Social intelligence mediates the protective role of resting-state brain activity in the social cognition network against social anxiety

Abstract Background Social intelligence refers to an important psychosocial skill set encompassing an array of abilities, including effective self-expression, understanding of social contexts, and acting wisely in social interactions. While there is ample evidence of its importance in various mental health outcomes, particularly social anxiety, little is known on the brain correlates underlying social intelligence and how it can mitigate social anxiety. Objective This research aims to investigate the functional neural markers of social intelligence and their relations to social anxiety. Methods Data of resting-state functional magnetic resonance imaging and behavioral measures were collected from 231 normal students aged 16 to 20 years (48% male). Whole-brain voxel-wise correlation analysis was conducted to detect the functional brain clusters related to social intelligence. Correlation and mediation analyses explored the potential role of social intelligence in the linkage of resting-state brain activities to social anxiety. Results Social intelligence was correlated with neural activities (assessed as the fractional amplitude of low-frequency fluctuations, fALFF) among two key brain clusters in the social cognition networks: negatively correlated in left superior frontal gyrus (SFG) and positively correlated in right middle temporal gyrus. Further, the left SFG fALFF was positively correlated with social anxiety; brain–personality–symptom analysis revealed that this relationship was mediated by social intelligence. Conclusion These results indicate that resting-state activities in the social cognition networks might influence a person's social anxiety via social intelligence: lower left SFG activity → higher social intelligence → lower social anxiety. These may have implication for developing neurobehavioral interventions to mitigate social anxiety.


