Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 May 1.
Published in final edited form as: Behav Ther. 2020 Sep 16;52(3):720–733. doi: 10.1016/j.beth.2020.09.003

Social Anxiety Disorder and Social Support Behavior in Friendships

Marilyn L Piccirillo 1, Michelle H Lim 1, Katya A Fernandez 1, Lauri A Pasch 2, Thomas L Rodebaugh 1
PMCID: PMC8382156  NIHMSID: NIHMS1723933  PMID: 33990245

Abstract

Interpersonal relationship quality is a strong predictor of health outcomes, and individuals with social anxiety disorder (SAD) report increased interpersonal impairment. However, there are few studies testing the effect of SAD on friendships and it is thus unclear whether there are behavioral differences that distinguish friendships in which a target individual has SAD from friendships in which the target individual does not have SAD. We tested for differences in the provision and receipt of support behaviors as a function of having a SAD diagnosis and accounting for comorbid depressive symptoms. Participants with SAD (n = 90) and their friends engaged in support conversations that were coded using the Social Support Interaction Coding System. Structural equation modeling revealed some differences between participants and friends when accounting for depression. Specifically, friends of participants with SAD and comorbid depression engaged in fewer positive helper behaviors than the friends of participants who did not have SAD or comorbid depression. Additionally, dyads in which the primary participant had SAD engaged in more off-task behaviors. Results suggest that SAD does not result in global interpersonal impairment, but that receipt of positive support behaviors from friends may differ as a function of SAD and comorbid depression. Interpersonal interventions aimed at increasing adaptive friendships and aspects of CBT that target subtle avoidance (e.g., safety behaviors) may be useful in facilitating more satisfactory relationships for these individuals.

Keywords: Social anxiety disorder, depression, social support behavior, friendships, interpersonal impairment

Satisfying friendships are associated with a stronger sense of well-being (Hartup & Stevens, 1997), greater emotional adjustment, higher levels of self-worth, social competence, and self-esteem (Bagwell, 2005; Hussong, 2000). Likewise, individuals who maintain successful friendships are more likely to utilize adaptive coping strategies and exhibit greater self-control, suggestive of a higher quality of life (Berkman, 1984; Schradle & Dougher, 1985). Further, there is evidence to suggest that social networks, or lack thereof (e.g., loneliness, social isolation) have an effect on earlier mortality (Giles et al., 2005; Steptoe et al., 2013).

Social anxiety disorder (SAD) is associated with fewer self-reported friendships (Vernberg et al., 1992) and romantic relationships (Hart, Turk, Heimberg, & Liebowitz, 1999; Leary & Dobbins, 1983; Schneier et al., 1994). A defining consequence of SAD is the avoidance of (or distress during) valued social interactions; thus, it is important to note that individuals with SAD often report low relationship satisfaction within the relationships that they maintain (Alden & Taylor, 2004; Rodebaugh, 2009).

Based on this cross-sectional research, it is often assumed that social anxiety causes interpersonal impairment; however, recent longitudinal studies using self-report measures have tested the directionality of this relationship and have provided evidence to counter this assumption. The findings of these longitudinal studies indicate that poor friendship quality and lower levels of perceived social support lead to increased social anxiety over time, but not vice versa (Rapee et al., 2015; Rodebaugh et al., 2015). Plausibly, interpersonal impairment may play an important role in determining the trajectory and severity of SAD.

An important aspect of interpersonal relationships involves the provision and receipt of social support. Little is known about how individuals with SAD provide and receive support, but interpersonal theories suggest that SAD is associated with maladaptive cognitive-behavioral biases and patterns in interpersonal relationships, which plausibly affect use of social support processes (Alden & Taylor, 2004). For example, comparison of individuals with SAD (or higher social anxiety) to individual without SAD (or who have lower social anxiety) reveals different behavioral patterns – lower social anxiety is associated with increased eye contact, a steadier tone of voice, longer speech turns, better tracking of conversations, and appropriate levels of self-disclosure to increase intimacy; in contrast, individuals with higher social anxiety and SAD exhibit these behaviors to a lesser extent (Fydrich et al., 1998). Researchers have largely focused on how individuals with social anxiety and SAD behave in romantic relationships (e.g., Heinrichs, 2003; Porter & Chambless, 2014); however, examining friendships may allow us to look at a larger, potentially wider, range of support behavior among individuals with SAD.

Moving beyond social anxiety alone, given the high rates of comorbidity between SAD and depression (Mineka et al., 1998), theories that address this comorbidity provide a useful model for evaluating the effect of interpersonal relationships on SAD and depression. For example, Epkins and Heckler (2011) suggest that negative interpersonal experiences (e.g., peer rejection) combined with poor-quality friendships predict social anxiety and depression. Researchers have rarely studied the effect of social anxiety and comorbid depression in adult friendships; however, the negative effects of depression alone on adult (romantic) relationships are well documented (for a review, see Mead, 2002). Additionally, researchers using the Social Support Interaction Coding System (SSICS) (Pasch & Bradbury, 1998), found that gender moderated the effect of negative affect on support behaviors within romantic relationships. That is, individuals who were high in negative affect were less likely to engage in positive support (i.e., helping) behaviors; however, women who were high in negative affect were also more likely to use negative behaviors in social interactions (Pasch, Bradbury, & Davila, 1997).

Interpersonal Behavioral Patterns Among Individuals With SAD

Notably, previous research on SAD and interpersonal behavior has almost exclusively relied on retrospective self-report and not behavioral indicators of interpersonal style. Thus, these results are likely subject to the well-studied negative self-referential biases that are especially pervasive among individuals with higher levels of social anxiety (Morrison & Heimberg, 2013; Moscovitch et al., 2009). A growing body of research suggests that individuals with higher levels of social anxiety are restricted in their interpersonal warmth as compared to individuals with lower levels of social anxiety (Alden & Taylor, 2004; Fernandez & Rodebaugh, 2011), which may translate to a more constrained pattern of social support within close relationships. An exemplary behavioral study by Meleshko & Alden (1993) examined social presentation style by manipulating a confederate’s level of self-disclosure within a small-talk conversation. Individuals with higher levels of social anxiety maintained a moderate level of self-disclosure, as opposed to matching the confederate’s level of self-disclosure, suggesting that individuals with higher levels of social anxiety adopt a more restricted and self-protective stance. This constrained presentation style may translate to restricted patterns of supportive behavior, such as decreased positive and increased neutral behaviors. An individual with higher levels of social anxiety may exhibit behavioral indicators of interpersonal constraint, such as providing fewer reassuring comments to a friend during an interpersonal interaction.

Perceptions And Patterns Of Support Among Friends Of Individuals With SAD

Researchers have suggested that individuals with SAD and those with higher levels of social anxiety report receiving less social support from friends and romantic partners (Cuming & Rapee, 2010; Porter & Chambless, 2014; Torgrud et al., 2004). However, additional evidence suggests that the level of social support provided by romantic partners (measured via self-report) does not differ as a function of the partner’s social anxiety (Dunkel-Schetter, & Bennett, 1990; Beck, Davila, Farrow, & Grant, 2006; Porter & Chambless, 2014). Furthermore, previous studies have demonstrated that friends of individuals with SAD are not more likely to have SAD themselves (Rodebaugh et al., 2014, 2015). It is plausible that individuals with higher levels of social anxiety may perceive less support to be available and may be more likely to interpret positive interactions through a negative lens, even when adequate support exists and is provided (Fernandez & Rodebaugh, 2011; Porter & Chambless, 2014). These studies suggest that the friends of individuals with SAD likely exhibit different interpersonal behavioral patterns from individuals with SAD, and may behave more similarly to friends of individuals without SAD.

