Shame and loneliness in social anxiety disorder: A comparison with panic disorder

Abstract

Shame and loneliness are two important emotions that have been suggested to play a maintaining role in social anxiety disorder (SAD). However, it is not clear whether emotional processes involving these emotions are predominantly experienced in SAD or shared with other anxiety disorders. In the present study we examined the temporal relationship between shame and loneliness on one hand and anxiety on the other among individuals with SAD and among individuals with panic disorder (PD). We used a daily diary design in which participants reported their emotional experiences every evening for 21 consecutive days. We found that individuals with SAD had higher levels of shame but not higher levels of loneliness compared to individuals with PD. We also found evidence for a bidirectional relationship among individuals with SAD, in which anxiety on a given day predicted shame and loneliness on the following day, and shame and loneliness on given day predicted anxiety on the following day. This relationship was not found for individuals with panic disorder. Our findings are consistent with a negative cycle of emotions that may play a role in SAD. Implications for emotional models of psychopathology and for treatment of SAD are discussed.

Social anxiety disorder (SAD) is a prevalent and impairing mental health condition characterized by an intense and persistent fear of being negatively evaluated by others, leading to significant distress and impairment (American Psychiatric Association, 2013). Lifetime prevalence estimates suggest that SAD affects approximately 12.1 % of the population, making it one of the most common anxiety disorders (Kessler et al., 2005). Individuals with SAD frequently experience difficulties in multiple life domains, including academic and occupational settings, where social demands are unavoidable (Aderka et al., 2012). Moreover, SAD has been associated with elevated risk for comorbid mood and substance use disorders (Stein & Stein, 2008) and increased suicidal ideation and behavior (Sareen et al., 2005). Consequently, SAD constitutes a significant clinical concern with far-reaching negative consequences across personal and societal levels.

Shame has been identified as a characteristic emotional experience in individuals with SAD, and has received considerable empirical and theoretical attention. Shame is a self-conscious negative emotion characterized by a global, negative self-assessment (e.g., seeing one’s self as flawed or inadequate), and triggered by a potential exposure of the flaws to others. This emotional state typically manifests in a desire to hide, escape, or disappear, reflecting a defensive response to social evaluation (Tangney & Dearing, 2002). Unsurprisingly, this emotion becomes particularly salient in social contexts, in which others can evaluate the self and flaws can be revealed. Because individuals with SAD often anticipate negative evaluation and interpret ambiguous social cues as signs of personal failure, they are especially vulnerable to experiencing shame (Clark & Wells, 1995; Heimberg et al., 2010). Thus, shame serves as a core emotion that plays an important role in SAD (Swee et al., 2021).

A growing body of research empirically supports the link between shame and SAD. In a systematic review, Swee et al. (2021) found robust evidence that individuals with SAD report elevated shame, particularly in social-evaluative contexts, compared to nonclinical samples. Moreover, shame has been shown to predict higher social anxiety severity and avoidance behaviors in SAD (Hedman et al., 2013). Experience sampling data also indicate that in SAD, shame and anxiety reinforce each other over time, creating a reciprocal cycle of emotional distress that is not observed in individuals without the disorder (Oren-Yagoda et al., 2024). Taken together, these findings suggest that shame may not only co-occur with SAD but may also play a key role in maintaining social anxiety.

Loneliness is another emotion that has received increasing attention in SAD. Loneliness refers to the perceived discrepancy between desired and perceived social relationships (Peplau & Perlman, 1982). Unlike objective social isolation, loneliness reflects subjective emotional dissatisfaction with social connectedness (Hawkley & Cacioppo, 2010). As a result, it is associated with various psychological and physiological difficulties, including sleep disturbances, depressive symptoms (Hawkley & Cacioppo, 2010), and even increased mortality (Holt-Lunstad et al., 2015). Importantly, loneliness can also serve as a maintaining factor in psychiatric conditions such as SAD, as individuals with SAD frequently avoid social contact due to fear of rejection or embarrassment (Alden & Taylor, 2010). The resulting social disconnection creates fertile ground for a self-perpetuating cycle: avoidance maintains loneliness, which in turn exacerbates social anxiety symptoms (Cacioppo et al., 2014). Notably, this pattern appears more pronounced in SAD than in other anxiety disorders, as individuals with SAD demonstrate greater loneliness even when controlling for general anxiety symptoms (Teo et al., 2018).