Introduction
Social intelligence refers to a set of psychosocial skills that encompass effectiv e self-expr ession, understanding social envir onments, and acting wisely in social interactions (Bar-On, 2006 ; Barnes and Sternberg, 1989 ;Petrides, 2011 ).As a crucial character strength in terms of positive psychology (Peterson and Seligman, 2004 ), social intelligence is beneficial to personal de v elopment and well-being (Avlaev, 2020 ;Azañedo et al., 2020 ).There is gr owing e vidence that social intelligence is pr otectiv e a gainst social anxiety, which is a popular mental health problem featured with fear/avoidance of social interaction and performance conditions (American Psychiatric Association, 2013 ).In particular, social anxiety is related to alterations in the process of self-social information, emotional and social loneliness, and social cognitive patterns (Alvi et al., 2022 ;Harrewijn et al., 2017 ;Wolters et al., 2023 ), all of which overlap with the concept of social intelligence (Petrides, 2011 ;Silv er a et al., 2001 ).There are stable negative associations of social intelligence with social anxiety in se v er al different samples (An and Kochanska, 2021 ;Chen et al., 2023 ;Hampel et al., 2011 ;Pickard et al., 2018).
Despite m uc h r esearc h r egarding social intelligence at the behavior al le v el (Walker and Foley, 1973 ), r elativ el y little is known about its neurobiological underpinnings.As a core aspect of social cognition (Emery et al., 2007 ), social intelligence has been expected to involve brain clusters belonged to social cognition network (SCN), e.g.pr efr ontal cortex, tempor al cortex, amygdala, and insula (Brothers, 1990 ;Frith, 2007 ;Kilford et al., 2016 ).An early functional magnetic resonance imaging (fMRI) research identified higher activations in the amygdala, inferior and superior frontal gyrus (IFG/SFG), superior and middle temporal gyrus (STG/MTG), cingulate gyrus , precuneus , and insula during a 'theory of mind' task, which tests social intelligence (Baron-Cohen et al., 1999 ).In a structural MRI research of healthy participants, higher social intelligence scores were related to greater gray matter volume (GMV) of bilateral caudate (Myznikov et al., 2021 ).In individuals with autism spectrum disorders, social information processing ability, an important dimension in social intelligence (Silv er a et al., 2001 ), has been related to functional connectivity between SFG and the anterior insula (Francis et al., 2019 ).This limited evidence indicates an important role for the SCN in social intelligence.Our first aim in the current resear ch w as to examine that relationship using resting-state brain activity in healthy individuals.
The SCN is also implicated in social anxiety.Meta-analyses hav e demonstr ated incr eased task-based fMRI acti vity in am ygdala, insula, IFG, and STG in SAD patients (Etkin and Wager, 2007 ), and increased GMV in prefrontal-temporal regions including SFG, IFG , STG , and MTG (Liu et al., 2022 ;Wang et al., 2021 ).There is growing evidence from resting-state studies in SAD of alterations in fr ontal r egions, suc h as lo w er activity in SFG and median cingulate gyrus (Brühl et al., 2014 ;Mizzi et al., 2022 ), possibly related to the impaired social cognitive processing.In the healthy population, br ain activ ations in the medial pr efr ontal cortex, temporal gyrus, and STG during a social norms processing task are positiv el y r elated to social anxiety (Bas-Hoogendam et al., 2020 ), whic h r eflects the r ole of these ar eas in self-r efer ential pr ocessing and social cognition, including understanding people's intentions from their actions.A voxel-based morphometry (VBM) research in healthy adolescents r e v ealed a positiv e link of social anxiety with right MTG GMV, which is an important structure for cognitiv e pr ocessing r egarding subjectiv e feeling and emotion (Wang et al., 2021 ).T his o v erla p between r egions associated with social anxiety and social intelligence suggests an underlying pathway from the SCN to social intelligence and social anxiety, although the specific mechanism linking brain features to behavior is unclear.Ther efor e, our second aim of this r esearc h was to use a br ain-personality-symptom fr ame work (Wang et al., 2021 ) to e v aluate whether social intelligence could mediate the linkage of SCN with social anxiety.
To explore the questions, we used resting-state fMRI (RS-fMRI) scanning and well-validated scales on social intelligence and social anxiety.We analyzed resting-state brain activity using the fractional amplitude of lo w-frequenc y fluctuation (fALFF) appr oac h (Zou et al., 2008 ); it has good validities and reliabilities (Gao et al., 2023 ;Li et al., 2022 ;Ma et al., 2023 ) and high specificities and sensitivities (Lv et al., 2018 ), and is widely emplo y ed to detect brain areas associated with behavioral constructs (Canario et al., 2021 ;Zou et al., 2008 ) and to identify brain activity changes among neur opsyc hiatric disorders (Ma et al., 2023 ;Qiu et al., 2019 ;Shang et al., 2016 ).Next, we implemented correlation analysis to explore the connections of social intelligence to voxel-wise fALFF across the whole brain, and confirmed this association with prediction analyses.Given the previous literature, we expected to detect this corr elation in SCN br ain r egions (e.g.SFG , IFG , STG , MTG , precuneus, amygdala, and insula).We then tested whether the brain areas associated with social intelligence were linked to social anxiety.Last, we carried out mediation analyses to test the indirect effect of social intelligence on the linkage of fALFF to social anxiety.
We studied students in the adolescent sta ge, a tr ansition period marked by changes in affection and cognition linked with structural and functional brain reorganization (Konrad et al., 2013 ;Foulkes and Blakemore , 2018 ).T here is a growing evidence of increasing social anxiety among adolescents, increasing their vulnerability to developing SAD (Haller et al., 2015 ;Miers et al., 2013Miers et al., , 2014 ) ). T hus , our work may throw light on the pr otectiv e function of social intelligence against social anxiety, and help the dev elopment of tar geted neur obehavior al interv entions to enhance this.

Participants
Our r esearc h enr olled 234 normal students, r ecentl y gr aduated from local public high sc hools, all nativ e Mandarin Chinese speakers who reported no history of neuropsychiatric illness.After three were excluded for incidentally discovered structural brain abnormalities, 231 participants (121 females) were included in the study.Each student was right-handed given the self-reports of Edinbur gh Handedness Inv entory (Oldfield, 1971 ) and gav e informed written consent before the testing, which was approved by the West China Hospital r esearc h ethics committee .T his dataset was collected as part of a larger project primarily investigating the neur al mec hanism underl ying personalities , academic success , and mental health in adolescent students (Pan et al., 2023b ;Wang et al., 2018 ).

Tromsø Social Intelligence Scale (TSIS)
This was assessed using the 21-item TSIS (Silv er a et al., 2001 ).It has three dimensions (i.e.social awareness, social skills, and social information pr ocessing), eac h comprising se v en statements.For each item, participants were asked to indicate ho w w ell a statement (e.g."I find people unpredictable") describes them, using a se v en-point Likert scale fr om 1 to 7. The total TSIS scor e (the most useful in empirical r esearc h (Savci et al., 2022 ;Swain et al., 2022 )) sums the ratings of each item, a higher score showing higher social intelligence .T his scale has good psychometric properties in adults (Silv er a et al., 2001 ) and adolescents (Gini, 2006 ), and the Chinese version has satisfactory validities and reliabilities (Guo et al., 2012 ;Ling et al., 2016 ;Zhou et al., 2014 ).Cronbach's Alpha for TSIS here was 0.89, evidencing good internal reliability.

