key: cord-0427141-35gfe7uz
authors: Kieckhaefer, Carolin; Schilbach, Leonhard; Bzdok, Danilo
title: Social belonging: Brain structure and function is linked to membership in sports teams, religious groups and social clubs
date: 2021-09-07
journal: bioRxiv
DOI: 10.1101/2021.09.06.459167
sha: 6e6aeb3194976aa7367ba3743aabd9ded4ab7127
doc_id: 427141
cord_uid: 35gfe7uz

Human behaviour across the life span is driven by the psychological need to belong, from kindergarten to bingo nights. Being part of social groups constitutes a backbone for communal life, and confers many benefits for physical and mental health. Capitalizing on neuroimaging and behavioural data from ~40.000 participants from the UK Biobank population cohort, we used structural and functional analyses to explore how social participation is reflected in the human brain. Across three different types of social groups, structural analyses point towards variance in ventromedial prefrontal cortex, fusiform gyrus and anterior cingulate cortex as structural substrates tightly linked to social participation. Functional connectivity analyses emphasized the importance of default mode and limbic network, but also showed differences for sports teams and religious groups as compared to social clubs. Taken together, our findings establish the structural and functional integrity of the default mode network as a neural signature of social belonging.

In the context of the COVID-19 pandemic with widespread loneliness and reduced 88 possibilities of participation due external circumstances, individual resilience gains 89 relevance from a public health standpoint. Being able to cope with and quickly recover 90 from a disaster is critically dependent on active belonging to a group within the 91 resilience process (Quinn et al., 2020) . A closer look at the biological manifestations of 92 social belonging and participation at the population-level is imperative, especially as 93 COVID-19 challenged not only regions or communities but affected entire populations. 94

Such confronted life events and overcome obstacles are important to be investigated 95

for the angle of psychopathology and clinical symptoms. 96

In the present population-based approach, we shift the focus towards factors that are 97 known to be key for successful resilience and mental well-being. Regarding how a 98 particular individual handles stressors, previous studies underlined the association 99

between the experience of control as well as resilient coping and the activation of the 100 medial prefrontal cortex (Maier & Watkins, 2010; Sinha et al., 2016) . Neuroimaging 101 data from real-life contexts offers important insight into social belonging and its many 102

wide-ranging consequences. One way to understand the broader link between social 103

interactions and its influence on brain architecture has been proposed by the social 104 brain hypothesis. The intensified need and intricacy of social relationships in humans 105 may have spurred refinement towards more complex representation of social bonds in 106 the brain (Dunbar, 2009 ). Brain systems that have long been described to be closely 107 implicated in social cognition processes involve the default mode network (DMN; Mars 108 et al., 2012). However, based on previous works, the neural basis of social participation 109

and belonging remain obscure. Taken together no human brain-imaging assessments 110 of participation in different group contexts exist, in part because such characteristics of 111

people's everyday social life have seldom been systematically acquired before the 112 emergence of large population datasets such as the UK Biobank cohort. Combining 113 structural and functional neuroimaging as well as demographic profiling with a 114

population-based approach, we investigated three particular forms of social 115 participation: sports teams, religious groups and social clubs. Using a population 116 cohort to examine and understand systematic variations of social belonging and 117 resilience help inform public-health decision making, which ultimately can foster 118

implementation and even interventions in practice. 119 120 121

Materials and Methods 122

Data resources 123

The UK Biobank is a prospective epidemiology resource that offers extensive 124 behavioural and demographic assessments, medical and cognitive measures, as well 125

as biological samples in a cohort of ~500,000 participants recruited from across Great 126

Britain (https://www.ukbiobank.ac.uk/). This openly accessible population dataset aims 127

to provide multimodal brain-imaging for ~100,000 individuals, planned for completion 128

in 2022. The present study was based on the recent data release from February/March 129 2020 that augmented brain scanning information to ~40,000 participants. 130

In an attempt to improve comparability and reproducibility, our study built on the 131 uniform data preprocessing pipelines designed and carried out by FMRIB, Oxford 132

University, UK (Alfaro-Almagro et al., 2018). We involved data from the ~40,000 133 participant release with brain-imaging measures of grey matter morphology (T1-134 weighted MRI [sMRI] ) and neural activity fluctuations (resting-state functional MRI 135

[fMRI]) from 48% men and 52% women, aged 40-69 years when recruited (mean age 136 54.9, standard deviation [SD] 7.5 years). Our study focused on regular social 137 engagement as captured by membership in social group (Bzdok & Dunbar, 2020; 138 Hawkley et al., 2003; Luhmann & Hawkley, 2016) . This self-reported item was based 139 on the following question: "Which of the following do you attend once a week or more 140 often?" (data field 6160). Our study focussed on three target groups: people reporting 141 engagement in sports teams, religious groups and social clubs. 142

Similar measures are found in widely used assessments of social embeddedness 143 (Cohen & Hoberman, 1983 approach allowed for the reliable identification and exclusion of problematic brain 169 scans, such as due to excessive head motion.

