The Anxious Mind: Examining the Conflicting Relationships Between Distinct Aspects of Social Anxiety and Social Cognition

1. Introduction

Social anxiety is characterized by a persistent and intense fear of social settings (Beck et al., 2005; Jefferson, 2001; Schneier & Goldmark, 2015; Stein & Stein, 2008). Critically, this fear is disproportionate to the situation and often detrimental to daily functioning, leading to extreme shyness, social withdrawal, and even panic attacks (Alomari et al., 2022). A clinical diagnosis of Social Anxiety Disorder (SAD) is met if symptoms persist for at least 6 months and cannot be explained by another disorder or medical condition (American Psychiatric Association, 1980, 1987, 1994, 2000, 2013, 2022).

Unfortunately, SAD and its symptoms are quite common. SAD is the most prevalent mental or emotional disorder, with a lifetime prevalence of up to 16% and an age of onset as early as preschool, though it more commonly emerges in middle childhood (Baartmans et al., 2020; Halldorsson & Creswell, 2017; Stuijfzand & Dodd, 2017; Rapee et al., 2019). SAD often develops into a chronic condition with severe and long-term consequences, such as poor mental health as well as dissatisfaction with, and/or lower achievement in, work, education, and relationships (Baartmans et al., 2020; Beidel & Alfano, 2011; Russell & Topham, 2012; Vilaplana-Pérez et al., 2021).

Symptoms of SAD vary widely among individuals and exist on a continuum ranging from severe – including incapacitating dread, panic attacks, and social withdrawal – to more moderate – including nervousness, worry, and gaze avoidance (Crişan et al., 2016; Dell’Osso et al., 2023; Stein et al., 2000, 2004). Importantly, social difficulties associated with social anxiety are not restricted to populations that meet the clinical threshold for a SAD diagnosis; research suggests that even subclinical symptoms, such as social discomfort or heightened emotional arousal, significantly impact quality of life (Davidson et al., 1994; Fehm et al., 2008; Filho et al., 2010; Kashdan & Roberts, 2004; Merikangas et al., 2002; Stein et al., 1994; Turner et al., 1986). For instance, both adults and children who exhibit social anxiety tend to be rated by their peers as less socially competent (Alden & Taylor, 2004; Meleshko & Alden, 1993), less likeable, and less warm (Creed & Funder, 1998; Papsdorf & Alden, 1998). Perhaps unsurprisingly, then, socially anxious individuals are reported as having fewer close friendships and romantic relationships compared to the general population as well as those with other anxiety disorders (e.g., Generalized Anxiety Disorder, Alden & Taylor, 2004).

Subclinical symptoms of social anxiety at any point put individuals at higher risk of reaching clinical levels of anxiety and secondary depression in the future (Beesdo et al., 2007; Bosman et al., 2019; Zhong et al., 2024); thus, understanding the emergence of social anxiety in childhood is of particular importance. Indeed, positive and stable social relationships are essential for good health outcomes in children (Birrell et al., 2025; Larkins, 2014; Peters et al., 2011), whereas a lack of social connection is associated with negative mental and physical health problems as well as poorer cognitive development; that said, individual outcomes may show high variability (Almeida et al., 2021; Batty et al., 2021; Holt-Lunstad, Smith, & Layton, 2010; Pearce et al., 2023). Failure to build stable social connections in childhood may lead to further social withdrawal, feelings of loneliness, and fewer opportunities to develop interpersonal skills, thereby exacerbating social anxiety symptoms (Ginsburg et al., 1998; Jefferson, 2001; Lacombe et al., 2024; Rapee et al., 2023; Reznick, 1989; Rubin & Mills, 1988). Although socially anxious individuals can experience peer difficulties at any age, they are particularly concerning in late childhood and adolescence, when peer acceptance and the presence of a supportive social network becomes increasingly crucial (Crone & Dahl, 2012; Rudolph, 2021).

Children’s social skills (or difficulties) are intimately tied to their social cognitive development (i.e., the range of cognitive processes which enable people to function within a social group) (Frith, 2008). In the current manuscript, we are specifically interested in the social cognitive processes involved in identifying and reasoning about the mental and emotional states of others, sometimes called ‘social perspective taking’ or ‘mental state understanding’. Importantly, certain aspects of these social cognitive processes have been linked to clinical and sub-clinical levels of social anxiety in both adults and children (Alvi et al., 2022; Briscoe et al., 2024; Eikelboom et al., 2025; Huang et al., 2026; Karaca Cengiz et al., 2025; McClure & Nowicki, 2001), but considerable debate remains over exactly which social cognitive processes are affected and how.

A large body of research in adults suggests that social anxiety is associated with poorer social cognition. For instance, studies show that socially anxious individuals often perceive ambiguous social cues as more negative or threatening, such as mislabelling neutral faces as angry (Alden et al., 2008; Beck et al., 2005; Gutiérrez-García & Calvo, 2017; Heuer et al., 2010; Maoz et al., 2016; Peschard & Philippot, 2017; Yoon et al., 2014). They may also have difficulty disengaging from negative social stimuli, like a look of disgust or a fearful tone of voice (Clark & Wells, 1995; Rapee & Heimberg, 1997). Other findings suggest that socially anxious individuals avoid engaging with negative social stimuli in the first place, and instead show an inappropriate amount of attention to the self and their internal states (Adamis et al., 2025; Mansell et al., 2003; Mellings & Alden, 2000; Poole & Henderson, 2022; see also Crozier, 1979 and Melchior & Cheek, 1990), that can lead to slower reaction times and poorer accuracy in identifying social cues (Tseng et al., 2017). Altogether, these findings suggest that biases in social attention and interpretation are heightened in socially anxious individuals. This may lead them to misinterpret, discount, or exaggerate important social feedback, resulting in poorer social cognition and limited capacity to reason about the mental states of others (Amir et al., 2003; Buckner et al., 2010; Liang et al., 2017; Liu et al., 2024; Vytal et al., 2013).

All that said, work examining links between social anxiety and social cognition has produced mixed findings. In contrast to the work reviewed above, some research shows no significant differences between socially anxious adults and healthy controls in various social cognitive tasks; including accuracy-based (de Jong & Martens, 2007; Folz et al., 2023; Lenton-Brym et al., 2018; Reinholdt-Dunne et al., 2009; Torro-Alves et al., 2016) and self-report measures (Ballespí et al., 2021; Nunes et al., 2018; for a meta-analysis see Pittelkow et al., 2021) of both emotion recognition and mental state reasoning. In contrast, other studies have demonstrated that socially anxious individuals are actually better in some social cognitive tasks such as recognizing others’ emotional states (Arrais et al., 2010; Joormann & Gotlib, 2006; Yuan et al., 2025); perhaps because socially anxious individuals develop a heightened sensitivity to the thoughts and feelings of others (Tibi-Elhanany & Shamay-Tsoory, 2011).

How do we reconcile these conflicting results? Previous studies have used a wide range of study designs to investigate the relationship between social anxiety and social cognition, it may be that these seemingly mixed findings reflect heterogeneity in the aspects of social anxiety under investigation, in the components of social cognition under investigation, or both, alongside differences in the types of measures employed to assess each construct. For example, levels of social anxiety are often evaluated based on ‘cognitive’ symptoms, such as feelings of social worry or nervousness about social settings and/or the fear of negative evaluation (Gao et al., 2025; Wong et al., 2016). On the other hand, social avoidance and social withdrawal are behavioral symptoms (Rodebaugh et al., 2025). Interestingly, although social avoidance is less common than general social worry or fear of negative evaluation (Chen et al., 2020), it may also be particularly detrimental to social cognitive ability, in that avoiding social interactions limits opportunities to learn about others’ mental states. Indeed, research has identified social avoidance as a significant predictor of future social anxiety, and some researchers consider it a central feature in the maintenance of SAD as well as the most likely to affect the processing of social cues (Hodson et al., 2008; Leigh & Clark, 2018; McClure & Nowicki, 2001; Rodebaugh et al., 2025; Schreiber et al., 2012; Stevens et al., 2014). Unfortunately, because previous research has not consistently differentiated between cognitive and behavioral components of social anxiety, it is difficult to understand their (potentially unique) relationships with social cognition.