Liebowitz Social Anxiety Scale (LSAS)
This was e v aluated using the 24-item LSAS (Liebowitz, 1987 ), depicting corresponding situations for each of which participants were asked to indicate, on a scale from 0 to 3, the frequency and degr ee r egar ding av oidance and fear.The total scor e sums the r atings for each item, higher scores representing greater social anxiety.LSAS shows satisfactory psychometric features (Baker et al., 2002 ;Heimberg et al., 1999 ;Oakman et al., 2003 ), and adequate validities and reliabilities among Chinese samples (He and Zhang, 2004 ;Liao et al., 2010 ;Yang et al., 2015 ).Cr onbac h's Alpha for LSAS here was 0.93, evidencing satisfied internal reliability.

Subjective Socioeconomic Status Scale (SSSS)
As socioeconomic status (SES) plays an important role in brain de v elopment (Hac kman and Far ah, 2009 ), we adjusted for SES assessed using a single-item scale, which is a diagram of a ladder using 10 rungs (Adler et al., 2000 ).The participants were required to choose a rung to indicate their par ents' situations.Compar ed with objectiv e measur es of SES, the SSSS is mor e pr edictiv e r egarding health-linked variables and has been well used among Chinese samples (Lai et al., 2020 ;Liu et al., 2023 ).

Data collection
MRI data were obtained from a Siemens (Erlangen, Germany) Trio 3.0 T MRI scanner, equipped with a 12-channel head coil.We obtained anatomical images (T1-weighted) using these parameters: 176 slices, flip angle 9  .We used foam pads and ear plugs to reduce head motions and noise perception; during resting scans, participants were indicated to lie still, to close their eyes but remain a wake , and to not thinking of things on purpose.

Data pr epr ocessing
Ima ges wer e inspected by a clinical r adiologist blind to the curr ent study; thr ee students wer e excluded giv en neur oanatomical alter ations.Ima ge pr epr ocessing, using SPM softwar e and the DPARSF toolbox (Chao-Gan and Yu-Feng, 2010 ), included: discarding the first 10 images to ensure signal stabilization; correcting slice timing and head motions; realignments; normalizing using 3 × 3 × 3 mm 3 resolutions; smoothing with an 8 mm full-width at half-maximum Gaussian kernel; removal of linear trends; and computing the mean frame-wise displacement (FD).Then we regressed out six head motioning parameters (Friston et al., 1996 ), as well as the signals of cer ebr ospinal fluid, white matter, and global mean.

fALFF calculation
We computed this measure using the method of Zou et al. ( 2008 ) via the DPARSF toolbox (Chao-Gan and Yu-Feng, 2010 ), which is based on the r esearc h of Zang et al. ( 2007 ).The detailed computing processing of this measure can be seen our pr e vious work (Zhang et al., 2022 ).

Sta tistical anal ysis fALFF-behavior correlation analysis
To detect the brain areas where resting activity was linked to social intelligence, we correlated individual social intelligence scores with voxel-wise fALFF in the brain, controlling for gender, a ge, FD, and famil y SES scor es.Further, we carried out conditionby-cov ariate inter action anal yses (P an et al., 2023b ) to test gender difference in the association between social intelligence and fALFF, with a ge, famil y SES, and FD as co variates .Gaussian random field theories were used to conduct corrections for the resulting map (Worsley et al., 1996 ;Eickhoff et al., 2006 ), with a voxelle v el thr eshold P < 0.01 and cluster-le v el thr eshold of P < 0.05, as widely applied with resting-state brain imaging research (Cox et al., 2012 ;Wang et al., 2022 ).We conducted these analyses with REST software (Song et al., 2011 ).

Confirmator y pr ediction analysis
As widely used in neuroimaging studies (Lai et al., 2020 ;Qin et al., 2014 ;Supekar et al., 2013 ;Wang et al., 2021Wang et al., , 2023 ; ;Zhang et al., 2022 ), this was performed to validate the stability of the fALFF-social intelligence connection.Sex, age, FD, and famil y SES wer e tr eated as the contr olling v ariables and the detailed pr ocedur e of this analysis can be seen in our pr e vious studies (Lai et al., 2020 ;Wang et al., 2021Wang et al., , 2023 ; ;Zhang et al., 2022 ).

