Social Interaction Anxiety in Developmental Prosopagnosia: Prevalence, Severity, and Individual Differences

Abstract

Objective

Although elevated social anxiety in developmental prosopagnosia (DP) has been reported in anecdotal and qualitative studies, the current study sought to better quantify the prevalence, severity, and moderators of social anxiety in a large DP sample.

Method

A total of 88 DPs and 58 controls completed the validated Social Interaction Anxiety Scale and assessments of face recognition, autism traits, personality (Big Five Inventory), and coping strategies.

Results

DPs reported greater social anxiety symptoms (M = 30.25) than controls (M = 17.19), with 44% of DPs above a probable clinical cutoff compared with 14% of controls. Exploratory analyses revealed that DPs’ social anxiety was more specific to items pertaining to mixing socially or interacting with unfamiliar people. Notably, several DPs reported minimal social anxiety, which was associated with being more extraverted and having fewer autism traits. A follow-up survey revealed that extraverted DPs more openly disclosed face recognition inabilities than introverted DPs, which may be a factor in their reduced social anxiety.

Conclusion

These results better quantify the potential serious psychosocial consequences of DP and highlight the importance of individual differences as well as targeted intervention.

INTRODUCTION

Individuals with developmental prosopagnosia (DP), a severe lifelong impairment in the ability to learn and recognize faces despite normal cognitive and visual functioning (Duchaine & Nakayama, 2006), have reported difficulty and distress regarding their social functioning (Murray et al., 2018; Yardley et al., 2008). Social functioning has been well established as an important predictor of physical health, happiness, and overall life satisfaction (Eisenberger & Cole, 2012; Hamarta, 2009; Holt-Lunstad et al., 2010). One important aspect of social functioning is social anxiety, which involves maladaptive affective and behavioral responses during the anticipation, avoidance, and post-avoidance of perceived social threats (Evans et al., 2019). Subclinical levels of social anxiety have been found to correlate with reduced physical health and life satisfaction as well as increased risk of psychopathology (Fehm et al., 2008). Clinical levels of social anxiety (i.e., social anxiety disorder, SAD) are associated with high rates of interpersonal stress, particularly impairment in close relationships (Davila & Beck, 2002; Schneier et al., 1994; Sparrevohn & Rapee, 2009), and increased social avoidance, which is itself associated with financial dependency and decreased educational and workplace functioning (Moitra et al., 2011; Santini et al., 2020; Schneier et al., 1992). Although previous research suggests that individuals with DP experience significant social anxiety (Yardley et al., 2008; Murray et al., 2018; though see Svart & Starrfelt, 2022, where a subset individuals with face recognition deficits reported being better at some social skills), evidence has been anecdotal or qualitative in nature with studies using smaller DP samples, a population known for its heterogeneity (Barton & Corrow, 2016; Dalrymple et al., 2014). Therefore, the extent of social anxiety in DPs remains unclear as well as if, how, and why individuals with DP may vary in social anxiety severity. The primary goal of the present study was to use a validated measure of social interaction anxiety to measure the prevalence and severity of social anxiety in a large DP sample, with a secondary goal to explore potential individual differences among DPs.

Several studies suggest that DP can pose serious consequences to social functioning and increased social anxiety. Yardley et al. (2008) conducted semi-structured interviews with 25 DPs, asking open-ended questions about how DP has affected participants’ lives, behavior, and others’ perceptions of them, as well as whether DP had led to any negative experiences or emotional difficulties. They found almost all DPs described recurrent and sometimes highly distressing social faux pas that often led to chronic anxiety surrounding social interactions, as well as feelings of embarrassment, guilt, fear, or failure. Yardley and colleagues (2008) suggested that “many” DPs reported serious, long-term consequences, such as social isolation, over-dependence on one or two friends or family members, limited employment opportunities, and decreased self-esteem, though no percentages are reported. Similarly, in a 2018 qualitative study of 50 DPs, Murray et al. (2018) reported that 90% of DPs mentioned negative social consequences due to face recognition difficulties in semi-structured interviews and open-ended questionnaires. Though prevalence rates are not reported, Murray and collegues (2018) suggest that there were “numerous” accounts that reflected patterns of social anxiety, including rarely introducing oneself, fearing coming off as disinterested or rude, and frequently avoiding social gatherings altogether. Similar negative social implications of DP have also been reported in semi-structured interviews with DP children and their parents Dalrymple et al., 2014). Consistent with these studies, a recent analysis of over 7.5 million de-identified electronic health records demonstrated high concordance rates of prosopagnosia (acquired or developmental), depression, and generalized anxiety with largely overlapping comorbidities across the three conditions, further highlighting the potential for serious clinical consequences of DP (Pressl et al., 2021). It may, however, be the case that not every DP struggles socially. A recent study of 115 individuals with self-reported face recognition problems demonstrated that about 10% of participants reported particular social strengths, such as knowing what other people feel, and why they feel it, or “intuitive profiling” (Svart & Starrfelt, 2022). Although these studies have collectively been very useful in identifying and calling attention to the many social and daily life effects of living with DP, more quantitative comparisons between DPs and neurotypical controls are needed. Moreover, studies thus far have not addressed why DPs experience varying degrees of social anxiety symptoms.

Individual differences in DP social anxiety could potentially be due to the severity of face recognition deficits (Davis et al., 2011), variance in autism spectrum disorder traits (ASD, Spain et al., 2020), or aspects of personality (Kaplan et al., 2015). First, there is evidence of a bidirectional relationship between face recognition and social anxiety. Previous research has shown that inducing physiological symptoms of anxiety can reduce face recognition performance (Attwood et al., 2015; Bolton & Robinson, 2017), whereas, as mentioned earlier, DP anecdotal reports suggest that face recognition challenges may lead to the development of chronic social stress and anxiety (Murray et al., 2018; Yardley et al., 2008). Further, in a study of 138 adults unselected for face recognition ability, Davis et al. (2011) found that poorer face identity recognition was correlated with increased social anxiety. Based on these studies, we predict that DPs with the highest social anxiety may be those with the greatest face recognition deficits.

