The pursuit of social acceptance: aberrant conformity in social anxiety disorder.

The defining pathological features of social anxiety disorder primarily concern the social landscape, yet few empirical studies have examined the potentially aberrant behavioral and neural patterns in this population using socially interactive paradigms. We addressed this issue by investigating the behavioral and neural patterns associated with social conformity in patients with social anxiety disorder. We recorded event-related potentials when healthy subjects (n = 19), and patients with social anxiety disorder (n = 20) made attractiveness judgements of unfamiliar others, while at the same time, being exposed to congruent/incongruent peer ratings. Afterwards, participants were asked to rerate the same faces without the presence of peer ratings. When compared with healthy controls, social anxiety disorder patients exhibited more positive attitudes to unfamiliar others and conformed more with peers-higher feedback. These behavioral effects were in parallel with neural responses associated with social conflict in the N400 signal, showing higher conformity to peers-higher feedback compared with peers-lower or peers-agree feedback among social anxiety disorder patients. Our findings provide evidence on the behavioral and neural patterns of social anxiety disorder during social interactions, and support the hypothesis that individuals with social anxiety disorder are more motivated to pursue social acceptance and possibly avoid social rejection.

Introduction

Social anxiety disorder is one of the most common anxiety disorders, with an estimated lifetime prevalence of ∼12% (Kessler ). A defining feature of social anxiety disorder is elevated anxiety and consequently the avoidance, of being criticized or rejected by others (Heimberg, 1995; American Psychiatric Association, 2013; National Collaborating Centre for Mental Health, 2013). With the advent of neuroscientific techniques such as functional MRI (fMRI) and event-related potential (ERP), the past decade has witnessed an increasing interest in exploring neural substrates underlying the abnormalities in socioemotional processing among individuals with social anxiety disorder (Etkin and Wager, 2007; Hattingh et al., 2012; Brühl ; Harrewijn ; Qi ). Although symptoms of social anxiety disorder are closely related to social encounters, few studies have examined behavioral and neural patterns of individuals with social anxiety disorder during interpersonal interactions. The current study addressed this gap by assessing the behavioral and neural correlates of social conformity in social anxiety disorder patients (Berns ; Klucharev ; Nook and Zaki, 2015).

Extant research has implicated the critical roles of social rejection/exclusion in developing and maintaining symptoms of social anxiety disorder (Voncken ; Levinson ; Cao ; Fung and Alden, 2017; Linardon ). Perceived social exclusion results in a pain-like experience (Eisenberger ), which are believed to constitute a primary source of social anxiety (Baumeister and Tice, 1990; Leary, 1990). Moreover, individuals with social anxiety disorder compared with healthy controls reported increased feelings of distress (Burklund ; Heeren ) and exhibited more prolonged recovery in response to social exclusion (Zadro ; Oaten ). These findings together suggest that patients with social anxiety disorder are particularly sensitive to social exclusions. Therefore, it is conceivable that the pursuit of social acceptance is significantly higher in social anxiety patients. Indeed, socially anxious individuals are at elevated risk for smoking and drinking due to peer pressure (Patton ; Stewart ).

Previous studies employing social conformity paradigms have consistently demonstrated that people often change their initial ratings during second rating session to align with group ratings. Importantly, however, not all social influence emit their effects equally, such that people are more prone to conforming with group opinions that are more matched with intrinsic opinions of one’s own (Chung ). At the neural level, the consensus between oneself and groups consistently engages the ventral striatum, a region important in valuation and reward processing. Conversely, disagreement with group norms evokes activity in the dorsal anterior cingulate cortex and anterior insula that are implicated in encoding aversive feelings and monitoring conflict (Montague and Lohrenz, 2007; Klucharev ; Campbell-Meiklejohn ; Nook and Zaki, 2015).