A previous study by Rodebaugh and colleagues (2014) tested the effect of generalized SAD on self and friend-rated relationship quality in a subsample of participants. Results demonstrated that participants with SAD were more likely to rate their friendship quality as poorer via self-report, compared to participants without SAD (NOSAD). Notably, friends of SAD participants reported that their friends with SAD experienced greater impairment, although this did not negatively impact the friend’s ratings of their relationship. Rodebaugh and colleagues (2014) concluded that SAD status had a significant effect on self-reported friendship quality, but the effect was only for participants with SAD; friends of SAD participants did not report such negative perceptions. Therefore, these differences may be the result of negative cognitive biases regarding the interpersonal relationships that are not shared by friends of SAD participants or NOSAD dyads. Studies utilizing objective procedures, such as behavioral coding, are needed to gain a more accurate picture of the effect of internalizing symptoms on friendships.

Present Study

This study advanced upon a previous study by using one of the same participant samples (Sample 2; Rodebaugh et al., 2014). We evaluated the relationship between generalized SAD1 status and social support behaviors within the context of friendships while accounting for depressive symptoms. The two social support conversations described previously were coded using the SSICS (Pasch & Bradbury, 1998), a behavioral coding system that measured the amount of positive, negative, neutral, and off-topic behavior. Both helper (providing support) and helpee (receiving support) behaviors were rated in these support conversations. Notably, as social anxiety and depression are highly comorbid (Mineka et al., 1998) and as depressive symptoms are also associated with significant interpersonal impairment (Mead, 2002; Pasch, Bradbury, & Davila, 1997), we examined the role of comorbid depressive symptoms as an additional predictor of support behaviors.

We aimed to examine differences in the social support behaviors among close friends in which the primary participant had SAD (versus those in which the primary participant did not). We hypothesized that participants with SAD would engage in fewer positive and more neutral and off-task support behaviors, in keeping with literature describing interpersonal constraint (Fernandez & Rodebaugh, 2011). Examining the behaviors coded by the SSICS, we expected that participants with SAD would utilize fewer reassuring or constructive problem-solving behaviors and make more comments that are not relevant to the conversation.

In keeping with the previously referenced study that used the SSICS to examine the effect of negative affect on support behaviors in romantic relationships (Pasch et al., 1997), along with the findings reviewed above concerning the comorbidity of SAD and depression, we also hypothesized that individuals with higher levels of depression would engage in fewer positive and more negative support behaviors. In regards to SSICS behaviors, we expected that participants with higher levels of depression would utilize fewer reassuring or constructive problem-solving behaviors and exhibit more criticizing or blaming behaviors when interacting with their friend.

Material and Methods

Participants

Individuals (N = 180) were recruited in three groups: those diagnosed with generalized SAD (SAD; n = 51), those who did not meet criteria for SAD (NOSAD; n = 39), and non-romantic friends of participants in both groups (n = 90). Participants in the SAD group were recruited from the St. Louis community through use of flyers posted in public settings and local clinics, as well as newspaper, TV, and Internet advertisements. Participants in the NOSAD group were recruited from a volunteer registry and were matched to SAD participants on gender, age, and race. All recruitment materials advertised a study focusing on relationships (the “Focus on Relationships” study), and primary participants were recruited based on probable SAD as well as low social anxiety (for the NOSAD group). Primary participants were told that the study concerned SAD, but friends were not provided this information by the study team. Both groups of participants were invited to bring a current, non-romantic friend to a second session. In an effort to increase external validity, there were no specifications on the sex of this friend, and a majority (77%) brought in a friend of the same sex. Participants (but not friends) first went through a phone screening process to rule out exclusion diagnoses, including active mania, psychosis, substance abuse, and imminent suicidality. Both participants and their friends were compensated $15 per hour.

Demographic characteristics of both participants and friends, as a function of SAD status, are presented in Table 1. The study sample was primarily middle-aged, female, and White, although over 40% of the sample identified as Black. The average duration of friendship was approximately ten years. Few friends had psychiatric diagnoses and there were no significant differences in the total number of current clinical diagnoses across friends brought in by SAD participants and friends brought in by NOSAD participants, t(85.13) = 1.23, p = .222, not assuming homogeneity of variance. Similarly, duration of friendship (r = .15) was not significantly correlated with SAD status.

Table 1.

Descriptive statistics for primary participants and friends

Primary Participants (n = 90) Friends (n = 90)
SAD (n = 51) NOSAD (n =39) PP-SAD (n = 51) PP-NOSAD (n = 39)
Age, years, M (SD) 40.57 (13.40) 38.62 (14.10) 38.62 (15.16) 41.23 (15.20)
Number of women (%) 35 (68.60%) 27 (69.20%) 34 (66.70%) 23 (59.00%)
Race (%)
 American Indian or Alaskan Native 1 (2.00%) -- -- --
 White 25 (49.00%) 22 (56.40%) 24 (47.10%) 19 (48.70%)
 Asian or Pacific Islander 2 (3.90%) -- 2 (3.90%) 1 (2.60%)
 Black 19 (37.30%) 16 (41.00%) 22 (43.10%) 17 (43.60%)
 Multiracial 4 (7.80%) 1 (2.60%) 1 (2.00%) 2 (5.10%)
 Other minority -- -- 2 (3.90%) --
Hispanic Ethnicity (%) 1 (2.00%) 2 (5.10%) 2 (3.90%) 1 (2.60%)
Same-sex friendship 38 (74.51%) 31 (79.49%)
Friendship duration in months, M (SD) 140.15 (123.97) 109.60 (93.02) -- --
BDI-II, M (SD) 19.06 (10.31) 4.47 (4.63) -- --
LSAS, M (SD) 92.98 (17.60) 10.33 (7.49) -- --
Depression (%) 20 (39.20%) 0 (0.00%) 4 (7.8%) 0 (0.00%)
Social anxiety disorder (%) 51 (100.00%) 0 (0.00%) 7 (13.70%) 1 (2.6%)
Other anxiety disorder (%)
 Panic disorder 9 (17.60%) 0 (0.00%) 1 (2.00%) 1 (2.60%)
 Agoraphobia 1 (2.00%) 0 (0.00%) 1 (2.00%) 2 (5.10%)
 Generalized anxiety disorder 7 (13.70%) 0 (0.00%) 5 (9.80%) 2 (5.10%)
 Obsessive-compulsive disorder 7 (13.70%) 0 (0.00%) 2 (3.90%) 1 (2.60%)
 Post-traumatic stress disorder 7 (13.70%) 0 (0.00%) 0 (0.00%) 2 (5.10%)
Total disorders, M (SD, Range) 1.98 (1.06, 1 – 5) 0.00 (0.00, 0 – 0) 0.0.41 (0.85, 0 – 3) 0.23 (0.53, 0 – 2)
SSICS Behavioral code percentage, M (SD)
Positive Helper 0.43 (0.13) 0.46 (0.12) 0.44 (0.13) 0.45 (0.13)
Positive Helpee 0.39 (0.13) 0.41 (0.12) 0.39 (0.13) 0.42 (0.11)
Negative Helper 0.03 (0.01) 0.02 (0.03) 0.03 (0.01) 0.02 (0.01)
Negative Helpee 0.03 (0.03) 0.02 (0.01) 0.03 (0.03) 0.02 (0.03)
Neutral Helper 0.38 (0.32) 0.38 (0.09) 0.37 (0.09) 0.39 (0.09)
Neutral Helpee 0.42 (0.08) 0.42 (0.08) 0.41 (0.09) 0.41 (0.07)
Off-Task Helper 0.17 (0.09) 0.15 (0.07) 0.17 (0.08) 0.15 (0.10)
Off-Task Helpee 0.17 (0.08) 0.15 (0.09) 0.17 (0.09) 0.14 (0.16)
Open in a new tab