A number of empirical studies have examined loneliness in SAD. For instance, individuals with SAD were found to have elevated loneliness compared to nonclinical controls (Oren-Yagoda et al., 2022). Furthermore, longitudinal data have supported the existence of a reciprocal relationship, in which loneliness and social anxiety predict one another over time (Lim et al., 2016). Similarly, experience sampling research has shown that for individuals with SAD, anxiety and loneliness tend to co-occur and reinforce each other on a daily basis, forming a maladaptive emotional cycle. This cyclical pattern was not observed in individuals without SAD (Oren-Yagoda et al., 2022).

Despite these insights, there are still gaps in the literature on shame and loneliness in SAD. First, previous studies have examined these emotions among either nonclinical samples, or have compared clinical samples of individuals with SAD to control groups of individuals without SAD. More broadly, although shame has been widely studied in the context of SAD, to date, no studies have directly compared levels of shame between individuals with SAD and those with other anxiety disorders. Similarly, the literature currently lacks studies examining loneliness across anxiety disorders. Such comparisons are essential for understanding the degree of specificity of emotions and emotional processes to SAD. Second, research on the temporal dynamics of shame and loneliness in SAD is still limited. Only a few studies have examined temporal emotional processes in which shame or loneliness affect anxiety over time in SAD (Oren-Yagoda et al., 2022; Oren-Yagoda et al., 2024) and no study that we are aware of has compared temporal processes in two anxiety disorders. This is important as it can help us understand whether an emotional process is predominantly experienced in SAD or is shared across anxiety disorders.

The present study addresses these gaps by comparing individuals with SAD to those with panic disorder (PD) using a 21-day daily diary methodology. This approach allows for the examination of emotional experiences in daily life, capturing both between-group differences and temporal relationships between emotions. Furthermore, by directly comparing individuals with SAD to those with panic disorder, this study enables a more precise understanding of whether shame and loneliness are predominantly linked to SAD or reflect broader transdiagnostic patterns.

We chose PD as the comparison group for two main reasons. First, shame has been found to be central in SAD but not in PD. Specifically, there is a large literature linking shame to social anxiety disorder (see Swee et al., 2021 for a review), but this link has not been documented in PD. Moreover, studies examining shame and its associations with different anxiety disorders have found strong associations of shame with SAD and generalized anxiety disorder (GAD), but not with PD (e.g., Fergus et al., 2010; Szentágotai-Tătar et al., 2020). There have also been studies showing that shame may play a maintaining role in social anxiety (e.g., Oren-Yagoda et al., 2024) but no such studies were found for PD. Thus, whereas both SAD and PD are anxiety disorders, shame may play a larger role in the former than the latter. Second, loneliness has been found to be associated with SAD and its maintenance more than has been found in PD. Specifically, in one study, individuals with SAD reported elevated levels of loneliness above and beyond general anxiety (Teo et al., 2018). In two additional studies, loneliness and social anxiety predicted each other significantly over time above and beyond cross-sectional and stability effects (Lim et al., 2016; Oren-Yagoda et al., 2022). Moreover, there are theoretical accounts of loneliness as a maintaining factor in social anxiety (Cacioppo et al., 2014). In PD, there is a single study that found an association between loneliness and PD in an epidemiological sample (Meltzer et al., 2013). However, there is no temporal evidence for loneliness as maintaining PD, nor theoretical models (e.g., it is not obvious that loneliness would increase panic attacks as panic attacks are usually triggered by highly arousing states). Thus, loneliness seems to play a more significant role in SAD compared to PD.

The current study tested three preregistered hypotheses (https://osf.io/gmn9d/overview?view_only=bf47924209be4764836f7ed37a9f92be). First, we hypothesized that individuals with SAD would report significantly higher levels of shame and loneliness than individuals with PD (H1). This prediction was grounded in prior evidence of heightened shame and loneliness in SAD and the theoretical significance of these emotions to SAD (Oren-Yagoda et al., 2022, 2024; Swee et al., 2021). Second, we hypothesized that for individuals with SAD, shame and anxiety would exhibit reciprocal temporal associations. Specifically, shame on a certain day was expected to predict greater anxiety on the following day, and vice versa (H2). We did not expect this pattern to emerge among individuals with PD (this is not a formal hypothesis but rather an expectation as one cannot formally hypothesize a null finding). Third, we hypothesized that for individuals with SAD, loneliness and anxiety would exhibit reciprocal temporal associations. Specifically, loneliness on a certain day was expected to predict greater anxiety on the following day, and vice versa (H3). We similarly did not expect this pattern to emerge among individuals with PD.