Adding to the complexity, studies vary in which components of social cognition they measured and how they measured them. For instance, emotion recognition, argued to be a relatively basic component of social cognition (Malle, 2021), relies on perception and attention and involves recognizing observable facial and prosodic (i.e., vocal) cues (Brothers & Ring, 1992; Ochsner, 2008; Frith & Frith, 2006; Fitzpatrick et al., 2018; Singer, 2006). In comparison, social perspective taking or mental state understanding requires making inferences and reasoning about broader, more complex, and largely unobservable mental states, including people’s beliefs, desires, knowledge, intentions, and hidden or ‘masked’ emotions (Corrigan, 1997; Malle, 2021).

A third complication involves the considerable variability in the types of tasks used to assess each of these different components of social cognition. This variation may account for, or at least contribute to, the mixed results regarding their association with social anxiety. For example, some behavioural tasks measure accuracy in mental state reasoning or emotion recognition (e.g., False Belief tasks or the Reading the Mind in the Eyes task; Baron-Cohen et al., 2016; Bauminger-Zviely, 2013), while others are self-report measures asking informants to subjectively reflect on their own (or someone else’s) social cognitive capacity (e.g., Interpersonal Reactivity Index or the Four Item Mentalizing Index; Davis, 1980; Clutterbuck et al., 2021). Scores from these different tasks may have distinct relationships with social anxiety. For instance, some research suggests that socially anxious individuals tend to ‘overmentalize’ in certain social situations, attributing more thoughts or intentions to another person than appropriate. For example, attributing a person’s silence to anger or quiet resentment rather than fatigue. Therefore, socially anxious individuals may report thinking about mental states more frequently (Öztürk et al., 2022; Pittelkow et al., 2021; Tibi-Elhanany & Shamay-Tsoory, 2011), but nevertheless be less accurate when doing so (Ballespí et al., 2019, 2021; see also Hunsche, 2025, for an interesting exception in individuals with autism spectrum disorder with comorbid social anxiety). On the other hand, individuals who ‘undermentalize’ (i.e., fail to notice or reason about mental states) in social situations, perhaps due to greater anxiety and avoidance (Chevalier et al., 2023), may rate themselves lower on self-report measures of mental state reasoning but perform accurately in certain lab-based tasks where the social stimuli are more restricted, constrained, or controlled. Thus, self-report measures are especially prone to confounding an individual’s accuracy in mental state understanding and their propensity, or tendency, to notice and reason about mental states.

Importantly, research examining the relationship between social anxiety and social cognition in children, though less extensively studied, mirrors the mixed results observed with adults. Some studies indicate poorer performance in socially anxious children, such as mislabeling emotional cues (Battaglia et al., 2004; McClure & Nowicki, 2001; Simonian et al., 2001); other studies show no difference (Mobach et al., 2022; Pequet & Warnell, 2021) or indicate greater motivation and precision in processing social information (Ale et al., 2010; Pequet & Warnell, 2021). For example, Ale and colleagues (2010) reported that children experiencing social anxiety are more vigilant in social settings and more accurate in labelling facial expressions. In contrast, in one of the few studies examining prosodic emotion recognition, McClure and Nowicki (2001) found that children aged 8 to 10 with higher levels of social anxiety and social avoidance were more likely to mislabel fearful and sad emotional cues. Notably however, a prosodic emotion recognition task was used, requiring children to identify emotions using only vocal cues. Given the absence of facial expressions, this type of emotion recognition task was arguably harder and more likely to capture subtle difficulties in socially anxious and avoidant children. The idea that deficits in emotion recognition might be more pronounced when examining complex rather than basic emotions is consistent with previous studies examining children with ASD (Rodgers, 2021) as well as children with higher levels of social anxiety (O’Toole et al., 2013).

Mixed results as to the link between social anxiety and social perspective taking arise not only from the task and its level of complexity, but also the target whose perspective is being taken. For instance, work has shown that children with higher social anxiety are significantly more likely than low-anxious people to use mental state terms when describing a close friend (e.g., talking about their friend’s beliefs and preferences rather than their actions or physical appearance), as compared to an acquaintance, which could indicate a link between social anxiety and a heightened focus on the thoughts and feelings of close others (Pequet & Warnell, 2021). However, when asked to identify the emotions or mental states of a hypothetical stranger (e.g., using standard tasks like Happe’s Strange Stories and the Reading the Mind in the Eyes Task), social anxiety did not relate to children’s perspective-taking performance (Pequet & Warnell, 2021).

Altogether, the mixed results outlined above suggest that a complete understanding of how social cognitive processes are related to social anxiety will require a multifaceted and multi-measure approach to measuring both social cognition and social anxiety. We anticipate that the conflicting findings arises from the inconsistency in measures used as well as genuine differences in the constructs being measured. The primary goal of the current study was to gain a clearer and more comprehensive understanding of the nature of the relationships between different components of social cognition and different aspects of social anxiety. The present study recruited children aged 6 to 12 for numerous reasons. First, middle to late childhood is a common onset age for social anxiety, and a time period that coincides with entering school, when SA can particularly impact a child’s daily functioning (Halldorsson et al., 2017, 2023; Ollendick & Hirshfeld-Becker, 2002). Second, the elementary period is when children make considerable progress in mental state understanding alongside increasing consideration of others’ mental states (Fu et al., 2023). Thus, this window of intense social cognitive development, combined with the emergence of significant SA symptomatology, has the potential to offer unique insights into how aspects of social anxiety interact with various social cognitive processes. Finally, research suggests a slight advantage in social cognitive development in girls throughout childhood (Baron-Cohen, 2002), as well as higher rates of social anxiety in girls (Asher et al., 2017; Essau et al., 1999; Merikangas et al., 2010). Therefore, examining this period may provide insight into the role of gender in the relationships between social cognition and social anxiety.

The current study implemented a multi-measure cross-sectional design with 6-12-year-olds, as better elucidating the relationship between SA and social cognition within this developmental window requires examining whether and to what degree different aspects of social cognition and social anxiety are related in individual children. Although a cross-sectional approach does not allow us to pinpoint the causal nature of any observed relationships, it helps to pave the way for future work that does. Equally importantly this work will shed critical light on the direction and magnitude of the relationships between different components of social cognition and different aspects of social anxiety. More concretely, children’s social cognition was measured using two behavioral measures examining social cognitive accuracy: one focusing on mental state understanding and the other on emotion recognition. In addition, we utilized a self-report (subjective) measure of children’s tendency to engage in social perspective taking. Finally, children’s parents completed a measure of their perception of their child’s mental state understanding and mentalizing tendencies, as revealed by their day-to-day behaviors. Specifically, the behavioral measures were a) the two most advanced items from the widely used 5-item Theory of Mind (TOM) scale assessing mental state understanding (i.e., the unexpected contents and hidden emotions tasks; chosen due to the age range of our participants; Wellman & Liu, 2004), and b) prosodic emotion recognition, using a subset of the voice trials from the Cambridge Mindreading Face-Voice Battery for Children (CAM-C; Golan et al., 2015; note that this task is (relatively) difficult and so might be particularly likely to reveal anxiety-related deficits in emotion recognition in this age range (see O’Toole et al., 2013; Rodgers, 2021). Self- and parent-report measures included c) a child-friendly modification of an adult self-report measure of mentalizing tendencies, the Four Item Mentalizing Index, (FIMI; Clutterbuck et al., 2021), and d) a parent-report measure of children’s abilities and tendencies toward mental state understanding, the Children’s Social Understanding Scale (CSUS; Tahiroglu et al., 2014).

Importantly, this design allowed us to assess whether and how different aspects of children’s social cognitive abilities relate to one another. We investigated different components of children’s social cognitive abilities using four different measures of social cognitive processes, focusing on those most commonly used in research comparing social anxiety and social cognition (i.e., emotion recognition and mental state understanding; see Baez, 2023). No previous research has used these specific social cognition measures together. In the absence of such data, we hypothesized that the four measures of social cognitive ability would be modestly correlated because they are generally perceived to reflect related constructs (sometimes conceptualized as two parts of one general ability, or as two separate abilities under the broader umbrella of social cognition; see Mitchell & Phillips, 2015). However, we did not expect strong correlations given the different components of social cognition being addressed and the different measurement approaches.