Note. SAD = Generalized social anxiety disorder; NOSAD = No social anxiety disorder; PP-SAD = Primary participant of friend is in SAD group; PP-NOSAD = Primary participant of friend is in NOSAD group. BDI-II = Beck Depression Inventory – II; LSAS = Liebowitz Social Anxiety Scale; Psychiatric diagnoses were assessed using the SCID-I/P for DSM-IV-TR for primary participants and the MINI 5.0 for DSM-IV for friends. Depression was assessed during the past month for primary participants and during the past two weeks for friends. Social anxiety disorder for primary participants was assessed over the lifetime based on a diagnosis of SAD and a score of 30 or higher on the LSAS. All other diagnoses were assessed over the past month, with the exception of generalized anxiety disorder, which was assessed over the past six months. Total disorders include current (i.e., past month) major depressive episode, bipolar I disorder, bipolar II disorder, psychotic symptoms, panic disorder, agoraphobia, generalized anxiety disorder (past six months), social anxiety disorder, and obsessive-compulsive disorder, and post-traumatic stress disorder; SSICS = Social Support Interaction Coding System; Codes are calculated as percentages of the total conversation. Estimates presented here are the mean and standard deviation of these percentages from across all imputed datasets.

Diagnostic Measures

A licensed clinical psychologist (TLR), post-doctoral fellow (MHL), and four graduate students in clinical psychology conducted diagnostic interviews. Graduate students conducting diagnostic interviews had received training in clinical assessment, including the SCID-IV-TR. Training was conducted and supervised by TLR. These procedures are briefly elaborated on below and are further described in a previous paper (Rodebaugh et al., 2014).

Structured Clinical Interview for DSM-IV (SCID-IV-TR)

The SCID-IV-TR (First, Spitzer, Gibbon, & Williams, 2002) is a semi-structured interview that was used to assess both current and selected lifetime psychiatric diagnoses. The SCID-IV-TR maps onto psychopathology as defined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth edition (DSM-IV; American Psychiatric Association, 1994) and is considered to be a gold-standard in diagnostic assessment. The SCID-IV-TR was used primarily to assess current internalizing symptomatology, including SAD. Past symptomatology was assessed in order to better define current diagnoses. Inter-rater reliability in this sample is discussed below.

M.I.N.I. International Neuropsychiatric Interview (MINI 6.0)

The MINI (Sheehan et al., 1998) is a short, structured interview with good reliability (κ =.88–1.00). Previous versions have compared favorably to the SCID (Sheehan et al., 1998). The MINI was used to assess friends of primary participants. Friends were excluded from the study if they met any of the exclusion criteria, including current mania, current psychosis, current substance abuse, and imminent suicidality. Notably, there were no inclusion criteria based on diagnoses for friends.

Liebowitz Social Anxiety Scale (LSAS)

The LSAS (Liebowitz, 1987) is a 48-item clinician-administered scale that uses a 4-point Likert-type scale to measure anxiety and avoidance of 24 social performance and interaction situations. The LSAS has demonstrated good internal consistency (α = .96, Heimberg et al., 1999). The LSAS was used as an additional criterion for determining SAD group status, such that scores at 60 and above suggest a probable clinical diagnosis of SAD and scores below 30 suggest no diagnosis of SAD (Mennin, Fresco, Heimberg, Schneier, Davies, & Liebowitz, 2002).

Diagnostic determination procedure

Primary participants with a diagnosis of SAD from the SCID-IV-TR, as well as a score of 60 and above on the LSAS, were determined to have SAD and were enrolled in the SAD group. Those individuals who did not meet diagnostic criteria for SAD on the SCID-IV-TR and had scores below 30 on the LSAS were enrolled in the NOSAD group. Individuals who did not meet either of these criteria were excluded. Friends were assessed using the MINI but were not selected based on diagnoses. That is, as long as friends did not meet for any of the exclusion criteria (e.g., current mania), they were allowed to participate as a partner of participants in either group. A sample of videos that included individuals assigned to SAD and NOSAD groups, as well as individuals who were not assigned to either group were reviewed. Agreement on primary participant diagnostic group using the SCID-IV-TR and the LSAS was 100%, such that κ = 1.00 (Rodebaugh et al., 2014).2

Self-Report Measures

Demographic Information

Demographic information was collected from participants and their friends.

Beck Depression Inventory – II (BDI-II)

The BDI-II (Beck, Steer, & Brown, 1996) is a 21-item self-report measure of depressive symptoms. The BDI-II was used to examine the effect of comorbid depressive symptoms on interpersonal behaviors. The internal consistency for the BDI-II was excellent (α = .94).

Coding System

Social Support Interaction Coding System (SSICS)

The SSICS (Pasch & Bradbury, 1998; Pasch et al., 1997) was used to code social support behaviors during the two social support tasks (see Procedure). This coding system has been used in several other studies exploring romantic relationships (e.g., Beck et al., 2006; Trombello, Schoebi, & Bradbury, 2011). It is important to acknowledge that the SSICS has not been previously used to study friendships. However, we do not know of any coding systems used in more than a single study that have been developed to evaluate friendships. Thus, we decided to use the original SSICS, as it is a commonly used coding system to evaluate social support processes within relationships.

In the coding system, both helper and helpee behaviors were rated for both participant and friend across two conversations. Helper behavior refers to support behaviors used by the partner in providing support to the helpee. Helpee behavior refers to support behaviors used by the partner who requested support on a specific topic. Helper behavior included: Positive Instrumental, Positive Emotional, Positive Other, Negative, Neutral, and Off-Task. Helpee behavior included: Positive, Negative, Neutral, and Off-Task. Participants were allowed to generate their own topic for the support conversations. For example, a dyad could choose to discuss how to help one of the partners reduce drinking. An individual in the helper role could engage in positive helper behavior by making a statement like, “Can you drink and not be drunk?”; whereas an individual in the helpee role could engage in positive helpee behavior by making a statement like “My plan was to stop drinking January 2nd”. A description of each code type is included in Table 2 and examples of phrases that exemplify each code can be found in Supplementary Table S1. We observed no differences in topics chosen across diagnostic group. Common topics selected included desire to lose weight, change careers, be more patient with others, or reduce stress. A list of topics can be found in Supplementary Tables S2.

Table 2.