In DP, variation in social anxiety severity may also be attributable to heterogeneity in the presence and severity of autism traits (Fry et al., 2023). ASD, a group of developmental disorders affecting communication and social behavior (DSM–5, American Psychiatric Association, 2013), has been found to be highly associated with face recognition difficulties (Hedley et al., 2011; Minio-Paluello et al., 2020). Likewise, the prevalence of DP among those with ASD is substantially higher (up to 12%¹, Minio-Paluello et al., 2020) than that of the general population — 1-3% (DeGutis et al., 2023). Even so, not all DPs have elevated autism symptoms, with a study of 43 DPs finding just over a third of participants to display traits of a broad autism phenotype, a subclinical set of characteristics that are qualitatively similar to features of autism (Fry et al., 2023; Landry & Chouinard, 2016). Many DP studies have not excluded those diagnosed with ASD (e.g., Yardley et al., 2008), though some have (e.g., Murray et al., 2018). Even when excluding DPs with probable ASD, DPs often have elevated subclinical levels of ASD (Fry et al., 2023). Studies consistently show a strong association between ASD and social anxiety, with clinical levels of social anxiety at a prevalence rate of as high as 72% among those with ASD, compared with 7%–13% in typically developing samples (Boulton & Guastella, 2021; Maddox & White, 2015; Spain et al., 2018). This increased social anxiety in ASD can be explained by both subtle and overarching difficulties with aspects of social interaction, such as generally being around others, non-verbal communication/emotion recognition, and uncertainty or changes to a routine (Halim et al., 2018; Spain et al., 2020; Trembath et al., 2012). Thus, it is possible that elevated social anxiety symptoms in some DPs may be driven by elevated autism traits.

Variation in social anxiety among DPs may also occur due to individual differences in personality dimensions. It is well established that lower extraversion and higher neuroticism predict greater social anxiety in unselected samples (Kaplan et al., 2015; Kotov et al., 2010; Newby et al., 2017; Shojaati et al., 2021). Neuroticism has shown to have a strong association with both stress and depression, which are important factors in the development of social anxiety (Brown et al., 1998; Farmer & Kashdan, 2015; Kendler et al., 2003), and has even been found to predict the onset of anxiety symptoms and disorders (Hayward et al., 2000; Watson et al., 1994). Although the literature on extraversion is more mixed, there is evidence for low extraversion to be associated with certain anxiety disorders, including SAD (Bienvenu et al., 2001; Watson et al., 2005; though see Gershuny & Sher, 1998). Personality differences between DPs and controls, however, have not yet been examined, nor has the contribution of these factors to DP social functioning or social anxiety.

To address the gaps in the existing literature, we quantified the prevalence and severity of social anxiety symptoms in a relatively large sample of DPs (N = 88 DPs and 58 neurotypical controls) using a validated measure of social anxiety, the Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998). The SIAS is a 20-item self-report measure of distress during social interactions with others. To determine if DP social anxiety is comparable to a clinical population, we also compared DP SIAS responses to a large previously reported sample with diagnosed social anxiety (N = 355, Carleton et al., 2009). Further, we conducted exploratory analyses of SIAS items to examine whether DPs' social anxiety generalizes across items, as is typical in clinical populations (Carleton et al., 2009), or is selective to certain situations. Finally, we hypothesized that not every DP would be affected to the same degree and that DP individual differences in social anxiety would be related to more severe face recognition deficits (Cambridge Face Memory Test, Duchaine & Nakayama, 2006; and Famous Faces Test, Mishra et al., 2019), higher autism traits (Autism Spectrum Quotient, AQ; Baron-Cohen et al., 2001), and aspects of personality (Big Five Inventory; John et al., 1991).

MATERIALS AND METHODS

Participants

A total of 88 DPs (40 in lab/48 online) and 58 control participants (all in lab) completed the study. Although several DPs were assessed in-person, the COVID-19 pandemic prompted testing to be moved online. Online participants were assessed by research staff over Zoom. Each task was completed in full screen on participants’ devices and began with a calibration exercise to ensure consistent stimulus resolution and size across the entire sample. To ensure minimal distraction and maximum similarity to our in-person protocol, the researcher turned off their video and microphone for the duration of each task. All instruction and face-to-face interaction remained consistent with in-person testing. DPs were recruited via our database of DPs who previously participated in laboratory studies, references from other research laboratories (e.g., Dr. Brad Duchaine, Dartmouth College, www.faceblind.org), advertising on the Massachusetts Bay Transportation Authority “T” system, and from our study listed on www.clinicaltrials.gov. To be eligible for the study, all participants first underwent a brief phone screen to ensure they reported symptoms consistent with DP. Exclusionary criteria included prosopagnosic symptoms after a brain injury, a history of any significant neurological disorders (e.g., Alzheimer’s Disease, multiple sclerosis), lifetime moderate or severe traumatic brain injury (TBI) or mild TBI within the last 6 months, musculoskeletal or sensory impairments, lack of English proficiency, psychiatric disorders including schizophrenia or bipolar disorder, or current dependence on alcohol or other substances. We additionally screened participants for low intellectual quotient (IQ) using the Wechsler Test of Adult Reading (WTAR; Wechsler, 2001), which provides an estimated verbal IQ level that is highly correlated with measures of verbal and full-scale IQ (Strauss et al., 2006). Individuals were not included if they had a WTAR-predicted IQ below 80. This did not exclude any of our participants.

All 88 DP participants reported lifelong face recognition deficits and were classified into major and mild DPs based on their Cambridge Face Memory Test (CFMT) and Famous Faces Memory Test (FFMT) scores. Our classification method is informed by a standard, validated approach that is currently used to diagnose continuously distributed neurocognitive disorders from the latest edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5, Sachdev et al., 2014). Based on poor reliability associated with using a single measure in diagnosis (Holdnack et al., 2017), the DSM-5 recommends that at least two objective validated measures within a domain are impaired (2 SD below the normative mean for major neurocognitive disorder, between 1 and 2 SD below the mean for mild) to receive a diagnosis. In this study, participants with z-scores ≤ −2 on 2 or more tests, in this case CFMT and FFMT, were classified as having a major impairment, whereas those with z-scores ≤ −1 on 2 or more tests were classified as mild; DeGutis et al., 2023). An additional 19 individuals reporting face recognition deficits were recruited and screened but failed to meet objective criteria for major or mild DP and were therefore excluded from participation in the study. The vast majority of participants (75/88) were found to be major DPs, whereas 13 were mild DPs. All 13 mild DPs had one z-score ≤ −2 and the other between −2 and − 1. Removing mild DPs had no appreciable effects on the key results. All DPs had normal or corrected-to-normal vision, scoring within the normal threshold on the Leuven Perceptual Organization Screening Test (< 3 subtests below the 10th percentile; Torfs et al., 2014). A total of 48 DPs returned to complete a follow-up survey.

A total of 75 control participants took part in the study. 17 participants were recruited and tested but excluded from analysis for having scored below 45 on the CFMT or evidence of low effort (near or below chance accuracy and/or responding very quickly). This left us with a sample of 58 controls. Subjects were recruited from the community primarily via flyers and the Harvard Decision Science lab. All participants had normal or corrected-to-normal vision and reported having never experienced difficulties with face recognition. All participants gave informed consent according to the Declaration of Helsinki and in compliance with either the Institutional Review Board of the VA Boston Healthcare System or Harvard University.