Children’s social anxiety was also measured in multiple ways. First, fear of negative evaluation (a specific ‘cognitive’ component of social anxiety; see above) was measured via a self-report scale, the Brief Fear of Negative Evaluation, (the BFNE; Leary, 1983). Children also completed a second self-report anxiety measure, the Social Worries Anxiety Index for Young children (i.e., the SWAIY; Stuijfzand & Dodd, 2017) that was modified to yield distinct scores for both social worry (cognitive) and social avoidance (behavioral). Finally, parents completed a parent-report version of each of the child-report measures. Parent-reports were utilized so that we could examine the correlation between these and children’s self-reports; however, given that prior studies find that children over the age of five can reliably report on their emotional and social well-being (Caqueo-Urízar et al., 2022; Kiel et al., 2015; Varni et al., 2007), and that children are in a better position to report on their internal states than their parents, we preregistered that wherever our child- and parent-reports were redundant (i.e., social worry, social avoidance, fear of negative evaluation, and social perspective taking; FIMI), we would utilize child-reports in our analyses so long as they correlated moderately well with parent-reports (r ≅ .30).

Akin to our predictions regarding social cognition, we expected the correlations between different aspects of social anxiety to be moderately correlated. Social worry, fear of negative evaluation, and social avoidance are all considered key factors of social anxiety whose measures are oftentimes combined into one overall score of social anxiety (Fresco et al., 2001; Watson & Friend, 2012; Wong & Rapee, 2016). Nonetheless, we did not expect to find overly strong correlations as they are distinct aspects of social anxiety and may vary from person to person (e.g., Gao et al., 2025; Rodebaugh et al., 2025; Wong et al., 2016).

Most importantly for our primary research question, we hypothesized that the relationships between social cognition and social anxiety would be measure dependent. Specifically, we expected that cognitive symptoms of social anxiety would predict lower scores on 3 measures of cognition (i.e., prosodic emotion recognition, the 2-item TOM scale, and the CSUS) but higher scores in self-reported social perspective-taking (i.e., FIMI), reflecting a heightened sensitivity to others’ thoughts and feelings as observed in prior research (Öztürk et al., 2022; Tibi-Elhanany & Shamay-Tsoory, 2011). We also predicted that the behavioral aspect of social avoidance would account for additional variability in each case, consistent with the theory that social avoidance plays an especially key role in the relationship between social anxiety and social cognition (Ghandchi et al., 2023; McClure & Nowicki, 2001; Stevens et al., 2014).

Finally, we controlled for age and gender in our analyses. Given there is substantial social cognitive development (Fu et al., 2023) as well as greater social awareness and peer evaluation within this age range (Poole et al., 2022), we predicted that age would positively correlate with children’s overall accuracy in social cognitive tasks and levels of social anxiety. In addition, we expected to find a small advantage for girls in children’s social cognitive performance as well as heightened social anxiety in girls, irrespective of age and consistent with the prior literature mentioned above (Asher et al., 2017; Baron-Cohen, 2002; Essau et al., 1999; Merikangas et al., 2010).

2. Materials and Methods

2.1. Participants

A total of 78 children (M = 8.15 years, SD = 1.61 years; 64.1% female, 2.6% missing gender) and their parents participated in this study. Although 97 participants were recruited, 19 were excluded for failure to complete at least 80% of items on each measure. Inclusion criteria included being, or being the primary caregiver of, a child aged 6 – 12 as well as being sufficiently fluent in English (defined as English being the primary language used at least 50% of the time). Our sample would be classified as an ‘unselected’ sample, rather than clinical or non-clinical, as participants were neither selected nor disqualified based one their psychopathology or clinical diagnosis.

Parent participants were 80.8% female (5.1% opted not to respond), and the majority identified as East Asian (48.7%) or European (29.5%). The remaining participants reported being East Indian (5.1%), Middle Eastern (2.6%), Latin/Central/South American (1.3%), mixed (2.6%), or other (3.8%); 6.4% opted not to respond. The sample was generally highly educated (80.7% of parents completed a bachelor’s degree or higher), and of mid to high economic status (e.g., 52.6% reported a family income of 100,000 per year or greater).

2.2. Measures

2.2.1. Social Cognition Measures

Prosodic Emotion Recognition: Children’s accuracy in prosodic emotion recognition (i.e., emotions conveyed through tone of voice) was measured using items from the Cambridge Mindreading Face-Voice Battery for Children (CAM-C; Golan et al., 2015). The CAM-C is a child-friendly adaptation of an adult battery designed to assess recognition of both simple and complex emotions in faces and voices. The child version uses static facial images (for ‘face trials’) and dynamic voice recordings (for ‘voice trials’) and has been validated for use with children aged 6 - 12, with a test-retest reliability for faces and voices ranging from 0.76 to 0.81 and an internal consistency of 0.74 for complex voices (Rodgers, 2021).

As mentioned above, previous studies have also shown that deficits in emotion recognition are more pronounced in studies using complex emotions (Rodgers, 2021) and more strongly related to social anxiety (O’Toole et al., 2013). Accordingly, the current study used a total of 10 items from the CAM-C that measured recognition of complex emotions conveyed solely through tone of voice, excluding items that were limited to basic emotions. We selected five complex vocal emotions that showed good reliability in previous studies in differentiating between typically developing children and those who struggle with social information processing, including unfriendly, amused, nervous, bothered, and jealous (Rodgers, 2021). Finally, for brevity we selected two (of three) items for each complex emotion based on item-total correlations in past work (Rodgers, 2021). In the current sample using only these 10 items, Cronbach’s alpha was lower than previous work at .51.

The Theory of Mind Scale (TOM scale): Children’s accuracy in understanding mental states was evaluated using 2 items from the 5-item version of the Theory of Mind scale (Wellman and Liu, 2004). In its original format, the TOM scale utilizes five tasks to evaluate a child’s understanding of others’ mental states, including desires, beliefs, knowledge, and emotions, particularly when these differ from the child’s own perspective or from objective reality. Wellman & Liu’s analyses revealed a clear developmental progression in children’s mental state understanding, such that items targeting simpler mental states showed ceiling effects past the age of six (see also Peterson et al., 2012). Therefore, the current study selected only the two most difficult tasks targeting complex mental states (i.e., Contents False Belief and Hidden Emotions). This measure was scored proportionally, meaning that participants’ scores reflected the proportion of theory of mind tasks they answered correctly (i.e., 0%, 50%, or 100%). Consistent with the scoring approach used by the authors in the original validation study, participants were required to pass a ‘memory’ check in the Contents False Belief task ensuring they paid attention and remembered the critical details of the story to have the trial included.

Children’s Social Understanding Scale (CSUS): The CSUS is a parent-report measure designed to evaluate a child’s understanding of others’ mental states, such as their beliefs, knowledge, desires, perceptions, and emotions by having parents report on a variety of behaviors they can observe in their child day-to-day.

The current study used the short-form of the CSUS, which has been validated in children ages 2-7 and demonstrated high test-retest reliability (r(29) = .88) and internal consistency (α = .89) as well as positive correlations with two False Belief measures (including the Contents False Belief task used in the current study and one other) adapted from Wellman and Liu’s TOM scale (Tahiroglu et al., 2014). The CSUS-Short Form has been used with children up to age 12 and reliably predicts multiple facets of social emotional health (Haddock & Birch, 2024). Cronbach’s alpha in our sample was high (α = .81).

Four Item Mentalizing Index (FIMI): The Four-Item-Mentalizing-Index (FIMI) is a 4-item self-report measure of one’s tendency to engage in social perspective taking (i.e., discerning the mental states of others within dynamic social contexts). It was originally validated in typically developing American adults (Clutterbuck, 2021) but has also shown cross-cultural reliability (Fekih-Romdhane et al., 2024). Previous analyses reported acceptable reliability for the FIMI, with a coefficient omega (ω) between 0.70 and 0.75 (Clutterbuck, 2021; Montgomery et al., 2024). As no validated self-reported measure of children’s perspective taking exists in the literature for our target age range, we modified the adult self-report measure for use as both a parent- and child-report measure. For example, we created a child-report FIMI by substituting more child-friendly wording (e.g., "I can usually understand another person's viewpoint, even if it differs from my own" was changed to “I can understand what others think even if it’s not the same as what I think”). In our sample, the coefficient omega for the child-report was moderately low at 0.53, indicating limited reliability, while the parent-report showed high reliability at 0.82.