Coefficient omega by condition for primary participants and friends

Social Support Code Description Coefficient omega 95% Credible Interval
Primary Participants
Helper
Positive Constructive problem solving or emotional reassurance and validation 0.92 0.82, 0.97
Negative Criticizing, blaming the helpee 0.85 0.69, 0.93
Neutral Supportive behavior related to the task, not otherwise accounted for 0.80 0.59, 0.91
Off-Task Behavior that is not related to the task 0.94 0.88, 0.97
Helpee
Positive Clearly and effectively stating the problem and requesting help 0.93 0.84, 0.97
Negative Demanding help, criticizing, blaming the helper 0.88 0.75, 0.94
Neutral Behavior related to the task, not otherwise accounted for 0.75 0.53, 0.89
Off-Task Behavior that is not related to the task 0.91 0.85, 0.95
Friends
Helper
Positive 0.92 0.86, 0.96
Negative 0.47 0.23, 0.73
Neutral 0.73 0.53, 0.87
Off-Task 0.93 0.87, 0.96
Helpee
Positive 0.90 0.80, 0.95
Negative 0.83 0.69, 0.91
Neutral 0.78 0.59, 0.90
Off-Task 0.94 0.88, 0.97
Dyads
Primary Participant Helper and Friend Helpee
Negative 0.84 0.77, 0.89
Neutral 0.83 0.55, 0.92
Off-Task 0.94 0.91, 0.96
Primary Participant Helpee and Friend Helper
Negative 0.92 0.75, 0.98
Neutral 0.84 0.72, 0.91
Off-Task 0.93 0.89, 0.95
Open in a new tab

Procedure

Data presented here were obtained as part of a larger study (see Rodebaugh et al., 2014), which was approved by an Institutional Review Board. This study examined the effect of SAD on generalized SAD and relationships by recruiting participants with and without SAD along with their friends and romantic partners for a three-session study. After an initial phone screen, primary participants came to the lab for their first session and consented to the study. Participants completed a diagnostic interview and self-report measures. Data presented here were obtained during the second session, in which primary participants were invited to bring a close friend to complete a diagnostic interview, self-report measures, and three conversation tasks that were then coded using behavioral coding schemes.

These conversations included two social support tasks and one conflict task. Instructions were provided for the first social support task and, immediately before beginning the first support conversation, the helpee was asked to select a topic or issue that he or she needed help managing or changing. Dyads were provided with a sheet of exemplar topics if they needed help choosing a topic. Participants were randomly selected to serve in either the helper (providing support) or helpee (receiving support) condition in the first support conversation. Participants and friends discussed the helpee’s topic during a 10-minute conversation. They then had 10-minute break before completing a 10-minute conflict task, discussing an issue the participant wanted to change in their relationship (data from this task were not analyzed in this study). Finally, after a second 10-minute break, they switched support conditions and completed a second 10-minute support conversation. No other individuals were in the room during either conversation, although participants were aware that the conversations were recorded for coding purposes. At the end of the experiment, participants were debriefed on the nature of the study.

Coding Procedure

Coders were trained using materials, including video examples, supplied by the fourth author (LAP). Training was provided by the final (TLR) and second authors (MHL), after which coders rated additional examples independently. TLR reviewed ratings from the example videos and feedback was given on potential problems. Video clips were assembled into a random order in two batches, as data collection was still in progress when coding began. Plans were made for all four raters to rate at least 10% of videos and for pairs of raters to rate another additional 13% of the videos at minimum. All videos were randomly assigned. To assess reliability early in the coding process, all coders rated the first four videos and pairs of coders rated the following six videos. All four coders rated a total of 16 videos and pairs of coders rated a total of 22 additional videos. The final author (TLR) reviewed reliability for the social support codes regularly and feedback was provided to ensure adequate adherence to the coding protocol. Coders rated a previously-coded clip together along with the second author (MHL) to reduce coder drift. All coding was completed within approximately 12 months.

The coders rated each individual on six different styles of providing social support (helper) or four different styles of receiving support (helpee) by recording the number of times that the helper engaged in each type of behavior. Coders were blind, in that they did not receive information on the individual’s diagnostic status or other information regarding the study. When queried at the end of coding, coders stated that they were not aware that the focus of the study was to compare interpersonal processes between individuals with SAD versus NOSAD. However, some coders reported that they believed that some participants might have been recruited based on level of (social) anxiety or depression, primarily because some participants mentioned such symptoms. When debriefed, the same coders expressed surprise that more than half of the dyads contained a participant diagnosed with SAD.

Coding Reliability

Due to low counts for some positive helper codes, all positive helper code values were summed to create a total positive helper code, as has typically been done in previous studies (Positive; cf. Trombello et al., 2011). Each code variable was divided by the number of turns (i.e., meaningful utterance or response) in the conversation and thus represented the percentage of the conversation that was rated as positive, negative, neutral, or off-task for each condition. Inter-rater reliability was measured using coefficient omega (McDonald, 1999). Coefficient omega has been argued to provide a more appropriate and accurate estimate of inter-rater reliability, as compared to alpha or interclass correlations, because coefficient omega makes fewer assumptions about the nature of the data and can be calculated alongside a confidence interval (Dunn et al., 2014). For our study, which used multiple coders, coefficient omega is better able to measure the amount that each coder differs from the other when calculating reliability, providing a more accurate reliability estimate than interclass correlations in a situation in which (a) some coders are better than others at coding and (b) which coders are best varies by code type.

Arguably, when using multiple coders, calculating omega hierarchical would provide the most meaningful reliability estimate, as omega hierarchical is better able to account for systematic differences among the coders as compared to coefficient omega. However, calculating omega hierarchical requires constructing a bifactor model. We attempted to construct a bifactor model but were not able to achieve model convergence, likely due to our lower sample size relative to the amount of missing data. Thus, we relied on coefficient omega, which we believed was an acceptable method for calculating inter-rater reliability and was superior compared to intraclass correlations in a case in which coders varied in their reliability, and that variation itself varied across code types. It would be reasonable to assume that omega hierarchical would estimate somewhat lower reliability than we report here.

It also appears that coders revealed different skills in assessing across codes (see Results). Reliability estimates for participant and friend positive helper and helpee codes and participant negative helper codes were comparable to the original SSICS reliability estimates that were calculated using intraclass correlations in previous studies (ICC) (Pasch & Bradbury, 1998). Notably, the guidelines and cut-offs for interpreting coefficient omega are similar to those for interpreting ICC. However, the estimates for all other codes were lower as compared to the original SSICS ICC values (Pasch & Bradbury, 1998).

Calculating reliability using ICC and by combining participant and partner values for codes has often been used in previous studies using the SSICS (c.f., Pasch & Bradbury, 1998; Sullivan, Pasch, Johnson, & Bradbury, 2010; Trombello et al., 2011). This method can help determine reliability as a function of code type. However, it does not account for the dependency between conversation partners, which violates standard assumptions for ICC (McGraw & Wong, 1996) and can result in inflation of reliability statistics. Thus, calculating coefficient omega for participant and friend codes separately is a more appropriate method for calculating reliability. Although reliability estimates varied across codes, all codes demonstrated estimates adequate to excellent omega values, as shown in Table 2.

The Negative, Neutral, and Off-Task codes were also averaged across primary participants and friends to create primary participant helper (i.e., friend was helpee) and primary participant helpee (i.e., friend was helper) dyadic codes. These dyadic codes demonstrated either similar or higher reliability to individual helper and helpee codes. Because dyadic codes resulted in higher reliability for certain codes that had initially demonstrated lower reliability (e.g., friend negative helper), dyadic codes were used for all non-positive behaviors (Table 2).

Data Analytic Method

Planned missingness

The coding procedure described above resulted in approximately 10% of interactions coded by all four coders, approximately 12% of interactions coded by two coders, and approximately 77% of interactions coded by one coder. This means that approximately 90% of interactions were missing at least one rating from one of the four coders. Previous coding studies have often not accounted for missing data but instead have implemented a combination of averaging ratings from multiple coders and using the single rating when only one coder was used. We believe the process of assigning coders to code fewer than all interactions results in planned missingness, which should be addressed through the use of appropriate missing data handling. Use of the Bayes estimator ultimately provides a parsimonious method for analyzing the data; it is one of several equivalent alternatives for handling missing data appropriately. To accommodate the planned, missing-at-random values of social support interaction codes rated by fewer than four coders, a Bayes estimator was used for model estimation.