Method

Participants

The sample included 73 participants and consisted of two groups: 44 participants who met diagnostic criteria for SAD and 29 participants who met diagnostic criteria for PD. On average, participants were 29.56 years old (SD = 9.52) with 42 women (57.5 %) and 31 men (42.5 %). The majority of participants reported studying as their main occupation (35 participants; 47.9 %), followed by work (23 participants; 31.5 %), and being unemployed (15 participants; 20.5 %). In terms of marital status, the majority of our sample were single (n = 59, 80.8 %), 9 were married (12.3 %), and 5 were divorced (6.8 %). Table 1 presents a comparison of demographic and clinical measures between the groups.

Table 1.

Demographic and clinical measures.

Varaiable	SAD (N = 44)	PD (N = 29)	Statistic	P
Age	28.89 (9.87)	30.59 (9.03)	t(71)=−0.74	.46
Gender			χ²(1) = 2.57	.11
Female	22 (50.0 %)	20 (69.0 %)
Male	22 (50.0 %)	9 (31.0 %)
Marital status			χ²(2) = 5.17	.08
Married	4 (9.1 %)	5 (17.2 %)
Divorced	1 (2.3 %)	4 (13.8 %)
Single	39 (88.6 %)	20 (69.0 %)
Romantic relationship	21 (47.7 %)	15 (51.7 %)	χ²(1) = 0.11	.74
Occuopational status			χ²(2) = 5.52	.06
Student	26 (59.1 %)	9 (31.0 %)
Employed	11 (25.0 %)	12 (41.4 %)
Unemployed	7 (15.9 %)	8 (27.6 %)
BDI-II	24.75 (11.86)	26.03 (13.31)	t(71)=-0.43	.67
SPIN	46.75 (10.59)	29.17 (14.54)	t(71)=5.97	001.>

Note. SAD = Social anxiety disorder; PD = Panic Disorder; BDI-II = Beck depression inventory – II; SPIN = Social Phobia Inventory.

Within the SAD group, comorbid diagnoses included major depressive disorder (n = 19, 43.2 %), panic disorder (n = 2, 4.5 %), obsessive–compulsive disorder (n = 2, 4.5 %), and generalized anxiety disorder (n = 5, 11.9 %). In the PD group, comorbid diagnoses included major depressive disorder (n = 11, 37.9 %), social anxiety disorder (n = 2, 6.9 %), obsessive-compulsive disorder (n = 2, 6.8 %), and generalized anxiety disorder (n = 3, 10.3 %).

To participate in the study, individuals were required to have a smartphone with an active data plan, as daily assessments were completed via mobile devices. Participants were not eligible to participate if they (a) were currently receiving psychological or psychiatric treatment for anxiety or depression, (b) endorsed frequent thoughts of suicide (defined as a score of 2 or 3 on item 9 of the BDI-II; Beck et al., 1996), or (c) showed evidence of psychotic symptoms according to section K of the Mini-International Neuropsychiatric Interview (MINI; Sheehan et al., 1998). These criteria were chosen to ensure that the emotional functioning observed in daily life reflected the untreated state of the disorders under investigation. Including participants who were undergoing treatment could have introduced uncontrolled variability stemming from the effects of therapy or medication. Moreover, including individuals with severe suicidality or psychosis would have presented ethical and clinical challenges, as these individuals typically require therapeutic intervention that falls outside the scope of the current research.

Procedure

Recruitment was conducted through online advertisements and social media platforms. Individuals interested in participating completed a number of online screening questions and individuals who met the inclusion criteria were contacted by telephone. A total of 298 individuals were contacted by telephone, and during the call, potential participants were informed about the study, and those interested were scheduled for a diagnostic interview. In total, 142 individuals attended the diagnostic interview and completed informed consent forms.