Complementing fMRI findings, recent ERP studies have demonstrated that the discrepancy between oneself and majority opinions induces ERP components previously implicated in error or conflict detection, such as feedback-related negativity (FRN) (Chen ; Kim ; Shestakova ; Schnuerch ; Schnuerch and Gibbons, 2015) and N400 (Huang ). For instance, disagreement with group opinions evoked more pronounced N400 component than agreement with group opinions, and amplitudes of the N400 were modulated as a function of levels of disagreement with normative opinions (Huang ). Moreover, the N400 was induced by counter-conformity choices that inherently resulted in disagreement with group opinions (Chen ). Therefore, the N400 component provides a promising measure for the sensitivity to disagreement with majority others.

In this study, we aimed to characterize behavioral and neural patterns of social anxiety disorder patients when they were engaged in social interactions, combining a social conformity paradigm with ERP technique (Chen ; Kim ; Shestakova ; Schnuerch ; Schnuerch and Gibbons, 2015). The conformity paradigm allowed for measures of prosocial tendency as initial attractiveness ratings (i.e. kindness to the presented people) as well as social conformity as changes in ratings in agreement with normative opinions (Klucharev ). We hypothesized that individuals with social anxiety disorder would report higher initial attractiveness ratings than healthy controls to establish increased positive reciprocity. We further hypothesized that social anxiety disorder patients would exhibit higher level of social conformity than controls, especially in the condition where their initial ratings were lower than those of group opinions. At the neural level, we expected that conflict-related ERP components such as N400 would be more pronounced in disagreement than agreement condition as well as more pronounced among social anxiety disorder patients than controls.

Materials and methods

Participants

Participants were 20 (11 females, mean age: 20.42 ± 0.77) right-handed adults who met the DSM-IV criteria for current social anxiety disorder and 19 (9 females, mean age: 20.85 ± 0.75) demographically matched healthy control with no history of any DSM-IV psychiatric disorders. Participants were recruited from a public mental health clinic at Harbin Medical University (DaQing Campus). All participants provided written informed consent in accordance with the Harbin Medical University Review Board guidelines. All participants were diagnosed using the validated Chinese translation of the Structured Clinical Interview for DSM-IV (SCID-IV; Ruying ). The interviewers were two clinical psychiatrists who received training for the administration of the SCID-IV. Inclusion criteria for the social anxiety disorder group included (i) a primary diagnosis of social anxiety disorder according to DSM-IV criteria, (ii) age between 18 and 25 and (iii) right handed. Exclusion criteria included (i) past or current diagnosis of schizophrenia and (ii) history of neurological disorders. Demographic data and the self-reported measures of the 39 participants in the two groups are presented in Table 1.

Table 1.

Demographic information of Participants in the social anxiety disorder and healthy control groups.

	Social anxiety disorder group (n = 20)	Healthy control group (n = 19)	t-test (df = 37)
Age in years (SD)	20.85 (.75)	20.42 (0.77)	1.77
Gender (% females)	55%	42.1%
IAS (SD)	52.25 (10.74)	28.53 (4.39)	2.998**
Trait anxiety (SD)	49.05 (14.00)	29.05 (4.61)	2.16**
State anxiety (SD)	51.05 (13.42)	33.05 (5.31)	3.01**

Note. SAD, social anxiety disorder; IAS, Interaction Anxiousness Scale.

**P < 0.005.

Stimuli

A set of 160 faces with neutral expressions (80 Chinese males, 80 Chinese females), selected from the CAS-PEAL-R1 Face Database (Gao ), were used as stimuli. We resized all face images to 360 × 480 pixels, and removed the salient features on the faces by Adobe Photoshop software. Furthermore, prior to the formal experiment, we recruited an independent sample of 20 participants (10 males, 10 females) to rate the attractiveness of the selected faces (from 1 = ‘not attractive at all’ to 5 = ‘extremely attractive’). The average facial attractiveness score was 2.09 (SD = 0.89), indicating that the attractiveness of the selected faces was moderate.