Objective face recognition assessments

These tests were implemented in either PsychoPy v1.85.4 or JavaScript (for CFMT) and run on a laptop computer (34.5 x 19.5 cm display, 1920 x 1080 pixels, 60 Hz) or were run online using the www.testable.org platform. Both in-lab and online participants were instructed to sit 60 cm from the computer screen and indicated their responses using either a keyboard or a computer mouse. Written and spoken instructions were provided.

Cambridge Face Memory Test

The CFMT (Duchaine & Nakayama, 2006) is a validated and widely used face memory task. The CFMT has shown to be highly reliable (e.g., Cronbach’s α = 0.89, Wilmer et al., 2010). During the learning phase, participants are shown each target face from three different viewpoints for 3 s each and are then asked to select which face they just viewed from a selection of three faces using the number keys. This learning trial repeats for each of the six target faces. After the learning phase, participants are shown all six target faces again for 20 s. They are then given 30 x 3-alternative-forced-choice trials where one was one of the six target faces they just learned. After these 30 test trials, participants again are shown all six faces for another 20 s, and the final 24 trials include Gaussian noise intended to obscure the facial features. The dependent variable is the total correct trials out of 72.

Famous Faces Memory Test

The FFMT is a test of familiar face recognition using a set of 20 front-view images of very famous celebrities from testmybrain.org (see Mishra et al., 2019) with hair, ears, and any accessories cropped out. Participants are prompted to provide the name of the face, or if they could not come up with the name, to provide other identifying information (e.g., movies they starred in). After they entered a response, the correct answer/name of the person was displayed on the screen, and they were prompted to click on either of the following: “I got it right”; “I got it wrong and I am familiar with this person”; or “I got it wrong and I am not familiar with this person.” Total score was calculated by taking the percent correct out of the total number of faces a participant reported being familiar with.

Self-report assessments

Prosopagnosia Index 20

The 20-Item Prosopagnosia Index (PI20; Shah et al., 2015) is a self-report measure of prosopagnosia traits designed to quantify one’s face recognition abilities in daily life. Participants respond to 20 statements describing their face recognition experiences (e.g., When people change their hairstyle, or wear hats, I have problems recognizing them) using a five-point likert scale from strongly disagree (1) to strongly agree (5). Five of the items are reverse-scored, and the total points are then summed to create a total measure score out of 100, with higher scores indicating greater face recognition difficulties. The measure has high internal consistency, with a Cronbach’s α of 0.84 (Shah et al., 2015).

Social Interaction Anxiety Scale

The Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998) is a self-report scale designed to measure the amount of distress an individual experiences during social interactions with others. The scale has high internal consistency (Cronbach’s α = 0.94) and test–retest reliability (r = 0.92) (Gomez & Watson, 2017; Mattick & Clarke, 1998). There are 20 items, each measured on a five-point Likert scale. Each item includes a statement about social interaction (e.g., “I am tense mixing in a group”). Individuals can respond whether they believe the characteristic is “not at all characteristic or true of me,” “slightly characteristic or true of me,” “moderately characteristic or true of me,” “very characteristic or true of me,” or “extremely characteristic or true of me.” Three of the items are reverse-scored, and the points are then summed to create a total measure score out of 80, with higher scores indicating greater social anxiety.

Autism Quotient Questionnaire

The Autism Quotient questionnaire (AQ; Baron-Cohen et al., 2001) is a 50-item self-report questionnaire designed to identify symptoms of ASD in adults. Participants are instructed to respond either “definitely agree,” “slightly agree,” “slightly disagree,” or “definitely disagree” to each item, with approximately half of the items designed to evoke an “agree” response from a neurotypical adult. The questionnaire covers the domains of social skills, communication, imagination, attention to detail, and attention switching. For the purposes of our study, the total score out of a maximum of 50 was used, with higher scores indicating more autism traits.

Big Five Inventory

The Big Five Inventory (BFI; John et al., 1991) is a self-reported questionnaire intended to measure the dimensions of one’s big five personality traits (agreeableness, conscientiousness, extraversion, neuroticism, and openness) (Goldberg, 1993). The version administered consisted of 44 items to which participants are instructed to respond on a five-point Likert scale. Responses include “disagree strongly,” “disagree a little,” “neither agree nor disagree,” “agree a little,” and “agree strongly.” Each item corresponds to one of the five personality traits, and responses are scored 1–5 (16/44 reverse scored), in which higher scores correspond to greater levels of that personality domain. The questionnaire yields a total score for each personality trait by averaging the responses to these items, with a maximum score of 5 for each domain. The subscale measures are highly reliable, with Cronbach’s α's ranging from 0.79 to 0.88 (agreeableness α = 0.82, conscientiousness α = 0.83, extraversion α = 0.88, neuroticism α = 0.84, and openness α = 0.79; Arterberry et al., 2014).

Prosopagnosia follow-up survey: Disclosure, responses to errors, and coping strategies

This follow-up survey was created after we performed our main analyses (see below) in order to further characterize differences between introverted vs. extraverted DPs in four domains: (1) disclosure of face recognition challenges, (2) coping strategies in social interaction, (3) responses to making a face recognition error, and (4) anticipation of social interaction. Each section included questions to which participants are instructed to respond on a five-point Likert scale. Responses include “never,” “rarely,” “sometimes,” “often,” and “always,” as well as a “not applicable” option, corresponding to how often participants used a particular strategy. Individual responses are then scored 0–4. See supplementary material online, Table S1 for a full list of survey items.

Statistical analysis approach

Group differences

We first performed independent samples t-tests between the DP and control groups’ total SIAS scores to determine if there was a significant group-level difference in social anxiety levels. We also conducted independent samples t-tests to explore group-level differences on the AQ, as autism is often correlated with anxiety symptoms. We subsequently ran an analysis of covariance (ANCOVA) to see if any group differences on SIAS remained after controlling for AQ. Finally, to measure the extent to which SIAS scores reached clinical levels, we calculated the percentage of each group that reported scores at or above a suggested clinical cutoff of 34 points to indicate probable SAD in the DSM-5 (Brown et al., 1997; Heimberg et al., 1992), and performed a chi-squared analysis to determine if these percentages were significantly different between groups.