Structural MRI: The sMRI data were acquired as high-resolution T1-weighted images 171 of brain anatomy using a 3D MPRAGE sequence at 1 mm isotropic resolution. 172

Preprocessing included gradient distortion correction (GDC), field of view reduction 173 using the Brain Extraction Tool (Smith, 2002) group average template "S1200_MSMAll".

Analysis of associations between social participation and grey matter patterns 202

Neurobiologically interpretable measures of grey matter volume were extracted in all 203 participants by summarizing whole-brain sMRI maps in Montreal Neurological Institute 204

(MNI) reference space. This feature generation step was guided by the topographical 205 brain region definitions of the widely used Schaefer-Yeo atlas comprising 100 parcels 206 (Schaefer et al. 2018) . The participant-level brain region volumes provided the input 207 variables for our Bayesian hierarchical modeling approach (cf. below). As a data-208 cleaning step, inter-individual variation in brain region volumes that could be explained 209 by variables of no interest were regressed out: body mass index, head size, average 210 head motion during task-related brain scans, average head motion during task-211 unrelated brain scans, head position and receiver coil in the scanner (x, y, and z), 212

position of scanner individuals with high and low social participation in each brain region. Instead of limiting 219 our results and conclusions to strict categorical statements, each region being either 220 relevant for differences in social participation or not, our analytical strategy aimed at 221 full probability distributions that expose how brain region volumes converge or diverge 222

in their relation to social participation as evidenced in the UK Biobank population. In a 223 mathematically rigorous way, our approach estimated coherent, continuous estimates 224

of uncertainty for each model parameter at play for its relevance in social participation.

Our study thus addressed the question "How certain are we that a regional brain 226

volume is divergent between high and low social participation individuals?". Our 227 analysis did not ask "Is there a strict categorical difference in region volume between 228 high and low social participation individuals?". 229

The elected Bayesian hierarchical framework also enabled simultaneous modeling of 230 multiple organizational principles: i) segregation into separate brain regions and ii) 231 integration of groups of brain regions in form of spatially distributed brain networks. 232

Two regions of the same atlas network are more likely to exhibit similar volume effects 233 than two regions belonging to two separate brain networks. Each of the region 234 definitions was pre-assigned to one of the 7 large-scale network definitions in the 235

Schaefer-Yeo atlas (Schaefer et al. 2018) or the collection of subcortical regions from 236

the Harvard-Oxford atlas (Desikan et al., 2006) , providing a native multilevel structure. 237

Setting up a hierarchical generative process enabled our analytical approach to borrow 238 statistical strength between model parameters at the higher network level and model 239 parameters at the lower level of constituent brain regions. By virtue of exploiting partial 240 pooling, the brain region parameters were modeled themselves by the hyper-241

parameters of the hierarchical regression as a function of the network hierarchy to 242 explain social participation. Assigning informative priors centered around zero provided 243

an additional form of regularization by shrinking coefficients to zero in the absence of 244 evidence to the contrary. We could thus provide fully probabilistic answers to questions 245 about the morphological relevance of individual brain locations and distributed cortical 246 networks by a joint varying-effects estimation that profited from several biologically 247 meaningful sources of population variation.

The model specification placed emphasis on careful inference of unique posterior 249

distributions of parameters at the brain network level to discriminate individuals with 250

(encoded as outcome 0) and without (outcome 1) a certain social group membership:

where sigma parameters estimated the overall variance across the p brain regions that 270 belong to a given atlas network, independent of whether the volume effects of the 271 respective constituent brain regions had positive or negative direction. As such, the 272 network variance parameters sigma directly quantified the magnitude of intra-network 273 coefficients, and thus the overall relevance of a given network in explaining regular 274 social participation based on the dependent region morphology measures. All regions 275

belonging to the same brain network shared the same variance parameter in the 276 diagonal of the covariance matrix, while off-diagonal covariance relationships were 277

zero.

