How Autism and Social Anxiety Shape Emotion Recognition and Confidence

Saanvi Talwar  –  saanvi2369@gmail.com 

August 4th, 2025

Edited by the YNPS Publications Team.

Abstract 

Facial emotion recognition is a basic component of human social cognition that enables individuals to deduce the affective state of another, predict intentions, and regulate social behavior in a way that is harmonious with this information. The integrity of this vital ability is, nevertheless, habitually compromised in people with high autistic traits and social anxiety traits, two dimensions that, while often comorbid, are subserved by divergent cognitive and affective mechanisms. The present study aimed to disentangle the differential impact of these trait dimensions on three important aspects of emotional processing: behavioral accuracy of emotion recognition, implicit facial mimicry responses assessed through facial electromyography (fEMG), and metacognitive sensitivity, operationalized as the degree to which participants’ confidence in their emotion recognition judgments correlated with their objective performance accuracy. Our findings clearly demonstrate that high autistic traits are linked to significant decreases in emotion recognition accuracy, particularly for fear expressions—a threat-detection-related emotion—and also with impaired metacognitive sensitivity, reflecting decreased capacity to validly monitor and evaluate the accuracy of one’s own social perception performance. In stark contrast, high social anxiety traits were not accompanied by deficiencies in recognition accuracy or facial mimicry patterns but were strongly associated with a general reduction of confidence for all emotion categories, suggesting a trait-like underconfidence or self-doubt in social cognitive processes. Moreover, the variability of facial mimicry of happiness and sadness was found as a facilitative condition for the accuracy of emotion recognition, but this beneficial effect was diminished at high levels of autistic traits, indicating that the embodied dimension of social cognition might be disrupted in individuals with higher autism-like features. Collectively, these findings illuminate the differential cognitive-affective processes by which autistic and social anxiety traits impact emotion processing and highlight the fundamental necessity for differential intervention targeting: metacognitive intervention aimed at self-efficacy and confidence is likely to be of most benefit

for social anxiety, whilst interventions aimed at enhancing embodied emotional processing and social-perceptual sensitivity are likely to be of most benefit for individuals with strong autistic traits. 

1. Introduction 

1.1 Facial Emotion Recognition as a Key to Social Interaction 

Facial emotion recognition is a fundamental component of human social cognition, and it is a critical mechanism by which people make inferences regarding other individuals’ emotional states, future behaviors, and intentions. The ability to recognize and interpret facial expressions helps people respond appropriately during social interaction and to facilitate cooperation, empathy, and successful communication. Neural structures of face emotion processing include regions such as the amygdala, which rapidly detects emotionally significant stimuli; the medial prefrontal cortex (mPFC), which plays a role in social and emotional integration; and sensorimotor cortices, via which embodied simulation may be mediated by facial mimicry. These systems interact to facilitate both rapid bottom-up processing of emotional content and slower top-down cognitive appraisal and regulation. Impairment or breakdown of these processes will usually give rise to social issues and emotional misperception and lie behind a broad range of neurodevelopmental and psychiatric disorders. Understanding, therefore, the processes of emotion perception and modulation becomes crucial to demystifying the path to social dysfunction and to creating successful treatments. 

1.2 Autistic Features: Social-Cognitive Impairments and Emotion Processing 

Autism Spectrum Disorder (ASD) is characterized by severe social interaction and communication problems, and is typically associated with restricted and repetitive behavior. Impaired emotion recognition, which can manifest as decreased accuracy in identifying expressions of emotion, particularly for subtle or complex emotions such as fear, sadness, or disgust, is one of the characteristic aspects of ASD. Research has consistently found that individuals with elevated autistic traits—whether in clinical groups with ASD or neurotypical high-scoring individuals on measures like the Autism Spectrum Quotient (AQ)—show deficits in face emotion recognition (Espelöer et al., 2023; Meyer et al., 2022). These impairments have been said to result from neural dysfunctions, such as in the medial prefrontal cortex (mPFC), which plays a role in uncertainty processing and regulating flexible emotional responses (South & Rodgers, 2017). This dysfunction is manifested as intolerance of uncertainty, in that autistic people are more distressed by ambiguity, which in turn may increase social anxiety and avoidance. In the same vein, deficits in emotion regulation (ER) and reduced self-consciousness regarding emotional states also underlie compromised social cognition, further limiting social competence and communication abilities (White et al., 2023; Meyer et al., 2022). Added to these are aberrant facial mimic patterns—irrepressible, unconscious facial muscle contractions to viewed expressions—that appear less informative or slower in autism, potentially compromising embodied emotional simulation (Mahalanobis et al., 2024). 

