalized to other cultural populations.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002545

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.