Item analysis

Before exploring group differences on individual items, we ran a repeated measures analysis of variance (ANOVA) to identify whether there was a significant interaction between diagnostic group and SIAS items. We followed this up with independent samples t-tests each SIAS item to determine where DPs and controls differed the most. We identified the items with the largest effect sizes as being the most different between groups and items that did not differ significantly as being most similar between groups. All p-values were corrected for multiple comparisons using a false discovery rate (FDR) of 0.05 (Benjamini & Hochberg, 1995). Although we considered performing a confirmatory factor analysis on the SIAS, we decided against this due to previous studies demonstrating either a single factor (Heidenreich et al., 2011; Rodebaugh et al., 2006; Safren et al., 1998) or two correlated factors (Safren et al., 1998) solutions, with some suggesting that these factors lacked consistency or validity (e.g., Gomez, 2016).

Individual differences

Next we calculated Pearson’s correlations between the SIAS and each of our hypothesized domains (face recognition ability, autism traits, personality) in our battery to identify what may be associated with higher levels of anxiety within DPs. For personality, due to there being multiple BFI subscales, we corrected for multiple comparisons using an FDR of 0.05. Any missing data were removed pairwise.

Because we found that extraversion was such a strong predictor of SIAS, we next split our DP group into introverts and extraverts based on a large normative sample (135,000 participants) of BFI data (Srivastava et al., 2003). Data from this sample were reported as transformed percentage of maximum possible (POMP) scores: a linear transformation of scores into a 0 to 100 scale, where 0 represents the minimum possible score and 100 represents the maximum. In order to replicate this transformation, we subtracted 1 from the scores yielded for each BFI personality trait and multiplied by 25. We then used the large sample’s reported mean for extraversion to perform a mean split on our sample, sorting participants with extraversion POMP scores below the mean into the introvert group and those above the mean into the extravert group. We then conducted independent samples t-tests of SIAS scores between introverted DPs and controls as well as between extraverted DPs and controls. Next, we ran an ANOVA with SIAS as the dependent variable, DP vs. control group as one factor, and introversion vs. extraversion as the other factor, to identify a possible interaction between extraversion scores and diagnostic group to explore the nature of the relationship between extraversion and social anxiety as affected by face recognition deficits. Finally, we ran a multiple regression across all introverts (DP and control) with SIAS as the dependent variable, using the measures that showed a significant zero-order SIAS relationship as predictors to see what best predicted anxiety across introverts.

Item analysis: Part 2

We performed the same item analysis described earlier (see section Item analysis) between the introverted DP group and controls, as well as the extraverted DP group and controls. In addition to comparing DPs to our control group, we also sought to compare DP responses to those of a clinical sample with diagnosed SAD. We compared the mean introverted and extraverted DP item responses to the clinical data reported in Carleton et al. (2009). Due to only means and SDs having been reported for their clinical sample, we compared the similarity of responses by identifying the items on which the clinical mean fell within the 95% confidence interval of our DP group means. We then identified the items on which one SD from the clinical mean fell within the same range. Finally, we calculated differences between the means from the introverted DPs and the clinical group, identifying the items with the greatest and smallest differences to explore the nature of DPs’ most heightened anxiety.

Follow-up survey

The goal of our follow-up survey was to examine coping strategy differences between DPs and to identify any trends in individual differences. We created the survey after we performed our main analyses to further characterize differences between introverted and extraverted DPs. We calculated an average score on each section for each participant, as well as average scores on each individual item across subgroups of interest (e.g., DPs with higher autism traits), and conducted independent samples t-tests between the average introverted and extraverted DP responses. We then ran independent samples t-tests on each individual item of the survey between introverted and extraverted DPs. All resulting p-values were then corrected for multiple comparisons, again using an FDR of 0.05.

Sample size justification

Our sample size was guided by previous studies comparing DP and control group differences (N = 76, 38 DP, DeGutis et al., 2012; N = 60, 30 DP, Fry et al., 2020; Barton et al., 2004; Jiahui et al., 2018). Due to the limited sample sizes in previous work studying social anxiety in DP (Murray et al., 2018; Yardley et al., 2008) and the heterogeneity in this population, we wanted to include a substantially larger sample than has previously been studied. By including these additional DPs, we could improve the sensitivity to detect differences across our battery of measures and increase the replicability of our findings. Using alpha of 0.05, power (1-β) of 0.8, 88 DPs and 58 controls allowed us to detect a small-to-medium DP vs. control group effect of Cohen’s d = 0.48. In terms of individual difference correlations within DP, using alpha of 0.05, power (1-β) of 0.8, 88 DPs allowed us to detect weak-to-moderate correlations of 0.29.

RESULTS

Participants

Our final sample consisted of 88 DPs (61 females, 2 nonbinary) and 58 controls (25 females) with a similar mean age of 41.13 (SD = 14.83) and 43.63 (SD = 12.58), respectively (see Table 1). Applying our DSM-5-informed approach for diagnosing neurocognitive disorder (≥ 1 SD below the mean on two or more tests for mild, ≥ 2 SD for major) to our face recognition measures (CFMT, FFMT), we classified 75 major DPs and 13 mild DPs (see DeGutis et al., 2023). As defined by our selection procedure, the entire DP group performed substantially worse than controls on the CFMT and FFMT (Table 1). The control group CFMT mean and SD were similar to the original Duchaine and Nakayama (2006) study and there were only three controls with scores of 70 or above, suggesting that controls with very high face recognition abilities were not driving DP vs. control group differences. Although the DP group did not differ significantly from controls in age, the control group had a higher proportion of males than the DP group (See Table 1). Though SAD has been found to be more common in females than males, many studies report no gender differences on the SIAS (Caballo et al., 2013; Heimberg et al., 1992). When looking across our entire sample (DPs and controls), we found no differences in SIAS scores across gender (M_female = 25.82, SD_female = 14.03; M_male = 24.71, SD_male = 15.88; p = .676). We also found that our DP sample had more highly educated individuals than our control group². As previous research has found high levels of social anxiety to be associated with decreased educational attainment (Kessler, 2003; Vilaplana-Pérez et al., 2021), this would, if anything, slightly decrease DP versus control group differences.

Table 1.

DP and control group means for demographics and face recognition assessments

Group	N	Age	Gender	AQ	CFMT	FFMT	PI20
DP	88	41.30 ± 15.02	61:25:2	21.73 ± 8.13	39.50 ± 5.03	0.33 ± 0.17	85.07 ± 8.14
Control	58	43.03 ± 12.85	28:30	18.38 ± 5.42	58.01 ± 7.55	0.74 ± 0.20	35.52 ± 7.87
p-value	-	.460	.009	.007	<.001	<.001	<.001
Cohen’s d	-	0.14	0.45	0.46	3.02	2.24	6.05

Note: Mean ± Standard Deviation, Gender = Female:Male:Non-binary, AQ = Autism Quotient, CFMT = Cambridge Face Memory Test, FFMT = Famous Faces Memory Test, PI20 = 20-Item Prosopagnosia Index.