1.3 Characteristics of Social Anxiety: Metacognitive Distortions and Social Perception Confidence

Social Anxiety Disorder (SAD), which is commonly comorbid with ASD, is characterized by widespread fear of negative judgment and pervasive social avoidance. Unlike in ASD, in which impairments are primarily manifested as emotion recognition accuracy, social anxiety characteristics tend to include a metacognitive bias in the form of low confidence or downestimation of one’s social cognitive abilities despite relatively preserved functioning (White et al., 2023; Bejerot et al., 2021). This divergence is a sign of a basic cognitive distortion in SAD, where individuals have adverse assumptions about their social skills and have a tendency for hyper-rumination on hyperperceptions of social failure, a process called post-event processing (White et al., 2023). Whereas facial mimicry and spontaneous emotion processing are less affected in social anxiety, the trait comes with reduced confidence in all emotion categories, reflecting internalized self-criticism affecting metacognitive sensitivity. The reduced confidence compromises social interaction and fosters avoidance behavior despite objective performance of intact capabilities in emotion perception (Mahalanobis et al., 2024). A comprehension of such performance-confidence dissociation is important to inform the development of interventions such as Metacognitive Training, which are aimed at enhancing biased self-assessment and social functioning in SAD. 

1.4 Embodied Emotion Recognition: The Face of Facial Mimicry 

Facial mimicry is an embodied simulation mechanism in which individuals automatically deploy facial muscles corresponding to observed emotional states, allowing emotion recognition through sensorimotor resonance (Dimberg, 1982; Mahalanobis et al., 2024). Facial mimicry is believed to be part of an efficient, bottom-up route of emotional processing that serves as a complement to slower, top-down cognitive appraisal. Theoretical models propose that mimicry enhances the speed and accuracy of emotion recognition by enabling the observer to ‘feel’ the emotion being observed inwardly, thereby improving affective understanding. Nevertheless, facial mimicry’s link with emotion recognition remains contentious. Meta-analyses yield inconsistent results, some showing mimicry to be able to predict better recognition for some but not all emotions (e.g., happiness and sadness), while others find no significant effects (Liu et al., 2023). Furthermore, there is individual variability in these effects. For autism, responses to mimicry are reduced, delayed, or less correlated with recognition accuracy, particularly for dynamic or briefly exposed stimuli (Espelöer et al., 2023; Meyer et al., 2022). This may be a diminished role of sensorimotor simulation routes in autistic emotion processing, with individuals relying more on other, knowledge-based top-down pathways. Social anxiety, on the other hand, does not typically disrupt mimicry per se but can influence its social function, for example, by amplifying mimicry of polite smiles as appeasement signals (Bejerot et al., 2021). Examining the subtle impact of autistic and social anxiety traits on mimicry will illuminate embodied underpinnings of social cognition. 

1.5 Metacognitive Sensitivity: Confidence Calibration in Emotion Recognition 

Metacognition is the capacity for self-monitoring and self-regulation of one’s own thought processes, such as the ability to evaluate one’s own performance on tasks such as emotion recognition. Metacognitive sensitivity, generally assessed using resources like the area under the type 2 ROC curve (AUROC2), is an index of the alignment between confidence judgments and actual performance. This alignment is essential in adaptive learning and social interaction, as it

allows for adjusting behavior according to the reliability of one’s social perceptions (Mahalanobis et al., 2024). It has been shown that those with greater autistic traits show reduced metacognitive sensitivity in recognizing emotions, an indicator of compromised awareness of one’s social cognitive functioning (Mahalanobis et al., 2024; White et al., 2023). In contrast, high social anxiety trait individuals exhibited lower confidence rather than worse calibration per se, i.e., they tend to underestimate their abilities overall but retain a decent ability to make correct/incorrect judgments. This dissociation indicates distinct metacognitive profiles associated with these traits and suggests intervention targets: improved confidence calibration in autism and adjustment of global negative self-beliefs in social anxiety. 