Bayesian estimation and modeling

Bayesian statistics were used to conduct study analyses as certain behavioral codes (e.g., negative codes) were non-normal and all behavioral codes contained missing observations, as discussed previously. Both issues can be accommodated using the Bayes method (overview provided in Muthén & Asparouhov, 2012; explanation of Bayesian estimation for missing data provided in Schafer, 1997). Thus, Bayesian statistics represented the ideal analytic method for our study analyses.3

All study analyses were conducted using Mplus, Version 8.0 (Muthén & Muthén, 1998–2017). Codes could not all be modeled in the same analysis because one code (e.g., positive) could be perfectly predicted by the remaining codes (e.g., nonpositive codes), which would produce nonpositive definite matrices. Positive codes were modeled using separate latent variable (e.g., participant positive helper); whereas indicators for negative, neutral, and off-task codes were modeled using dyadic latent variables (e.g., participant negative helper and friend negative helpee) to improve reliability as discussed previously. Correlation pathways were included between participant and friend support variables to account for the dyadic nature of the data. All models were run at least twice, with the second or additional model specifying twice the number of iterations to confirm model convergence. As we did not have empirical results to draw from to set priors, we used the default (i.e., uninformative) priors provided by Mplus. This means that the priors specified in these models did not have a biasing effect on final estiamtes (with the exception of biasing variances towards being positive). The potential scale reduction (PSR) factor and Kolmogorov-Smirnov test were used to evaluate model convergence. PSR values lower than 1.1 represented acceptable convergence (Gelman, 2003). Additionally, the Kolmogorov-Smirnov test was used to evaluate the goodness-of-fit of the posterior chains identified in the model. Notably, relative fit indices are irrelevant for smaller sample sizes (i.e., N < 300) and thus, we evaluated the posterior predictive p-value to determine whether the models constructed for this study exhibited adequate fit (i.e., p > .05), as a p-value of less than .05 is recommended as the standard by which to reject a model for poor fit (Asparouhov & Muthén, 2020). All models displayed posterior predictive p values above .05, suggesting adequate fit. Predictors were considered significant if the 95% credible interval, which represents the range of values observed in the posterior, did not include 0. This is generally consistent with α level of .05 in that it indicates some confidence that the effect is not equal to zero.

Post-hoc model

An Actor-Partner Interdependence Model (APIM) was used in a follow-up analysis to evaluate the effect of participant and friend psychiatric diagnoses on positive support behaviors. Such a model was feasible for this variable because it was available with sufficient range for both participant and friend. A single APIM model was constructed for positive behaviors using Bayesian estimation. Predictors were considered significant if the 95% credible interval did not include 0 (consistent with α level of .05).

Results

Creation of the Structural Equation Model

Bivariate correlations were examined between predictors and support behaviors, as well as participant and friend support behaviors. Results are available in Supplementary Tables S3 and S4. Two models were estimated, one model for positive codes and a second model for non-positive codes. Positive behaviors included participant helper and helpee, as well as friend helper and helpee codes. Non-positive behaviors included dyadic negative helper (i.e., participant negative helper, friend negative helpee), negative helpee, neutral helper, neutral helpee, off-task helper, and off-task helpee codes. Given the considerable rates of comorbidity between depression and SAD, we wanted to determine the effect of SAD in conjunction with depression. Additionally, code reliability appeared to differ as a result of condition (i.e., being helper or helpee first). Thus, models were constructed using SAD status, condition, and BDI-II score, as well as the respective two-way interactions, and the three-way interaction.

How Do Positive Support Behaviors Differ As A Function Of SAD And Comorbid Depression?

A model was constructed examining the effect of SAD diagnosis on positive support behaviors. The three-way interaction between SAD status, condition, and BDI-II score was not significant. However, the two-way interaction between SAD status and BDI-II score was significant, B = 0.14, p = .005, 95% CI [0.04, 0.25]. Probing this interaction revealed that friends of NOSAD participant with higher depression used more positive helper behaviors, B = 0.14, p < .001, 95% CI [0.06, 0.25]. In contrast, friends of SAD participants with higher depression used fewer positive helper behaviors, B = −3.03, p = .002, 95% CI [−0.94, −5.01]. No other probed effects were statistically significant. These results suggested that individuals with SAD and higher depression received less positive support from their friends, whereas individuals with NOSAD and higher depression received relatively more positive reactions from their friends.

Additionally, the two-way interaction between BDI-II score and condition significantly predicted participant helpee behaviors, B = −0.08, p = .020, 95% CI [−0.18, −0.004]. Probing this interaction revealed that participants with higher depression who were helpees in the first conversation used more positive helpee behaviors, B = 0.08, p = .003, 95% CI [0.02, 0.15]. There were no other significant probed effects. Standardized estimates are included in Figure 1 and Table S5.

Figure 1.

Figure 1

Open in a new tab

SEM model demonstrating standardized estimates for SSICS codes

Note. The model on the left illustrates participant and friend postivie helper and helpee codes. The model on the right displays estimates for dyadic negative, neutral, and off-task helper and helpee codes. The model includes estimates for SAD status, BDI, condition (i.e., whether the participant was helper or helpee first), as well as the two-way interaction between SAD status and BDI (SAD x BDI), the two-way interaction between SAD status and condition (SAD x Condition), the two-way interaction between BDI and condition (BDI x Condition), and the three-way interaction between SAD status, BDI, and condition (SAD x BDI x Condition). Observed participant and friend positive helper and helpee codes were used to estimate latent outcome variables. Pathways between observed codes and latent variables are not included for simplicity. The estimates presented here for SAD status, Condition, and interactions including these variables are y-standardized (i.e., a partially-standardized coefficient). Thus, the estimate represents the increase in standard deviation of the outcome for every one-unit increase in the predictor. The estimates presented for the BDI are fully standardized. These estimates differ from the unstandardized estimates that appear in the main text. SAD = Generalized social anxiety disorder; BDI = Beck Depression Inventory – II. SAD status is coded 1 = SAD and 0 = NOSAD. *p < 0.05.

How Do Negative, Neutral, And Off-task Support Behaviors Differ As A Function Of SAD And Comorbid Depression?

A second model using the same predictors was constructed to examine the effect of SAD on negative, neutral, and off-task dyadic behaviors. Neither the three-way interaction nor the two-way interactions were significant. SAD status significantly predicted greater off-task helper behaviors, B = 1.14, p = .024, 95% CI [2.32, 0.01]. Additionally, depression significantly predicted fewer neutral helpee behaviors, B = −0.06, p = .014, 95% CI [−0.12, −0.01]. There were no other significant predictor effects in this model. Standardized estimates are included in Figure 1 and Table S5.

Modeling The Effect Of Friends’ Diagnoses On Support Behaviors: Post-hoc APIM Model

An APIM model was constructed to evaluate whether the effect of SAD participants with higher depression receiving less positive support was (a) due to number of psychological problems, rather than depression and SAD in particular and (b) a general effect for all participants (including friends), as opposed to an effect stemming only from the primary participants. We tested the effect of total number of participant and friend clinical diagnoses on positive support behaviors. Participant pathways were included in which participant total diagnoses predicted both participant (i.e., actor) and friend (i.e., partner) helper and helpee support behaviors. Friend pathways were included in which friend total diagnoses predicted friend (i.e., actor) and participant (i.e., partner) helper and helpee support behaviors. No effects were statistically significant.