Experimental procedure

This study employed a social conformity task popularly used in prior research on social influence (Klucharev ; Zaki ). Participants were informed that they were taking part in a study on facial attractiveness and that a group of 200 university students had already participated in this study to rate the attractiveness for a series of faces. Participants then rated the attractiveness of the same faces (initial ratings) and, following the ratings of themselves, would be shown the average rating from 200 other peer participants (normative feedback). Unbeknownst to participants, the normative feedback was generated by a pseudorandom algorithm that resulted in three experimental conditions (Zaki ; Huang ): on ∼30% of the trials, normative ratings were same as the participant’s ratings (peers-agree condition); on ∼30% of the trials, normative ratings were one to three points higher than the participant’s ratings (peers-higher condition); whereas on the remaining approximately 30% of trials, normative ratings were one to three points lower than participant’s ratings (peers-lower condition). Using an adaptive algorithm, we kept the overall ratio of ‘lower’ and ‘higher’ normative ratings approximately equal across the experiment. All face stimuli were randomized across conditions and participants. Approximately 30 min after completing the initial ratings, participants underwent an unexpected (unannounced) behavioral session to rerate the attractiveness of the same 160 faces that they rated in the initial rating session. Unlike the initial rating session, participants were not informed with the normative ratings in the second rating session.

Stimulus presentation and behavioral data collection were implemented using E-Prime software (Version 2.0, Psychology Software Tools, Inc.). On each trial of the initial rating session, participants viewed a face presented at the computer screen and evaluated the degree to which they perceived the face to be attractive on an 8-point scale ranging from 1, unattractive, to 8, very attractive. Participants completed their ratings for each face within 6 s. Their own rating was then highlighted with a blue rectangle frame to the chosen number for 0.5 s. Afterwards, the normative rating (i.e. how attractive the previous group of 200 participants had found that face) was presented (by a red rectangle frame) on the scale for 2 s (Figure 1a).

Fig. 1.

Experimental procedure and behavioral results. (a) Experimental procedure on each trial. (b) Initial ratings as a function of group. (c) Behavioral updates as a function of group and condition. (d) Conformity scores as a function of group and condition. SAD, social anxiety disorder; HC, healthy controls; ns, not significant; *P < 0.05; ***P < 0.005.

Electroencephalographic recording and preprocessing

Electroencephalographic (EEG) data were recoded while participants were completing the initial rating session. Participants sat comfortably in an electrically shielded room ∼80 cm from a computer screen. The EEG data were recorded with a 64-channel NeuroScan system (NeuroScan Inc, Herndon, VA). Raw EEG data were sampled at 1000 Hz/channel, with impedances < 5 kΩ. Online recordings were referenced to the nose, and re-referenced offline to the average bilateral mastoids. Vertical electrooculograms were recorded supra- and infra-orbitally at the left eye. Horizontal electrooculograms were recorded by electrodes at the left and right orbital rims. The online continuous data were digitized with a bandpass filter of 0.05–100 Hz.

EEG data were filtered with a low pass of 30 Hz (24 dB/oct) off-line. Epochs were locked to the normative feedback, beginning 200 ms before feedback onset to 600 ms after. Trials exceeding the threshold of ± 80 μV were excluded from further analysis. Trials without a response were excluded from both behavioral and EEG average analyses. To examine the effect of social influence, trials from three conditions (i.e. peers-agree, peers-higher, and peers-lower) were respectively averaged, and a −200 to 0 ms baseline was used to perform a baseline correction.

Statistical analysis: behavioral data

Initial ratings

Participant’s initial ratings on the face stimuli were compared between social anxiety disorder and healthy control groups with a two-sample t-test.

Behavioral updates

Social conformity was calculated as changes of attractiveness ratings between second and initial ratings on the same faces (Klucharev ). In order to control for the overall changes in ratings across initial and second sessions, we first calculated the mean rating for each rating session and participant. Afterwards, the distance of each rating from the mean was computed for each trial, and the difference in the mean-corrected scores between two sessions was calculated as a ‘behavioral update’ for each face stimulus (Sharot , 2012). A mixed 3 × 2 analysis of variance (ANOVA) on the behavioral updates was applied with Feedback (peers-higher vs peers-lower vs peers-agree) as a within-subjects factor and Group (social anxiety disorder vs healthy control) as a between-subjects factor.