2. Methods 

2.1 Participants and Trait Assessment 

The sample of the study included 60 adults who were sampled from university and community settings with a high female prevalence (about 70%). Participants varied on a wide range of autistic and social anxiety traits to cover dimensional variation in a largely non-clinical sample. Inclusion criteria were lack of active psychiatric diagnosis, but self-rated level of traits varied broadly, with nearly half the sample exceeding clinical cutoffs on the Liebowitz Social Anxiety Scale (LSAS) for social anxiety traits, indicating potential subclinical or mild social anxiety disorder symptomatology. The Autism Spectrum Quotient (AQ) was employed to assess autistic traits dimensionally. This is in accordance with the Research Domain Criteria (RDoC) system, which emphasizes transdiagnostic, dimensional ratings rather than strict diagnostic categories (Cuthbert & Insel, 2013). Through targeting trait continua, the research aimed to make clear how varying levels of autistic and social anxiety traits on their own and in interaction affect emotion recognition and metacognitive processes. 

Prior to data analysis, rigorous screening for multivariate outliers was performed through the use of Mahalanobis distance measures to identify participants with unusual combinations of attributes or response patterns that could warp correlations. To ensure that the bivariate association was not contaminated by outliers, three participants were excluded from autistic trait and AUROC2 metacognitive sensitivity analyses, and two were excluded from social anxiety trait analyses. This prudent screening added stability and interpretability to statistical findings. Furthermore, the study collected demographic information and accounted for age and sex effects in all analyses because of acknowledged sex differences in emotion processing and ASD and SAD incidence rates (McClure, 2000; Lai et al., 2015). 

2.2 Stimuli and Experimental Paradigm 

The paradigm employed in this research was emotion recognition from dynamic video recordings of natural facial displays of six emotions: happiness, sadness, fear, anger, surprise, and neutral. Dynamic, as opposed to static, pictures were employed to maximize ecological validity because natural emotion perception under natural conditions typically involves transient, unfolding displays (Krumhuber et al., 2013). Each clip was approximately 2 seconds long, a presentation

time adequate to register both automatic and controlled processing stages, but perhaps too short to engage prolonged social attention biases typical of social anxiety (Horley et al., 2003). 

Stimuli were pre-tested via pilot testing for recognizability and intensity consistency. Each emotion was presented three times in randomized blocks without instruction to imitate, so spontaneous facial muscle activation patterns could be measured free of priming effect. Rest or neutral expression baseline periods were interposed for EMG normalization. Repeated presentation was a risk for learning or habituation effects, but this design permitted effective signal averaging and larger statistical power in detecting facial mimicry responses. 

2.3 Facial Electromyography (fEMG) Setup and Processing 

Electrodes were positioned over the corrugator supercilii and zygomaticus major muscles to measure activation related to negative and positive affect, respectively, as in a great deal of prior psychophysiological research (Dimberg, 1982; Larsen et al., 2003). Signals were sampled at 1000 Hz, bandpass filtered between 20–450 Hz, and rectified. Data were baselined and corrected using pre-stimulus intervals and standardized within participants through z-scoring to control for variations in baseline tension of the muscles. The time window of analysis measured rapid mimicry responses from 500–1500 ms following stimulus onset, directed towards implicit, automatic motor resonance rather than voluntary production (Oberman et al., 2007). 