After constraining actor and partner paths for each support behavior in an effort to conserve power, no paths were statistically significant. When main effects for SAD, depression, and the interaction between SAD and depression were added the model, results were substantively identical to results from the positive behavior models. This suggests that the effect of friends providing fewer positive helper behaviors to their partner with SAD and higher depression was not accounted for by the friend’s own psychopathology.

Discussion

Recent research has indicated that interpersonal impairment may drive future social anxiety symptomatology (Rapee et al., 2015; Rodebaugh et al., 2015). This study provides findings from a behavioral coding study of support behaviors within friendships – a previously unstudied type of relationship in this context. We examined how SAD and comorbid depressive symptoms affected support behaviors within friendships. Overall, results suggested that a diagnosis of SAD alone did not result in global interpersonal impairment; however, results clearly indicated SAD status, depression, and their interaction significantly impacted positive support behaviors. Probing this interaction revealed that friends offered less support to individuals with SAD who were also depressed, as compared to individuals with NOSAD and higher depression. Results also suggested that SAD status and depression appeared to have an effect on off-task helper and neutral helpee dyadic behaviors, respectively. Finally, the total number of clinical diagnoses did not significantly predict participant or partner’s behavior.

Notably, results point to the role of depressive symptoms in conjunction with SAD that may affect social support behaviors, although in limited contexts (e.g., fewer positive support behaviors from friends). Results from the positive behaviors model suggest that friends were less likely to utilize positive support behaviors (e.g., constructive problem-solving or reassurance) when they interacted with their friend (the primary participant) who had SAD and higher depressive symptoms; however, this effect was not seen for friends of participants with NOSAD and higher depressive symptoms. These findings provide behavioral evidence that is in keeping with well-documented negative effects of depression on relationships (Mead, 2002). For example, Pasch and colleagues (1997) found that husbands were less likely to provide positive support to wives with higher levels of negative affect (Pasch et al., 1997).

Furthermore, we found no evidence of interpersonal constraint (i.e., fewer positive and more neutral interpersonal behaviors) in this study. It is possible that interpersonal constraint may manifest in interactions with strangers, acquaintances, or newly forming friendships, as compared to the longer-lasting friendships studied here. Indeed, the previous study using this sample found that SAD participants who endorsed lower quality friendships were more likely to be younger and were more likely to have reported on newer friendships (Rodebaugh et al., 2014). Thus, it seems possible that different dynamics are present in newer friendships, rather than more established friendships. However, we did find limited evidence of interpersonal avoidance, in that individuals with SAD were more likely to utilize off-task helper behaviors, which may be conceptualized as a form of subtle avoidance of more constructive support behaviors. Thus, it is possible that the friends of individuals with SAD are more likely to tolerate or prefer certain interpersonal styles as compared to friends of NOSAD individuals.

Our results are somewhat similar to a previous study in which the authors reported that partners perceived similar levels of support from romantic partners, regardless of their own social anxiety symptoms (Porter & Chambless, 2014). Additionally, level of depressive symptoms affected both men and women’s perceptions of intimacy within the romantic relationship. In this study, SAD diagnosis, in conjunction with higher depression, significantly predicted the friend’s use of fewer positive support behaviors across the two interactions. This effect was not related to the friends’ number of clinical disorders, as measured in the post-hoc APIM model. Additionally, this effect was not seen for SAD, without comorbid depression. Relatedly, although individuals with SAD are more likely to describe their relationships negatively, even after accounting for depressive symptoms (Rodebaugh et al., 2014), we found little behavioral evidence of negative behaviors from participants or friends based on SAD status alone. Thus, self-reported differences may be attributable to a combination of negative cognitive biases, more negative reactions from friends when the person with SAD also has additional disorders (i.e., depression), or a combination of these factors.

Results from this study should be interpreted within the limitations of our study design. This study took place within a laboratory setting, which limits external validity. Moreover, this study was conducted as part of a larger investigation into the effects of SAD on interpersonal relationships. Thus, additional study tasks (e.g., the conflict task discussed earlier) that were not examined here affected individual’s performance in this study’s tasks. These effects were accounted for in our statistical models, although still represent limitations for external validity. Notably, the SSICS has primarily been used to examine heterosexual, romantic relationships and to our knowledge, has not been previously used for the study of friendships. We selected the SSICS for use due to its face validity and demonstrated validity in another type of close relationship (romantic relationships), but specific data on the use of the SSICS in friendships would provide greater assurance of its validity in this context. For example, there were limited instances of non-positive behaviors between participants and their friends. Although it is plausible that such behaviors are simply uncommon in friendships (compared to romantic relationships) it could also be true that the SSICS coding system may not have adequately captured how such behaviors manifest specifically within friendships. Future work could examine the use of non-positive behaviors within friendships for individuals with SAD, depression, and other clinical symptoms. Additionally, the SSICS was designed to quantify verbal exchanges (including nonverbal factors such as tone), rather than purely non-verbal behaviors (such as rate of eye contact). Thus, to the extent that there are useful nonverbal indicators that may differentiate interpersonal interactions between SAD dyads from NOSAD dyads, the SSICS would not have captured these nonverbal cues. However, there are no other empirically supported coding systems designed to study friendships, let alone friendships of individuals with SAD, and thus, the SSICS represented the best option for this study. In regard to nonverbal indicators, a previous study using this same sample focused solely on gaze avoidance. This previous study found that participants with GSAD exhibited increased gaze avoidance during conflict-primed conversations (Langer, Lim, Fernandez, & Rodebaugh, 2016). That is, in this previous study there were no significant differences in gaze avoidance during the first social support conversation, suggesting that – similar to verbal indicators – there may be few differences in the use of nonverbal behaviors in support conversations across GSAD and NOSAD participants.

A limitation that should be kept in mind when interpreting results was that our study sample was primarily female. Although rates of SAD are higher in women versus men (Kessler et al., 2012), our sample included few men. This limited our ability to examine gender differences in support behaviors, although descriptive statistics of support behaviors across gender is provided in Supplementary Table S6. Likewise, most participants selected a same-sex friend, which limits generalizability of these results to female same-sex friendships. Finally, we are not able to make conclusions about support behaviors in individuals with SAD who lack close friendships, given that this sample only included individuals who were able to identify and bring in a close friend.

Despite these limitations, our study design also exhibits key strengths. Namely, we recruited a clinical population of individuals with SAD and collected behavioral data from both participants and their friends. Much of the previous literature has focused on collecting retrospective self-report data from individuals and their romantic partners and these data provides behavioral evidence of an additional, potentially more accessible, type of relationship for individuals with SAD.

Conclusion

The behavioral evidence presented in this study suggests that there were limited effects for SAD on support behaviors. In contrast, friends of individuals with SAD and higher depression were less likely to exhibit positive helper behaviors, suggesting further evidence of the negative effect of depressive symptoms on relationships. That is, although we did not observe changes in behavior on the part of the primary participants with SAD and higher comorbid depression, the differential behavior of their friends plausibly suggests that a longer-term process has led to their friends adopting these less helpful behaviors. Interpersonal therapy interventions to address the likely maladaptive behavioral patterns used by individuals with higher depression may be useful in improving friendship quality. Given the limited behavioral findings for SAD specifically in this study, it may be that differential reports on friendship quality among SAD participants are the result of negative cognitive biases, rather than dysfunctional behavioral patterns. These results can inform interpersonal interventions for psychotherapy. For example, it may be useful to target cognitive biases using cognitive restructuring in cognitive-behavioral therapies for SAD, rather than addressing interpersonal behaviors specifically, at least among people with SAD who have established friendships. That is, given the general lack of behavioral differences in SAD friendships, it may be more useful to target the cognitive biases that influence how individuals with SAD perceive their friendships. However, further work is needed to evaluate how individuals with SAD utilize off-task helper behaviors (i.e., a potential form of subtle interpersonal avoidance) and consider whether CBT treatment for SAD would be useful in addressing these potential safety behaviors. Finally, additional behavioral studies examining the effects of SAD on new or developing friendships would help to illustrate the dynamic relationship of SAD on support processes and the trajectory of this effect on relationship formation and maintenance.