During the diagnostic session, participants completed a series of self-report questionnaires and a structured clinical interview conducted by trained graduate students in clinical psychology. Interviews were supervised by a licensed clinical psychologist with expertise in anxiety disorders. Diagnoses of SAD and PD, as well as other comorbid anxiety and mood disorders, were established using the Anxiety and Related Disorders Interview Schedule for DSM-5 (ADIS-5; Brown & Barlow, 2014). Psychotic symptoms and suicidality were further assessed using the MINI (Sheehan et al., 1998). Clinical severity ratings ranging from 0 to 8 were assigned for each diagnosis, with a score of 4 or above considered to meet threshold for diagnosis.

Following the diagnostic session, participants completed a 21-day daily diary protocol. Each evening at 8:00 PM, participants received a text message with a link to an online survey. They were instructed to complete the survey by 11:00 PM. The daily diary assessed emotional experiences and social interactions during the past 24 h. Each daily survey took approximately six minutes to complete.

Measures

Self-report measures (administrated during the diagnostic session)

Social anxiety symptoms

The Social Phobia Inventory (SPIN; Connor et al., 2000) was used to assess social anxiety. It is a 17-item self-report measure rated on a 5-point Likert scale (0 = not at all to 4 = extremely), encompassing three domains: fear (e.g., social situations, negative evaluation, embarrassment), avoidance (of social interactions and performance situations), and physiological symptoms (e.g., blushing, sweating, heart palpitations). The SPIN demonstrated high internal consistency in the present study (α = 0.97).

Panic disorder symptoms

The Panic Disorder Severity Scale – Self-Report (PDSS-SR; Houck et al., 2002) was used to assess the severity of panic disorder. It is a 7-item self-report measure rated on a 5-point Likert scale (0 = none to 4 = extreme), assessing the frequency and distress of panic attacks, anticipatory anxiety, avoidance of situations and physical sensations, and impairment in work and social functioning. The measure has demonstrated good internal consistency in the present study (α = 0.79).

Depressive symptoms

The Beck Depression Inventory – II (BDI-II; Beck et al., 1996), was used to assess the severity of depressive symptoms. The BDI-II is a widely used 21-item questionnaire, rated on a 4-point Likert scale (0 = not at all to 3 = severe). In the current sample, internal consistency was high (α = 0.92).

Daily diary measures (administered daily for 21 consecutive days)

Emotions

Each evening over the 21-day period, participants reported on their emotional experiences for that day. A modified version of the Positive and Negative Affect Schedule – Expanded Form (PANAS-X; Watson & Clark, 1994) was used for this purpose. Participants rated the intensity of eight positive emotions (e.g., excitement, calmness) and eight negative emotions (e.g., shame, anxiety) on a 6-point scale ranging from 0 (“not at all”) to 5 (“extremely”). For the present study, analyses focused on the negative emotions of loneliness, anxiety, and shame.

Analytic strategy

All analyses were conducted using hierarchical linear modeling (HLM) to account for the nested structure of repeated daily assessments (Level-1) within participants (Level-2). All models included random intercepts, and random slopes were specified when Level-1 predictors varied across time. Models were estimated using restricted maximum likelihood (REML) with an unstructured covariance structure, which makes no a priori assumptions about the data (Tabachnick & Fidell, 2018). For temporal associations (Hypotheses 2 and 3), lagged models were used, in which the predictor variable from day t was used to predict the outcome variable on day t + 1. Categorical independent variables were effects-coded to facilitate interpretation of coefficients. We initially ran models with random slopes for all independent variables. However, as the variance of slopes was nonsignificant in all models, we removed these random slopes based on common guidelines (see Tabachnik & Fidell, 2018).

Results

Missing data

The present study included 73 individuals who completed a daily diary for 21 days (1533 total measurements). Of these measurements, 90 (5.9 %) were missing. Thus, participants completed 94.1 % of their daily diary measurements. Little’s Missing Completely at Random (MCAR) test was nonsignificant, indicating that the pattern of missing data was not significantly different from a random distribution of missingness (χ² = 2.73, df = 3, p = .44). Considering the small amount of missing data, and the lack of a systematic bias (i.e., data were MCAR), we based our analyses on complete data (i.e., 1443 measurements).