2.2.2. Social Anxiety Measures

Social Worry (SWAIY-modified): Social worry, a cognitive component of social anxiety, was measured using a modified version of the Social Worries Anxiety Index for Young Children (SWAIY), a 10-item parent-report measure modified from the Social Worries Questionnaire (SWQ; Spence, 1995) to assess anxiety in children aged 4-8 (Stuijfzand & Dodd, 2017). To ensure that our measure was appropriate for the full age range of our sample, we used all 10 items from the SWAIY, plus two additional items from the SWQ appropriate for older children (i.e., “Gets worried about eating in public” and “Gets worried about using public toilets”). Finally, the wording was changed to reflect a child-report rather than a parent-report. This left us with a final set of 12 items (SWAIY-modified; see supplemental material - Section S1), with answers ranging from 0 (Not at all like me) to 2 (A lot like me). The N/A option from the original survey was removed to ensure the highest amount of usable data.

Social Avoidance (SWAIY-avoidance): Social avoidance scores, a behavior component of social anxiety, were measured by removing the word avoid in each item of the original SWAIY and adapting it into a new ‘yes’ or ‘no’ follow-up question. For example, the item “Avoids or gets worried about going to parties or play dates” became “I get worried about going to parties or play dates” to create a two-part question: the first assessing the child’s level of social worry about parties or playdates and the second asking a yes or no question about whether that social worry would lead them to avoid the situation altogether, (e.g. ‘avoid it all together?’, SWAIY-avoidance; see supplemental material – Section S1). This allowed us to examine levels of social avoidance separately from social worry, to determine whether social avoidance impacts social cognitive processes independently from worry alone.

The 10-item SWAIY has shown high internal consistency (> .80), test-retest reliability (r = .87), and convergent validity (r > .50) (Spence, 1995). In our sample, Cronbach’s alpha was .79 for the parent report and .77 for the child report for the 12 items measuring social worry. For the follow-up question on social avoidance (SWAIY-avoidance), Cronbach’s alpha was .78 for the parent report and .47 for the child report. To alleviate concerns regarding shared method variance and multicollinearity, we ensured the correlation between these two constructs did not exceed the recommended threshold of r = .80 (Mason & Perreault, 1991; de Winter, 2025). The correlation between these two measures was r = .239.

Fear of Negative Evaluation (FNE): Levels of social anxiety were also assessed through parental reports of children’s fear of being negatively evaluated in social settings, using the Brief Fear of Negative Evaluation scale (BFNE; Leary, 1983). A child-report version was created by simplifying the language to accommodate participants’ age and comprehension level (e.g., “I am frequently afraid of other people noticing my shortcomings” became “I am afraid that other people will notice my mistakes”).

In previous work, the BFNE scale’s test-retest reliability (ICC = .71) and Cronbach’s alpha (α = .82) for socially phobic students were both high (Tavoli et al., 2009). In our sample, Cronbach’s alpha was .93 for the parent report, and .88 for the child report.

Table 1. Description of measures and examples items.

Measure	Description & Example Items
Cambridge Mindreading Face-Voice Battery for Children (CAM-C) (Golan et al., 2015)	10 item battery measuring children’s accuracy in prosodic emotion recognition. Respondents listen to a voice recording and are presented with four options, scoring 1 point for each correct response. Minimum score = 0 and Maximum score = 10. Sums of scores used in analysis Voice recording: “I wish I didn’t have to do it.” 1. Disbelieving; 2. Bothered (correct); 3. Liked; 4. Impressed
Theory of Mind Scale (TOM scale; Wellman and Liu, 2004) *Child behavioural task	Two task battery measuring children’s mental state understanding, including False beliefs and Hidden emotions. Minimum score = 0 and Maximum score = 2. Higher scores indicated greater mental state understanding, and sums of scores were used in analysis. False Belief “Peter has never ever seen inside this Band-Aid box [that child knows contains a car]. What does Peter think is in the box?” 1 – Band-Aids (correct); 2 – Toy Car Hidden Emotions “How did Matt really feel, when everyone laughed?” 1 – Happy; 2 – Okay; 3 – Sad “How did Matt try to look on his face?” 1 – Happy; 2 – Okay; 3 – Sad
Children’s Social Understanding Scale (CSUS; Tahiroglu et al., 2014) *Parent report	18 item parent-report measure assessing their child’s understanding of mental states with answers ranging from 1 - 4. Minimum score = 4 and Maximum score = 72. Higher scores indicate greater mental state understanding, and sums of scores were used in analysis Example item: My child talks about differences in what people like or want (e.g., “you like coffee, but I like juice”)
Four Item Mentalizing Index (FIMI; Clutterbuck et al., 2021)	4 item child- and parent-report measures assessing their tendency to engage in social perspective taking with answers ranging from 1 - 4. Minimum score = 4 and Maximum score = 16. Higher scores indicate greater fear of negative evaluation, and sums of scores were used in analysis. Example item: I/My child sometimes try/tries to understand my/their friends better by imagining what they are thinking or feeling.
Social Worries Anxiety Index for Young Children – Social Worry and Social Avoidance (SWAIY; Spence, 1995)	12 item child- and parent-report measure assessing levels of social worry and social avoidance. Social worry answers ranged from 0 - 2. Minimum score = 0 and Maximum score = 24. Social avoidance answers were coded as a binary ‘yes’ or ‘no’ response. Minimum score = 0 and Maximum score = 24. Higher scores indicate greater levels of social worry and avoidance, and sums of scores were used in analysis. Example items: I/My child get/s worried about going to parties or playdates.
Brief Fear of Negative Evaluation (BFNE; Leary, 1983)	8 item child-report measure assessing their fear of negative evaluation with answers ranging from 1 - 5. Minimum score = 8 and Maximum score = 40. Higher scores indicated increased fear of negative evaluation, and sums of scores were used in analysis. Example item: I/My child am/is afraid that other people will notice my/their mistakes.

Table 1. Description of measures and examples items.

Measure	Description & Example Items
Cambridge Mindreading Face-Voice Battery for Children (CAM-C) (Golan et al., 2015)	10 item battery measuring children’s accuracy in prosodic emotion recognition. Respondents listen to a voice recording and are presented with four options, scoring 1 point for each correct response. Minimum score = 0 and Maximum score = 10. Sums of scores used in analysis Voice recording: “I wish I didn’t have to do it.” 1. Disbelieving; 2. Bothered (correct); 3. Liked; 4. Impressed
Theory of Mind Scale (TOM scale; Wellman and Liu, 2004) *Child behavioural task	Two task battery measuring children’s mental state understanding, including False beliefs and Hidden emotions. Minimum score = 0 and Maximum score = 2. Higher scores indicated greater mental state understanding, and sums of scores were used in analysis. False Belief “Peter has never ever seen inside this Band-Aid box [that child knows contains a car]. What does Peter think is in the box?” 1 – Band-Aids (correct); 2 – Toy Car Hidden Emotions “How did Matt really feel, when everyone laughed?” 1 – Happy; 2 – Okay; 3 – Sad “How did Matt try to look on his face?” 1 – Happy; 2 – Okay; 3 – Sad
Children’s Social Understanding Scale (CSUS; Tahiroglu et al., 2014) *Parent report	18 item parent-report measure assessing their child’s understanding of mental states with answers ranging from 1 - 4. Minimum score = 4 and Maximum score = 72. Higher scores indicate greater mental state understanding, and sums of scores were used in analysis Example item: My child talks about differences in what people like or want (e.g., “you like coffee, but I like juice”)
Four Item Mentalizing Index (FIMI; Clutterbuck et al., 2021)	4 item child- and parent-report measures assessing their tendency to engage in social perspective taking with answers ranging from 1 - 4. Minimum score = 4 and Maximum score = 16. Higher scores indicate greater fear of negative evaluation, and sums of scores were used in analysis. Example item: I/My child sometimes try/tries to understand my/their friends better by imagining what they are thinking or feeling.
Social Worries Anxiety Index for Young Children – Social Worry and Social Avoidance (SWAIY; Spence, 1995)	12 item child- and parent-report measure assessing levels of social worry and social avoidance. Social worry answers ranged from 0 - 2. Minimum score = 0 and Maximum score = 24. Social avoidance answers were coded as a binary ‘yes’ or ‘no’ response. Minimum score = 0 and Maximum score = 24. Higher scores indicate greater levels of social worry and avoidance, and sums of scores were used in analysis. Example items: I/My child get/s worried about going to parties or playdates.
Brief Fear of Negative Evaluation (BFNE; Leary, 1983)	8 item child-report measure assessing their fear of negative evaluation with answers ranging from 1 - 5. Minimum score = 8 and Maximum score = 40. Higher scores indicated increased fear of negative evaluation, and sums of scores were used in analysis. Example item: I/My child am/is afraid that other people will notice my/their mistakes.