Supplementary Material

1

Highlights:

  • We examined support behaviors among dyads in which one partner had SAD

  • Partners with SAD and higher depression received less positive support from friends

  • There were minimal effects of SAD and depression on the participant’s behavior

Acknowledgments

This research was by a grant from the National Institute of Mental Health [MH090308] to Thomas L. Rodebaugh and by a grant from the National Institutes of Health [UL1 RR024992] to Washington University. Marilyn L. Piccirillo was supported by the National Institute of Mental Health [F31 MH 115641]. The authors do not have any conflicts of interest to declare. We wish to thank Richard Zinbarg for his guidance regarding our use of coefficient omega.

Versions of this manuscript have been presented at a national conference. Portions of the data used in this manuscript have been previously published.

Footnotes

1

This study was conducted using DSM-IV diagnostic criteria, which allows for the specification of a generalized subtype of social anxiety disorder, which is not included in DSM-5. We will use the term SAD for simplicity.

2

This kappa value reflects the agreement on diagnostic group for the full sample reported on in the previous study.

3

We used structural equation modeling with the MLR estimator and standard fit indices as well as imputed datasets in a previous version of this manuscript. Results differed slightly in that the original results suggested that SAD status and comorbid depression had a significant effect on specific negative behaviors with results varying across imputed datasets. Critical feedback from Timothy A. Brown and anonymous reviewers led us to use Bayesian statistics in this version of the manuscript.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Alden LE, & Taylor CT (2004). Interpersonal processes in social phobia. Clinical Psychology Review, 24, 857–882. 10.1016/j.cpr.2004.07.006 [DOI] [PubMed] [Google Scholar]
  2. American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders-IV (DSM-IV) (4th ed.). American Psychiatric Association. [Google Scholar]
  3. Asparouhov T, & Muthén B (2020). Advances in Bayesian model fit evaluation for structural equation models. Structural Equation Modeling, 1–36. 10.1080/10705511.2020.1764360 [DOI] [Google Scholar]
  4. Bagwell CL (2005). Friendship quality and perceived relationship changes predict psychosocial adjustment in early adulthood. Journal of Social and Personal Relationships, 22(2), 235–254. 10.1177/0265407505050945 [DOI] [Google Scholar]
  5. Beck AT, Steer RA, & Brown GK (1996). Manual for the Beck Depression Inventory - II. Psychological Corporation. [Google Scholar]
  6. Beck GJ, Davila J, Farrow S, & Grant D (2005). When the heat is on: Romantic partner responses influence distress in socially anxious women. Behaviour Research and Therapy, 44(5), 737–748. 10.1016/j.brat.2005.05.004 [DOI] [PubMed] [Google Scholar]
  7. Berkman LF (1984). Assessing the physical health effects of social networks and social support. Annual Review of Public Health, 4, 413–432. 10.1146/annurev.pu.05.050184.002213 [DOI] [PubMed] [Google Scholar]
  8. Cuming S, & Rapee RM (2010). Social anxiety and self-protective communication style in close relationships. Behaviour Research and Therapy, 48(2), 87–96. 10.1016/j.brat.2009.09.010 [DOI] [PubMed] [Google Scholar]
  9. Dunkel-Schetter C, & Bennett TL (1990). Differentiating the cognitive and behavioral aspects of social support. In Sarason BR, Saradon IG, & Pierce GR (Eds.), Social Support: An Interactional View (pp. 267–296). Wiley. [Google Scholar]
  10. Dunn TJ, Baguley T, & Brunsden V (2014). From alpha to omega: A practical solution to the pervasive problem of internal consistency estimation. British Journal of Psychology, 105(3), 399–412. 10.1111/bjop.12046 [DOI] [PubMed] [Google Scholar]
  11. Fernandez KC, & Rodebaugh TL (2011). Social anxiety and discomfort with friendly giving. Journal of Anxiety Disorders, 25(3), 326–334. 10.1016/j.janxdis.2010.10.005 [DOI] [PubMed] [Google Scholar]
  12. First MB, Spitzer RL, Gibbon M, & Williams JBW (2002). Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition (SCID-I/P). Biometrics Research, New York State Psychiatric Institute. [Google Scholar]
  13. Fydrich T, Chambless DL, Perry KJ, Buergener F, & Beazley MB (1998). Behavioral assessment of social performance: A rating system for social phobia. Behaviour Research and Therapy, 36(10), 995–1010. 10.1016/S0005-7967(98)00069-2 [DOI] [PubMed] [Google Scholar]
  14. Gelman A (2003). Bayesian Data Analysis (2nd ed.). Chapman & Hall. [Google Scholar]
  15. Giles LC, Glonek GFV, Luszcz MA, & Andrews GR (2005). Effect of social networks on 10 year survival in very old Australians: The Australian Longitudinal Study of Aging. Journal of Epidemiological Community Health, 59(7), 574–579. 10.1136/jech.2004.025429 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hart TA, Turk CL, Heimberg RG, & Liebowitz MR (1999). Relation of marital status to social phobia severity. Depression and Anxiety, 10(1), 28–32. [DOI] [PubMed] [Google Scholar]
  17. Hartup W, & Stevens N (1997). Friendship and adaptation in the life course. Psychological Bulletin, 121(3), 335–370. 10.1037/0033-2909.121.3.355 [DOI] [Google Scholar]
  18. Heimberg RG, Horner KJ, Juster HR, Safren SA, Brown EJ, Schneier FR, & Liebowitz MR (1999). Psychometric properties of the Liebowitz Social Anxiety Scale. Psychological Medicine, 29(1), 199–212. 10.1017/S0033291798007879 [DOI] [PubMed] [Google Scholar]
  19. Heinrichs N (2003). Does a partner matter? European Congress of Behavioural and Cognitive Therapy. [Google Scholar]
  20. Hussong A (2000). Perceived peer context and adolescent adjustment. Journal of Research on Adolescence, 10(4), 391–415. 10.1207/SJRA1004_02 [DOI] [Google Scholar]
  21. Kessler RC, Petukhova M, Sampson NA, Zaslavsky AM, & Wittchen H-U (2012). Twelve-month and lifetime prevalence morbid risk of anxiety and mood disorder in the US. International Journal of Methods in Psychiatric Research, 21, 169–184. 10.1002/mpr.1359 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Langer JK, Lim MH, Fernandez KC, & Rodebaugh TL (2017). Social anxiety disorder is associated with reduced eye contact during conversation primed for conflict. Cognitive Therapy and Research, 41(2), 220–229. 10.1007/s10608-016-9813-x [DOI] [Google Scholar]
  23. Lawrence E, Bunde M, Barry RA, Brock RL, Sullivan KT, Pasch LA, White GA, Dowd CE, & Adams EE (2008). Partner support and marital satisfaction: Support amount, adequacy, provision, and solicitation. Personal Relationships, 15(4), 445–463. 10.1111/j.1475-6811.2008.00209.x [DOI] [Google Scholar]