H1: group differences in shame and loneliness

To examine differences in shame and loneliness between individuals with SAD and PD, we estimated two hierarchical linear models (HLM) in which the dependent variables were shame and loneliness (both Level-1 continuous variables). The independent variable in both models was group (SAD vs. PD; Level-2 categorical variable). We found a significant effect of group on shame (B = −0.268, SE = 0.109, t = −2.468, df = 74.675, p = .016, 95 % CI = −0.485 to −0.052). Specifically, individuals with SAD reported higher levels of shame (M = 2.56, SD = 1.39) compared to individuals with PD (M = 1.99, SD = 1.28). In contrast, there was no significant group effect on loneliness (B = −0.128, SE = 0.140, t = −0.917, df = 74.772, p = .362, 95 % CI = −0.407 to 0.150), indicating that levels of loneliness did not significantly differ between the SAD (M = 2.71, SD = 1.57) and PD (M = 2.48, SD = 1.52) groups albeit being in the direction of the hypothesis (see Fig. 1). Thus, we found partial support for our first hypothesis.

Fig. 1.

Estimated means of shame and loneliness among individuals with SAD and PD.

Note. SAD = Social Anxiety Disorder; PD = Panic Disorder; n.s. = non-significant; error bars represent standard errors. * p < .05.

H2: prospective reciprocal relations between shame and anxiety

To examine reciprocal relations between shame and anxiety we estimated an HLM model in which anxiety at time t was the independent variable (Level-1 continuous variable). The dependent variable was shame at time t + 1. Put differently, we examined whether anxiety on a given day predicted shame on the following day along the course of our 21 daily measurements. We also examined the reverse model in which shame at time t was the independent variable (Level-1 continuous variable) and anxiety at time t + 1 was the dependent variable. Put differently, we examined whether shame on a given day predicted anxiety on the following day along the course of our 21 daily measurements. We ran both of these models for individuals with SAD and with PD separately.

For individuals with SAD, we found that anxiety at time t significantly predicted shame at time t + 1 (B = 0.098, SE = 0.034, p = .004, 95 % CI = 0.032 to 0.165). This effect was positive indicating that the more anxiety experienced at time t, the more shame was experienced at time t + 1. We also found that shame at time t significantly predicted anxiety at time t + 1 (B = 0.082, SE = 0.038, p = .031, 95 % CI = 0.008 to 0.156). This effect was positive indicating that the more shame experienced at time t, the more anxiety was experienced at time t + 1. Taken together, these analyses suggest that for individuals with SAD, a deleterious cycle emerges between anxiety and shame, such that each emotion predicts the other over time. This is in line with our second hypothesis (Fig. 2).

Fig. 2.

Prospective reciprocal relations.

For individuals with PD, we found that anxiety at time t did not significantly predict shame at time t + 1 (B = 0.018, SE = 0.035, p = .600, 95 % CI = −0.050 to 0.087). In addition, when examining the reverse model, we found that shame at time t did not significantly predict anxiety at time t + 1 (B = 0.096, SE = 0.052, p = .064, 95 % CI = −0.006 to 0.198). Taken together, these analyses suggest that for individuals with PD, neither anxiety nor shame predict subsequent changes in each other (Fig. 2) .1

H3: prospective reciprocal relations between loneliness and anxiety

To explore the reciprocal relationship between loneliness and anxiety, we estimated an HLM model where anxiety at time t served as the independent variable (Level-1 continuous variable), and loneliness at time t + 1 was the dependent variable. In other words, we tested whether anxiety on a given day predicted loneliness on the following day across our 21 daily measurements. We also examined the reverse model, where loneliness at time t was the independent variable (Level-1 continuous variable), and anxiety at time t + 1 was the dependent variable. Put differently, we tested whether loneliness on a given day predicted anxiety on the subsequent day throughout the same period. We ran both models separately for individuals with SAD and PD.

For individuals with SAD, we found that anxiety at time t significantly predicted loneliness at time t + 1 (B = 0.102, SE = 0.034, p = .003, 95% CI = 0.035 to 0.169). This positive effect indicates that experiencing more anxiety at time t is associated with increased loneliness at time t + 1. Additionally, we found that loneliness at time t significantly predicted anxiety at time t + 1 (B = 0.074, SE = 0.038, p = .049, 95% CI = 0.001 to 0.149), such that greater loneliness at time t was linked to higher anxiety on the next day. Overall, these findings indicate that for individuals with SAD, anxiety and loneliness predict each other as part of a cycle of negative emotions that may play a role in SAD. This is in line with our third hypothesis (Fig. 2).

