Abstract
Unselfishness is one of the admired facilitators for human group
endeavors, especially in times of urgent calls for global collaboration.
Despite its importance, the neural dynamics behind its formation is
scarcely understood. With 26 triads interacting as turn-taking pairs in
a coordination game, we investigated reciprocal interactions in this
tri-fMRI hyperscanning experiment. The critical role of the right
temporal-parietal junction (rTPJ) was examined by adopting both time-
and frequency-domain analyses. For the former, in the successful versus
failed “reciprocity” contrast, brain regions associated with the
mirror neuron system (MNS) and the mentalizing system (MS) were
identified. In addition, the differences of connectivity between the
rTPJ (seed region) and the abovementioned network areas (e.g., the right
Inferior Parietal Lobule, rIPL) were negatively correlated with the
individual reward. These results both verified the experimental design,
which favored ‘reciprocal’ participants/triads with larger gains, and
supported the opposition of rTPJ (other-) vs. rIPL (self-concerned)
areas during successful social exchanges. Furthermore, the cerebral
synchronization of the rTPJs emerged between the interacting pairs, and
the coupling between the rTPJ and the right Superior Temporal Gyrus
(rSTG) was found between those interacting simultaneously with others of
the same group. These coherence findings not only echoed our previous
findings, but also reinforced the hypotheses of the rTPJ-rTPJ coupling
underpinning simultaneous collaboration and the rTPJ-rSTG coupling for
decontextualized shared meaning emergence. Taken together, these results
support two of the multi-functions (other-concerning and
decontextualizing) subserved by the rTPJ, and highlight its interaction
with other self-concerning brain areas in reaching co-benefits.
Keywords: fMRI hyperscanning, Psychophysiological Interactions (PPI),
coherence, collaboration, reciprocity
INTRODUCTION
Imagine you are in class and four people are arranged as a group to
complete a task. You start to talk with the one next to you, and then
maybe the one sitting facing you. After a run of talking, you may find
yourself having better interactions with certain partners; meanwhile it
is just too hard to keep your conversation flowing with someone.
Somehow, each of your partners makes you feel something, and this leaves
an impression of how you decide to react to them later on. Normally, the
better interaction you have, the more willing you are to help with the
task. By the definition of ‘help,’ it actually means you give and take
in achieving an end, mutually. Phenomena of mutual interaction,
experienced or expressed by each of two or more people or groups about
the other, are ubiquitous. Social scientists may wonder if there is a
difference that can be observed between pairs who are doing the same
task at the same time but interacting with others. If yes, what is the
neural difference between the partner who is sitting in front of you and
talking to you, and the one who is sitting next to you and talking to
another? How does your brain react to ‘good’ partners and ‘bad’
partners? As for neuroscientists, investigating social interactions in
various contexts and revealing their neural correlates has never been
more profound and implicative enough. In the present study, we are
attempting to answer these questions regarding social interaction with
three brains communicating pairwisely.
Hyperscanning, simultaneously measuring the brain activity of multiple
brains, allows the investigation of intra- and inter-brain neural
relations in real-time dynamics (Czeszumski et al., 2020; Hari &
Kujala, 2009; Scholkmann et al., 2013). Social neuroscience has
developed research designs in hyperscanning in modalities other than
fMRI, such as magnetoencephalography (MEG) (Holmes et al., 2023;
Mayseless et al., 2019), electroencephalography (EEG) (Haresign et al.,
2022; Turk et al., 2022), and functional near-infrared spectroscopy
(fNIRS) (Nguyen et al., 2021; Zhang et al., 2023). Despite multiple
brains interacting in sync, dyadic interactions and computations retain
their fundamental roles in methodology and in research on the brain
processes of social science. To progress toward real-life interactions,
recent neuroscientific fMRI hyperscanning studies are mostly with dyads
interacting with close-by or internet-connected scanners. Tasks to
capture the emergent dynamics of simultaneous dual brain interactions
include face-to-face interactions (such as gaze behavior) (Koike et al.,
2019; Miyata et al., 2021), joint grips (Abe et al., 2019), coordination
games (Goelman et al., 2019; Špiláková et al., 2019; Stolk et al., 2014;
Wang et al., 2023; Yoshioka et al., 2021), natural events such as movie
viewing (Schmälzle & Grall, 2020), and reciprocity in the ultimatum
game (Shaw et al., 2018; Sperduti et al., 2014). Both the mirror neuron
system (MNS) (Iacoboni & Dapretto, 2006) and the mentalizing system
(MS) (Frith & Frith, 2006; Saxe, 2006) are indicated as playing vital
roles in social interactions, especially in collaboration (Wang et al.,
2018). For example, the Superior Temporal Gyrus (STG), the Inferior
Parietal Lobule (IPL), the Inferior frontal gyrus (IFG), and the
Precentral Gyrus, are among the brain networks during imitation (Heyes,
2001; Rizzolatti, 2005), an indispensable ingredients in positive
rippling effects (Akgün et al., 2015; Barry, 2009); the TPJ and
Precuneus are part of a mentalizing and default-mode network for
interpreting cues of others (Hyatt et al., 2015; Li et al., 2014; Mars
et al., 2012). Yet, as an old saying goes, “two’s company, three’s a
crowd,” tri-MRI hyperscanning provides an unprecedented opportunity
into the intricacies of group dynamics, given its varieties of three
bi-directional, or even tri-directional, communications among group
members.
