Introduction
Depression is characterized by persistent feelings of sadness or hopelessness and a loss of pleasure or interest in activities (World Health Organization, 2023b). Social anxiety involves an intense fear or worry of being humiliated, embarrassed, or rejected in social situations (World Health Organization, 2023a). Both mental health conditions are highly comorbid and prevalent, imposing a significant burden globally (World Health Organization, 2017). Theoretical frameworks suggest that both conditions are caused and maintained by distorted patterns of interpretation (Beck & Haigh, 2014; Chen et al., 2020). Individuals with depression or social anxiety tend to infer more negative and/or fewer positive interpretations of ambiguous social situations (“negative/less positive interpretation bias”; Hirsch et al., 2016; Huppert et al., 2007). Additionally, they exhibit a tendency to inadequately update initial negative interpretations based on new disconfirmatory experiences (“negative interpretation inflexibility”; Everaert et al., 2018). Importantly, these biased and inflexible negative interpretations prospectively predict these forms of psychopathology (Everaert et al., 2021; Songco et al., 2020).
Theories have elaborated on various socio-affective mechanisms by which distorted interpretations might fuel depressive and/or social anxiety symptoms (Ginat-Frolich et al., 2024; Hammen, 2006; Wittenborn et al., 2016). Systemic models of depression (Wittenborn et al., 2016) propose that overly negative interpretations encourage dysfunctional behaviors that erode social ties. It is postulated that the stress resulting from such poor social interactions may, in turn, consolidate biased information-processing, thereby establishing a vicious cycle involving cognitive and social factors. Moreover, a recent integrative model of social anxiety (Ginat-Frolich et al., 2024) suggests that individuals with elevated levels of social anxiety interpret social cues negatively, reinforcing negative self-views during social interactions. These distorted cognitions not only impact one’s self-concept but also influence how others perceive them, leading to difficulties in developing and maintaining satisfying relationships. This vicious cycle of negative beliefs may intensify social anxiety, making it increasingly challenging to effectively engage in social interactions. According to another influential perspective, the interpersonal stress generation hypothesis, individuals with depression and social anxiety exhibit certain behaviors that provoke negative reactions from others, thereby increasing interpersonal stress and exacerbating symptoms of depression and social anxiety (Hammen, 2006).
To date, research has primarily focused on intrapersonal processes, linking distorted interpretations to enhanced emotional reactivity (Joormann et al., 2015), distorted daily affect (Puccetti et al., 2023), and maladaptive emotion regulation (Everaert et al., 2020). However, little is known about how distorted interpretations are linked to key social processes in psychopathology. This is an important gap because interpretations of ambiguity often occur within interpersonal contexts (Alden & Taylor, 2004; Everaert, 2021). In these contexts, it is unclear how biased or inflexible interpretations might derail socio-affective functioning difficulties relevant to depression and social anxiety.
The limited literature on this topic suggests that distorted interpretations might promote depression and social anxiety through their impact on both intrapersonal and interpersonal dimensions of socio-affective difficulties, including rejection sensitivity (Downey & Feldman, 1996), interpersonal emotion regulation (Hofmann, 2014), negative social interactions (Hammen, 2006), and social integration (Santini et al., 2015). Given their central role in theory and research, this study focused on a selection of risk factors capturing these dimensions of socio-affective difficulties. The following sections describe the relation between depression and social anxiety and each of these domains as well as how they could be modulated by distorted interpretations.
Rejection Sensitivity
Depression and social anxiety are characterized by rejection sensitivity – the tendency to readily perceive and anxiously anticipate social rejection within close relationships (Berenson et al., 2009). Rejection sensitivity develops as a result of early maladaptive attachment experiences (e.g., rejection and abandonment) which are carried through later relationships (Bowlby, 1980). Individuals who anxiously anticipate rejection may excessively worry about acceptance by others, leading to social withdrawal and loneliness (London et al., 2007). The interplay between rejection sensitivity, interpretation bias, and psychopathology is likely complex. Rejection sensitivity may increase depression and social anxiety by promoting negative interpretations of ambiguous social situations (Downey & Feldman, 1996; Normansell & Wisco, 2017), which may themselves encourage rejection sensitivity (Bronstein et al., 2022). Furthermore, inflexible negative interpretations may perpetuate rejection sensitivity by allowing interpretations consistent with rejection to persist even when these interpretations are no longer warranted.
Interpersonal Maladaptive Emotion Regulation
Depression and social anxiety feature emotion regulation difficulties in social interactions and relationships (Hofmann, 2014). Theories hypothesize that negative feedback-seeking (the tendency to focus on information that confirms pre-existing negative beliefs about oneself) and excessive reassurance-seeking (the tendency to repeatedly ask for reassurance about one’s self-worth) foster a social environment conducive to depression and/or social anxiety (Hames et al., 2013; Valentiner et al., 2011). Supporting this hypothesis, individuals with negative belief systems tend to focus on negative feedback and excessively seek reassurance, setting the stage for social rejection and psychopathology (Hames et al., 2013).