  24. Leary MR, & Dobbins SE (1983). Social anxiety, sexual behavior, and contraceptive use. Journal of Personality and Social Psychology, 45(6), 1347–1354. 10.1037/0022-3514.45.6.1347 [DOI] [PubMed] [Google Scholar]
  25. Liebowitz MR (1987). Social phobia. Modern Problems in Pharmacopsychiatry, 22, 141–173. [DOI] [PubMed] [Google Scholar]
  26. McDonald RP (1999). Test theory: A unified treatment. Lawrence Erlbaum. [Google Scholar]
  27. McGraw KO, & Wong SP (1996). Forming inferences about some intraclass correlation coefficients. Psychological Methods, l(1), 30–46. 10.1037/1082-989X.1.1.30 [DOI] [Google Scholar]
  28. Mead DE (2002). Marital distress, co-occurring depression, and marital therapy: A review. Journal of Marital and Family Therapy, 28(3), 299–314. 10.1111/j.1752-0606.2002.tb01188.x [DOI] [PubMed] [Google Scholar]
  29. Mennin DS, Fresco DM, Heimberg RG, Schneier FR, Davies SO, & Liebowitz MR (2002). Screening for social anxiety disorder in the clinical setting: Using the Liebowitz Social Anxiety Scale. Journal of Anxiety Disorders, 16(6), 661–673. 10.1016/S0887-6185(02)00134-2 [DOI] [PubMed] [Google Scholar]
  30. Mineka S, Watson D, & Clark LA (1998). Comorbidity of anxiety and unipolar mood disorders. Annual Review of Psychology, 49, 377–412. 10.1146/annurev.psych.49.1.377 [DOI] [PubMed] [Google Scholar]
  31. Morrison AS, & Heimberg RG (2013). Social anxiety and social anxiety disorder. Annual Review of Clinical Psychology, 9, 249–274. 10.1146/annurev-clinpsy-050212-185631 [DOI] [PubMed] [Google Scholar]
  32. Moscovitch DA, Orr E, Rowa K, Gehring Reimer S, & Antony MM (2009). In the absence of rose-colored glasses: Ratings of self-attributes and their differential certainty and importance across multiple dimensions in social phobia. Behavior Research and Therapy, 47(1), 66–70. 10.1016/j.brat.2008.10.007 [DOI] [PubMed] [Google Scholar]
  33. Muthén B, & Asparouhov T (2012). Bayesian SEM: A more flexible representation of substantive theory. Psychological Methods, 17(3), 313–335. 10.1037/a0026802 [DOI] [PubMed] [Google Scholar]
  34. Muthén B, & Muthén L (1998–2017). Mplus, version 8.0. Muthén & Muthén. [Google Scholar]
  35. Pasch LA, & Bradbury TN (1998). Social support, conflict, and the development of marital dysfunction. Journal of Consulting and Clinical Psychology, 66(2), 219–230. 10.1037/0022-006X.66.2.219 [DOI] [PubMed] [Google Scholar]
  36. Pasch Lauri A, Bradbury TN, & Davila J (1997). Gender, negative affectivity, and observed social support behavior in marital interaction. Personal Relationships, 4(4), 361–378. 10.1111/j.1475-6811.1997.tb00151.x [DOI] [Google Scholar]
  37. Porter E, & Chambless DL (2014). Shying away from a good thing: Social anxiety in romantic relationships. Journal of Clinical Psychology, 70(6), 546–561. 10.1002/jclp.22048 [DOI] [PubMed] [Google Scholar]
  38. Rapee RM, Peters L, Carpenter L, & Gaston JE (2015). The Yin and Yang of support from significant others: Influence of general social support and partner support of avoidance in the context of treatment for social anxiety disorder. Behaviour Research and Therapy, 69, 40–47. 10.1016/j.brat.2015.03.012 [DOI] [PubMed] [Google Scholar]
  39. Rodebaugh TL (2009). Social phobia and perceived friendship quality. Journal of Anxiety Disorders, 23(7), 872–878. 10.1016/j.janxdis.2009.05.001 [DOI] [PubMed] [Google Scholar]
  40. Rodebaugh TL, Fernandez KC, & Levinson CA (2012). Testing the effects of social anxiety disorder on friendship quality across gender and ethnicity. Cognitive Behaviour Therapy, 41(2), 130–139. 10.1080/16506073.2012.661451 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Rodebaugh TL, Lim MH, Fernandez KC, Langer JK, Weisman JS, Tonge NA, Levinson CA, & Shumaker EA (2014). Self and friend’s differing views of social anxiety disorder’s effect on friendships. Journal of Abnormal Psychology, 123(4), 715–724. 10.1037/abn0000015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Rodebaugh TL, Lim MH, Shumaker EA, Levinson CA, & Thompson T (2015). Social anxiety and friendship quality over time. Cognitive Behaviour Therapy, 44(6), 502–511. 10.1080/16506073.2015.1062043 [DOI] [PubMed] [Google Scholar]
  43. Schafer J (1997). Analysis of incomplete multivariate data. Chapman & Hall. [Google Scholar]
  44. Schneier FR, Heckelman LR, Garfinkel R, Campeas R, Fallon BA, Gitow A, Street L, Del Bene D, & Liebowitz MR (1994). Functional impairment in social phobia. The Journal of Clinical Psychiatry, 55(8), 322–331. [PubMed] [Google Scholar]
  45. Schradle SB, & Dougher MJ (1985). Social support as a mediator of stress: Theoretical and empirical issues. Clinical Psychological Review, 5(6), 641–661. 10.1016/0272-7358(85)90039-X [DOI] [Google Scholar]
  46. Sheehan D, Janavs J, Baker R, Harnett-Sheehan K, Knapp E, & Sheehan M (1998). M.I.N.I. Mini International Neuropsychiatric Interview University of South Florida. PMID: 9881538 [Google Scholar]
  47. Steptoe A, Shankar A, Demakakos P, & Wardle J (2013). Social isolation, loneliness, and all-cause mortality in older men and women. Proceedings of the National Academy of Sciences of the United States of America, 110(15), 5797–5801. 10.1073/pnas.1219686110 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Sullivan KT, Pasch LA, Johnson MD., & Bradbury TN (2010). Social support, problem solving, and the longitudinal course of newlywed marriage. Journal of Personality and Social Psychology, 98(4), 631–644. 10.1037/a0017578 [DOI] [PubMed] [Google Scholar]
  49. Torgrud LJ, Walker JR, Murray L, Cox BJ, Chartier M, & Kjernisted KD (2004). Deficits in perceived social support associated with generalized social phobia. Cognitive Behaviour Therapy, 33(2), 87–96. 10.1080/16506070410029577 [DOI] [PubMed] [Google Scholar]
  50. Trombello JM, Schoebi D, & Bradbury TN (2011). Relationship functioning moderates the association between depressive symptoms and life stressors. Journal of Family Psychology, 25(1), 58–67. 10.1037/a0022138 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Verhofstadt LLL, Lemmens GMD, & Buysse A (2013). Support-seeking, support-provision and support-perception in distressed married couples: A multi-method analysis. Journal of Family Therapy, 35(3), 320–339. 10.1111/1467-6427.12001 [DOI] [Google Scholar]
  52. Vernberg EM, Abwender DA, Ewell KK, & Beery SH (1992). Social anxiety and peer relationships in early adolescence: A prospective analysis. Journal of Clinical Child Psychology, 21(2), 189–196. 10.1207/s15374424jccp2102_11 [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1723933-supplement-1.docx (67.7KB, docx)

RESOURCES