Mapping onto large-scale brain networks
As depicted in our pr e vious r esearc h (Liu et al., 2023 ;Pan et al., 2023a ), we implemented this to map the detected brain regions onto se v en k e y netw orks: visual netw ork, v entr al attention netw ork, somatomotor netw ork, affective netw ork, central executive network, dorsal attention network, and default mode network (DMN) (Yeo et al., 2011 ).

Mediation analysis
By using the SPSS macro PR OCESS (Hay es, 2017 ), we performed this to c hec k the indir ect effect of social intelligence on the linkage of intrinsic brain activity to social anxiety.In the main analysis, the predict variable ( X ) was resting brain activity, the mediator variable ( M ) was social intelligence, and the outcome variable ( Y ) was social anxiety; the indirect effect is computed as the product of path a (relationship between X and M ) and path b (relationship between M and Y after controlling for X) (Baron and Kenny, 1986 ).The indirect effect measured the mediation, and to estimate its significance we used bootstr a pping pr ocedur es (Pr eac her and Ha yes , 2008 ), in whic h 5000 bootstr a p sampling was used to create 95% confidence interval (CI); if a CI did not contain 0, the indirect effect was significant at P < 0.05.Sex, age, FD, and famil y SES wer e tr eated as nuisance variables.To test the directionality of these relations we built an alternative mediation model in which social intelligence was the X , social anxiety the Y , and r esting-state br ain activity the M .

Behavioral results
The descriptive statistics are shown in Table 1 .All measures might be normally distributed, with kurtosis and skewness between −1 and + 1 (Marcoulides and Hershberger, 1997 ).TSIS total scores wer e highl y positiv el y corr elated with its three-component dimension scores (social information processing: r = 0.84, P < 0.001; social skills: r = 0.87, P < 0.001; social awareness: r = 0.82, P < 0.001), and thus was used as the single measure of social intelligence.Social intelligence did not differ between genders [ t (229) = 0.01, P = 0.99] or correlate with age ( r = −0.03,P = 0.55), but sho w ed a positive correlation with family SES ( r = 0.22, P < 0.01).There was a negative correlation ( r = −0.34,P < 0.001) between social intelligence and social anxiety.

Brain regions associated with social intelligence
Corr elation anal ysis with voxel-wise fALFF (contr olling for gender, age, FD, and family SES) found that social intelligence was related to fALFF in two clusters: positiv el y in the right MTG (Fig. 1 A and B, Table 2 ) and negativ el y in the left SFG (Fig. 2 A and B, Table 2 ).Pr ediction anal yses (with gender, a ge, FD, and famil y SES as the covariates) confirmed the stability of these relationships for both right MTG ( r [pr edicted, observ ed] = 0.22, P < 0.05) and left SFG ( r [pr edicted, observ ed] = 0.16, P < 0.05).In short, higher social intelligence is associated with lo w er SFG activity and higher left MTG activity.Mor eov er, condition-by-cov ariate inter action anal yses r evealed no significant clusters for the interacting effects of social intelligence with gender.
Mapping the MTG onto the large-scale brain networks (Fig. 1 C), the most voxels were in the DMN [relative distribution (RD) 37.22%] and affective network (RD 17.54%).Mapping the SFG onto the large-scale intrinsic functional connectivity atlas (Fig. 2 C), the most voxels were in the central executive network (RD 49.70%) and DMN (RD 15.62%).

Rela tions betw een social anxiety and br ain activity in clusters associated with social intelligence
Having extracted the mean fALFF in these two brain regions we found a positive connection with social anxiety for left SFG ( r = 0.21, P < 0.01), but no correlation for right MTG ( r = −0.12,P = 0.06).T hus , lo w er le v els of social anxiety ar e r elated to lo w er left SFG activities.

Social intelligence links SFG brain activity and social anxiety
Putting together the results of the abo ve , we found that lo w er SFG activities are linked to both lo w er social anxiety and higher social intelligence.We wer e primaril y inter ested in the causal links leading to social anxiety.Applying the brain-personality-symptom analysis described in the Method section, we found that social intelligence (the M in the main model) sho w ed significant mediation effects on the connection of the left SFG activity (the X ) to social anxiety (the Y ) [the indirect effect = 0.084, 95% CI = (0.036, 0.141), P < 0.05], accounting for gender, age, FD, and family SES (Fig. 3 ).By contrast, in the alternate model, left SFG activity (now the M ) did not mediate the link of social intelligence (now the X ) to social anxiety (the common Y ) [the indirect effect = −0.034,95% CI = ( −0.104, 0.007), P > 0.05], accounting for sex, age, FD, and family SES.Thus this evidence supports the model in whic h r esting-state SFG activity in the social cognition network impacts social anxiety via social intelligence, not the alternate model in which social intelligence affects social anxiety via SFG activity.