Moreover, depression and social anxiety are characterized by maladaptive interpersonal responses to negative and positive emotions. Research indicates that depression and anxiety are associated with co-rumination (repeated discussion of personal problems with close others; Spendelow et al., 2017) and depression is linked to co-dampening (repetitive pattern of discussion with others that downregulates positive emotions; Bastin et al., 2018). Both forms of self-disclosure during social interactions encourage a focus on negative affect and limit pleasurable experiences that otherwise may alleviate depression/social anxiety. While research has yet to examine interpretation processes in relation to interpersonal feedback-seeking and responses to affect, recent eating disorder research suggests that biased/inflexible interpretations may affect various interpersonal emotion regulation processes (Bronstein et al., 2022).
Negative Social Interactions
Individuals suffering from depression or social anxiety behave in ways that elicit negative reactions from others, which in turn increase depression or anxiety levels (Hames et al., 2013; Hammen, 2006). Research has extensively supported this stress generation hypothesis (Rnic et al., 2023). Indeed, it is plausible that negative interpretations of social situations and inflexible revision based on disconfirmatory experiences trigger inappropriate emotions and behaviors (e.g., feeling rejected, social withdrawal) that provoke negative reactions from others. Frequent negative social interactions may further contribute to biased and inflexible interpretations, erode relationships, and promote psychopathology.
Social Integration
Individuals with depression and social anxiety experience difficulties in forming and maintaining social relationships (Alden & Taylor, 2004; Santini et al., 2015). Research suggests that chronic depression is associated with smaller social networks (Santini et al., 2015), and social anxiety is linked to fewer friends, dating/sexual partners, and marriages (Alden & Taylor, 2004). In addition to these structural features of social relationships, loneliness plays a critical role in depression and social anxiety. This subjective experience of isolation, resulting from the discrepancy between a person’s desired and actual social relationships, predicts changes in depression and social anxiety (Lim et al., 2016). Both subjective and objective markers of social integration may be linked to distorted interpretations. Previous research testing this notion indicates that negatively biased interpretations may not directly cause loneliness in adolescents, but rather maintain feelings of loneliness, even after controlling for depression and anxiety (Lau et al., 2021). The potential influence of interpretations on social integration is likely indirect. That is, biased and inflexible interpretation processes may prompt behaviors (e.g., excessive reassurance-seeking) that elicit negative social interactions (e.g., conflicts with close others). Frequent occurrences of such social stressors may lead to the dissolution of relationships, reduce a person’s social networks, diminish feelings of connectedness, and elicit depressive and social anxiety symptoms over time.
This Study
This study aimed to elucidate the role of distorted interpretations of ambiguity in key socio-affective functioning difficulties implicated in depression and social anxiety. By adopting a data-driven approach, the study investigated unique associations between cognitive and socio-affective processes in relation to symptoms of both depression and social anxiety. This analytic approach not only tests the preregistered hypotheses (see below) but also identifies potential relations that were not hypothesized, thereby laying the groundwork for future studies and expanding our theoretical understanding. Through this examination, this study sought to bridge the traditionally separate literatures on cognitive and interpersonal mechanisms underlying depression and social anxiety, thereby informing more comprehensive and integrative theoretical models of risk mechanisms for these mental health conditions.
The preregistered hypotheses (osf.io/7c6zp) were as follows: Higher levels of negative interpretation bias and negative interpretation inflexibility, as well as lower levels of positive interpretation bias, are related to greater depression and social anxiety symptom severity (hypothesis 1). More biased and inflexible negative interpretations and fewer positive interpretations are related to difficulties in socio-affective functioning (rejection sensitivity, negative feedback-seeking, excessive reassurance-seeking, co-rumination, co-dampening, negative social interactions, and loneliness) (hypothesis 2a). Increased positive interpretations and decreased negatively biased and inflexible interpretations are associated with greater social network diversity, size, and embeddedness (hypothesis 2b).
In addition to the preregistered hypotheses, this study explored pathways from interpretation bias and inflexible negative interpretation to symptom severity of depression and social anxiety via various socio-affective functioning difficulties (hypothesis 3). Drawing on recent conceptual contributions (Everaert, 2021; Ginat-Frolich et al., 2024; Wittenborn et al., 2016), the general hypothesis that socio-affective risk factors would mediate the relation between interpretation processes and psychopathology was tested for specific socio-affective risk factors.
Method
Participants and Sampling Strategy
A total of 311 participants enrolled via Amazon’s Mechanical Turk (MTurk) in November 2019. MTurk provides an online platform with access to large crowdsourced samples suitable for research collecting mental health data (Chandler & Shapiro, 2016). To ensure sufficient variability in depression and anxiety levels, MTurk ads invited individuals who were “feeling blue, down, or anxious lately”. Sixteen participants with any missing data were excluded, resulting in a total sample of 295 (Mage = 36.97, SD = 11.02, 54% male). The final sample (n = 295) and the sample with missing values (n = 16) did not differ with respect to key study variables. Table 1 provides demographic information.