Conformity scores

To establish an even closer association between normative ratings and individual behavior, we performed a correlation analysis between (i) the magnitude of the conflict (i.e. the difference value between participants’ own and group ratings) and (ii) the subsequent changes in the perceived facial attractiveness for each participant (i.e. the difference between the participant’s follow-up ratings and their initial ratings). The correlation coefficients (Pearson’s r) were computed across all trials within each participant, and for peers-higher/peers-lower condition, respectively. The correlation analysis cannot be performed for the peers-agree condition since the conflict value was always zero in this condition. We then transformed raw r scores to z scores through Fisher’s r-to-z transformation to as ‘conformity scores’. Higher conformity scores indicated a stronger within-individual tendency to shift follow-up ratings toward group ratings on a trial-by-trial level. A mixed 2 × 2 ANOVA on the conformity scores was applied with Feedback (peers-higher vs peers-lower) as a within-subjects factor and Group (social anxiety disorder vs healthy control) as a between-subjects factor.

Statistical analysis: ERP data

The amplitude of the N400 was measured as the mean amplitude in a time window of 300–500 ms post-onset of the feedback (i.e. normative ratings) over frontocentral electrodes (Fz, F1, F2, FC1, FC2 and FCz). The resulting N400 amplitude was analyzed with a mixed 3 × 2 ANOVA with Feedback (peers-higher vs peers-lower vs peers-agree) as a within-subjects factor and Group (social anxiety disorder vs healthy control) as a between-subjects factor.

Correlation analyses

Bivariate (Pearson’s r) correlations were computed to determine associations among dispositional (anxiety scores), behavioral (attractive ratings), and ERP (amplitudes) measures. All statistical analyses for behavioral and ERP data were conducted by IBM SPSS Statistics version 18.

Results

Behavioral results

On average, social anxiety disorder group (M = 3.92, SE = 0.21) reported higher ratings of facial attractiveness than healthy control group (M = 2.74, SE = 0.24) during the initial rating session (t37 = 3.74, P < 0.005, Cohen’s d = 0.98, Figure 1b).

A 3 (Feedback: peers-higher vs peers-lower vs peers-agree) × 2 (Group: social anxiety disorder vs healthy control) ANOVA on the behavioral updates yielded a significant main effect of Feedback (F2, 74 = 35.64, P < 0.0005, ηp2 = 0.49), indicating that participants’ follow-up ratings shifted in the direction of normative ratings (i.e. social conformity). Furthermore, a Feedback × Group interaction was significant (F2, 74 = 5.19, P < 0.01, ηp2 = 0.12), demonstrating differences in behavioral updates for the social anxiety disorder and healthy control groups (Figure 1c). In particular, social anxiety disorder group (M = 0.14, SE = 0.038) exhibited stronger conformity to group ratings than healthy control group (M = 0.033, SE = 0.039) in the peers-higher condition (P < 0.05, Figure 2a); whereas social anxiety disorder group (M = −0.22, SE = 0.080) showed lower conformity to group ratings than healthy control group (M = −0.51, SE = 0.082) in the peers-lower condition (P < 0.05, Figure 2b). In the peers-agree condition, there was no significant difference between social anxiety disorder (M = 0.023, SE = 0.039) and healthy control (M = 0.12, SE = 0.040) groups in the changes of ratings (P > 0.05).

Fig. 2.

Individual differences in social conformity. (a and b) Behavioral updates across participants in both groups in the peers-high and peers-low conditions, respectively. (c and d) Conformity scores across participants in both groups in the peers-high and peers-low conditions, respectively. SAD, social anxiety disorder; HC, healthy controls.