Whereas other studies document delayed onset of mimicry or reduced amplitude of mimicry in ASD groups (McIntosh et al., 2006; Stel et al., 2010), temporal delays were not, in fact, measured here due to the study’s sampling design. Reduced magnitude or altered timing, however, could be critical in the detection of anomalous embodiment of emotion in clinical manifestations. The design also excluded confounds like variation in subjects’ baseline facial expressiveness or movement artifacts by excluding noisy EMG trials. 

2.4 Behavioral Measures: Recognition of Emotions and Metacognitive Judgments 

Participants identified the emotion from a forced-choice alternative following each video presentation and provided confidence ratings on a continuous scale from “not confident at all” to “extremely confident.” The dual response paradigm permitted calculation of objective accuracy and subjective confidence and facilitated analysis of metacognitive sensitivity—how well confidence ratings match actual performance. 

Metacognitive sensitivity was assessed using the area under the type 2 receiver operating characteristic curve (AUROC2), a measure that has traditionally been assumed to be dissociable from response bias in metacognitive awareness (Fleming & Lau, 2014). AUROC2 ratings near 0.5 represent chance-level metacognition, whereas ratings near 1 represent near-perfect calibration of confidence and accuracy. This allowed the study to separate under- or overconfidence per se (metacognitive bias) from genuine metacognitive efficiency or sensitivity. Statistical analysis of how trait levels correlated with accuracy and metacognitive sensitivity controlled for demographics and removed outliers in order to preserve parametric assumptions. 

2.5 Statistical Modeling and Hypothesis Testing

Accuracy of emotional recognition was modelled with generalized linear mixed-effects models with logistic link functions to control for repeated measures by participant and varying stimuli. Fixed effects were facial muscle activity (corrugator and zygomaticus), levels of traits (AQ and LSAS), and their interactions to determine if the strength of mimicry-accuracy relationships varied by trait level. Random intercepts for participants and stimuli controlled for individual differences and stimulus-specific difficulty. Interaction significance was tested using likelihood ratio tests between nested models. Odds ratios and confidence intervals were utilized to present effect size and direction. 

Correlational tests probed correlations among traits and metacognitive sensitivity (AUROC2), excluding Mahalanobis outliers in order to maintain normality for analyses that were parametric in nature. Statistical tests were two-tailed with alpha = 0.05. Predicted value plots and simple slopes analysis were employed for the purposes of graphically displaying significant interactions among mimicry effects on recognition accuracy and different low, medium, and high trait values. 

3. Results 

3.1 Differential Impact of Social Anxiety and Autistic Traits on Emotion Recognition 

The study set out to prove that autistic traits exert a pervasive negative impact on accuracy in the recognition of emotions between emotion categories, with a particularly strong deficit for the recognition of fearful expressions. This is consistent with immense ASD literature highlighting abnormalities in the processing of threat-related cues necessary for social cognition and adaptive interpersonal interaction (Baron-Cohen et al., 1997; Harms et al., 2010). Disorders in fear recognition may be a consequence of atypical amygdala function and reduced sensitivity to socially relevant cues such as the eyes (Pelphrey et al., 2002). 

Notably, identification of sadness demonstrated a divergent pattern, and relative accuracy analyses at higher levels of autistic traits showed modest improvements, which could suggest compensatory cognitive strategies or intact processing systems for some affective states (Brewer et al., 2016). Social anxiety tendencies failed to covary with objective accuracy, as theory would expect under the view that SAD is primarily defined by cognitive bias and affective dysregulation rather than fundamental perceptual deficits (Clark & Wells, 1995; Heinrichs & Hofmann, 2001). 

3.2 Metacognitive Sensitivity and Confidence: Trait-Specific Profiles 

One important observation was the strong reduction of metacognitive sensitivity (AUROC2) with more autistic traits, which indicates that these participants were less able to align confidence judgments with recognition performance. This is an indication of a dissociation between subjective experience and performance monitoring, and this could be impairing adaptive social learning and integration of feedback (Williams et al., 2018). This reduced metacognitive monitoring in autism is accompanied by earlier research suggesting lower introspective awareness of social cognitive ability and may be a contributing factor in difficulties with social communication and interaction.