Table 1
Demographic information of the participants
Characteristic | % |
|---|---|
Ethnicity | |
White | 69.15 |
Hispanic, Latino/a, or Spanish origin | 3.73 |
Black or African American | 14.58 |
Asian | 5.08 |
Native American or Alaska Native | 1.70 |
Middle Eastern or North African | 0.34 |
Multiracial | 5.42 |
Education | |
Some high school | 2.03 |
High school graduate diploma or equivalent | 20.34 |
Trade/technical/vocational school | 14.92 |
Bachelor’s degree | 47.46 |
Master’s degree | 12.88 |
Professional degree | 1.69 |
Doctorate degree | 0.68 |
Marital status | |
Single or never married | 43.39 |
Married or domestic partnership | 47.80 |
Widowed/Widower | 1.69 |
Divorced | 5.42 |
Separated | 0.68 |
Missing | 1.02 |
The sample size was determined using G*Power 3.1 (Faul et al., 2009). A minimum sample size of 250 was needed to obtain 0.80 power (α = 0.05) to detect small-to-medium effects in the regression analyses. Additionally, simulation studies indicate a minimum sample of 250 participants to power network analyses to detect edges that are significantly different from zero (Epskamp et al., 2018). The minimum sample size was increased by 25% to account for potential missing data and responses of low data quality.
Data Quality Requirements
Following recommendations for research using crowdsourced samples (Chandler & Shapiro, 2016), several steps were taken to ensure high data quality and power. First, MTurk participants were required to have a history of providing good-quality responses (i.e., an acceptance ratio of ≥ 98% across at least 500 submissions). Second, three attention questions were used to discriminate attentive from inattentive participants. Nineteen participants failed one of the three attention questions. None of the participants failed two or three attention questions. Sensitivity analysis with and without these participants produced nearly identical results. The full sample was considered in the final analysis. Third, a captcha was included to prevent bots from starting the survey. Finally, duplicate IP addresses and suspicious geolocations were automatically blocked and IP addresses were verified for consistency with a US location through CloudResearch (an online platform that provides additional services for researchers using MTurk). Note that our previous studies following these recommendations have produced replicable results (Everaert et al., 2018, 2020).
In addition to these data quality requirements, the R-package careless (Version 1.2.1; Yentes & Wilhelm, 2018) was used to identify participants with highly patterned responses potentially indicative of careless responding. The intra-individual response variability (IRV) of the emotional Bias Against Disconfirmatory Evidence task data (described below) was examined to detect degraded response quality. Individuals with an IRV score that was below (or above) the third (or first) quartile of scores by more than 1.5 times the interquartile range (IQR) were marked. Five participants (three with low IRV, two with high IRV) with highly patterned responses were identified. Sensitivity analyses with and without these five participants produced highly similar results for the preregistered regression analyses and follow-up network analysis. Therefore, the participants with high and low IRV scores were retained in the final analyses.
Procedure and Measures
Participants gave informed consent in accordance with the Yale University Institutional Review Board. Participants first completed a demographic questionnaire followed by the interpretation bias and inflexibility measure. Participants then completed the questionnaires in randomized order. The survey lasted on average 57 min (SD = 33.04) to complete. Participants were debriefed and remunerated ($8).
Interpretation Bias and Inflexibility
Interpretation bias and inflexibility were measured using the emotional Bias Against Disconfirmatory Evidence (BADE) task (Everaert et al., 2018; Everaert et al., 2021; Everaert et al., 2020). The emotional BADE task consists of scenarios that describe daily life situations (Everaert et al., 2018). Scenarios are social, emotional, and self-referential to capture depression and social anxiety-related concerns regarding rejection and personal failure. Participants were instructed to imagine each scenario as if it were happening to them at that moment.
Each scenario consists of three statements which were presented in series. Each statement provided additional information about the unfolding social situation. After viewing each statement, participants rated the plausibility of four interpretations of the information acquired thus far in that scenario on a 21-point rating scale (1 = poor to 21 = excellent). The task requires participants to revise their interpretations for a given scenario by integrating the disconfirmatory information provided by each of the latter two statements (see Fig. 1). There are two types of scenarios in the emotional BADE task: disconfirming-the-negative scenarios (requiring revisions of negative interpretations) and disconfirming-the-positive scenarios (requiring revision of positive interpretations). Given previous research (e.g., Everaert et al., 2018) suggesting that metrics of interpretation inflexibility derived from disconfirming-the-negative scenarios may be more strongly related to psychopathology symptoms, this study used a short form of the emotional BADE task that administered 12 disconfirming-the-negative scenarios randomly interleaved with 4 disconfirming-the-positive scenarios.
Fig. 1
Schematic depiction of emotional BADE task scenario: “Your Birthday”. Across scenarios, the potential interpretations fell into three categories: Absurd interpretations remained implausible after all statements. Lure interpretations were initially most plausible but became less plausible after the third statement. True interpretations were initially less plausible than the Lure interpretations but became the most plausible after the third statement
×
![]()
As in prior work (Everaert et al., 2021), task metrics were scored as follows: Negative interpretation inflexibility (Absurd1 + Absurd2 + Absurd3 + LureA3 + LureB3 − True3), positive interpretation bias (True1 + True2 + True3), and negative interpretation bias (LureA1 + LureB1 + LureA2 + LureB2). In these formulas, numbers in variable names represent the statement after the rating was made, and the remainder of the name denotes the category of explanation being rated. Outlier analysis was performed before computing the three emotional BADE task metrics (see Supplement 1). Importantly, a validation study supported the reliability and convergent validity of the emotional BADE task (Deng et al., 2022). In support of their validity, emotional BADE task parameters correlate as expected with corresponding bias/flexibility parameters derived from a picture-based belief revision task (Deng et al., 2022). In this study, the internal consistency was McDonald’s ωtotal = 0.99 for Absurd interpretations, ωtotal = 0.99 for Lure interpretations, and ωtotal = 0.87 for True interpretations.