A 2 (Feedback: peers-higher vs peers-lower) × 2 (Group: social anxiety disorder vs healthy control) ANOVA on the conformity scores revealed a significant main effect of Feedback (F1, 37 = 14.58, P < 0.0005, ηp2 = 0.28), revealing that conformity scores were higher in the peers-lower condition (M = 0.32, SE = 0.044) than the peers-higher condition (M = 0.088, SE = 0.037). In addition, a Feedback × Group interaction was significant (F1, 37 = 10.96, P < 0.005, ηp2 = 0.23, Figure 1d), indicating that social anxiety disorder group (M = 0.19, SE = 0.052) exhibited higher conformity scores than healthy control group (M = −0.01, SE = 0.053) in the peers-higher condition (P < 0.05, Figure 2c); while social anxiety disorder group (M = 0.22, SE = 0.061) showed lower conformity scores than healthy control group (M = 0.42, SE = 0.063) in the peers-lower condition (P < 0.05, Figure 2d).

Correlations

Initial ratings exhibited positive correlations with behavioral updates in peer-higher conditions for both social anxiety disorder group (r = 0.64, P < 0.005) and healthy control group (r = 0.62, P < 0.005) (Figure 3a). In contrast, Initial ratings showed negative correlations with conformity scores in peer-lower condition for both social anxiety disorder group (r = −0.47, P < 0.05) and healthy control group (r = −0.50, P < 0.05) (Figure 3b).

Fig. 3.

Correlations between initial ratings and social conformity. (a) Correlations between initial ratings and behavioral updates in peers-higher condition within both groups. (b) Correlations between initial ratings and conformity scores in peers-lower condition within both groups. SAD, social anxiety disorder; HC, healthy controls.

ERP results

The ANOVA of N400 amplitude revealed a significant social influence × group interaction effect, F2, 74 = 3.34, P < 0.05, ηp2 = 0.08 (Figure 4). Post-hoc comparisons indicated that peers-higher feedback evoked more negative N400 amplitude than peers-agree feedback among social anxiety disorder group (P < 0.01) but not among healthy control group (P > 0.05). Furthermore, N400 amplitude induced by peers-higher feedback tended to be more negative among social anxiety disorder group than healthy control group (P = 0.078).

Fig. 4.

ERP findings. Upper panel: the grand average ERPs evoked by peers-higher, peers-lower, and peers-agree feedback for both social anxiety disorder and healthy control groups were illustrated. Lower panel: the voltage topographies for the N400 component in each group and condition were illustrated. SAD, social anxiety disorder; HC, healthy controls.

Discussion

The key symptoms of social anxiety disorder mainly concern interpersonal interactions, however, far few studies have examined behavioral and neural patterns of social anxiety disorder in social and interactive contexts. Combining ERP technique with a commonly used social conformity paradigm, our study explored the neural dynamics underlying the social behavior patterns of social anxiety disorder in an interactive context. We demonstrated that social anxiety disorder patients compared with healthy controls reported enhanced attractiveness ratings during initial rating, implicating higher motivations to establish positive reciprocity among individuals with social anxiety disorder than controls. We further revealed differential patterns of social conformity among social anxiety disorder patients and healthy participants; such that social anxiety disorder patients were more subject to social influence when their initial ratings were lower than group ratings (i.e. peers-higher), whereas healthy controls were more subject to social influence when their initial ratings were higher than group ratings (i.e. peers-lower). Underlying these behavioral effects were differential patterns of N400 responses among social anxiety disorder patients and healthy controls. In particular, among social anxiety disorder patients, amplitudes of N400 were more pronounced for peers-higher condition compared with other conditions, whereas among healthy controls no reliable differences were identified. Together, our findings provide evidence regarding behavioral and neural patterns of how social influence is heightened in social anxiety disorder.

We first demonstrated that individuals with social anxiety disorder expressed more positive attitudes towards unfamiliar others than healthy controls, manifesting as higher attractiveness ratings during the initial rating session. In daily life, praising others is a common and effective way to develop and maintain positive interpersonal relationship across a variety of scenarios (Al-Ghamdi, 2017; Blizzard ; Davies and Fafchamps, 2017; Luerssen ). Furthermore, our analysis indicated that social anxiety disorder patients changed their ratings more than healthy controls when normative ratings are higher than one’s own ratings (peers-higher); in contrast, social anxiety disorder patients changed their ratings less than controls when normative ratings are lower than one’s own ratings (peers-lower). These findings suggest that social anxiety disorder patients did not simply exhibit more conformity to the opinions of the majority, but instead balanced positive reciprocity motivation and conformity during social interactions.