Probabilistic posterior distributions for all model parameters were estimated for the 279 hierarchical models. Our Bayesian approach could thus simultaneously appreciate 280 grey matter variation in segregated brain regions as well as in integrative brain the time series of whole-brain fMRI signals, obtained in the absence of an externally 295 structured experimental task, were summarized by averaging for each brain region in 296 the atlas. The approach yielded the functional coupling signature of the whole cortex 297 as a 100 x 100 region coupling matrix for each participant. The ensuing region-region 298 coupling estimates underwent standardization across participants by centering to zero 299 mean and unit scaling to a variance of one (cf. next step). Inter-individual variation in 300 the functional coupling strengths between brain regions that could be explained by 301

variables of no interest were regressed out in a data-cleaning step (analogous to sMRI 302 analysis): body mass index, head size, average head motion during task-related brain 303 scans, average head motion during task-unrelated brain scans, head position as well 304

as receiver coil in the scanner (x, y, and z), position of scanner table, and data 305 acquisition site, as well as age, sex and age-sex interactions. 306

We then sought the dominant coupling regime -signature or "mode" of population 307

covariation -that provides insight into how functional variability in 100 brain regions 308

can explain regular social participation. Partial least squares (PLS) was an ideal 309 analytical method to decompose the obtained 100 x 100 fingerprint matrix of functional 310 couplings with respect to social participation. The variable set X was constructed from 311 the lower triangle of the participants' functional coupling matrices. The target vector y 312 encoded more socially engaged participants as +1 and participants without a given 313 social group membership as -1. PLS involves finding the matrix factorization into k low-314 rank brain representations that maximize the correspondence with our social trait of 315

interest. PLS thus identified the matrix projection that offered the maximal covariance 316 between sets of region couplings in the context of participant reports of social 317 participation.

In other words, the extracted functional coupling mode identified the driving linear 319

combinations of cortical brain connections that featured the best correspondence to 320 regular social participation. Concretely, positive (negative) modulation weights 321 revealed increased (decreased) correlation strengths, relative to average functional 322

coupling. This is because the computed functional connectivity estimates were initially 323 normalized to zero mean and unit variance across participants. For example, a 324 functional connectivity input into PLS of 0 denoted the average functional coupling 325 strength in our UK Biobank sample, rather than an absence of functional connectivity 326

between the region pair. The derived pattern of PLS weights, or canonical vectors, thus 327

indicated deviations from average functional coupling variation in our cohort. Moreover, 328

the variable sets were entered into PLS after a confound-removal procedure (cf. 329 above). 330

Next, we assessed the statistical robustness of the resulting dominant PLS mode of 331 functional coupling deviations related to social participation in a non-parametric 332 permutation procedure, following previous research (Miller et al. 2016 ). Relying on 333 minimal modeling assumptions, a valid empirical null distribution was derived for the 334

Pearson's correlation between low-rank projections of the dominant mode resulting 335 from PLS analysis. In 1,000 permutation iterations, the functional connectivity matrix 336 was held constant, while the social participation labels were submitted to random 337

shuffling. The constructed surrogate datasets preserved the statistical structure 338

idiosyncratic to the fMRI signals, yet permitted to selectively destroy the signal 339

properties that are related to social participation (Efron, 2012) . The generated 340 distribution of the test statistic reflected the null hypothesis of random association 341 between the brain's functional coupling and regular social participation across 342 participants. We recorded the Pearson's correlations rho between the perturbed low-343 rank projections in each iteration. P-value computation was based on the 1,000 344

Pearson's rho estimates from the null PLS model.

Demographic profiling analysis of the brain correlates of social participation 347

We finally performed a profiling analysis of the brain regions that were most strongly 348 associated with regular social participation. Based on our results in brain structure, we 349

carried out a rigorous test for multivariate associations between our top regions and a 350

diverse set of indicators that exemplify the domains of a) basic demographics, b) 351 personality features, c) substance-use behaviours, and d) social network properties 352

(for details see https://www.ukbiobank.ac.uk/data-showcase/). The set of behavioural 353

variables and the set of brain measures were z-scored across participants to conform 354

to zero mean and unit variance. The brain variables were submitted to the top 10 355

(sMRI) of brain measures that were identified as most important in the context of social 356 participation (cf. above). In the case of brain structure, the target brain regions were 357 selected based on the (absolute) modes of the Bayesian posteriors of marginal 358 parameter distributions at the region level (cf. above). In the case of brain function, the 359 target brain connections were selected based the (absolute) effect sizes from the 360 dominant PLS mode (cf. above). 361

Using the two variable sets of brain and behaviour measurements, we then carried out 362 a bootstrap difference analysis of the collection of target traits in individuals with high 363 versus low social participation (Efron & Tibshirani, 1994 At the network level, in participants who attend sports teams at least once every week, 393 the highest explanatory relevance across spatially distributed brain regions emerged again found to be prominent (Table 1 ). In addition, among all three groups, similar 477 regions tended to come to the fore, including parahippocampal and fusiform gyrus, 478

anterior cingulate cortex, temporal and prefrontal cortex (Fig. 2) .

Functional brain correlates of social participation 481

Next, fMRI data from our UK Biobank cohort were examined to investigate possible 482 deviations in functional coupling fingerprints related to weekly engagement in sports 483 teams, religious groups and social clubs. The cortex-wide functional connectivity 484

profiles of each participant were submitted to a multivariate pattern-learning algorithm 485

that identified a collection of reliable positive and negative shifts in network connectivity 486

in the context of social participation (p < 0.05). In so doing, the single most coherent 487 pattern of deviation for the functional connectome of participants within each of the 488 three groups was identified (Fig. 3) .