Psychopathology Measures
Depressive Symptom Severity
Depressive symptom severity over the last two weeks was measured using the Beck Depression Inventory–II (BDI-II; Beck et al., 1996). Participants rate 21 items on a 4-point scale (range: 0–3). For example, participants use the following scale to indicate their level of sadness: 0 = I do not feel sad, 1 = I feel sad much of the time, 2 = I am sad all the time, 3 = I am so sad or unhappy that I can’t stand it. Scores are summed to produce total scores (theoretical range: 0–63). The BDI-II has good overall reliability and validity (Dozois et al., 1998). The internal consistency of the BDI-II measured in the current study was excellent (Cronbach’s alpha \(\:\alpha\:\) = 0.95; McDonald’s ωtotal = 0.96).
Social Anxiety Symptom Severity
Social anxiety symptoms were measured using the Severity Measure for Social Anxiety Disorder (Social Phobia) – Adult (SAD-D; Craske et al., 2013). The SAD-D consists of 10 items rated on a 5-point scale. For example, participants rate the item “During the past 7 days, I have felt anxious, worried, or nervous about social situations” on a scale from 0 = never to 4 = all of the time. The total score is the sum of all items. Higher scores indicate greater social anxiety symptom severity (range: 0–40). Research examining the psychometric properties of this questionnaire has shown high test-retest reliability (Knappe et al., 2014). In addition, the SAD-D showed good internal reliability, concurrent validity, and divergent validity measured in an adult community sample (Rice et al., 2021). The internal consistency of the SAD-D measured in the current study was excellent (\(\:\alpha\:\) = 0.94; ωtotal = 0.96).
Rejection Sensitivity Measure
The 9-item adult version of the Rejection Sensitivity Questionnaire (A-RSQ; Berenson et al., 2009; Downey & Feldman, 1996) was used to assess associated anxiety and expected rejection by close others. The A-RSQ consists of 9 hypothetical social situations where rejection is possible. For example, in response to the situational description “You approach a close friend to talk after doing or saying something that seriously upset him/her”, participants rate the following items: “How concerned or anxious would you be over whether or not your friend would want to talk to you?” (1 = very unconcerned to 6 = very concerned) and “I would expect that he/she would want to talk with me to try to work things out” (1 = very unlikely to 6 = very likely). Rejection sensitivity scores are the product of anxiety ratings and reverse-scored expectations of acceptance (range: 1–36). The total rejection sensitivity score is the mean of the rejection sensitivity scores for the 9 situations. Prior research has shown good construct validity (Berenson et al., 2009). The internal consistency assessed in the current study was good (\(\:\alpha\:\) = 0.79; ωtotal = 0.84).
Interpersonal Emotion Regulation Measures
Negative Feedback-Seeking
The Feedback Seeking Questionnaire (FSQ; Swann et al., 1992) was used to index participants’ interest in receiving positive or negative feedback from close others in five domains (intellectual, social, musical/artistic, athletic abilities, and physical attractiveness). Each domain consisted of three positive (e.g., “What about your friend makes you think s/he would be confident in social situations?”) and three negative items (e.g., “In terms of social competence, what is your friend’s worst asset?”). Participants were asked to choose two items which they would most like to have their closest same-sex friend answer about them. A negative feedback score was computed by counting the number of negative items selected, with higher scores indicating greater negative feedback-seeking (range: 0–10). The reliability across the five domains was acceptable (\(\:\alpha\:\) = 0.71; ωtotal = 0.74).
Excessive Reassurance-Seeking
The 4-item Reassurance Seeking subscale of the Depressive Interpersonal Relationship Inventory (DIRI-RS; Joiner et al., 1992) was used to measure participants’ tendency to excessively seek reassurance within close relationships and the individuals’ perception of others’ reactions to their reassurance-seeking. For each item, participants rate the extent to which they excessively seek reassurance or how they perceive others’ reaction to their reassurance-seeking behavior. For example, the item “Do you frequently seek reassurance from the people you feel close to as to whether they really care about you?” is rated on a scale ranging from 1 = not at all to 7 = very much. A sum-score is computed (range: 4–28). The DIRI-RS demonstrated good construct validity in samples of clinically depressed, anxious, and non-clinical participants (Cougle et al., 2012; Joiner & Metalsky, 2001). In this study, the internal consistency was excellent (\(\:\alpha\:\) = 0.95; ωtotal = 0.96).
Co-Rumination
The Co-Rumination Questionnaire was used to assess the tendency to co-ruminate with close same-sex friends (CRQ; Rose, 2002). Across 27 items, participants were asked to rate the extent to which each statement described them on a 5-point scale. For example, “When one of us has a problem, we talk to each other about it for a long time” (1 = not at all true to 5 = really true). A sum score is computed (range: 27–135), with higher scores indicating greater use of co-rumination. The CRQ provided an excellent internal consistency in prior work (Rose, 2002). The reliability measured in this study was excellent (\(\:\alpha\:\) = 0.97; ωtotal = 0.97).