Developmental Prosopagnosic vs. control group differences in social interaction anxiety

We first examined DP vs. control group SIAS scores and found there was a significant difference (t(102) = −5.74, p < .001, Cohen’s d = 0.99; see Fig. 1), with the DP mean being substantially higher than that of controls (M_DP = 30.25, SD_DP = 14.20; M_TD = 17.19, SD_TD = 11.02). DPs also had significantly higher AQ scores than controls (M_DP = 21.73, SD_DP = 8.13; M_TD = 18.38, SD_TD = 5.42; p = .007). Importantly, the significant group difference in SIAS remained after controlling for AQ in an ANCOVA (F(2, 126) = 20.53, p < .001), suggesting that group differences in social anxiety were not only due to elevated autism traits.

Fig. 1.

DP versus control social interaction anxiety scale (SIAS) distributions. Note. SAD = social anxiety disorder. y-axis represents normalized density, meaning DP and control distributions are displayed with the same area under the curve for the sake of comparison. DP, developmental prosopagnosia.

We then determined how many DPs met the criteria for a clinical level of anxiety on SIAS, using the clinical cutoff of 34³ (see Methods). Within the DP sample, 39 of the 88 participants (44%) had scores ≥34, a significantly higher percentage than controls above the cutoff, 8 of the 58 (14%, similar to other reports in normative samples, e.g., Ruscio et al., 2008; χ² (1, N = 146) = 14.92, p < .001). Notably, this shows that over half of DPs did not demonstrate clinical levels of social anxiety, a finding contrary to the existing DP literature (Murray et al., 2018; Yardley et al., 2008). This, along with the large overlap between DPs and controls (Fig. 1), suggests considerable social anxiety heterogeneity in DPs. Although some DPs demonstrate clinical levels of social anxiety others may not suffer in this area to the same degree or even at all.

Social interaction anxiety scale item analysis

Though DPs showed significantly elevated SIAS scores compared to controls, it remains unclear whether certain items on the SIAS drove this effect or rather responses across the entire scale were elevated, as has been found in individuals diagnosed with SAD (Caballo et al., 2013; Šipka et al., 2023, see below). To test this, we performed a group (DP vs. Control) × SIAS item repeated-measures ANOVA and found a significant interaction (F(2, 19) = 6.31, p < .001). We followed this up with independent samples t-tests (FDR corrected for multiple comparisons), finding significant DP vs. Control differences on 15 of 20 items (see Supplementary material online, Table S2). The items with the largest effects involve mixing socially or interacting with familiar and unfamiliar people. Four items related to interacting with potentially familiar people in unconstrained settings showed large effect sizes: Q6 (I tense up if I meet an acquaintance in the street; p < .001, q = 0.003, Cohen's d = 0.83), Q7 (When mixing socially, I am uncomfortable; p < .001, q = 0.003, d = 0.95), Q9 (I am at ease meeting people at parties, etc.; p < .001, q = .003, d = 1.02), and Q20 (I am unsure whether to greet someone I know only slightly; p < .001, q = 0.003, d = 1.03). The next highest effect sizes were found on Q16 (I am nervous mixing with people I don’t know well; p < .001; q = 0.003, d = 0.78) and Q19 (I feel tense mixing in a group; p < .001; q = 0.003, d = 0.68).

In contrast, DPs did not differ from controls on items largely involving one-on-one interactions that do not depend on face recognition: Q3 (I become tense if I have to talk about myself or my feelings; p = .791, q = 0.791, d = 0.04), Q8 (I feel tense if I am alone with just one other person; p = .249, q = 0.326, d = 0.19), Q11 (I find it easy to find things to talk about; p = .087, q = 0.109, d = 0.29), and Q14 (I have difficulty talking to attractive persons of the opposite sex; p = .339, q = 0.703, d = 0.08). Together, these results suggest that DPs’ elevated levels of social anxiety may be more specific to situations that are affected by challenges with face recognition rather than being generalized to all aspects of social interaction.

Predicting individual difference in developmental prosopagnosic social interaction anxiety

We next investigated predictors of individual differences in DP social anxiety, focusing on face recognition ability, autism traits, and personality measures. Because social anxiety and face recognition are associated in controls (Davis et al., 2011), we predicted that DPs with more severe face recognition deficits would have higher levels of social anxiety. We found little-to-no correlation between SIAS and CFMT (r = −0.03, p = .782) or FFMT scores (r = −0.02, p = .853, See Fig. 2; see Supplementary material online, Fig. S1 for all task correlations). This could be due to DPs’ restricted CFMT range from low scores being a diagnostic requirement, though when comparing SIAS scores between our major and mild DPs, there was still no significant difference observed (t(86) = 8.05, p = .216). SIAS and PI20 scores, however, were significantly correlated (r = 0.25, p = .020), but this is likely because the PI20 includes a few items related to social anxiety (e.g., Q11-Anxiety about face recognition has led me to avoid certain social or professional situations). In controls, we similarly did not find significant correlations between SIAS and objective face recognition (CFMT r = −0.26, p = .056; FFMT r = −0.14, p = .284) but found a significant correlation between SIAS and self-reported face recognition ability (r = 0.31, p = .017, see Supplementary material online, Fig. S2; see Supplementary material online, Fig. S3 for pooled sample correlations). We next examined autism traits on the AQ and found that greater autism traits in DP were robustly associated with higher social anxiety on the SIAS (r = 0.49, p < .001). A very similar AQ/SIAS relationship was found in controls (r = 0.49, p < .001).

Fig. 2.

Measures associated with the severity of social anxiety in DPs and controls. Note. SIAS = social interaction anxiety scale, AQ = autism quotient; PI20 = 20-item prosopagnosia index. * = q < 0.05, ** = q < 0.001, where q = Benjamini–Hochberg adjusted p-value. DP = developmental prosopagnosia.