The regular attendees of sports teams showed wide-ranging deviations in intra-490 network connectivity of default and limbic network. The somatomotor network revealed 491 negative coupling effects in intra-network connectivity. An increase in inter-network 492 connectivity was dominated by connections from the DMN, but also featured the limbic 493

and frontoparietal control network's functional ties to the somatomotor network. 494

Strengthened coupling was detected between the DMN and most other examined 495 large-scale functional networks, including visual network, somatomotor network, 496 frontoparietal control network and limbic network. The frontoparietal control network 497 was found to exhibit enhanced coupling links especially with both the visual and 498 somatomotor networks. A decrease in functional coupling strength was observed for 499 the somatomotor network, the visual network and the dorsal attention network. 500

People participating in religious groups in turn were especially characterized by a 501

compounding of within-network functional connections within the DMN, limbic network 502

and to some extent also in the frontoparietal control network. These three neural 503 network systems showed enhanced within-and between-network functional 504 connectivity patterns. In contrast, the dorsal attention network and the visual network 505

both showed reduced within-network connectivity strengths. Connectivity strengths 506 between regions of the DMN and the neural circuits of the limbic, frontoparietal control 507 and dorsal attention network were significantly increased. Furthermore, we identified a 508 decrease in connectivity strengths between the visual network and several other large-509 scale networks, including dorsal attention network, frontoparietal control network, 510 salience network as well as the somatomotor network to a reduced extent. 511

In contrast to these two types of social participation, social participation in social clubs 512

did not lead to a salient increase of functional connectivity strengths within most of the 513 aforementioned networks. Instead, a relevant decrease in functional connectivity was 514 noted within DMN and limbic networks as the single most coherent pattern of deviation. 515

The DMN in turn showed enhanced functional coupling with the somatomotor and the 516 dorsal attention networks. 517

As such, among all three types of social participation, the default and the limbic network 518 stuck out in the overall collection of observed deviations in region-region coupling 519

strengths. This insight was evidenced by enhanced connectivity patterns in active 520 members of sports teams and religious groups, and by diminished connectivity 521 patterns in social club participants. Additionally, reminiscent of our general findings 522 from the structural analyses (cf. above), the default and limbic network played a 523 prominent role among the systematic shifts of between-network coupling.

Demographic profiling analyses of social participation 526

In the final set of analyses, we have linked grey matter volume deviation in the most 527 relevant identified regions (top 10%) to behavioural and sociodemographic data via a 528 multivariate pattern analysis in each of the three groups of social participation (Fig. 4) . 529

In terms of consistent findings across groups, the time spent watching television ranked 530 highest in all three separate analyses: sports teams (mean = -0.39, 5/95% confidence social participation showed similar convergence for the type and extent of the 541 consumption of alcohol intake (amount of alcohol drunk on a typical drinking day, 542 alcohol intake frequency) and tobacco use (past tobacco smoking, current tobacco 543 smoking). To a varying extent, different psychological conditions also showed a high 544 concordance within all three groups of social participation. These psychological 545

conditions included including loneliness, mood swings, neuroticism, fed-up feelings, 546

suffer from "nerves", tense, miserableness, irritability and sensitivity. In sum, charting 547 relevant brain-behaviour associations revealed a high concordance among all three 548

forms of social participation on sociodemographic and behavioural level for family 549 structure, alcohol and tobacco consumption. 550 551

Discussion 552

Experiencing times of unmet social desire and not being able to fulfil one's need to 553 belong can be a wake-up call and highlights the pivotal importance of community and 554 social exchange. Consequently, exploring the neurobiological substrates of social 555 participation and their ties to physical and mental health is imperative. In our present 556 investigation, the DMN and the limbic system were placed in the center of robust brain 557 manifestations across three examined types of social participation in ~40,000 UK 558

Biobank participants: sports teams, religious groups, and social clubs. First, our 559 structural analyses demonstrated the importance of prefrontal and cingulate cortex. 560