Co-Dampening
The use of co-dampening within same-sex close-friend relationships was measured with the 9-item Co-Dampening subscale of the Co-Dampening and Co-Enhancing Questionnaire (CoDEQ; Bastin et al., 2018). On each item, participants indicate how often they interact with their closest friend in a certain way using a 4-point scale. For example, participants rate the item “When one of you feels glad or happy (for example, because something nice that has happened) and you talk about this, then we also talk about times that we were not as happy” on a scale from 1 = almost never to 4 = almost always. The total score was calculated by adding together the ratings for each item of the scale (range: 9–36). In previous studies, the co-dampening subscale had high internal consistency (Bastin et al., 2018). In this study, the internal consistency of the Co-Dampening subscale was excellent (\(\:\alpha\:\) = 0.93; ωtotal = 0.94).
Negative Social Interactions Measure
Negative social interactions were measured using the 24-item Test of Negative Social Exchange (TENSE; Finch et al., 1999). Participants indicate how often they experienced different types of negative social interactions in the past month, including anger, insensitivity, manipulation, ridicule, impatience, and rejection. For example, the item “During the last month, indicate the frequency with which someone argued with me” is rated on a 0 = not at all to 9 = frequently scale. As recommended by Finch et al. (1999), a total score of negative social interactions was computed, with higher scores indicating more negative social interactions (range: 0–216). The TENSE has good reliability and validity (Finch et al., 1999; Ruehlman & Karoly, 1991). In this study, the internal consistency was considered excellent (\(\:\alpha\:\) = 0.98; ωtotal = 0.98).
Social Integration Measures
Loneliness
Participants’ self-report feelings of loneliness were assessed using the 5-item NIH Toolbox Loneliness scale (Salsman et al., 2013). Participants rate the extent to which they felt alone in the past month (1 = never to 5 = always). Example items: “I felt left out” and “I felt lonely”. A higher total score (range: 5–25) indicates greater perceived loneliness. The NIH Toolbox Loneliness scale has demonstrated good reliability and validity in an adult community sample (Salsman et al., 2013). The reliability measured in the current study was excellent (\(\:\alpha\:\) = 0.94; ωtotal = 0.95).
Social Network
The Social Network Index (Cohen et al., 1997) was used to measure participants’ network diversity, network size, and embedded networks. This questionnaire measures participation in 12 different social relationships (e.g., spouse, relatives, close friends, family member, parents, etc.). The different metrics derived from the scale represent the total number of people with whom the participant has regular contact (network size), the number of social roles in which the participant has regular contact with at least one person (network diversity), and the number of different network domains in which a participant is active (embedded networks). Example items are: “How many close friends do you have? (meaning people that you feel at ease with, can talk to about private matters, and can call on for help)” (0 = 0 to 7 = 7 or more) and “How many of these friends do you see or talk to at least once every 2 weeks?” (0 = 0 to 7 = 7 or more).
Data-Analytic Plan
Statistical analyses were performed using R (Version 4.1.1; R Core Team, 2013). The preregistered statistical plan (osf.io/7c6zp) included a series of multiple regression models to investigate direct and unique relations between emotional BADE components (negative interpretation inflexibility, negative interpretation bias, and positive interpretation bias) and other study variables (see hypotheses 1, 2a, and 2b). The models included metrics of negative interpretation inflexibility, negative interpretation bias, and positive interpretation bias which were simultaneously entered as predictors. A total of 12 models were tested, one for each criterion variable: Depressive symptoms, social anxiety symptoms, rejection sensitivity, negative feedback-seeking, excessive reassurance-seeking, co-rumination, co-dampening, negative social interactions, loneliness, network diversity, network size, and embedded networks. Bonferroni correction was applied to reduce risk of Type I error (i.e., false positives), resulting in a significance threshold of 0.05 / 12 = 0.00417. Regression model parameters were estimated using bootstrapping procedures using the R-package boot (Version 1.3–28; Canty & Ripley, 2016). Data relevant to each model were bootstrapped 5000 times and 99.583% confidence intervals were generated. Confidence intervals that do not overlap with zero suggest statistically significant effects.
For hypothesis 3, psychometric network analysis was used to examine indirect relations between study variables parsimoniously. The advantage of network analysis (over regression models) is that relations between a large number of study variables can be simultaneously studied, allowing inferences about their unique (i.e., not accounted by other variables in the network) direct and indirect relations. This would not be possible when looking at all variables separately (Haslbeck & Waldorp, 2018).
Psychological networks are abstract models consisting of a set of nodes that represent the study variables and a set of edges that represent statistical relationships between nodes accounting for all other nodes in the network (Epskamp & Fried, 2018). The network included negative interpretation inflexibility, negative and positive interpretation bias, depression and social anxiety symptom severity, rejection sensitivity, negative feedback-seeking, excessive reassurance-seeking, co-rumination, co-dampening, negative social interactions, loneliness, and three social network indices (diversity, size, and embeddedness).
The network model was estimated as a Gaussian Graphical Model (GGM) and visualized using the R-package qgraph (Version 1.9.2; Epskamp et al., 2012). The GGM was estimated based on partial correlation coefficients (Epskamp & Fried, 2018). Data was transformed prior to estimating the network model using the huge.npn function from the huge R-package (Liu et al., 2009). This nonparanormal transformation adjusts for non-normality by using cumulative distributions, which encode the probability that a variable is below a certain threshold, thereby transforming the distribution of the observed variable into that of the latent normally distributed variable (Epskamp & Fried, 2018). The graphical least absolute shrinkage and selection operator (GLASSO) was used to regularize the GGM by pulling edge weights toward zero (Friedman et al., 2007). To estimate a parsimonious and interpretable network, the GLASSO tuning parameter (lambda) was set to 0.5, prioritizing avoidance of Type I errors (Foygel & Drton, 2010) to remove spurious edges while optimizing the number of true edges. The most important or influential constructs in the network (i.e., centrality analyses) as well as the accuracy and stability of the network estimates were examined (see Supplement 2). Note that stability analyses supported the robustness of the network findings (correlation stability coefficient = 0.67; which is well above the recommended thresholds, see Epskamp et al., 2018).