For individuals with PD, we found that anxiety at time t did not significantly predict loneliness at time t + 1 (B = 0.035, SE = 0.040, p = .376, 95% CI = −0.043 to 0.113) nor did loneliness at time t significantly predict anxiety at time t + 1 (B = 0.074, SE = 0.045, p = .101, 95% CI = −0.015 to 0.162). Overall, these analyses suggest that for individuals with PD, neither anxiety nor loneliness predict subsequent changes in each other (Fig. 2).

Additional analyses – depression

To ensure our findings were not due to depression, we conducted a number of additional analyses. First, we compared levels of depression between the SAD and PD groups and found that they were not significantly different (t₍₇₁₎=−0.43, p = .67). Specifically, the average BDI score for the SAD group was 24.75 (SD = 11.86) and the average BDI score for the PD group was 26.03 (SD = 13.31). Similarly, in the SAD group, 43.2 % of individuals had MDD, and in the PD group 37.9 % of individuals had MDD. This difference was not statistically significant (χ²₍₁₎ = 0.19, p = .66). Thus, both groups had comparable levels of depression and MDD. This suggests that differences between the groups found in hypotheses 2 and 3 cannot be explained by levels of depression and/or MDD (as both groups have similar levels). In addition, we examined the correlations between BDI scores and levels of shame and loneliness in the SAD and PD groups. We found that all correlations were small in magnitude and nonsignificant (r = 0.13 to 0.32, all ps > 0.05).

Additional analyses – cross-diagnoses

Two individuals from the SAD group had comorbid PD and two individuals from the PD group had comorbid SAD. Thus, we repeated the analyses while excluding these individuals who had cross-diagnoses. We found that the pattern of findings was identical and evidence for a bi-directional relationship between shame/loneliness and anxiety emerged only in the SAD group.

Additional analyses – concurrent associations

One potential alternative explanation for our findings is that the prospective associations found (for the SAD group but not the PD group) were simply the result of differences between the two groups in concurrent associations. That is, the SAD group could potentially have stronger concurrent associations (i.e., when both emotions are measured on the same day) compared to the PD group. If this is the case then the concurrent associations could potentially lead to our temporal findings (in which one emotion predicts the other on the following day). To examine this alternative explanation, we conducted additional analyses that focus on concurrent associations between emotions, and on group differences in concurrent associations.

We found evidence for a concurrent positive association between loneliness and anxiety (B = 0.239, SE = 0.030, df = 62.176, t = 8.011, p < .001, 95% CI = 0.179 to 0.299) as well as a concurrent positive association between shame and anxiety (B = 0.229, SE = 0.030, df = 37.533, t = 7.709, p < .001, 95% CI = 0.169 to 0.289). Importantly, neither the concurrent association between loneliness and anxiety (B = 0.002, SE = 0.030, df = 62.176, t = 0.080, p = .937, 95% CI = −0.057 to 0.062) nor the concurrent association between shame and anxiety (B = −0.053, SE = 0.030, df = 37.533, t = 1.795, p = .081, 95% CI = −0.113 to 0.007) significantly differed between the SAD and PD groups. This suggests that the concurrent associations are unlikely to have accounted for the prospective associations reported in H2 and H3 (as the concurrent associations are similar for the two groups).

Discussion

The present study aimed to examine shame and loneliness among individuals diagnosed SAD and PD, using a 21-day daily diary design in which participants reported their emotional experiences every evening for 21 consecutive days. This allowed us to examine both overall levels of shame and loneliness in each group, as well as the day-to-day, bidirectional relationships between these emotions and anxiety. The study was designed to test three main hypotheses: (1) that individuals with SAD would report higher average levels of shame and loneliness than individuals with PD; (2) that in SAD, shame and anxiety would predict one another across days, forming a reciprocal dynamic not observed in PD; and (3) that a similar reciprocal pattern would emerge between loneliness and anxiety in SAD, but not in PD. The results partially supported Hypothesis 1 and fully supported Hypotheses 2 and 3.