Discussion
We set out to examine the brain bases (in terms of regional spontaneous brain activity) of social intelligence, and the potential mediating role of social intelligence in linking spontaneous brain activity to social anxiety.We sho w ed that impr ov ed social intelligence was related to lower fALFF in the left SFG and higher fALFF in the right MTG (and vice versa), and that social intelligence mediated the positiv e linka ge between the left SFG fALFF and social anxiety.T his study ma y be the first to define resting-state neurological markers of social intelligence, and throws light on the neur obehavior al mec hanism by whic h social intelligence pr otects against social anxiety.We discuss these points next.

Neur al correla tes of social intelligence
The two regions whose spontaneous activity linked to social intelligence make sense given what is known from other structural and  functional brain studies.First, social intelligence was positiv el y associated with fALFF in right MTG .The MTG , bounded dorsally by the STG/superior temporal sulcus and ventrally by the inferior temporal gyrus/inferior temporal sulcus (Jabbour et al., 2004 ), is a cor e r egion in the SCN (Div eica et al. , 2021;Fernández et al. , 2018;Xu et al., 2019;Yun et al., 2017), involved in processing social signals related to sound and emotion (Feng et al. , 2018 ;K uhnke et al., 2023 ;Sabatinelli et al., 2011 ).fMRI r esearc h has shown MTG activation in theory of mind tasks (Baron-Cohen et al., 1999 ;Diveica et al., 2021;Schurz et al., 2017), and a positive connection between MTG activation and empathy (Immordino- Yang et al., 2009 ;Kédia et al., 2008 ;Mercadillo et al., 2011 ;Moll et al., 2007 ).Individuals with autistic traits show activation of MTG when processing negative emotion (Yu et al., 2020 ), whic h ar e a type of social information (Garrido, 2020 ).In addition, "mentalizing" (sometimes r eferr ed to as "thinking about thinking") a social situation recruits MTG (Veroude et al., 2012 ).A recent meta-analysis reported MTG activation in self-related understanding and perception (e .g. being a ware of, obtaining knowledge about, or making judgments to w ar d the self) (Lobo et al., 2023).Understandings of self-others' beliefs and emotions and complex social situation information are important dimensions of social intelligence (Kosmitzki and John, 1993 ).
Second, social intelligence was negativ el y linked with fALFF of the left SFG, another core region in the SCN (Tuerk et al., 2020 ).The SFG is activated by the tasks of theory of mind (Baron-Cohen et al., 1999 ).Children with autistic spectrum disor der sho w negativ e r elations between left SFG GMV and social communication ability (Cheng et al., 2023 ).The le v el of social information processing, a core component of social intelligence (Silv er a et al., 2001 ), is predicted by the functional connectivity of SFG and anterior insula (Francis et al., 2019 ).The SFG is involved in social cognition (Chen et al., 2018 ), self-awar eness (Goldber g et al., 2006), and emotional regulation (Frank et al., 2014).In a social-cognitive task, functional connectivity in the left SFG is increased in the othersituation compared with the self-situation (Ribeiro da Costa et al., 2022).Recent meta-analysis has revealed SFG involvement in af-filiation and attachment as well as understanding and perception of self and others constructs in social processes (Lobo et al., 2023).

Correlates of social anxiety
The moder ate negativ e behavior al corr elation we observ ed r egarding social intelligence and social anxiety is consistent with pr e vious r eports (Ashbaugh et al. , 2005 ;Hampel et al. , 2011 ;Voncken and Bögels, 2008 ).We also identified a positive relation of social anxiety to fALFF in left SFG.Again, this neural correlate makes sense .T her e ar e man y r eports of SFG functional/structur al alterations in SAD patients (Hamilton et al., 2015 ;Liu et al., 2022 ;Qiu et al., 2015 ;Wang et al., 2018 ) and healthy people with increased social anxiety (Smith et al., 2019 ;Kim et al., 2023).These include resting state studies: SAD patients show increased intra-network functional network connectivity in the anterior DMN (mainly the SFG) in contrast to healthy controls (Zhang et al., 2023 ); in families geneticall y enric hing for SAD, social anxiety co-segregates with the functional connectivity in the dorsal attention network including SFG (Bas-Hoogendam et al., 2021 ) and altered functional connectivity of SFG and anterior cingulate gyrus, a k e y circuit of SCN has been suggested as a dia gnosticall y useful biomarker in SAD (Cui et al., 2017 ).