Though DP autism traits were a robust predictor, we found the strongest predictor of DPs’ SIAS scores were personality measures from the BFI. We first compared DPs and controls across the five personality dimensions (FDR correcting for multiple comparisons) and found similar means across dimensions, with DPs only trending toward being less extraverted than controls (p_extraversion = 0.017, q_extraversion = 0.085; all other ps and qs > 0.05; see Supplementary material online, Table S3 for full results). Within the DP group, after multiple comparison correction, we found that higher SIAS scores were strongly correlated with lower extraversion (r = −0.57, p < .001, q = 0.005) and higher neuroticism (r = 0.27, p = .010, q = 0.025), but only showed a trend with lower agreeableness (r = −0.22, p = 0.039, q = 0.073). We found little-to-no correlation between conscientiousness or openness and SIAS in our DP group (r = −0.09, p = .436, q = 0.436; r = 0.13, p = .223, q = 0.279, respectively). Considering the BFI extraversion measure has a reliability of 0.88 (Arterberry et al., 2014) and the SIAS itself has a reliability of 0.94, making the upper bound of a correlation between the two 0.91, our data suggests a very strong relationship between extraversion and SIAS scores. Controls similarly showed a strong relationship between SIAS scores and extraversion (r = −0.52, p < .001, q = 0.003), but their social anxiety also appeared to have significant, moderate-to-strong associations with all four other personality traits, a trend not seen in DPs (see Supplementary material online, Figs S1 and S2). Interestingly, neuroticism had numerically the strongest correlation to SIAS scores in our control sample (r = 0.59, p < .001).

Finally, to examine the relative contribution of measures, we ran multiple regression models with the predictors being each of the variables that showed a significant zero-order association with SIAS. In DPs, we found that extraversion was the strongest predictor of unique variance in SIAS scores (β = −0.47, p < .001), followed by AQ (β = 0.33 p < .001) and neuroticism (β = 0.21, p = .012), whereas PI20 scores did not predict unique variance. The total model predicted 50.5% of the variance in SIAS scores (R² adjusted = 0.48). When running a similar analysis in controls (multiple regression with the variables that significantly predicted unique variance of SIAS scores in controls), we found neuroticism and extraversion were significant unique predictors, though not AQ (see Supplementary material online, Table S4).

Social anxiety in introverted vs. extraverted developmental Prosopagnosics

Considering that extraversion was the strongest predictor of DP social anxiety and that the extravert vs. introvert dichotomy is a widely used (Zelenski et al., 2012) and could be a useful framework for DPs, we split our sample based on normative data from a large BFI study (Srivastava et al., 2003). There were 52 introverted and 36 extraverted DPs, compared with 24 introverted and 34 extraverted controls (χ² (1, N = 146) = 4.39, p = .036). Both introverted DPs and controls had greater mean SIAS scores (M_DP = 36.37, SD_DP = 11.95; M_TD = 24.88, SD_TD = 13.18) compared to their extraverted counterparts (M_DP = 21.42, SD_DP = 12.55; M_TD = 14.79, SD_TD = 10.66; p_DP < 0.001, p_TD = 0.003). Though the DP group showed a larger introvert/extravert SIAS difference than controls (Cohen’s d_DP = 1.23, d_TD = 0.86), the extraversion x group interaction was not significant. Notably, the distribution of extraverted DPs SIAS scores was very similar to that of the full control sample, whereas the introverted DPs showed elevated social anxiety (see Fig. 3). Of all the DPs above the SIAS clinical cutoff, 76.9% were introverts (with a similar proportion of 75.0% in controls).

Fig. 3.

SIAS scores in introverted vs. extraverted DPs and controls. Note. SIAS = social interaction anxiety scale. DP groups defined by scores on the extraversion subscale of the big five inventory. DP, developmental prosopagnosia.

Next, using published SIAS item results of individuals diagnosed with SAD (Carleton et al., 2009⁴), we determined which SIAS items introverted DPs reached clinical levels (Fig. 4). We identified only three items where the clinical SAD group’s mean fell within the 95% confidence interval of introverted DP mean: Q6-I tense up if I meet an acquaintance on the street (M_DP-I = 1.85, 95% CI[1.49, 2.20]; M_Clinical = 2.15), Q9-I am at ease meeting people at parties, etc. (M_DP-I = 3.21, 95% CI[2.95, 3.47]; M_Clinical = 3.26) and Q20-I am unsure whether to greet someone I know only slightly (M_DP-I = 2.48, 95% CI[2.12, 2.84]; M_Clinical = 2.37). In contrast, introverted DPs showed substantially reduced social anxiety compared to the SAD group on several items: Q8-I feel tense if I am alone with just one other person (M_DP-I = 0.62, M_Clinical = 1.86), Q14-I have difficulty talking to attractive persons of the opposite sex (M_DP-I = 0.83, M_Clinical = 2.63), and Q15-I find myself worrying that I won’t know what to say in social situations (M_DP-I = 1.81, M_Clinical = 2.95). This shows that, in circumstances not requiring face recognition abilities, even the most socially anxious introverted DPs did not reach clinical levels of social anxiety. It is also noteworthy that both the analyses comparing DPs with controls and DPs with a clinical sample identified very similar items as being particularly heightened or reduced for DPs. This further emphasizes that social anxiety experienced by DPs can reach clinical levels but appears to be more situation specific rather than generalized to all social interactions.

Fig. 4.

SIAS item responses by extraversion and diagnostic group. Note. Items arranged in order of greatest to least introverted DP response. The clinical group represents a sample of 355 individuals with a diagnosis of social anxiety disorder (SAD; Carleton et al., 2009). No clinical data reported for item 5. Item numbers referred to in the text have been highlighted in boxes. DP, developmental prosopagnosia.

DP follow-up survey on disclosure, coping strategies, and social anxiety related behaviors

To investigate what could be driving the introvert/extravert DP social anxiety differences we observed, 48 DPs returned (35 female, M age = 43.70, SD = 15.08, 23 extraverts and 25 introverts) to take an optional short follow-up survey (see Supplementary material online, Table S5 for full survey results). The survey consisted of four sections: (1) disclosure of face recognition challenges, (2) coping strategies in social interaction, (3) responses to making a face recognition error, and (4) anticipation of social interaction. First, when averaging across all disclosure items, we found that introverted DPs reported lower rates of disclosure compared to extraverted DPs (t(46) = 3.30, p = .002, Cohen’s d = 0.95). Disclosure rates were consistently lower in introverted versus extraverted DPs across all 10 survey items (See Fig. 5), with the strongest effects observed for item 1.1 (my partner and/or immediate family; p = .003, q = 0.015, Cohen’s d = 0.92) and item 1.4 (my casual friends; p < .001, q = 0.010, Cohen’s d = 1.14). When examining disclosure as a continuous variable, we found a significant correlation between overall disclosure and extraversion (r = 0.47, p < .001), and a weaker correlation between disclosure and SIAS (r = −0.27, p = .060).

Fig. 5.

Follow-up survey disclosure item responses by extraversion group. Note. Items can found in Supplementary Methods (Table S1). Disclosure frequency = mean group response. Participants responded to each item on a five-point likert scale of “never”, “rarely,” “sometimes,” “often,” and “always.” responses were then scored 0–4, respectively. DP introvert/extravert groups defined by scores on the extraversion subscale of the Big Five Inventory. DP = developmental prosopagnosia.