Results
Sample Characteristics
BDI-II scores represented almost the full range of symptom severity: 120 participants reported minimal symptoms (range: 0–13), 38 reported mild symptoms (range: 14–19), 59 reported moderate symptoms (range: 20–28), and 78 reported severe depressive symptoms (range: 29–63) (Beck et al., 1996). On the SAD-D, a total of 203 participants reported minimal to moderate symptoms (range: 0–18) and 92 reported severe social anxiety symptoms (range: 19–40) (Rice et al., 2021). Table 2 provides Spearman’s correlations between all study variables.
Table 2
Spearman’s rho correlation and edge weights partial correlation matrix
M (SD) | 1. | 2. | 3. | 4. | 5. | 6. | 7. | 8. | 9. | 10. | 11. | 12. | 13. | 14. | 15. | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1. Negative interpretation inflexibility | 6.5 (11.02) | 0.04 | 0.04 | 0.00 | 0.00 | 0.00 | 0.00 | 0.05 | 0.00 | 0.33 | 0.08 | 0.00 | 0.09 | 0.00 | 0.00 | |
2. Negative interpretation bias | 46.85 (12.50) | 0.29*** | 0.03 | 0.00 | 0.14 | 0.01 | 0.05 | 0.04 | 0.17 | 0.00 | 0.05 | 0.00 | 0.00 | 0.00 | 0.00 | |
3. Positive interpretation bias | 34.36 (7.15) | 0.33*** | 0.25* | 0.00 | 0.00 | − 0.21 | 0.00 | 0.14 | 0.10 | 0.06 | 0.04 | 0.00 | 0.18 | 0.00 | 0.00 | |
4. Depression | 19.16 (13.52) | 0.25*** | 0.29** | 0.11 | 0.35 | 0.07 | 0.06 | 0.00 | 0.00 | 0.00 | 0.12 | 0.35 | 0.00 | 0.00 | − 0.02 | |
5. Social anxiety | 12.77 (10.30) | 0.40*** | 0.42*** | 0.20*** | 0.75*** | 0.09 | 0.00 | 0.17 | 0.10 | 0.10 | 0.13 | 0.17 | 0.00 | 0.00 | 0.00 | |
6. Rejection sensitivity | 10.56 (4.72) | 0.08 | 0.12* | − 0.25*** | 0.35*** | 0.35*** | 0.00 | 0.02 | 0.00 | 0.00 | 0.05 | 0.12 | − 0.05 | − 0.07 | 0.00 | |
7. Negative feedback-seeking | 3.21 (2.19) | 0.21*** | 0.22*** | 0.06 | 0.26*** | 0.21*** | 0.08 | 0.00 | 0.00 | 0.12 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | |
8. Reassurance-seeking | 11.04 (7.47) | 0.50*** | 0.34*** | 0.39*** | 0.44*** | 0.59*** | 0.17** | 0.14* | 0.05 | 0.24 | 0.18 | 0.03 | 0.09 | 0.00 | 0.00 | |
9. Co-rumination | 77.29 (24.68) | 0.34*** | 0.39*** | 0.40*** | 0.31*** | 0.45*** | 0.00 | 0.18** | 0.49*** | 0.22 | 0.00 | 0.00 | 0.14 | 0.03 | 0.01 | |
10. Co-dampening | 15.63 (6.45) | 0.63*** | 0.32*** | 0.34*** | 0.49*** | 0.58*** | 0.13* | 0.30*** | 0.66*** | 0.55*** | 0.19 | 0.01 | 0.00 | 0.00 | 0.00 | |
11. Negative social interactions | 100.07 (55.76) | 0.45*** | 0.35*** | 0.27*** | 0.63*** | 0.66*** | 0.28*** | 0.21*** | 0.62*** | 0.39*** | 0.62*** | 0.24 | 0.01 | 0.00 | 0.00 | |
12. Loneliness | 13.56 (5.52) | 0.31*** | 0.26*** | 0.09 | 0.73*** | 0.66*** | 0.38*** | 0.18** | 0.45*** | 0.20*** | 0.44*** | 0.64*** | − 0.07 | − 0.02 | 0.00 | |
13. Network diversity | 5.30 (2.59) | 0.37*** | 0.17** | 0.48*** | 0.00 | 0.17** | − 0.27*** | 0.09 | 0.37*** | 0.47*** | 0.32*** | 0.22*** | − 0.08 | 0.47 | 0.05 | |
14. Network size | 18.23 (14.52) | 0.29*** | 0.09 | 0.34*** | − 0.04 | 0.10 | − 0.28*** | 0.08 | 0.23*** | 0.41*** | 0.24*** | 0.15** | − 0.10 | 0.80*** | 0.65 | |
15. Number of embedded networks | 1.92 (1.20) | 0.24*** | 0.09 | 0.24*** | − 0.06 | 0.07 | − 0.22*** | 0.06 | 0.19* | 0.35*** | 0.20*** | 0.09 | − 0.11 | 0.65*** | 0.85*** |
Regression Analysis
Table 3 shows the results of the fitted regression models. Results of the regression models showed that the three emotional BADE components explained significant amounts of variance in all criterion variables. With respect to hypothesis 1, inflexible negative interpretations and negative interpretation bias were significantly associated with both depression and social anxiety. However, positive interpretation bias was not significantly related to depression or social anxiety levels.