Parts of the ventromedial prefrontal cortex (vmPFC) emerged as particularly relevant 561

for participation in all three social groups. Second, in our analyses of functional 562 coupling fingerprints, the DMN and limbic system also emerged consistently among all 563 three groups as the cornerstone of the neural bases indicative of social embeddedness 564 via group membership. Third, the combination of neuroimaging with behavioural and 565 sociodemographic data showed a high consistency among all three types of 566 participation for alcohol and tobacco consumption as well as certain psychological 567

states. Hence, participants of sport teams, religious groups and social clubs do not 568 only show similarities on the neural level, but also on the level of substance use 569

behaviour and mood. 570

The close involvement of the DMN in social and affective processes has been reported The anterior cingulate cortex (ACC) and parahippocampal gyrus are often attributed to 597

the limbic system. The parahippocampal gyrus revealed additional grey matter volume 598 within sports team and social club participants. The ACC showed significant volume 599 deviations for religious groups and social clubs but none for sport teams. For religious 600 groups, ventral parts of ACC showed positive volume effects, while dorsal parts 601

showed negative volume effects. For social club members, dorsal and ventral volume 602 effects for cingulate cortex differed between the two hemispheres. Given that social 603 behaviour is potentially most uniquely developed in the human species (Frith & Frith, 604 2010) it may come as no surprise to find brain asymmetry features, as hemispheric 605 asymmetry is also exceptionally well developed in the human brain (Hartwigsen et al., 606 2021). The regional findings match the notable interindividual variability of the limbic-607 trait associations at the network-level of our Bayesian hierarchical analyses. Only 608 recently, grey matter volume increase in the dorsal and perigenual ACC was linked to 609 social affective benefit (Gan et al., 2021) . The community-based cohort study recorded 610

daily-life social contacts and affective valence via smartphone for one week and 611

combined the data with neuroimaging. Applied to the current results, this might indicate 612 less social affective benefit from the membership in religious groups. 613

The fusiform gyrus is classically associated with processing information from other 614

people's faces. These neural processes assist in recognizing a person identity from Previous studies compared the neuronal activation of fusiform gyrus and posterior 622

superior temporal sulcus through trustworthiness, attractiveness, emotion and age 623

judgments (Bzdok et al., 2012; Oosterhof & Todorov, 2008) . While posterior superior 624

temporal sulcus was associated with trustworthiness judgments, the fusiform gyrus 625

was recruited by attractiveness judgments. In our study we found a volume increase 626

of the fusiform gyrus for participants of sports teams and social clubs, but none for 627 participants of religious group. With respect to the posterior superior temporal sulcus 628

we found volume decrease within the group of sports team attendees and volume 629

increase within attendees of religious groups. This suggests that processing of facial 630

properties of other individuals may be of greater importance in social clubs and sport 631 teams than in religious groups. 632

Those regions with the most pronounced volume deviations in our study have been 633

widely connected to characteristics of social participation such as social network size.

A previous functional neuroimaging study with resting-state functional magnetic 635 resonance imaging found a positive association between social network size and the 636 connectivity strength between the amygdala and superior temporal sulcus, as well as 637 that between fusiform gyrus and to vmPFC respectively (Bickart et al., 2012). A related 638 study by Bickert and colleagues (2011) linked the human social network size to brain 639 structure of caudal inferior temporal sulcus, medial frontal cortex and ACC. The 640

number of social contacts within one's online social network (Facebook) has previously 641 been linked to cortical volume deviation in the posterior superior temporal sulcus, as 642 well as middle temporal gyrus (Kanai et al., 2012) . Extending these results from human 643

to non-human primates, macaques with a larger social network showed an increase in 644 grey matter volume in rostral prefrontal cortex, ACC and superior temporal sulcus 645 (Sallet et al., 2011) . Our analysis regarding social participation highlighted several 646

regions that were mentioned in the context of social network size. This indicates, that 647 social participation is linked to a larger social network. 648

Only recently, the relevance of the DMN in the context of a lack of social connection, 649

in other words loneliness has been highlighted (Spreng et al., 2020) . Based on the 650 results in Spreng and colleagues (2020) and those of our study, we can say that 651 overlapping networks are implicated both for social isolation and social participation. 652

Results included similar volume deviations in regions for participants feeling lonely, 653

such ACC, posterior superior temporal sulcus and fusiform gyrus. Even on the network 654 level, the relevance of DMN within lonely participants was alike the results within the 655 groups of social participation. Hence, loneliness and social participation seem to tap 656 on similar brain networks and may even be viewed as opposite ends of a continuum. 657

Besides loneliness as a subjective form of social isolation, social support reflects an 658 objective form of social isolation. 659

Social support is another key construct closely related to social participation. The 660

construct of social support comprises closest friends and family, with a direct impact 661 on health and well-being (Dunbar, 2018) . Perhaps counterintuitive at first glance, 662 results for social support and variance in brain regions showed less concordance to 663

our results for social participation (Schurz et al., 2021) . Although both assess lifestyle 664

phenotypes that are related to forms of social cognition, only the effect of the limbic 665 network and the volume deviation in the ACC appear to be more closely borne out by 666 our present results. We speculate that the relevance of DMN and limbic system for 667 social participation might derive from two sides. On the one hand, subjective aspects 668 may be rooted in the DMN and reflect components like the sense of belonging to a 669 group, hence less loneliness. Similarly, objective aspects of social participation may 670 be linked to the limbic system and are possibly referring to a larger social network, 671

providing social support.