2.3. Procedure

Participants were recruited online through social media (Facebook and Instagram), posting the study’s promotional material, and a local database of interested families from [location blinded for review]. All communication with parents before the study session was done via email. Once participation was confirmed and a child’s Zoom session was scheduled, parents received online consent forms for themselves and their child, as well as the parent-report survey to complete prior to their child’s participation. Child participation was done over Zoom with a trained female researcher in a 30-45 minute Zoom session. This study was conducted in the same session as a separate study (see OSF for separate study; 10.17605/OSF.IO/AMGX6); therefore, brevity in measures was an important consideration. Consent to participate and be recorded via Zoom was re-confirmed before the researcher began the study.

Researchers were provided with a script to maintain consistency between sessions. Children were asked to answer questions about each scene or indicate on a Likert-scale how much certain statements sounded like them (see Appendix B). Items and scenarios for each scale were read aloud. Participants were also told they could stop their session at any time and skip any questions that they did not want to answer. At the end of the session, parents and their children were given the opportunity to provide feedback on the study and to choose between receiving a $10.00 Amazon gift card or entering a draw (1 in 25 chance) to win a $50.00 gift card as compensation.

2.4. Ethics Approval

This study was approved by the UBC Behavioural Research Ethics Board.

2.5. Power Analysis

An a priori power analysis was conducted using GPower to determine the required sample size for detecting a medium effect in a hierarchical regression with three predictors and two control variables (two-tailed, α = .05, power = .80). Our minimum target sample size was 77 participants to detect a medium effect.

2.6. Analytic Plan

Our analytic plan was preregistered on the Open Science Foundation (see https://osf.io/muxvn/).

3. Results

3.1. Preliminary Analyses

Participant responses were summed for each of our measures, with higher scores indicating greater levels of social anxiety and better performance on social cognitive measures. One exception was the TOM scale, for which total scores were calculated using the proportion of tasks that participants passed both the attention check and the mental state understanding probe. We calculated means, standard deviations, and maximum scores for each variable (see Table 2).

3.1.1. Age

We computed zero-order correlations between age and gender and our measures of social cognition and social anxiety to test the expected positive relationships with girls and older participants (see Table 2). As predicted, age was significantly correlated with higher accuracy on the two behavioral social cognition tasks (prosodic emotion recognition, r(76) = .379, p = <.001; the TOM scale, r(76) = .252, p = .026). However, against expectations age did not correlate with either child- or parent-reported social cognitive ability (i.e., FIMI, CSUS).

Age was also positively correlated with most of our cognitive measures of social anxiety, though this correlation only reached significance in the child-report measure of the fear of negative evaluation, r(76) = .143, p = .032, and the parent-report measure of social worry, r(72) = .256, p = .022. No age differences emerged in the behavioral measure of social avoidance.

3.1.2. Gender

In contrast to the prediction that girls would have somewhat higher scores than boys on social cognition and anxiety, only one significant relationship emerged with gender. The parent-report measure of children’s tendency to engage in social perspective taking (FIMI) was positively correlated with gender, indicating that parents felt their daughters were better able to take others’ perspectives than their sons, r(70) = .238, p = .044. Critically, however, children themselves did not show or report any gender differences in either mentalizing or social anxiety.

3.1.3. Child- and Parent-Report Correlations

Next, we compared child- and parent-reports for our measures of social anxiety and found that all measures including social worry, r(71) = .259, p = .027, social avoidance, r(53) = .302, p = .025, and the fear of negative evaluation, r(72) = .281, p = .015, were significantly and positively correlated. We elected to use the child reports in our main analyses, as the correlations between child- and parent-reports either met or were very close to our predetermined threshold (r ≥ .30). These modest correlations between children and parent reports are consistent with the broader literature on children’s mental health (e.g., Caqueo-Urízar et al., 2022; De Los Reyes & Kazdin, 2005). For comprehensiveness and comparisons, we also conducted the main analyses using the parent-report data (see supplemental material – Section S2). Results from these supplemental analyses generally followed the same pattern with exceptions noted within.

We also compared the child- and parent-report measures of social perspective taking (FIMI) and found no relationship, r(70) = .03, p = .83. Nonetheless, we again chose to use the child-report in our main analyses, both to maintain consistency and given the internalized and subjective nature of the constructs of interest.

Table 3. Zero-order correlations for age in months, gender, social anxiety and social cognitive measures.

Variable		1	2	3	4	5	6	7	8	9
1. Age months		-
2. Gender	r	-.077
	p	.510	-
	N	76
3. CAM-C	r	.379**	.098
	p	<.001	.402	-
	N	78	76
4. TOM	r	.252*	-.181	.133
	p	.026	.118	.245	-
	N	78	76	78
5. CSUS (PR)	r	-.136	-.052	-.045	.007
	p	.247	.661	.704	.955	-
	N	74	74	74	74
6. FIMI (CR)	r	.129	-.01	.083	.045	.032
	p	.260	.930	.472	.696	-.784	-
	N	78	72	78	78	74
7. SWAIY-M (CR)	r	-.017	.064	-.071	-.050	-.041	-.062
	p	.886	.587	.540	.668	.731	.593	-
	N	77	75	77	77	73	77
8. SWAIY-A (CR)	r	.119	.026	-.084	.050	.010	-.218	.450**
	p	.308	.825	.472	.666	.934	.059	<.001	-
	N	76	74	76	76	72	76	76
9. BFNE (CR)	r	.243*	.081	.140	.191	-.260*	-.038	.564**	.414**
	p	.032	.486	.222	.093	.025	.740	<.001	<.001	-
	N	78	76	78	78	74	78	77	76

*p < .05, **p < .01 .

Table 3. Zero-order correlations for age in months, gender, social anxiety and social cognitive measures.

Variable		1	2	3	4	5	6	7	8	9
1. Age months		-
2. Gender	r	-.077
	p	.510	-
	N	76
3. CAM-C	r	.379**	.098
	p	<.001	.402	-
	N	78	76
4. TOM	r	.252*	-.181	.133
	p	.026	.118	.245	-
	N	78	76	78
5. CSUS (PR)	r	-.136	-.052	-.045	.007
	p	.247	.661	.704	.955	-
	N	74	74	74	74
6. FIMI (CR)	r	.129	-.01	.083	.045	.032
	p	.260	.930	.472	.696	-.784	-
	N	78	72	78	78	74
7. SWAIY-M (CR)	r	-.017	.064	-.071	-.050	-.041	-.062
	p	.886	.587	.540	.668	.731	.593	-
	N	77	75	77	77	73	77
8. SWAIY-A (CR)	r	.119	.026	-.084	.050	.010	-.218	.450**
	p	.308	.825	.472	.666	.934	.059	<.001	-
	N	76	74	76	76	72	76	76
9. BFNE (CR)	r	.243*	.081	.140	.191	-.260*	-.038	.564**	.414**
	p	.032	.486	.222	.093	.025	.740	<.001	<.001	-
	N	78	76	78	78	74	78	77	76

*p < .05, **p < .01 .

3.2. Main Analyses

3.2.1. Correlations Between Social Cognitive Processes

We computed partial correlations between our four social cognitive variables to examine the degree of association between prosodic emotion recognition, mental state understanding, and self-reported social perspective taking. We found little to no overlap between these measures, after controlling for age in months (see Table 4).