Consistent with our first hypothesis, individuals with SAD reported significantly higher levels of shame compared to those with PD. This finding aligns with prior research highlighting shame as a central emotion in the experience and maintenance of SAD (Hedman et al., 2013; Swee et al., 2021). Shame in SAD is thought to arise from heightened concerns about negative evaluation, internalized self-criticism, and the perceived exposure of flaws in social contexts (Clark & Wells, 1995; Tangney & Dearing, 2002). In contrast, individuals with PD tend to experience anxiety in response to internal bodily sensations that are interpreted as dangerous or life-threatening (McNally, 2002). These episodes typically involve less self-conscious social evaluation that often triggers shame (Saboonchi et al., 1999). Alternatively, it is possible that individuals with PD experience shame as a result of their panic attacks (especially those that occur in public or while in presence of others) but those panic attacks do not occur as often as social interactions occur for individuals with SAD, and this may lead to lower average levels of shame among individuals with PD.

In contrast to our first hypothesis, no significant difference in loneliness emerged between individuals with SAD and those with PD. Thus, as opposed to our expectation that loneliness would be elevated in SAD but not in PD, the finding suggests that loneliness may be similarly elevated in both disorders. One possible reason for this finding is that different processes contribute to loneliness in the two disorders. In SAD, loneliness may stem from a sense of disconnection from others and fears of social inadequacy, driven by chronic avoidance of social situations (Alden & Taylor, 2010; Oren-Yagoda et al., 2022). In PD, on the other hand, social withdrawal is often associated with agoraphobia, a disorder commonly comorbid with panic disorder (Palardy et al., 2018). Agoraphobia involves anxiety about being in places or situations where escape might be difficult or help is unavailable in the event of a panic attack, frequently resulting in complete or partial avoidance of these situations (APA, 2013). Thus, although the underlying mechanisms may differ, both groups could be vulnerable to loneliness due to their avoidance. These findings highlight the need to further explore the processes leading to loneliness in different anxiety disorders.

In support of our second hypothesis, individuals with SAD demonstrated significant reciprocal associations between shame and anxiety over time. Specifically, anxiety on a given day predicted shame on the following day, and shame on a given day predicted anxiety on the following day. These findings replicate and extend prior experience sampling research (Oren-Yagoda et al., 2024), and indicate that a mutually reinforcing dynamic between these emotions in SAD can be found using daily diary methodology as well. This reinforcing cycle may reflect the fact that individuals with SAD often experience shame about their own anxiety symptoms, particularly when those symptoms are visible or may attract negative attention from others (Heimberg et al., 2010). For example, physiological symptoms such as blushing or trembling may be interpreted by individuals with SAD as evidence of social failure, thereby triggering shame. That shame, in turn, can increase one’s perceived inferiority and ineptitude, thus leading to anxiety about future social interactions, and perpetuating the cycle.

Consistent with our second hypothesis, no bidirectional relationship between shame and anxiety was found among individuals with PD. Thus, for individuals with PD, neither anxiety nor shame on a given day predicted the other emotion on the following day. This supports the notion that shame may not play a role in the maintenance or exacerbation of anxiety in PD as it does in SAD. One potential reason for this is that in PD, anxiety is primarily linked to catastrophic misinterpretations of physical sensations (e.g., impending insanity, fear of an imminent heart attack; McNally, 2002). In such moments, immediate survival concerns may be pronounced, potentially overriding social or self-conscious emotions such as shame. Alternatively, it is possible that individuals with PD experience shame as a result of their panic attacks (especially those that occur in public or while in presence of others) but those panic attacks do not occur as often as social interactions occur for individuals with SAD, and this may lead to lower average levels of shame among individuals with PD.

Our third hypothesis was also supported. Specifically, in the SAD group, loneliness and anxiety also predicted each another on consecutive days, forming a reciprocal emotional cycle. These findings are consistent with previous research suggesting that loneliness may serve a role in maintaining social anxiety (Oren-Yagoda et al., 2022). One possible pathway that can lead from anxiety to loneliness involves feelings of inauthenticity. Specifically, the more social anxiety individuals feel, the more likely they are to use safety behaviors (Piccirillo et al., 2016), and both social anxiety and safety behaviors have been found to lead to feelings of inauthenticity among individuals with SAD (Asher & Aderka, 2021; Plasencia et al., 2016). When one feels inauthentic, the desire to interact with others is greatly reduced, and one may feel isolated and lonely. A possible pathway from loneliness to anxiety may involve perceptions of inferiority. Specifically, feeling lonely and perceiving one’s social connections to be insufficient and inadequate, can easily lead to perceptions of self-inferiority (“No one wants to talk to me or hang out with me. Something must be wrong with me.”). Perceptions of inferiority have been theorized to lead to social anxiety (e.g., Gilboa-Schechtman et al., 2024).