On the conceptual level, social participation is not only used in various ways as a term 673 but also measured in sometimes diverging ways (Chang & Coster, 2014) , with 674 constituent aspects consisting of a social role (e.g., daughter/son, friend, club member) 675

and a social task (e.g., work environment, school). To date, the construct of social 676 participation is predominantly discussed in the context of rehabilitation and healthy 677

aging, being extensively investigated as an outcome measure ( In addition to a number of commonalities in brain structure and function, we also found 688 differences between participants of sports teams, social clubs and religious groups. 689

These differences might depend on the types of regular experienced interaction with 690 the people who tend to be at the periphery of one's social circles. For regular physical 691 activity, positive effects on brain development and cognition in adolescence are 692

presented (Herting & Chu, 2017) . Herein reviewed neuroimaging studies reported 693 volume increase in hippocampus and lingual gyrus was related to aerobic training. 694

Improved cognition due to physical activity included executive functions, cognitive 695 flexibility and inhibitory control. A review study attributes an increase in grey matter 696 volume to physical activity for all brain regions except for superior temporal gyrus and 697 fusiform gyrus (Batouli & Saba, 2017). However, in our study we found volume 698 deviation for these two regions in sport team members as well. This in fact may be 699 driven by social aspects of sport participation rather than physical activity. 700

For groups related to religious beliefs and spirituality, altered functional coupling 701 patterns in the DMN have been reported and was discussed in the context of mystical 702

and "insight" experiences (van Elk & Aleman, 2017). Differences in social processing 703 between religious and nonreligious participants were found but controversial discussed 704 at the same time, with special attention to peer influence and membership in religious 705

groups (Grafman et al., 2020) . Furthermore, no consistent grey matter volume 706 differences for religiosity and mystical experiences were found in a recent voxel-based 707 morphometry study (van Elk & Snoek, 2020). A meta-analytic study found no positive 708 longevity effects based on individuals' beliefs but still suggests positive health effects 709

from simply belonging to a religious group (Shor & Roelfs, 2013 results concerning the impact of the particular nature of the investigated groups. 728

Reported positive effects on health and cognition as well as deviations in brain 729

structures might primarily derive from the recurring engagement with a group per se. 730

In our study, we reported minor differences in brain structure and function among all 731 three examined groups of social participation. However, similarities tended to dominate 732 on structural and functional level in our collective findings. 733

Within our results, convergence was also shown across the different demographic 734 profiling analyses. These brain-phenotype associations showed high similarity within 735 most factors among participants of sports teams, religious groups and social clubs.

Highest overlap across all three groups was observed in the factors television 737 consumption, number of siblings, health-related lifestyle behaviour and psychological 738

conditions. Although the direction of the revealed associations has no single answer in family size stand out. The individual consumption of television of adolescents was 748 associated with the TV consumption of their peer group (Fletcher, 2006) . Furthermore, 749

siblings foster social competence, especially in a young age (Downey et al., 2015) . 750

More specific, perspective taking can be improved by siblings but findings vary and 751 depend on the family context (Sang & Nelson, 2017) . Again, despite of differences in 752 the participated activity, commonalities exceed among the investigated groups in the 753 demographic profiling analyses. 754

Membership to social groups scaffolds human life in society. Extending previous 755 experimental neuroscience evidence, our investigation shows that brain substrates of 756 social participation are interrelated with health-related concepts like social support and 757 psychological well-being at the population level. Among all three examined types of 758 groups, we identified the DMN and limbic network as central for social participation. 759

Both highlighted networks gained further relevance in the context of belonging to a 760 group, as aspects of everyday life participation were studied in a population cohort and 761

could be related to additional demographic and everyday-life information. In a 762

comprehensive demographic profiling analysis, we here find concrete benefits of social 763

participation in groups such as reduced substance use and improved psychological 764

well-being. 765

Overall, our collective findings could be taken to suggest that the concrete type of 766 social participation may be of less importance than the regular attendance itself. 767

Looking for a way to harness the positive effects of social participation, this calls for 768 accessible forms of routine interventions in cohesive social groups, sometimes 769 described as 'social prescribing'. This is all the more important in periods of social 770 isolation in which a lack of social participation takes its toll on mental health. 

health practices among 13-and 15-year-old 836 adolescents: Results from the health behaviours in school-aged children study in 837 italy

Engagement and participation for health equity

ALE meta-analysis on facial judgments of trustworthiness 845 and attractiveness

Hierarchical region-848 network sparsity for high-dimensional inference in brain imaging

The Neurobiology of Social Distance

Analysing brain networks in 854 population neuroscience: A case for the Bayesian philosophy