The current findings dovetail with the assertion that people incorporate social information in a biased way, depending on the extent to which social information and one’s own opinions are matched (Chung ). For instance, risk-averse individuals tend to conform more with safe decisions of others, whereas risk-seeking individuals tend to conform more with risky decisions of others (Chung ). Likewise, people usually have positive self-evaluations (Alicke and Govorun, 2005; Beer and Hughes, 2010). Accordingly, people process self-relevant social feedback in a positively biased way, i.e. changing self-evaluations more toward desirable than toward undesirable feedback (Korn , 2014).

In light of these findings, the patterns of conformity behaviors observed in the current study resonate with observations of initial ratings. That is, more conformity to peers-higher feedback among social anxiety disorder patients could be attributed to their more positive attitudes to the presented faces during initial rating session. In a similar vein, more conformity to peers-lower social feedback among healthy controls could be attributed to their more negative attitudes to the presented faces. This conjecture is corroborated by correlational findings that initial ratings were positively correlated with behavioral changes in the peers-higher condition, but were negatively correlated with behavioral changes in the peers-lower condition.

Our ERP findings provided further insights into the neural dynamics underlying the behavioral findings. Specifically, peers-higher feedback induced more pronounced response in a conflict-related ERP component (i.e. N400) among social anxiety disorder patients but not among healthy controls. The N400 component has been found specifically sensitive to violations of social norms, suggesting that the component might serve as a social deviance marker in the human brain (Mu ). Complementing the current findings, recent ERP studies have indicated that disagreement with other induced increased N400 deflection (Chen ; Huang ; Schnuerch ), especially when one’s own ratings were lower than group ratings (Huang ). A possible explanation of these findings is that individuals, especially those with social anxiety disorder, might feel ashamed and guilty about rating others as less attractive than the average, which might be more unacceptable than the situation when their initial ratings were above group ratings (see also Huang ). This is consistent with the hypothesis that individuals with social anxiety disorder compared with controls had stronger motivations to establish positive reciprocity by giving unfamiliar others higher attractiveness evaluations (Weeks ). In brief, our ERP findings suggest that ratings lower than group ratings were encoded as a more aversive event among social anxiety disorder patients than healthy controls, which presumably drove higher conformity in this context for individuals with social anxiety disorder compared with controls.

Several limitations related to the current study should be noted. First, one could argue that higher attractiveness ratings among social anxiety disorder patients than controls might reflect differences in impression forming processes rather than motivations for positive reciprocity between two groups. This alternative is unlikely, considering that individuals with social anxiety disorder often overestimate social threat from faces (Gutierrez-Garcia and Calvo, 2016). For instance, relative to non-anxious individuals, socially anxious individuals often interpret neutral facial expressions in a more negative manner (Mobini ). Second, the current study did not identify changes in conflict-related FRN or N400 components among healthy controls, but those effects have been reported in previous studies (Chen ; Kim ; Shestakova ; Huang ; Schnuerch ; Schnuerch and Gibbons, 2015). The reasons for the discrepancies between current and previous findings should be explored in future studies.

In summary, our work investigated behavioral and neural patterns of social anxiety disorder during social interactions and identified a plausible motivation to give positive evaluation to others and show prosocial conformity among individuals with social anxiety disorder. When compared with healthy controls, social anxiety disorders exhibited more positive attitudes to unfamiliar others during initial rating and were more willing to incorporate peers-higher feedback during a second evaluation. Those behavioral effects were parallel with neural responses in the N400, which was more pronounced in response to peers-higher feedback compared with peers-lower or peers-agree social feedback among social anxiety disorder patients. Our findings provide evidence on the behavioral and neural patterns of social anxiety disorder during social interactions and support the hypothesis that individuals with social anxiety disorder may show more prosocial motivation to pursue social acceptance or avoid social rejection. Our findings also endorse social neuroscience as a promising approach to shed light on the neuropsychological mechanisms of social anxiety disorder.