T he media ting role of social intelligence
Thus there is a connection among higher social intelligence, lo w er left SFG activity, and lo w er anxiety (and vice versa).If, as we hypothesize, these ar e causall y linked, our mediation anal ysis giv es the direction of causation: social intelligence plays a mediating role between fALFF in the left SFG and social anxiety.This suggests that resting-state activity in the social cognition network might influence a person's social anxiety via social intelligence: lo w er left SFG activity → higher social intelligence → lo w er social anxiety.

Limitations
First, a cross-sectional design cannot draw definitive causal conclusions.Longitudinal studies need to be done.Second, although these behavioral measures are widely used and have satisfied reliabilities and validities (Majid et al., 2022 ;Pepe et al., 2021 ), the selfreport aspect may lead to response bias (Lyu and Bolt, 2022 ).Objectiv e behavior al measur es ar e needed in futur e r esearc h.Third, the participants are healthy high sc hool gr aduates, a population to vulnerable to social anxiety (Bruce et al., 2005 ), thus our results may not generalize to other samples.Futur e r esearc h should recruit participants with more diversity in age, education, occupation, and mental illness.Last, in our study only SFG and MTG were associated with social intelligence, not other cor e br ain r egions in the SCN such as the amygdala, IFG, and STG.T his ma y be due to our use of fALFF, which can only reflect local brain function.Future studies could usefully take a network approach (Hacker et al., 2013 ;Lin et al., 2023;Yeo et al., 2011 ).

Conclusion
This r esearc h extends pr e vious inv estigations by identifying a functional brain marker of social intelligence and r e v ealing a potential "brain-personality-symptom" pathway to protect social anxiety .Specifically , we found that social intelligence was supported by spontaneous activities in the right MTG and left SFG and r e v ealed indir ect effects of SFG activity on social anxiety via social intelligence .T his r esearc h pr ovides an insight into the neurobiological bases linked to social intelligence, and may have significance for underl ying neur opsyc hological markers for the early detection and prevention of social anxiety in adolescents, and for pr e v entiv e and ther a peutic neur obehavior al interv entions (Kaminska et al., 2020 ;Paes et al., 2013 ) to reduce the social anxiety of adolescents and impr ov e their mental health.

Figure 1 :
Figure 1: Brain region is positively linked to social intelligence.( A ) Brain images reveal that social intelligence is positively linked with fALFF in the right MTG after adjusting for gender, age, head motion, and family SES.( B ) Scatter plots demonstr ate the corr elation between social intelligence and fALFF in the MTG ( r = 0.28, P < 0.001).( C ) Plot shows the similarity of co-activation pattern of right MTG to large-scale functional networks.Abbr e viations: L, left; R, right; DMN, default mode network; CEN, central executive network; DAN, dorsal attention network; VAN, v entr al attention network; SMN, somatomotor network; VN, visual network; AFN, affective network.

Figure 2 :
Figure 2: Brain region is negatively linked to social intelligence.( A ) Brain images reveal that social intelligence is negatively linked with fALFF in the left SFG after adjusting for gender, age, head motion, and family SES.( B ) Scatter plots demonstr ate the corr elation between social intelligence and fALFF in the SFG ( r = −0.27,P < 0.001).( C ) Plot shows the similarity of co-activation pattern of left SFG to large-scale functional networks.Abbr e viations: L, left; R, right; DMN, default mode network; CEN, central executive network; DAN, dorsal attention network; VAN, v entr al attention network; SMN, somatomotor network; VN, visual network; AFN, affective network.

Figure 3 :
Figure 3: Mediation analysis.Social intelligence mediates the effect of left SFG activity on social anxiety.Standardized r egr ession coefficients are presented in the path diagram.Gender, age, head motion, and famil y SES ar e contr olled for in the model.* * * P < 0.01, * P < 0.05.

Table 1 :
Means, SD, ranges, and correlations of age and behavioral constructs.

Table 2 :
Brain regions associated with social intelligence.