The modular neuroarchitecture of social judgments on faces

Conceptualizing the construct of participation in 861 adults with disabilities

The correlation between gray matter volume and perceived social 865 support: A voxel-based morphometry study

Do I like this person?" 868 A network analysis of midline cortex during a social preference task

Social participation reduces 871 depressive symptoms among older adults: An 18-year longitudinal analysis in 872

Social participation, sense of community and social well being: A study 875 on American, Italian and Iranian University students

Positive Events and Social Supports as 878 Buffers of Life Change Stress

Social disconnectedness, perceived isolation, 881 and health among older adults

An automated labeling 889 system for subdividing the human cerebral cortex on MRI scans into gyral based 890 regions of interest

Reliability and validity of single-item self-892 reports: With special relevance to college students` alcohol use, religiosity, 893 study, and social life

Social participation as an 895 indicator of successful aging: An overview of concepts and their associations 896 with health

Number of Siblings and Social

Skills Revisited Among American Fifth Graders

The Anatomy of Friendship

Large-scale inference: empirical bayes methods for estimation, 904 testing, and prediction (p

An introduction to the bootstrap

Organized 907 Group Activity as a Protective Factor Against Adolescent Substance Use

What is resilience: an affiliative neuroscience approach

Neural correlates of 912 integrated self and social processing

Social interactions in adolescent television viewing. Archives of 915 Pediatrics and Adolescent Medicine

Toolkit on social participation

The social brain: Allowing humans to boldly go where no 923 other species has been

Neural Correlates of Affective Benefit From Real-life Social 927 Contact and Implications for Psychiatric Resilience

Bayesian data analysis

The Neural Basis of 931

Religious Cognition

ICA-based artefact 935 removal and accelerated fMRI acquisition for improved resting state network 936 imaging

How does hemispheric specialization 938 contribute to human-defining cognition? Neuron

How can I connect with 941 thee? Let me count the ways

Loneliness in Everyday Life: Cardiovascular Activity, Psychosocial Context, and 945 Health Behaviors

Exercise, cognition, and the adolescent brain

The multifaceted role of ventromedial prefrontal 950 cortex in emotion, decision-making, social cognition, and psychopathology

Social relationships and 953 mortality risk: A meta-analytic review

Specific default 956 mode subnetworks support mentalizing as revealed through opposing network 957 recruitment by social and semantic FMRI tasks

Similarity in 960 functional brain connectivity at rest predicts interpersonal closeness in the social 961 network of an entire village

Perceiving social 965 interactions in the posterior superior temporal sulcus

Improved optimization for 969 the robust and accurate linear registration and motion motion correction of brain 970 images

A global optimisation method for robust affine 972 registration of brain images

Online social network size 974 is reflected in human brain structure

The impact ofsocial activities, social networks, social 978 support and social relationships on the cognitive functioning of healthy older 979 adults: A systematic review

10,000 social brains: sex 983 differentiation inhuman brain anatomy

Social connectedness, 985 mental health and the adolescent brain

Are loneliness and social isolation associated with 989 cognitive decline?

Participation in the occupations of everyday life

Trilogy of body imaginary: Dance/movement therapy for a 994 psychiatric patient with depression

The effect of social participation on the subjective and 997 objective health status of the over-fifties: Evidence from SHARE

Damage to the left ventromedial prefrontal cortex impacts 1001 affective theory of mind

A 1004 longitudinal study on social support, social participation, and older Europeans' 1005 Quality of life. SSM -Population Health

Accomplishment level and 1008 satisfaction with social participation of older adults: Association with quality of life 1009 and best correlates

Inventory and analysis 1012 of definitions of social participation found in the aging literature: Proposed 1013 taxonomy of social activities

Ventromedial prefrontal volume predicts understanding of others and social 1017 network size

Age Differences in Loneliness from Late 1020 Adolescence to Oldest Old Age

Social engagement mediates the 1023 relationship between participation in social activities and psychological distress 1024 among older adults

Role of the medial prefrontal cortex in coping 1026 and resilience

Mentalizing skills do not differentiate believers from non-believers, but 1030 credibility enhancing displays do

On the relationship between the "default mode network" and the 1034 "social brain

Inclusion of community in self 1037 scale: a single-item pictorial measure of community connectedness

Multimodal population brain imaging in 1043 the UK Biobank prospective epidemiological study

Gray-matter expansion of social brain 1046 networks in individuals high in public self-consciousness

1049 Longitudinal associations between team sport participation and substance use in 1050 adolescents and young adults

Loneliness and meaning in life are 1053 reflected in the intrinsic network architecture of the brain

Loneliness and Health in Older 1056 Adults: A Mini-Review and Synthesis

The functional basis of face evaluation

Participation and social participation: Are 1063 they distinct concepts?