In contrast, social anxiety features were associated with lower confidence ratings overall but not with diminished metacognitive sensitivity, suggestive of preserved ability to monitor performance but a pervasive negative bias in self-judgment. This underconfidence is consistent with cognitive-behavioral models of SAD emphasizing suboptimal social competence and general fear of evaluative rejection (Rapee & Heimberg, 1997). Importantly, the fall in confidence did not impair the optimistic correlation between accuracy and confidence, showing maintained metacognitive discrimination amidst a general downward bias in self-judgment. 

3.3 Facial Mimicry Patterns and Modulation by Traits 

EMG data showed that facial mimicry, as expressed through corrugator and zygomaticus muscle activity, positively predicted recognition accuracy of sad and happy expressions, respectively, offering support for the embodiment theory of emotion recognition whereby sensorimotor simulation contributes to emotion understanding of others (Niedenthal, 2007). This facilitation effect of mimicry was, however, diminished in individuals with more autistic characteristics. Specifically, the association of corrugator activity and recognition of sad faces attenuated with increasing autistic traits, suggesting a less salient contribution of embodied simulation processes in this group. 

Similarly, zygomaticus activity was predictive of more recognition of happiness, but this relationship was also reduced in individuals with higher autistic traits. These results are in line with hypotheses that ASD involves altered or diminished embodiment of emotional cues, perhaps due to disruptions in mirror neuron systems or abnormal interoceptive processing (Gallese et al., 2004; Bird et al., 2010). Social anxiety traits did not influence the magnitude of mimicry or accuracy association much, although zygomaticus-accuracy coupling in response to happy faces was slightly disrupted, which could be indicative of social motivational processes or avoidance 

3.4 Emotion-Specific and Contextual Nuances 

Contrary to the predictions of some of the earlier research, no negativity bias was significant for social anxiety traits in overall accuracy outcomes, yet relative accuracy suggested a trend toward improved identification of angry expressions. This concurs with meta-analyses illustrating inconsistent evidence for enhanced threat bias in SAD in non-ecologically valid or interactive situations (Bar-Haim et al., 2007). 2-second stimulus presentation and absence of true social interaction may have mitigated the elicitation of attentional biases or threat hypervigilance characteristic of SAD (Horley et al., 2003). The heterogeneity of emotion processing changes and the need for finely nuanced examination by emotion category are further indicated by emotion-specific autistic trait patterns, including the differential impact on fear and sparing/enhancement of sadness recognition. 

4. Discussion 

4.1 Detailed Profiles of Emotion Processing Changes

This systematic investigation provides strong evidence that autistic and social anxiety features influence bottom-up and top-down processes of emotion recognition differently. Stronger autistic traits were associated with impaired accuracy in recognition, particularly for fear, along with lower metacognitive sensitivity and lower facial mimicry utilization. These findings are in favor of models proposing that autism involves both perceptual and introspection deficits on social cognition (Pellicano & Burr, 2012; Lawson et al., 2017). Mimicry attenuation would suggest disruption of the automatic embodied simulation processes, possibly forcing reliance on more effortful cognitive appraisal channels that do not fully replace speedy emotion processing demands. 

By contrast, social anxiety is characterized by predominantly affected subjective confidence with minimal interference with accuracy or mimicry, indicating a metacognitive bias instead of an invariant perceptual or bodily deficit. The preserved metacognitive sensitivity with lowered confidence fits with cognitive-behavioral accounts emphasizing maladaptive self-evaluation and fear of criticism as core processes in SAD (Clark & Wells, 1995). This distinction emphasizes the utility of developing clinical treatments designed to address particular aspects of social cognitive impairment. 

4.2 Theoretical and Clinical Implications 

Our findings validate the theoretical explanation of emotion recognition as a dual-route model of bottom-up sensorimotor simulation and top-down metacognitive appraisal. The trait-specific dissociation between these routes in autism suggests intervention techniques like facilitating embodied emotional simulation through mimicry training and enhancing metacognitive monitoring through explicit criticism and reflective practice (Russo-Ponsaran et al., 2021). For social anxiety, cognitive restructuring of negative self-schemas and building confidence may be best. Interventions with metacognitive training, mindfulness-based approaches, or biofeedback for recalibration of internal body awareness may yield synergistic improvements (White et al., 2023). 