3.2.2. Mental State Understanding (CSUS) Predicted by Social Anxiety

Hierarchical regression analyses predicting mental state understanding as assessed by the CSUS parent-report (Table 5) revealed that age and gender accounted for approximately 2.2% of its variability, F(2, 70) = .80, p = .45. Introducing levels of social anxiety – assessed through levels of social worry and fear of negative evaluation – accounted for an additional 6.7% of the variation in children’s mental state understanding (CSUS), and this change in R² was not significant, F(1, 66) = .588, p = .446. Adding levels of social avoidance accounted for only an additional 0.8% of children’s mental state understanding (CSUS), and this change in R² was not significant, F(1, 66) = .588, p = .446.

Furthermore, fear of negative evaluation (BFNE) emerged as a unique and significant predictor of mental state understanding as assessed by the CSUS parent-report measure. A higher fear of negative evaluation was associated with lower ratings of parent-reported mental state understanding, t(68) = -2.05, p = .044.

3.2.3. Mental State Understanding (2-Item TOM Scale) Predicted by Social Anxiety

Hierarchical regression analyses predicting mental state understanding as assessed by the TOM scale (Table 6) revealed that age and gender accounted for approximately 9.7% of its variability, F(2, 72) = 3.85, p = .026. Introducing levels of social anxiety – assessed through levels of child-reported social worry and fear of negative evaluation – accounted for only an additional 5.0% of the variation in children’s mental state understanding (TOM scale), and this change in R² was not significant, F(2, 70) = 2.07, p = .134. Adding levels of social avoidance accounted for zero additional variation in children’s mental state understanding (TOM scale).

Furthermore, in Model 3, child-reported fear of negative evaluation (BFNE) again emerged as a unique and significant predictor of mental state understanding as assessed by the TOM scale, with a higher fear of negative evaluation associated with greater accuracy in mental state understanding, t(68) = 2.17, p = .033. Notably, this relationship is in the opposite direction of that between the fear of negative evaluation and the CSUS as reported above.

3.2.4. Social Perspective Taking (FIMI) Predicted by Social Anxiety

Hierarchical regression analyses predicting child-reported social perspective taking (Table 7) revealed that age and gender accounted for 1.4% of its variability, F(2, 72) = .52, p = .60. Introducing levels of social anxiety – assessed through levels of social worry and fear of negative evaluation – accounted for an additional 2.7% of the variation in children’s self-reported perspective taking (FIMI), and this change in R² was not significant, F(2, 70) = .978, p = .381. Adding levels of child-reported social avoidance accounted for an additional 5.5% of the variation in children’s self-reported perspective taking (FIMI), and this change in R² was significant, F(1, 68) = 4.18, p = .045. That is, increased self-reported avoidance behavior predicted poorer self-reported perspective taking.

3.2.5. Prosodic Emotion Recognition (CAM-C) Predicted by Social Anxiety

Hierarchical regression analyses predicting prosodic emotion recognition (Table 8) revealed that age and gender accounted for 15.8% of the variability in emotion recognition, F(2, 72) = 6.73, p = .002. Introducing levels of social anxiety – assessed through the cognitive measures of social worry and fear of negative evaluation – accounted for only an additional 1.5% of the variation in children’s prosodic emotion recognition, and this change in R² was not significant, F(2, 70) = .621, p = .540. Further, adding levels of social avoidance accounted for an additional 1.7% of the variation in children’s prosodic emotion recognition, and this change in R² was also not significant, F(1, 68) = 1.41, p = .239.

In summary, these analyses reveal that, when combined, our cognitive measures of social anxiety – social worry and fear of negative evaluation – did not account for a significant amount of variation in our social cognitive measures above and beyond age and gender. However, our measure of fear of negative evaluation (BFNE) proved to be a unique and significant predictor of certain components of social cognition. Greater child-reported fear of negative evaluation was reported as a significant predictor of poorer mental state understanding, as assessed by the CSUS parent-report. Counterintuitively, child-reported fear of negative evaluation was also a significant predictor of improved actual performance in mental state understanding tasks, as assessed by the 2-item TOM scale. Adding levels of social avoidance did not account for a significant amount of variation in prosodic emotion recognition (CAM-C) or mental state understanding. However, social avoidance emerged as a significant predictor of additional variance in children’s self-reported social perspective taking (FIMI), with higher social avoidance associated with lower self-reported social perspective taking. See Table 9 for an overview of the main results from the regression analyses.

3.3. Exploratory Analyses

We controlled for age in all of the regression analyses. However, we were also interested in understanding these relationships across development. Therefore, we conducted exploratory analyses using the same hierarchical regressions among younger and older subgroups of our original sample (using the median split technique, median age = 105 months), acknowledging this would limit our power to detect small or medium sized effects. The child-report BFNE remained a unique predictor for mental state understanding in the younger group, with greater fears of negative evaluation predicting greater accuracy on the 2-item TOM scale, t(32) = 2.03, p = .051, but lower ratings on the parent-report of children’s mental state understanding CSUS, t(30) = -2.83, p = .008. This pattern of results was not observed in the older group. In comparison, social avoidance failed to reach significance as a predictor of self-reported perspective-taking (FIMI) in either age group, but showed a stronger relationship in the older group, (1, 30) = 3.03, p = .092 (see supplemental material – Section S3).

4. Discussion

The current study was designed to gain a clearer and more comprehensive understanding of the nature of the relationships between different components of social cognition and social anxiety in children. With this goal in mind, we included multiple measures of social anxiety and social cognition that address different aspects of each, in an attempt to capture both accuracy and subjective assessments of social cognitive abilities.

We hypothesized that the four measures of social cognitive ability would be correlated, but not strongly, given the aforementioned distinctions in measurement type and components of social cognition, including prosodic emotion recognition, mental state understanding, and self-reported perspective taking. Similarly, we expected our three measures of social anxiety to show moderate correlations. Consistent with these expectations, we also hypothesized that the relationships between social cognition and social anxiety would be measure dependent. Specifically, we expected that cognitive symptoms of social anxiety would predict lower scores on 3 measures of social cognition (i.e., prosodic emotion recognition, the 2-item TOM scale, and the CSUS) but higher scores in self-reported social perspective-taking (i.e., FIMI), reflecting a heightened sensitivity to others’ thoughts and feelings as observed in prior research (Öztürk et al., 2022; Tibi-Elhanany & Shamay-Tsoory, 2011). We also predicted that the behavioral aspect of social avoidance would account for additional variability in each case, consistent with the theory that social avoidance plays an especially impactful role in the relationship between social anxiety and social cognition (Ghandchi et al., 2023; McClure & Nowicki, 2001; Stevens et al., 2014).

Our findings support the overarching hypotheses that the nature, and even direction, of the relationships between social cognition and anxiety, depend on the measurement type and cognitive component under investigation. However, results were only partially consistent with our measure-specific hypotheses. First, preliminary analyses showed that aspects of social anxiety were moderately correlated, as expected. Surprisingly, however, analyses revealed weak or non-existent correlations among the four measures of social cognition, which failed to support our prediction of weak to moderate correlation. This minimal overlap among the social cognition measures underscores the idea that the concept of social cognition is incredibly broad and cannot be adequately captured using one measure. As a result, it is unsurprising that associations between social cognition and other constructs (e.g., social anxiety) were directionally varied depending on the specific component or type of measure under investigation.

Age was significantly and positively correlated with greater accuracy in prosodic emotion recognition (CAM-C) and the 2-item TOM scale. This aligns with prior research examining the development of social cognition and its substantial improvement through early childhood (Wellman, 2014). These associations lend credence to the validity of these measures by indicating that the 2-item TOM scale and prosodic emotion recognition tasks do, in fact, capture the expected age-related changes in social cognitive accuracy. To our knowledge, no one has previously used a reduced-item version of the CAM-C. Thus, an additional contribution of this work is the utility of this abbreviated measure to capture age-related changes in prosodic emotion detection in a young sample, while minimizing the fatigue and attentional concerns posed by the full 54 item version (Golan et al., 2015). Age was also positively and significantly correlated with child-reported fear of negative evaluation (BFNE), consistent with the substantial increase in social awareness shown to occur through middle to late childhood (Crone & Dahl, 2012; Rudolph, 2021), supporting the sensitivity and accuracy of our fear of negative evaluation measure.