Orbital prefrontal cortex volume predicts social network size: An imaging study of 1067 individual differences in humans

Community 1070 resilience and well-being: an exploration of relationality and belonging after 1071 disasters

Using second-person neuroscience to elucidate 1073 the mechanisms of social interaction

Social 1076 Network Size Affects Neural Circuits in Macaques

Probabilistic programming in 1079 python using PyMC3

The effect of siblings on children's social skills 1081 and perspective taking. Infant and Child Development

Somewhere I belong: Long-term increases in 1085 adolescents

Local-Global Parcellation of the Human 1090

Introspective Minds: Using ALE meta-analyses to study 1094 commonalities in the neural correlates of emotional processing, social & 1095 unconstrained cognition

Minds at rest? Social cognition as the default mode of cognizing and its putative 1099 relationship to the "default system

Variability in Brain Structure and Function Reflects Lack 1103 of Peer Support

Characterization of empathy deficits following prefrontal brain damage: The role 1106 of the right ventromedial prefrontal cortex

The Differential Impact of Social Participation and 1109 Social Support on Psychological Well-Being: Evidence From the Wisconsin 1110 Longitudinal Study

The longevity effects of religious and nonreligious 1113 participation: A meta-analysis and meta-regression

Dynamic neural activity 1116 during stress signals resilient coping

Social participation and healthy ageing: An 1120 international comparison using SHARE data

Fast robust automated brain extraction

Accurate, robust, and automated longitudinal and cross-1126 sectional brain change analysis

The default network of 1130 the human brain is associated with perceived social isolation

Population variability in social brain morphology for social support, 1134 household size and friendship satisfaction

Regional gray matter volume is associated with 1138 empathizing and systemizing in young adults

Sense of Community and Community 1141 Participation: A Meta-Analytic Review

Exercise and 1144 substance use among american youth

Mechanisms linking social ties and support to physical and 1148 mental health

The relationship between individual differences in 1154 gray matter volume and religiosity and mystical experiences: A preregistered 1155 voxel-based morphometry study

Neural responses to visually 1158 observed social interactions

Functional parcellation of 1161 the default mode network: a large-scale meta-analysis

Active ageing: a policy framework

The contribution of club participation to 1167 adolescent health: Evidence from six countries

Segmentation of brain MR images through 1170 a hidden Markov random field model and the expectation-maximization 1171 algorithm

Definition and characterization of an extended social-affective default network. In 800Brain Struct Funct (Vol. 220, Issue 2). prefrontal cortex showed no volume deviation in right hemisphere of (a) sport teams 1210 participants. A ventromedial prefrontal gray matter decrease was shown in (b) religious 1211 groups (light green) and a substance increase in participants of (c) social clubs 1212(yellow). The anterior cingulate cortex showed no volume deviation in (d) sports team 1213 members in members of (e) religious groups in the left hemisphere. A volume decrease 1214in the anterior cingulate cortex was instead shown in members of (f) social clubs (light 1215 green). The parahippocampal area showed a volume increase within the group of (g) 1216 sports teams (yellow) and (i) social clubs (yellow), while no volume deviation was 1217shown for (h) religious groups. For the lingual gyrus, positive and negative volume 1218 effects were shown for (j) sports teams, and no volume effects for (k) religious groups. 1219Negative volume effects were shown for (l) social clubs. Within the temporal lobe of leading mode identified positive (red) and negative (blue) shifts in network connectivity 1237 using a pattern-learning algorithm, with statistical significance at p < 0.05 (one-sided 1238 test) using nonparametric permutation testing. The intra-and inter-network connectivity 1239is depicted for the participants of (a) sports teams, (b) religious groups and (c) social 1240clubs. L/R refers to left and right hemisphere. The default and limbic network showed 1241an increase in connectivity strengths in members of sport teams and religious groups 1242 and a decrease in members of social clubs. 1243 1244 those with little such exchange of personal events. The computed differences in brain-1256behavior associations between both groups (i.e., dominant canonical vector entries) 1257were gathered across the 1,000 perturbed realizations of our original dataset to obtain 1258 faithful bootstrap intervals. These estimates of uncertainty directly quantified how 1259group-related deviations vary in the wider population. Asterisks indicate statistical 1260 relevance based on excluding zero between the 5/95% quantiles of the bootstrap 1261 distribution. These brain-behavior associations showed great correspondence for 1262 regular intake of alcohol and tobacco as well as multifaceted aspects of psychological 1263well-being across participants of (a) sports teams, (b) religious groups and (c) social 1264clubs. Further consistent findings across all three forms of social participants included 1265 the time spent watching television, number of siblings and further mental health 1266conditions. The boxplot whiskers depict the interquartile range. 1267 1268 a b cMembers of sports teams Members of religious groups Members of social clubs