FIGURE 4. The comparisons of inter-regional couplings.(a) The coherence spectra between rTPJseed-rTPJp and (b) The rTPJseed-rSTG in the conditions of the within-group interacting pairs (WIP), within-group non-interacting pairs (WNP), within-group permuted pairs (WPP), and between-group permuted pairs (BPP). The brain maps show the seed (left) and the target (right) regions on the 36th frequency bin (out of 324/2 = 162 bins, or the target frequency 0.0556 Hz, or 1/18 s), under the coherence threshold of 0.215 and the cluster threshold of 40 voxels (FWE p < .05 corr.). This bin corresponds to the average trial frequency, or the concatenated event frequency (beta series built by combining 9 betas, every 2 s/ 1 TR, after the trial feedback time). The spectra of the WIPs are shown in green, WNPs in yellow, and the BPPs in black. The independent two-sample t-tests are applied among every possible pair at the 36th frequency bin. (c) The rTPJseed to whole-brain coherence maps of the WIP and WNP in the two stages of feedback periods, respectively. In the first stage of feedback (planning stage), the WIP and WNP have bilateral Lingual Gyri linked to processing vision and episodic memory. In the second feedback (decision stage), more couplings with the rTPJ can be observed in the WIP, while the rTPJ-STG coupling of the WNP outstands, compared to the WNP map in Stage 1. (d) When the reciprocal pairs are interacting with each other, their left TPJ and right IFG are coupled with the rTPJseed in the second feedback. (e) The non-reciprocal pairs yield rTPJ- amygdala and –left IFG couplings when interacting together.
Discussion
Collaboration through reciprocity is evident in many social contexts, as humans share and connect with each other from interactions. In the present study, we investigated whether brain circuitry was identified as relevant for reciprocal collaborations through social learning in a coordination game. First, our behavioral results showed that successful collaboration was correlated with reciprocity, and the reciprocal participants and pairs earned the most reward. This reflects that the utmost successful interaction of the present study lies in reciprocity. Second, in the neural level, the MNS and the MS involving coordination-, reward-, and decision-related regions known for cooperation tasks were identified. Many abovementioned regions were both found in our time-domain (i.e., GLM and PPI) and frequency-domain (i.e., coherence) analyses. Furthermore, the rTPJ was suggested to play its dual roles in social interaction with the brain regions related to other/self-concern (intrapersonally) and meaning-emergence (interpersonally). In sum, given its critical placement in mentalizing, the rTPJ is likely to participate in multiple functions through the present three-person hyperscanning.
‘ Our GLM results confirmed that significant activity in the mirror neuron related regions, including the Precentral Gyrus, the Posterior Cingulate, the IFG, and the TPJ when successful coordination was achieved. The MNS has been known for imitating actions as shared communication (Schmidt et al., 2021). In previous review studies, activity in the MNS is often significantly recruited in joint action (Mu et al., 2018; Pacherie & Dokic, 2006; Sebanz & Knoblich, 2009). In human social cognition, the MNS plays its crucial role of understanding of action and intentional agency (Gallese et al., 1996; Rizzolatti, 2005; Rizzolatti et al., 2001) and the emergence of language (Arbib & Bota, 2003; Arbib, 2002; Mu et al., 2018; Rizzolatti & Arbib, 1998). In addition, regions related to execution and observation of actions, such as the IFG, the Middle Frontal Gyrus, and the Medial Frontal Gyrus (Molenberghs et al., 2012; Takeuchi et al., 2013), were also involved in succeeding reciprocity. Since our task is a revised coordination game, far more complicated than merely imitating actions, successful dyadic interactions lie in careful decision-related processes. One study also suggests that the mirror neuron system may not just help participant mirror actions but process intentionally complementary or opposing actions (Campbell et al., 2018). In order to achieve reciprocity for final utmost reward, previous trial interactions should be observed and taken into consideration for the following trial. In particular, the role of the left IFG in action understanding and processing intention is examined (Pobric & Hamilton, 2006) as a necessity in making a perceptual judgment about people’s actions. This may result in the recruited activity of the IFG, especially, found in all of our three analyses. Lastly, activity associated with reward/reinforcement in the Caudate increased as the dyads were mutually benefited in a row. This explains how the pairs commonly reacted to reward when the successful results were revealed during the second feedback stage and continued to follow reciprocal interaction. We further ran correlations between the [successful trials > failed trials] condition and the three indices (i.e., reward, greed, and reciprocity), and the right Sub-gyral in the motor cortex was strongly correlated with an individual’s reciprocity. The Sub-gyral has been suggested for its role in coordination (Swinnen et al., 2010; Ullén et al., 2003). Not surprisingly, a person’s successful interaction with the other, which often leads to reciprocity in the present experiment, lies in good coordination in this motor-related area.