Despite dyadic hyperscanning’s early onset (Montague et al., 2002), it
was not until 2020 that we saw the first three-person fMRI hyperscanning
study (Xie et al., 2020). In it, twelve triads engaged in a drawing task
with collaborative and independent phases. Both GLM and intersubject
correlation analyses (Hasson et al., 2004; Nummenmaa et al., 2018)
indicated the critical role of rTPJ in the triadic collaborative
interaction. As a likely second, in the present study we adopt both the
time-domain, including GLM contrasts and Psychophysiological
Interactions, or PPI, (O’Reilly et al., 2012), and frequency-domain
coherence analysis (Wang et al., 2023) to reveal neural substrates of
triadic social interactions. For the former, the aim is to investigate
the neural differences between successful and failed reciprocity. We
then adopted the successful and failed trials with reciprocity in the
second stage of feedback time as our target trials. For the latter, we
extend our prior work (Wang et al., 2023), a frequency-domain coherence
analysis, into a 3-person internet-based hyperscanning (with the 4th
sitting outside doing the same behavioral task) context (Sebanz et al.,
2006; Vesper et al., 2016). Here, we focus on the stage of revealing
whether the collaboration is successful (i.e., Stage 2 Feedback) because
it is the shared period when the pairs check the results, build up the
trust, and plan for the following trials. Note that there might be some
nuances regarding unsuccessful, but not a failure to try to cooperate,
and thus, may be underwritten by cooperative brain processes. This is
not within the scope of the present study. Additionally, pairs win in
two formats: the dyads could be kept in one dominant and the other
submissive, or they could reciprocate with each other, which wins most
in the end according to the experimental design. Lastly, when the
inter-brain coherences were reported in the hyperscanning literature,
the control conditions were mostly done by permuting individuals from
different pairs (because of the exclusive within-pair interactions in
dyads). In the present tri-fMRI study, however, the analyses of
interpersonal coherence can be separated into several pairing
combinations (see Methods for details), rendering at least three
possible interpersonal couplings, thereby yielding multiple constraints
of possible explanations.
What neural substrates, alone or together, contribute to meaningful
interaction/coordination still holds scientists’ interest. From one
brain to three, questions regarding subtle differences in mutual
interaction can be empirically better understood. In summary, besides
the technical advances in the tri-MRI implementation, reciprocity in
collaboration and its meaning shared among dyads are elucidated in three
analyses. These results generally suggest brain areas related to two
essential networks in social interaction, the MNS and the MS. The
present study plays an important contributor to the neuroscientific
understanding of group collaboration, with implications for structuring
co-benefits.
METHODS
Participants
Twenty-six triads of fMRI participants (NMale = 42;
NFemale = 36) were recruited from National Cheng Kung
University (NCKU), National Taiwan University (NTU), and National
Chengchi University (NCCU). NCKU is situated in southern Taiwan
(Tainan), and NTU and NCCU in northern Taiwan (Taipei). All participants
are native Taiwanese speakers, with normal or corrected-to-normal
vision, and no history of psychiatric or neurological disorders.
Participants gave informed consent and adhered to the guidelines and
regulations approved by the NCKU Governance Framework for Human Research
Ethics https://rec.chass.ncku.edu.tw/en, with the case number 106-254.
Experimental task
The experimental task was a revised coordination game (Farrell, 1988).
Four players were playing at the same time, three inside the scanners
and one outside for the behavior data only (Figure 1a). The participants
took random turns playing with each other as a pair. The player roles,
such as Player A, were fixed. Within a run, there were three 6-trial
blocks, e.g., Block 1: A&B/C&D, Block 2: A&C/B&D, and Block 3:
A&D/B&C. (Figure 1b), in which any assigned pair (e.g., Block 1:
A&B/C&D) completed 6 trials in a row. There were 6 runs of such 3
(blocks) x 6 (trials) combinations, with the random order of pairings as
3! = 6 (e.g., 123, 132, 213, 231, 312, and 321; one order for each run).
Each run took about 7 to 8 minutes.
Each trial lasted about 19 seconds, containing two stages: the possible
decision stage (preplay) and the actual decision stage (Figure 1c).
Before each trial started, the screen showed with whom they would be
interacting, e.g., “Interacting with Player B.” Next came the first,
or communication, stage (Stage 1 Decision) , where any given
participant told the other player his/her choice by clicking the
leftmost (‘1’) or the second leftmost key (‘2’), representing X and Y.
The reward combination of (X, Y) was (1, 6), (6, 1), or (0, 0),
suggesting that either one would choose the reward, 1 (or 6), and the
other chooses 6 (or 1), or neither of them would get any reward. About 6
seconds later, the possible choices of both players were shown on the
screen (Stage 1 Feedback) . What they chose would be framed with
a dotted rectangle, with one’s own choice in white and the other’s in
another color. It is important to note that the possible/communicating
choice in Stage 1 did not predicate the later actual choice.
Participants in the second stage (Stage 2 Decision) , after
making their final decisions, would see their responses framed with a
solid lined rectangle, with one’s own choice in white and the other’s in
another color (Stage 2 Feedback) , for up to 3 seconds. Three
combinations to run statistics were Within-group Interacting Pairs
(WIPs), Within-group Non-interacting Pairs (WNPs), and Between-group
Permuted Pairs (BPPs) (Figure 1d).