The findings from our regression analyses lend critical support for the overarching hypotheses that the relationships between social cognition and social anxiety are measurement dependent. When combined, our cognitive measures of social anxiety, social worry and fear of negative evaluation did not significantly predict variation in any of the social cognitive measures. However, on its own, fear of negative evaluation emerged as a unique and significant predictor. Specifically, children’s fears of negative evaluation predicted lower scores on the parent-report measure of mental state understanding (i.e., the CSUS) but greater accuracy in a behavioural task of mental state understanding (i.e., the 2-item TOM scale). Interestingly, adding social avoidance to these models did not explain additional variability above and beyond age, gender, and our cognitive measures of social anxiety. However, when predicting a child’s self-reported social perspective taking (i.e., the FIMI), social avoidance accounted for additional variation above and beyond age, gender, and our cognitive measures of social anxiety. In this case, higher levels of social avoidance predicted somewhat lower self-reported social perspective taking. That is, children who were more likely to report that they avoided specific social situations altogether also tended to report that they were not very good at perspective taking (e.g., do not ‘find it easy to guess what someone else is feeling or thinking’ or ‘don’t often imagine other perspectives to better understand them’). Conversely, our measure of prosodic emotion recognition was not significantly predicted by any of our measured aspects of social anxiety after controlling for age and gender.

The finding that fear of negative evaluation plays an important role in the relationship between social anxiety and social cognition is consistent with previous research highlighting fear of negative evaluation as a key cognitive feature of social anxiety. It is also consistent with work showing its important role in both maintaining Social Anxiety Disorder and mediating the relationship between social anxiety and other cognitive outcomes (Briscoe et al., 2024; Fredrick et al., 2020; Kocovski & Endler, 2000; Morrison et al., 2024; Sigurvinsdottir et al., 2021). In the current work, we found significant relationships between fear of negative evaluation and mental state understanding, but in opposite directions, depending on the measure (i.e., the TOM scale versus the parent report CSUS). This differential pattern presents an interesting contradiction (or an apparent one). It may be that fear of negative evaluation increases a child’s vigilance or effort, and resulting accuracy, in behavioural measures of social cognition, aligning with research suggesting that social anxiety increases one’s awareness and sensitivity to social cues (Ale et al., 2010, Arrais et al., 2010; Joormann & Gotlib, 2006; Pequet & Warnell, 2021; Tibi-Elhanany & Shamay-Tsoory, 2011; Yuan et al., 2025). At the same time, fear of negative evaluation may decrease children’s motivation, or propensity, to engage in mental state understanding (and frequency of doing so) in their daily lives, for fear of what others may be thinking. In other words, while socially anxious children may have the tools to accurately engage in mental state understanding when asked directly, they may be less inclined to use them in natural settings, thus engaging in mental state understanding less frequently. This idea aligns with our findings that children with higher levels of social anxiety performed with greater accuracy on the TOM scale (behavioural task) but received lower ratings on the CSUS (parent-report assessing mental state understanding). Additionally, these findings align with studies suggesting an important distinction between state- and trait-like mentalizing, wherein socially anxious individuals struggle to mentalize (i.e., reason about the mental state of others) in social settings which they find stressful (i.e., state-like) (Chevalier et al., 2023). Yet, they may perform accurately in certain environments such as a research lab where the social stimuli are more constrained, controlled, or simplistic or where they are interacting with a single attentive and encouraging researcher. Finally, it may be important to note that parents are limited in their objectivity and observations of children’s engagement in mental states reasoning, particularly when they are in school and interacting with their peers. Thus, parent-reports may not fully capture children’s understanding, and may therefore differ from measures such as structured observation by a trained researcher.

It is also worth acknowledging that these measures of mental state understanding (i.e., TOM scale and parent-report CSUS) differ not only in their objective versus subjective nature, but also in terms of breadth. That is, the behavioral measure of mental state accuracy specifically addressed children’s understanding of only two mental states: false beliefs and hidden emotions. In contrast, the other measure (i.e., CSUS) asked parents to reflect on their child’s observable engagement with a more diverse array of mental state reasoning that included beliefs and emotions as well as intentions, desires, preferences, and knowledge states. Moreover, the parent measure assesses children’s behaviour over the course of recent weeks and across multiple social situations. As such, it is unclear whether these conflicting relationships emerged because of differences in measurement breadth, objectivity, or both. To help clarify the specific relationships between fear of negative evaluation and social cognitive processes, future work should consider replicating and expanding upon this work using a broader accuracy measure (e.g. Happe’s Strange Stories; Happé, 1995) and/or another subjective measure of mental state understanding (e.g., the Interpersonal Reactivity Index; Davis, 1980).

Overall, these findings confirmed the hypothesis that measurement type impacts the observed relationship between social cognition and social anxiety. While social avoidance showed no relationship with either behavioural or parent-reported measures of social cognition, it was the strongest predictor of children’s self-reported social perspective taking, with greater social avoidance predicting less social perspective taking as measured by the FIMI. Although the FIMI is often framed as a measure of mentalizing ability (Bertrams et al., 2024; Clutterbuck, 2021; Fekih-Romdhane et al., 2024), there has been relatively little discussion of its limitations as a subjective self-report tool. It seems likely that the FIMI captures an individual’s perception of how often or easily they engage in mental state reasoning in their daily life which may or may not reflect their actual accuracy or ability, especially given the wording of the items (e.g., “I sometimes try to understand my friends better by imagining how things look from their perspective”). This could explain why the FIMI showed weak correlations with our other social cognitive measures and why its only relationship was with the child’s self-reported levels of social avoidance; that is, their willingness to participate in social events. This makes intuitive sense, in that children who don’t find perspective-taking easy, or do not think they are very good at it, would be more likely to avoid social situations. Of course, it is equally plausible that avoiding social situations is detrimental to children’s developing social perspective taking ability (and/or their perceptions of their social perspective taking ability) (Ghandchi et al., 2023; McClure & Nowicki, 2001; Rodebaugh et al., 2025). We suspect this finding reflects a bidirectional causal relationship, but understanding the nature and development of this relationship will necessitate further research (e.g. longitudinal designs).

The conflicting relationships in our results, and in the broader literature, may also be partially explained by the use of different informant types. We found only moderate or ‘close-to-moderate’ correlations between child- and parent-report on the anxiety measures. That these correlations were not larger, coupled with a negligible association between parent- and child-reports of perspective-taking (FIMI), raises the question of the source of these discrepancies. Do the lack of correlations reflect a child’s limitations in reporting on their own thoughts and behaviours, the parent’s limitations in accurately judging their child’s social cognitive skills and feelings of anxiety, or both?

On the one hand, parents rated girls as having stronger mental state understanding than boys, yet no relationship emerged between gender and the children’s performance on either of our two behavioral measures of mental state understanding. It is thus possible that parents are simply biased in believing that their girls are more sensitive to the emotions and perspectives of others than are their boys. Notably, children’s self-reports of their own social perspective taking also did not reflect any gender differences, consistent with our behavioural measures but not the parent-report. Similarly, the parent-report of children’s mental state understanding was not significantly correlated with age, which is inconsistent with the age-related changes observed in our behavioral measures. In contrast, children’s self-reports on social avoidance did predict their self-reported social perspective taking. This lends further credence to arguments that children’s reports might be more accurate than parent reports in this age range (Caqueo-Urízar et al., 2022; Kiel et al., 2015; Varni et al., 2007).