4.3 Limitations and Future Research Directions 

While deriving from a dimensional trait approach increases ecological validity, utilizing a non-clinical sample limits direct applicability to diagnosed ASD and SAD populations, who will present with less homogeneous and more severe profiles. The sample’s female predominance limits generalizability based on reported sex differences in emotion processing and trait expression (Lai et al., 2015; McClure, 2000). The absence of real social interaction and repeated static laboratory stimulation may have reduced ecological validity and the expression of trait-related biases. Subsequent research should involve dynamic, interactive paradigms, multi-method assessments including neuroimaging and physiological interoception measures, and longitudinal approaches to explain developmental trajectories. Clinical samples and gender-balanced cohorts will be vital to maximizing the translational relevance of such findings. 

5. Conclusion

This study delineates subtle and differential autistic and social anxiety features that influence emotion recognition processes in adults. Autistic features correlated with impaired fear recognition, metacognitive sensitivity disruption, and diminished facial mimicry contributions, consistent with core disruptions in bottom-up and top-down social cognitive process mechanisms. Social anxiety features, impairing neither objective performance nor mimicry, elicited systematic underconfidence in social cognitive decisions, consistent with expectations under maladaptive cognitive bias. These findings advance understanding of social cognitive heterogeneity on trait dimensions and underscore the necessity of individually tailored intervention tactics that target specific component processes. Future investigation that balances ecological validity, neurobiological markers, and clinical samples will be necessary to optimize outcomes for individuals with autism spectrum and social anxiety disorders. 

6. Acknowledgments 

The writer is also keen on expressing their sincere appreciation to researchers whose early work explained the intricate connection between stress and working memory in adolescents. Acknowledgement is also due to colleagues and mentors for useful comments and criticisms that shaped this review. 

7. References

1. Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., & van IJzendoorn, M. H. (2007). Threat-Related Attentional Bias in Anxious and Nonanxious Individuals: A Meta-Analytic Study. Psychological Bulletin, 133(1), 1–24. 

2. Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001). The Autism-Spectrum Quotient (AQ): Evidence from Asperger Syndrome/High-Functioning Autism, Males and Females, Scientists and Mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. 

3. Bird, G., Silani, G., Brindley, R., White, S., Frith, U., & Singer, T. (2010). Empathic Brain Responses in Insula Are Modulated by Levels of Alexithymia But Not Autism. Brain, 133(5), 1515–1525. 

4. Brewer, R., Biotti, F., Catmur, C., Press, C., Happé, F., Cook, R., & Bird, G. (2016). Can Neurotypical Individuals Read Autistic Facial Expressions? Atypical Production of Emotional Facial Expressions in Autism Spectrum Disorders. Autism Research, 9(2), 262–271. 

5. Clark, D. M., & Wells, A. (1995). A Cognitive Model of Social Phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier (Eds.), Social Phobia:

Diagnosis, Assessment, and Treatment (pp. 69–93). Guilford Press. 

6. Cuthbert, B. N., & Insel, T. R. (2013). Toward the Future of Psychiatric Diagnosis: The Seven Pillars of RDoC. BMC Medicine, 11(1), 126. 

7. Dimberg, U. (1982). Facial Reactions to Facial Expressions. Psychophysiology, 19(6), 643–647. 

8. Espelöer, K., et al. (2023). Intolerance of Uncertainty as an Intermediate Factor in Anxiety Development in Adults with Autism Spectrum Disorder. Journal of Autism and Developmental Disorders, 53(5), 2345–2361. 

9. Fleming, S. M., & Lau, H. C. (2014). How to Measure Metacognition. Frontiers in Human Neuroscience, 8, 443. 

10. Gallese, V., Keysers, C., & Rizzolatti, G. (2004). A Unifying View of the Basis of Social Cognition. Trends in Cognitive Sciences, 8(9), 396–403.