There were no significant overall differences found on overall strategy use in social interactions (t(46) = 1.42, p = .158, Cohen’s d = 0.41), though introverted and extraverted DPs did differ on behaviors following making a face recognition error (t(46) = 3.00, p = .005, Cohen’s d = 0.87). In particular, introverted DPs reported a higher frequency of ruminating on an error for days or longer (p = .005, q = 0.012, Cohen’s d = 0.96), self-criticism in response to the event (p = .005, q = 0.015, Cohen’s d = 0.86), and worrying that they have offended someone or changed the way they are viewed (p = .009, q = 0.018, Cohen’s d = 0.81). We also found a significant difference in social anticipation behavior (t(46) = 2.62, p = .012, Cohen’s d = 0.76), with introverts showing greater consideration of avoiding social events due to their face recognition difficulties (p = .006, q = 0.018, Cohen’s d = 0.83) and increased worrying about potentially making recognition errors (p = .020, q = 0.030, Cohen’s d = 0.63).

DISCUSSION

The primary goal of this study was to better understand the prevalence and severity of social anxiety in DP with a secondary goal to better understand heterogeneity of social anxiety in DP. To this end, we administered a validated, self-report measure of social interaction anxiety (SIAS) to a large sample of DPs (N = 88) and typically developed controls (N = 58) as well as a battery of face recognition assessments, Autism Quotient questionnaire (AQ), and Big Five personality Inventory (BFI) to examine potential contributions to DPs’ social anxiety. We compared DP responses to both our control group and a large dataset of individuals diagnosed with SAD (Carleton et al., 2009) to further examine the clinical severity of the social anxiety experienced by DPs. Our findings demonstrate that, as a group, DPs experienced significantly higher levels of social anxiety than controls with 44% of our DP sample reporting SIAS scores at or above a probable cutoff for SAD. In contrast, only 14% of control participants reported SIAS scores reaching this threshold. Although DPs collectively exhibited elevated social anxiety, we observed significant heterogeneity, with not all DPs reporting elevated social anxiety. We found the most affected DPs were more introverted, neurotic, and had higher autism traits. Further, DPs’ higher SIAS scores appeared to be more driven by specific items pertaining to situations that could be affected by challenges with face recognition (e.g., meeting up with people at social gatherings). Together, these results provide both important insights into the potential consequences of DP beyond face recognition as well as clinical and treatment implications.

The current study is the first to quantify the prevalence and degree of social anxiety in DPs. The most prominent study of social anxiety in DP (Yardley et al., 2008) conducted semi-structured interviews with 25 DPs, and while they found all participants reported frequent social interaction difficulties, they did not quantify the severity or prevalence of these difficulties. The interviews and questionnaires administered by Murray et al. (2018) expanded on these reports by including a larger sample of 50 DPs, finding the vast majority report social challenges, including anxiety that leads to avoiding certain social situations and excessive worrying about potential embarrassment from mixing people up. However, the measures administered and qualitative analyses conducted in these studies did not quantify the severity or compare DP results to either normative or clinical samples. Our approach better quantified the degree of social anxiety in DPs compared with controls and identified its possible specificity to certain situations rather than being globally elevated across situations as observed in generalized social anxiety. We also showed that the prevalence of significant social anxiety was 44% of our DP sample, in addition to there being broad individual differences across DPs.

Perhaps our most novel finding is the large individual difference in the degree of social anxiety experienced by DPs, a surprising result considering that previous studies have suggested most DPs suffer from social anxiety (Dalrymple et al., 2014; Murray et al., 2018; Yardley et al., 2008). Although the mean DP SIAS score was nearly twice that of controls, we found that only 44% of DPs reported scores at or above the suggested cutoff for probable SAD, leaving over half with scores that would not be indicative of clinical or even particularly elevated levels of social anxiety. Although only one study has highlighted a small proportion of individuals with impaired face recognition that reported aspects of social adeptness rather than dysfunction (Svart & Starrfelt, 2022), our findings suggest a wide range of social anxiety experienced by those with face recognition deficits: some who may experience anxiety levels comparable to those with SAD, others who may not experience social anxiety at all, and many who fall somewhere in between. This heterogeneity differs from ASD, a population with much higher prevalence rates of comorbid social anxiety (e.g., 72%; Boulton & Guastella, 2021).

Importantly, we were able to investigate the factors associated with individual differences in social anxiety, helping us to better understand who may struggle the most. We did not find an association between objective face recognition performance and social anxiety in DPs or controls. There was a correlation between self-reported face recognition difficulties and SIAS scores in DPs and controls, though this is likely due to the survey containing a few items related to social anxiety. Although the ability to detect and process information from faces may be critical to social functioning (Avery et al., 2016), we only found the effect of face recognition on social anxiety to exist between groups, not within DPs or controls. This suggests that the severity of DPs’ face recognition deficits is not associated with the severity of their social anxiety symptoms. We did, however, find evidence to support the claim that autism traits are associated with higher social anxiety in both DPs and controls. The relationship between autism and social anxiety has been studied extensively, with overlapping symptoms and high rates of comorbidity between the disorders (Simonoff et al., 2008; White et al., 2009). Our finding that DPs with higher AQ scores reported higher SIAS scores is in line with this literature as well as previous studies demonstrating elevated autism traits in samples with only a primary diagnosis of SAD (Cath et al., 2008; Puleo & Kendall, 2011). Although we found autism traits to correlate with social anxiety in both DPs and controls, we found it predicted unique variance in social anxiety only in DPs. This could be due to a more restricted range of AQ scores in our control group, or perhaps autism traits have a greater effect on social anxiety in DP due to the increased frequency of awkward or uncomfortable situations DPs encounter.

In DPs and controls, personality was found to be an even stronger predictor of social anxiety than autism traits, with extraversion specifically showing the strongest association with DPs’ social anxiety severity. Although introverted DPs displayed elevated levels of social anxiety, even reaching clinical levels of anxiety on certain SIAS items, extraverted DPs were highly similar to controls. It is not surprising that introversion was predictive of social anxiety, as extensive research has shown there to be a relationship between the two (Bienvenu et al., 2001; Costache et al., 2020; Heiser et al., 2003). It is, however, notable that extraversion was the most predictive of any variable when it comes to social anxiety in DPs and was a unique predictor in DPs but not controls, a result that begs the question of why extraversion has such a strong relationship with social anxiety in DP. It is possible that extraversion is protective against social anxiety, as it has shown to be associated with less rumination and self-criticism (Naragon-Gainey & Watson, 2012; Stroud et al., 2015). It may be the case, however, that extraverted DPs exhibit specific behaviors that help to decrease social anxiety. From our follow-up coping strategy survey, we found extraverts to be significantly more open about their DP diagnosis, disclosing more to family members, partners, and casual friends, as well as to strangers. It is possible that this disclosure alleviates some anxiety regarding fear of judgment, embarrassing oneself, or offending others due to face recognition errors. We also found extraverts to be less self-critical, avoid social interaction less frequently, and ruminate for shorter periods of time after making a face recognition error, all of which may contribute to less anxiety in social situations. It may also be the case that being comfortable disclosing face recognition difficulties is an effect of lower social anxiety rather than extraversion, though disclosure correlated much higher with extraversion than SIAS. In any case, it is important to note that though some of these approaches/behaviors could be suggested to introvert DPs to help reduce social anxiety, for some it could be challenging to implement. For example, some introverted DPs may find it more distressing to disclose their prosopagnosia to casual friends than to not. Still, it may be beneficial for DPs that experience distress surrounding disclosure to practice disclosing in the settings our SIAS item analysis highlighted as being the least anxiety-inducing, such as one-on-one interactions.