On the other hand, the parent-report measure (CSUS) assessing how often a child reasons or speaks about the minds of others revealed a negative relationship with child-reported fear of negative evaluation. This aligns with prior research in which the CSUS positively predicted social skills, and negatively predicted social emotional difficulties, in children (aged 3 to 7; Haddock & Birch, 2024). This relationship was somewhat attenuated in samples ranging in age from 3 to 12 years compared to samples with a narrower and younger age range (ranging from 3 - 7 years), raising the possibility that the CSUS’s ability to capture individual differences in older children may be more limited (Haddock & Birch, 2024). Again, these measurement differences underscore the challenges of relying on one measure, or one informant-type, when assessing social cognition. There are undoubtedly pros and cons to both informant types; for example, parent-reports may offer greater detail relative to child self-reports, while the child self-reports provide greater insight into their internalized, and therefore less observable, thoughts and feelings. Evidently, further research is needed to determine whether one type, child- or parent-report, shows greater reliability and accuracy in assessing certain aspects of social anxiety and social cognition, or whether the two complement each other in important ways.

In our exploratory analyses comparing the younger (6-8 years) and older (9-12 years) subgroups of our sample we found that the significant relationships between fear of negative evaluation and mental state understanding (i.e., positively related to the 2-item behavioural TOM task and negatively related to the parent-report CSUS), were limited to the younger half of the sample. This may suggest that children who possess an early sensitivity to criticism, may be less inclined to consider the minds of others in their daily lives for fear of being negatively evaluated, which could negatively impact their future social interactions and relationships. However, as this pattern did not hold in the older half of our sample, it is possible that, through social experience and exposure, children develop the necessary social skills so that the fear of negative evaluation no longer affects their ability (or willingness) to reason about the minds of others. It may be important for clinicians working with young children to address misconceptions surrounding the judgements of others to support social cognitive development. For example, interventions aimed at reducing fears of negative evaluation may prove especially helpful. Future research using a longitudinal design could also examine whether early developing social anxiety predicts or interacts with later social cognitive processing and how those influence children’s social skills and difficulties over time.

Conversely, the negative relationship between social avoidance and self-reported social perspective taking (FIMI) was only found in the older half of the sample. This finding raises the idea that social avoidance has a cumulative effect on certain components of social cognition (Hu, 2024), such that its influence is potentially stronger, or more likely to be observed, in older children. This aligns with prior work indicating social avoidance as a key predictor and maintenance of social anxiety (Ghandchi et al., 2023; McClure & Nowicki, 2001; Rodebaugh et al., 2025). However, these age-related findings warrant cautious interpretation, as splitting the original sample into two smaller groups limits the statistical power and reliability of these analyses. Future research would benefit from replications in targeted age samples and longitudinal designs.

The current study also offers novel methodological contributions. It is the first study to our knowledge to administer the FIMI as both a child-report and a parent-report (rather than as an adult self-report). The parent-report showed acceptable reliability with a Cronbach’s alpha of .74 and a significantly positive correlation with the well-validated parent-report measure of children’s mental state understanding (CSUS). This indicates that the parent-report FIMI may be a useful tool to measure a child’s propensity for social perspective taking. The child-report FIMI, on the other hand, showed limited internal consistency, yet was significantly related to children’s ratings of social avoidance. This suggests that the child-report FIMI may be a useful measure to capture children’s desire to engage in social perspective taking, and may be especially useful in relation to, or when examining, children’s tendency to avoid social situations. Future research should consider conducting Exploratory and Confirmatory Factor Analyses on the child and adult FIMI self-reports to evaluate factor structure and test measurement invariance across groups. Finally, this study used a novel measure of social avoidance (i.e., adding a follow-up ‘yes’ or ‘no’ question to each item on the SWAIY). Its internal consistency was low with a Cronbach's alpha of .47, but it nonetheless was positively and significantly correlated with our other measures of social anxiety, offering a novel way to distinguish the cognitive and behavioral components of anxiety that are not differentiated in the original SWAIY.

The current study found no association between prosodic emotion recognition and different aspects of social anxiety. Thus, our findings did not align with our expectations, nor with prior research indicating better (e.g., Folz et al., 2023; Mobach et al., 2022) or worse (e.g., Peschard & Philippot, 2017; Yuan et al., 2025) performance. This includes the limited number of studies specifically examining prosodic emotion recognition in children and adults (using the Diagnostic Analysis of Nonverbal Accuracy Tests), that reported poorer accuracy and slower reaction times in socially anxious individuals (McClure & Nowicki, 2001; Tseng et al., 2017). It is possible that our use of only 10 items to measure prosodic emotion recognition accounts for the differences between our results and earlier work, especially in light of the lower internal consistency of our 10-item version compared to the full battery (Rodgers, 2021). It is also possible that more interesting differences in our measures of emotion recognition and social avoidance contributed to these inconsistencies. For instance, McClure and Nowicki (2001) assessed recognition of simple emotions such as anger, fear, sadness, and happiness, whereas our measure included only complex emotions such as unfriendly, jealous, and amused. Additionally, their anxiety measures captured children’s social avoidance in both general and novel social situations, and their findings were specific to children who showed higher levels of general social avoidance. Our measure did not distinguish between avoidance of general versus novel situations possibly limiting its sensitivity or specificity.

Our assessment of prosodic emotion recognition may have also suffered some notable limitations, particularly in the voice recordings used in the CAM-C. The measure was designed and validated in the UK; therefore, the recordings were done with British English actors, and some of the children in our study (located in Canada) may have struggled to pick up on the emotions being expressed because of the unfamiliar accents. Given these limitations we remain agnostic about how to interpret the emotion recognition results and recommend further research to examine the relationships between emotion recognition and aspects of social anxiety.

Finally, our study was somewhat limited in its sample size. Our power analysis indicated that 78 would be adequate to detect a medium effect in each of our regression analyses, and past reviews on social anxiety and social cognition have suggested a medium effect size at minimum (Alvi et al., 2022). That said, the current work ran multiple regression models which would benefit from larger sample sizes to correct for multiple tests, minimizing the risk of bias and inflated p-values (Jafari & Ansari-Pour, 2019). Future research would benefit from a replication of this work, using larger and broader samples.

5. Conclusions

In sum, our findings reveal that social cognition is a broad construct encompassing a number of cognitive processes that may show little overlap and varied associations with other constructs, such as social anxiety, depending on the component of social cognition under investigation. We found that greater fear of negative evaluation was a significant predictor of increased accuracy in mental state understanding, at least in terms of reasoning about false beliefs and hidden emotions (measured using the TOM scale). In comparison, greater fear of negative evaluation was a significant predictor of lower parent-reported mental state understanding in their child’s observable day to day behavior (measured using the CSUS). Including measures of social avoidance accounted for additional variance in children’s self-reported social perspective taking (measured using the FIMI). Our findings also highlight the importance of distinguishing between one’s mental state understanding (i.e., their ability, or capacity, to accurately reason about others’ mental states) and their propensity to do so (i.e., the ease and frequency with which they think about other’s mental states). Although these often align (e.g., an individual with a high propensity will tend to be more accurate in their reasoning), they are importantly dissociable as discussed in the aforementioned examples of over- and under-mentalizing. Unfortunately, many existing social cognition measures confound these two components. A more complete understanding of the distinction between children’s competence and their propensity may be especially useful when predicting children’s social skills and social-emotional well-being. Together, our results suggest that a complete understanding of social cognitive ability requires a multifaceted approach with a variety of measures. Future research would benefit from developing and validating additional social cognition measures for school-age children to fully capture important differences in these social cognitive components. Researchers should also refrain from using broad terms such as “social anxiety” or “social cognition” to summarize previous results without clarifying the specific constructs and measures involved.

This research is particularly relevant today, as a generation of young children experienced a substantial reduction in social interactions due to the COVID-19 pandemic, which a growing body of work suggests negatively impacted social cognitive development (Alan & Turkum, 2024; Rodriguez-Monge et al., 2023) and increased levels of social anxiety (Peçanha et al., 2025). Unfortunately, the question of experiencing another pandemic as disruptive as COVID-19 may not be a matter of if, but when (King, 2025). It is more important than ever to understand which factors support or obstruct effective social cognition and exacerbate anxiety, given that poor social competency in childhood is linked to social exclusion, loneliness, poor mental health, lower academic achievement, and even criminality (Bullock, 1988; Frogner et al., 2022; Vahedi et al., 2012; Webster-Stratton & Lindsay, 1999). Thus, research examining social cognition and how factors such as social anxiety or social avoidance affect both children’s capacity and motivation to engage in social cognition is of great importance for fostering children’s social-emotional wellbeing.