The loneliness-anxiety maintenance/exacerbating cycle did not emerge among individuals with PD. For these individuals, anxiety did not predict subsequent loneliness, nor did loneliness predict future anxiety. One possible explanation is that individuals with PD often seek out close others for support when experiencing anxiety (e.g., enlist the help of a “safe person”; Carter et al., 1995), which could buffer against feelings of loneliness. This is in contrast to individuals with SAD for whom seeking social support may be a double-edged sword with potential benefits but also potential risks such as elevated social anxiety experienced toward the helper. In line with this explanation, a systematic review found that among individuals with panic disorder, higher perceived social support was consistently associated with lower symptom severity (Palardy et al., 2018). Thus, social support may play a protective role in PD, potentially counteracting the development of loneliness, and disrupting the reciprocal cycle observed in SAD. When considering the opposite directionality (loneliness to anxiety), it is possible that individuals with PD may not experience anxiety as a result of loneliness because their anxiety is primarily experienced in response to high-arousal states that are associated with heightened physiological activity (Brown & McNiff., 2009). Loneliness however, is a low-arousal emotion and therefore may not trigger anxiety among individuals with PD. Importantly, much future research is needed to explore and understand loneliness in PD.

Our findings have several important implications for clinical practice. First, the reciprocal cycles observed between shame and loneliness on the one hand, and anxiety on the other may serve as diagnostic indicators or markers for SAD. For instance, many individuals with SAD experience panic attacks in social situations. This typically requires diagnosing clinicians to conduct a differential diagnosis between SAD and PD. Observing elevated shame could potentially be a marker of SAD, as well as observing individuals experiencing shame as a result of their anxiety. Second (but along the same lines), clinicians can consider assessing emotions such as shame and loneliness when treating individuals with SAD. For instance, monitoring of emotions is common intervention/practice in cognitive behavior treatment. Typically, clients and clinicians focus on monitoring anxiety. However, monitoring additional emotions such as shame and loneliness may provide valuable information on the emotional processes that can play a role in the maintenance of social anxiety. Third, clinicians can use interventions that target shame and loneliness in order to help clients with social anxiety improve. Such strategies may serve to weaken the self-reinforcing emotional cycles that maintain social anxiety.

Several limitations of our study should be acknowledged. First, although our daily diary design with repeated measurements helped us to examine directionality, our design cannot support causal inferences as no manipulation of emotions was conducted. Future studies can utilize lab-based experimental designs to manipulate emotions. However, it is important to note that examining emotions in participants’ daily lives is also a strength of the present study, as it enhances external validity, and this can complement lab-based experimental studies that maximize internal validity. Second, all measures relied on self-report measurement, which is subject to a number of biases. Future studies can use physiological or behavioral assessments to complement self-report. Third, our sample included fewer individuals with PD than with SAD, which may have affected the ability to detect more subtle effects in the PD group. Future studies can examine larger samples of individuals with PD as well as ensure equal sample sizes for different diagnostic groups. Fourth, emotions were sampled once per day, which may have failed to capture higher resolution changes occurring throughout the day. Moreover, a single end-of-day measurement could introduce recall bias. Thus, future studies may benefit from incorporating momentary assessments at multiple times every day. Fifth, using a single item to measure emotions may capture state-level changes but may not fully reflect broader relational aspects. Future studies can examine more comprehensive measures of shame and loneliness. Sixth, our study focused on PD as the comparison group but future studies can examine comparison groups with other disorders such as GAD, which more closely resemble SAD in terms of shame and loneliness, and which may serve as a more difficult test for our hypotheses. Finally, as depression may potentially affect shame, loneliness, and anxiety, future studies should continue to evaluate its role in emotional processes occurring in SAD.

Despite these limitations, the present study makes a unique contribution to the literature by being the first to examine the daily temporal dynamics of shame, loneliness, and anxiety across two anxiety disorders. The findings demonstrate that among individuals with SAD, shame and loneliness may interact with anxiety in mutually reinforcing ways over time – and these emotional patterns were not observed in PD. These results emphasize the importance of specific emotional processes in SAD and highlight the need for clinical attention to shame and loneliness in assessment and treatment of the disorder.

Declaration of competing interest

The authors declare that there were no conflicts of interest with respect to the authorship or the publication of this article.