Our PPI analyses indicated individual differences in successful and failed reciprocity trials in functional connectivity between the rTPJ (seed region) and the other regions, which are similar to our GLM results. Interestingly, lower functional connectivity was found with the rTPJ, which is known for a role in theory of mind for mentalizing others’ intentions (Saxe & Kanwisher, 2003). Our results may elucidate how the rTPJ interacts with other regions associated with the MNS and MS. One hypothesis is that the rTPJ may fire more strongly and thus suppress the other self-related regions. Findings of the previous studies on the MNS and MS suggest a complementary or collaborative role of the two networks during social exchanges (Sperduti et al., 2014; Van Overwalle & Baetens, 2009; Wang et al., 2018), while Frith and Frith (2006) concluded that the MS is superior to the MNS for communicative intent. In our coordination game, the task is not simply about winning cooperatively, but reciprocally. Pairs do not just imitate each other’s behavior. They have to figure out how to reach mutual benefits and meanwhile how to gain more reward for themselves. This mentalizing process may recruit more activity of the rTPJ and less activity of self-related regions. One cannot be too greedy or too generous (or submissive), for both of the strategies result in less amount of final reward than acting reciprocal. This may explain why regions related to self-related aspects of experience, including the Precuneus, insula and IPL (Cabanis et al., 2013; Chiao et al., 2009; Harada et al., 2020; Ray et al., 2010; Shi et al., 2021), exhibited negative correlations with the rTPJ.
Further PPI analyses were conducted to investigate the relationship between the functional connectivity and the total reward of individuals. Regions associated with coordination/MNS (the rPG), attention reorientation, and self-perception (the rAG and rIPL) (Corbetta et al., 2008; Igelström & Graziano, 2017; Mitchell, 2008) were shown lower functional connectivity with the rTPJseed when an individual’s total reward increased in the contrast of successful trials versus failed trials. In other words, the bigger the contrast between the rTPJ and the ROIs mentioned above exhibits, the less total reward one earns. As discussed above, the rPG for imitation in the MNS and the IPL for self-other differentiation may be more suppressed when the rTPJ needs to be more resource-allocated for mentalizing for others to achieve reciprocity. Moreover, the AG and IPL are activated especially in predicting valid cues (Corbetta et al., 2002; Vossel et al., 2006), which results in less effort in successful trials for the cues as predicted. A gradient descent functional connectivity between these ROIs and the rTPJ negatively correlated with the total reward was also found among the dynamics of the subgroups (submissive > dominant > reciprocal) (see Figure 3f and the Supplementary Figure S3). That is, a bigger ROI-rTPJ contrast but a smaller amount of the total reward was yielded among the submissive participants. This can be interpreted that those who acted submissive might be less coordinated and less attentive than the reciprocal participants, and suppressing their desire in reward might exhaust their mind when they were fulfilling others (higher activity of the rTPJ in mentalizing others).