Table 3
Bootstrapped regression models with interpretation processes as predictors and socio-affective functioning or psychopathology measures as the criterion variable
Regression model | B | SE | R2 | 99.583% CI | |
|---|---|---|---|---|---|
LL | UL | ||||
Depressive symptoms | 0.14 | 0.04 | 0.26 | ||
Intercept | 10.72 | 4.66 | -2.20 | 24.80 | |
NII | 0.18 | 0.05 | 0.04 | 0.30 | |
NIB | 0.23 | 0.06 | 0.03 | 0.40 | |
PIB | -0.09 | 0.12 | -0.46 | 0.25 | |
Social anxiety symptoms | 0.33 | 0.18 | 0.48 | ||
Intercept | 1.23 | 3.26 | -7.90 | 11.20 | |
NII | 0.20 | 0.03 | 0.11 | 0.28 | |
NIB | 0.24 | 0.04 | 0.12 | 0.36 | |
PIB | -0.27 | 0.08 | -0.26 | 0.22 | |
Rejection sensitivity | 0.16 | 0.05 | 0.30 | ||
Intercept | 16.22 | 1.54 | 11.60 | 20.85 | |
NII | 0.03 | 0.01 | -0.01 | 0.07 | |
NIB | 0.10 | 0.02 | 0.03 | 0.17 | |
PIB | -0.29 | 0.04 | -0.41 | -0.17 | |
Negative feedback-seeking | 0.09 | 0.02 | 0.19 | ||
Intercept | 3.30 | 0.75 | 1.27 | 5.60 | |
NII | 0.03 | 0.01 | 0.01 | 0.05 | |
NIB | 0.03 | 0.01 | -0.01 | 0.06 | |
PIB | -0.04 | 0.02 | -0.12 | 0.02 | |
Reassurance-seeking | 0.48 | 0.33 | 0.61 | ||
Intercept | 1.11 | 2.00 | -4.93 | 7.30 | |
NII | 0.19 | 0.02 | 0.12 | 0.25 | |
NIB | 0.09 | 0.03 | 0.01 | 0.19 | |
PIB | 0.13 | 0.06 | -0.05 | 0.30 | |
Co-rumination | 0.30 | 0.19 | 0.41 | ||
Intercept | 25.26 | 7.08 | 5.20 | 46.37 | |
NII | 0.25 | 0.07 | 0.07 | 0.45 | |
NIB | 0.52 | 0.11 | 0.20 | 0.86 | |
PIB | 0.75 | 0.22 | 0.13 | 1.40 | |
Co-dampening | 0.64 | 0.48 | 0.73 | ||
Intercept | 9.82 | 1.63 | 5.45 | 14.61 | |
NII | 0.22 | 0.01 | 0.18 | 0.26 | |
NIB | 0.07 | 0.02 | 0.01 | 0.14 | |
PIB | 0.04 | 0.05 | -0.10 | 0.17 | |
Negative social interactions | 0.37 | 0.21 | 0.51 | ||
Intercept | 26.29 | 16.20 | -15.48 | 72.11 | |
NII | 1.16 | 0.15 | 0.69 | 1.56 | |
NIB | 1.04 | 0.23 | 0.36 | 1.70 | |
PIB | 0.51 | 0.47 | -0.81 | 1.82 | |
Loneliness | 0.15 | 0.05 | 0.27 | ||
Intercept | 11.70 | 1.90 | 6.51 | 17.34 | |
NII | 0.09 | 0.02 | 0.04 | 0.13 | |
NIB | 0.09 | 0.03 | 0.01 | 0.07 | |
PIB | -0.08 | 0.05 | -0.22 | 0.07 | |
Network diversity | 0.37 | 0.22 | 0.51 | ||
Intercept | 1.24 | 0.95 | -1.60 | 3.70 | |
NII | 0.05 | 0.01 | 0.02 | 0.07 | |
NIB | -0.01 | 0.01 | -0.03 | 0.03 | |
PIB | 0.12 | 0.02 | 0.05 | 0.18 | |
Network size | 0.13 | 0.04 | 0.26 | ||
Intercept | 7.54 | 5.25 | -8.20 | 22.90 | |
NII | 0.18 | 0.05 | 0.04 | 0.34 | |
NIB | -0.04 | 0.07 | -0.25 | 0.16 | |
PIB | 0.34 | 0.13 | -0.06 | 0.71 | |
Embedded networks | 0.11 | 0.02 | 0.23 | ||
Intercept | 1.24 | 0.50 | -0.20 | 2.70 | |
NII | 0.02 | 0.01 | 0.01 | 0.03 | |
NIB | -0.01 | 0.01 | -0.02 | 0.02 | |
PIB | 0.02 | 0.01 | -0.02 | 0.05 | |
Consistent with hypothesis 2a, inflexible negative interpretations were positively and uniquely associated with almost all markers of socio-affective functioning, namely negative feedback seeking, reassurance-seeking, co-dampening, and negative social interactions, and loneliness. With respect to interpretation biases, negative interpretation bias was positively related to rejection sensitivity, reassurance-seeking, co-rumination, co-dampening, negative social interactions, and loneliness. Positive interpretation bias was negatively related to rejection sensitivity. However, in contrast to the hypotheses, positive interpretation bias was positively related to co-rumination.