Evaluating a large sample of DPs on the SIAS also allowed us to evaluate whether the social anxiety experienced by DPs would reach the level of individuals with SAD. To be diagnosed with SAD, one must demonstrate persistent, intense fear or anxiety about specific social situations or fear of being judged negatively, embarrassed, or humiliated (DSM-5, American Psychiatric Association [APA], 2013). This fear/anxiety must lead to avoidance of social situations (or enduring them with intense distress), be out of proportion for the given situation, interfere with daily living, and cannot be better explained by another medical condition. Our results, along with reports from Yardley et al. (2008) and Murray et al. (2018), suggest that many DPs have some features of SAD. However, DPs mostly endorsed items that were specific to the challenging social situations they could encounter rather than more broadly endorsing items across the SIAS scale, like the SAD individuals. For this reason, DPs should not be regarded as having generalized SAD. DPs’ higher SIAS scores could also be in part due to the measure itself, as it includes many items about meeting acquaintances or socializing in group settings, in contrast to other self-report social anxiety measures (e.g., LSAS; Liebowitz, 1987). In sum, as was similarly concluded in Yardley et al. (2008), DP appears to be a risk factor for developing some key features of SAD, though a SAD diagnosis is not appropriate in most cases due to DPs’ social dysfunction/distress being explained by severe face recognition deficits.

Even so, it is clear that many DPs suffer from elevated levels of social anxiety, particularly compared to typically developing controls. Experiencing elements of SAD, whether or not the anxiety warrants a clinical diagnosis, still can have major affects on one’s well-being. Previous research has found that those who experience some features of SAD without reaching a clinical threshold still have much poorer mental and physical health compared to the general population (Davidson et al., 1994). In fact, their health looked to be closer to that of a population with diagnosed SAD. Further, the social avoidance that often results from even mild levels of social anxiety is associated with adverse life outcomes, including early mortality (Fiori et al., 2006; Holt-Lunstad et al., 2010; Seppala et al., 2013). With social anxiety being associated with increased risk of psychopathology, such as anxiety and affective disorders, as well as substance abuse and overall poorer quality of life (Wong et al., 2012), it is critical to address these elevated levels among DPs. Elevated social anxiety in DPs may in fact be key to the high comorbidity rates between prosopagnosia, depression, and generalized anxiety reported in Pressl et al. (2021). Future work must further examine the broader effects of DPs’ social anxiety as well as potential intervention techniques targeted to DPs with severe social anxiety (e.g., group therapy, bringing together often isolated DPs to share their experiences). We can also develop therapeutic interventions based on the coping strategies or mindsets of DPs that do not report elevated social anxiety levels (e.g., implement cognitive-behavioral techniques, Butler et al., 2006, of reducing rumination and self-criticism). Our results provide a strong justification for further investigating the effects of disclosing one’s prosopagnosia diagnosis. Future work should experimentally test the effects of disclosure among DPs as a potential treatment for social anxiety, as well as what may influence comfort levels regarding disclosure. There is also variation in how long our participants have been aware of prosopagnosia as a condition. It is possible that participating in our study or obtaining an official diagnosis influences DPs’ self-image and level of comfort in disclosing their face recognition challenges, as several of our participants have noted that knowing they have a diagnosis of prosopagnosia has unburdened them greatly (see Dingfelder, 2019). A potential long-term aim could be to explore changes in social anxiety and coping strategies after diagnosis or participation in prosopagnosia studies.

Though the results of the current study are compelling, they also have limitations. First, to participate in our study, participants must overcome barriers that could be very uncomfortable or anxiety-inducing for someone with severe social anxiety, including reaching out to our lab via email, talking over the phone for screening, and completing assessments with research personnel, either in-person or over Zoom. Our sample is therefore likely composed of individuals who are not as socially anxious, with the most severely or generally socially anxious DPs perhaps not accounted for. We also administered only one validated self-report measure of social anxiety to our sample. It would be important to further explore DPs’ social anxiety on multiple scales that probe a broader array of situations and allow for additional comparisons to both control and clinical populations. Additionally, although the current study used the AQ as a brief measure to gain insight into general autism traits, it would be important for future work to explore how social anxiety manifests in DPs with well-characterized autism. A study using a sample of DPs with proper ASD diagnoses (e.g., made with an ADOS/ADI, by a multidisciplinary team, etc.) would provide a more accurate picture of the interaction between the disorders beyond what we can learn from the AQ. Finally, we were only able to compare DP responses to clinical group means rather than individual data points. Without having access to individuals’ data, we are not able to precisely calculate group SIAS differences. Comparing DPs to a full dataset of both control and clinical populations would be an important future direction.

In sum, using a large sample of 88 DPs, we found that DPs experience elevated social anxiety that is more situation specific than generalized, with personality being a major determinant of symptom severity. Notably, there was a great deal of heterogeneity in social anxiety reported by DPs, with those most affected demonstrating higher autism traits and lower extraversion. DPs with the lowest reported levels of social anxiety also reported more open disclosure of face recognition deficits. These results better quantify the potential serious psychosocial affects of DP while highlighting the importance of individual differences and developing targeted interventions.

FUNDING

This work was supported by the National Eye Institute [R01EY032510] awarded to J.D.

CONFLICT OF INTEREST

The authors report there are no competing interests to declare.

AUTHOR CONTRIBUTIONS

Leah Kirsch (Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Writing—original draft, Writing—review & editing), Travis C Evans (Conceptualization, Writing—review & editing), Regan Fry (Conceptualization, Data curation, Investigation, Methodology, Writing - review & editing), Alison Campbell (Visualization, Writing—review & editing), Joseph DeGutis (Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing—original draft, Writing—review & editing).