Regarding our frequency-domain results, two neural couplings of interest, the rTPJ-rTPJ and rTPJ-rSTG, were found highly consistent with social interactions. The first connectivity of the rTPJ-rTPJ emerging in the interacting pairs can be seen to draw on constant tracking of the other’s intention, mentalizing, and continuous communications for achieving success in collaboration. This coupling also implies common operation with attention (Igelström & Graziano, 2017) and social attributions (Saxe & Kanwisher, 2003). Recently, the rTPJ has been suggested to act as an interface of self- and other-related or external- or internal-triggered information processing (Bzdok et al., 2013). In particular, the rTPJp is associated with social reasoning in a theory-of-mind experiment design (Krall et al., 2015; Kubit & Jack, 2013). Previous experiments invoking face-to-face joint attention with one participant in autism spectrum disorder (Redcay et al., 2012), and in gaze-following of monkeys (Kamphuis et al., 2009) and humans (Laube et al., 2011), or two participants (Abe et al., 2019; Bilek et al., 2015; Koike et al., 2019), have shown activation of the rTPJp for extra attention to track social cues. In the present study, this rTPJ-rTPJp neural synchrony based on social information integration and social behavior formation is in line with the literature regarding the coordination of self- and other-behavior (Carter & Huettel, 2013; Corbetta et al., 2008; Geng & Vossel, 2013). We further compared this feedback in the decision stage (Stage 2 Feedback) to that in the planning stage (Stage 1 Feedback). The results again strongly support our original and previous results that the rTPJ-rTPJ and -rSTG couplings can be found only in the second/decision/final feedback, not in the feedback in the planning stage. Another interesting result is that similar couplings (the rTPJ coupled with the bilateral Lingual Gyri, linked to processing vision and episodic memory) were found among the pairs who were interacting (WIPs) and those who were doing the same task in the same group as the WIP but with other partners (WNPs). In addition, the right Precuneus, involved in self-referential processing, imagery, and memory, was identified in sync with the rTPJ only in the WIP in the planning stage, compared with the WNP also in the planning stage and the WIP in the decision stage.
Our results of the rSTG’s roles in social cognition and perception are in good agreement with the existing knowledge. A few hyperscanning fMRI studies, which also implemented the interpersonal coherence measures, reported similar findings. For example, Wang et al. (2023) demonstrated the similar rTPJ-rTPJp coupling when participants competed for the whole reward (i.e., thus highly attentive), the rTPJ-rSTG coupling when participants collaborated and split the reward (97.5% trusting the partner, thus less attention demand required). In the Token-Coordination task (Stolk et al., 2014), similar rSTG-rSTG coherence between highly collaborative pairs, implicates such coupling as the foundation of meaning (Stolk et al., 2016). In the present study, being able to compute among three pairing groups (i.e., interacting, non-interacting, and permuted), we would like to propose the idea of decontextualization of the rTPJ-rSTG coupling as the shared meaning for the common good. When pairs are in the context of interacting, their rTPJs synchronize to mentalize each other. While they are interacting with their pairmates, the rTPJs of theirs are coupled with the rSTGs of the other two groupmates who are interacting with their pairmates at the same time. The shift of the rTPJ-rTPJ coupling between pairmates (while interacting with each other) and the rTPJ-rSTG coupling between groupmates (while interacting with another in the same group), compared to the baseline (the permuted results), highlights the formation of decontextualization in the rTPJ-rSTG coupling. As trials go on, the rTPJ-rSTG coupling can only be explained to be for shared goal orientation, or simply understanding the game rules. While more resources may be allocated in the rTPJ for thinking about the real interacting partner, the interpersonal rTPJ-rSTG coupling is found in sync between groupmates despite the same coordination game and the simultaneous interacting time among all of them. In short, we would like to refer to the rTPJ-rSTG coupling as the understanding of the shared meaning. Here the interpretation of this coupling can be referred as to the meaning of the common good, which is that ‘reciprocity leads to better individual and group gain.’ This neural substrate serves as a mutual understanding of abstraction, which is relevant to how human languages were formed and shared among a group of people as human beings are the symbolic species (Deacon, 1997).
By using the behavioral data, we classified subjects into three types (reciprocal, dominant, and submissive). We further calculated the differences in coherence values when subjects of different types were paired or when subjects of uniform type were paired. This may help to understand the mental processing of different types of subjects yielding different oscillation synchronization. The TPJ was found between the reciprocal pairs, and the IFG was identified among the reciprocal pairs and the non-reciprocal types (dominant and/or submissive). These brain regions are often associated with for action understanding and processing intention as we discussed a lot above. Compared with the reciprocal pairs, the non-reciprocal pairs had negative emotion arousal in their right amygdala in sync with the rTPJ. This is an interesting finding which echoes our real-life scenario that there may be bad blood or hard feelings between the non-reciprocal pairs.
To date, the analysis for spatio-temporal dynamics of hyperscanning data does not have standard computational models to follow (Kelsen et al., 2022). Still, we tested if the evoked neural responses in the frequency domain were temporally coherent. The results were all found insignificant, suggesting that the effect size of frequency-domain coherence is relatively miniscule, rendering the transformation into the time domain almost non-existent. Small or null temporal effects in the ROIs were derived from the analyses of the frequency domain. For example, in the interacting pairs, the rTPJseed-rTPJp time-course correlation was 0.039 and its beta correlation was 0.005. Besides more spatio-temporal analysis comparisons, we suggest that interpersonal connectivities may be better for frequency-domain analysis, while the time domain analysis is better for intra-personal correlations. Moreover, two implications are concluded: 1) publication bias and 2) future methodological requirements for more than two brains. First, due to different temporal and spatial resolution of tools (e.g., EEG and fMRI), analyses vary. Take the fMRI studies for example. Most studies adopt time domain analyses, such as GLM, PPI, and multivariate pattern analysis because of fMRI’s poor temporal resolution for frequency domain analysis (e.g., coherence). Comparisons with temporal-spatial analysis are barely tested, not to mention being reported, for its unfavorable insignificant results. Some may propose to collect two kinds of data from EEG-fMRI or MEG-fMRI. Again, the scanner thresheld within 3 mm head motion flexibility may result in the ecological setting issue. Therefore, it still remains incomplete but explorable between spatio-temporal methods. Second, our limitation here is that the triad data were not analyzed in a triad method. Instead, the contrasts and connectivities between pairs were employed. We acknowledge that we have not been able to come up with new analyses to test the correlations and connectivities among three participants. As we all know, there will be a long way to validate the three- or even four-way hyperscanning results by adopting more than pairwise correlations. In fact, even though dyadic fMRI hyperscanning has been adopted for more than two decades, the methods are still diverse and debatable. Besides, previous studies employed EEG and fNIRS with multiple participants interacting in sync and also adopted pairwise methods, such as coherence. Interpersonal couplings are always averaged across pairs; there is no difference statistically to average across triads. For future possibilities, analyses can be run by three or more brains simultaneously, as the results can support more for interpreting multi-brain interactions (Babiloni & Astolfi, 2014; Hamilton, 2021).
Aside from individual’s mind and brain states to characterize social neuroscience in the past, the burgeoning works of such hyperscanning, both in implementation and analysis perspectives, provide novel insights into this much needed (in both theory and applications) field of study. Reviewing histories and looking ahead, we are constantly reminded that the future of human races, facing the challenges of global crises in environment, welfare, resource management and so forth, will only be solved by collective actions. The prerequisites of collaborations among organizations of various scales, from families to countries, are assurances of mutual attitudes throughout the processes. In cases where any sign of disconfirming evidence arises, the social reciprocity might fall and collapse along a downward spiral. This is why in addition to the protection from supervising organizations or sound policies, players’ mutual guarantee through trust and reciprocity plays a vital role in long-term social mutual benefits.