Examining hypothesis 2b, inflexible negative interpretations were positively associated with all social network metrics, namely network diversity, network size, and embedded networks. However, the observed relations were opposite to the hypothesized effects. Negative interpretation bias was not related to social network features of diversity, size, and embeddedness. As hypothesized, positive interpretation bias was positively related to network diversity, but not other metrics.
Network Analysis
Figure 2 depicts the regularized partial correlation network structure. Table 2 shows the edge weights partial correlation matrix representing the association between two nodes, controlling for the influence of all other nodes. Testing hypothesis 3, negative interpretation inflexibility had several indirect relations with depression and social anxiety symptoms. Negative interpretation inflexibility had a positive and unique relation with negative social interactions, which was in turn related to the severity of depression and social anxiety symptoms. In addition, negative interpretation inflexibility was related to co-dampening and excessive reassurance-seeking, which were both, in turn, linked to social anxiety symptom severity. Yet, contrary to the hypotheses, rejection sensitivity, negative feedback-seeking, co-rumination, perceived loneliness, and social network indices did not connect negative interpretation inflexibility with psychopathological symptoms.
Fig. 2
Regularized partial correlation network structure. Blue edges (lines connecting two nodes) represent positive associations. Red edges represent negative associations. Thicker edges denote stronger partial correlations between nodes (the measured constructs)
×
![]()
Negative interpretation bias was indirectly linked to social anxiety symptom severity via rejection sensitivity, co-rumination, excessive reassurance-seeking, and negative social interactions. Negative interpretation bias was also indirectly linked to depressive symptom severity via rejection sensitivity, negative social interactions, and negative feedback-seeking. Contrary to our expectations, no indirect links via co-dampening, perceived loneliness, and social network indices emerged.
Finally, positive interpretation bias was indirectly related to both depression and social anxiety symptom severity through its negative association with rejection sensitivity. Positive interpretation bias was also positively associated with excessive reassurance-seeking, co-dampening, and co-rumination, which were all related to the severity of social anxiety symptoms. Moreover, positive interpretation bias was positively related to negative social interactions, which was in turn related to both symptom severity of depression and social anxiety. In contrast to the hypotheses, no indirect links between positive interpretation bias and psychopathology symptoms via negative feedback-seeking, perceived loneliness, and social network characteristics were found.
The network plot also provides additional insights into hypotheses 1 and 2. In contrast with the regression models, the network plot (accounting for socio-affective functioning difficulties) did not show direct and unique relations between inflexible negative interpretations and self-reported psychopathology symptoms. Moreover, the network plot revealed that negative interpretation bias was directly and uniquely related to social anxiety symptom severity, but not to depressive symptoms when accounting for socio-affective functioning difficulties. Consistent with findings from the regression analyses, no direct associations between positive interpretation bias and symptoms of depression or social anxiety emerged.
Moreover, the network plot revealed that negative interpretation inflexibility was directly related to higher levels of reassurance-seeking, co-dampening, negative social interactions as well as network diversity. Negative interpretation bias was related to higher levels of rejection sensitivity, negative feedback-seeking, reassurance-seeking, co-rumination, and negative social interactions. Finally, positive interpretation bias was negatively related to rejection sensitivity, but positively related to reassurance-seeking, co-rumination, co-dampening, negative social interaction, and network diversity. These observed associations suggest that various links with socio-affective difficulties found in the regression models disappear after accounting for all variables considered in this study.
Discussion
This study extends previous examinations of intrapersonal correlates of biased/inflexible interpretations to an interpersonal context. Using an exploratory approach, it addresses a critical gap in the literature by identifying potential socio-affective pathways linking distorted interpretations to depression and social anxiety. These new insights provide directions for future hypothesis-driven research. The findings support the general hypothesis that distorted interpretations are related to socio-affective functioning difficulties, forming indirect pathways that connect these interpretations to symptoms of depression and social anxiety. The observed indirect pathways align with cognitive-interpersonal theories (Ginat-Frolich et al., 2024; Wittenborn et al., 2016), suggesting that overly negative interpretations of social interactions can lead to difficulties in interpersonal functioning, which in turn, exacerbate depression or social anxiety by hindering effective navigation of social environments. The findings are also consistent with a broader stress generation perspective, which posits that individuals with depression and social anxiety exhibit behaviors (e.g., excessive reassurance-seeking) that generate interpersonal stress, thereby perpetuating symptoms (Hammen, 2006).
Consistent with these theories, the regression models and network plot collectively suggest that individuals who have difficulty revising negative interpretations based on positive social experiences may tend to more frequently co-dampen positive emotions, experience more negative social interactions, and seek excessive reassurance about their self-worth. Individuals who experience more negative social interactions are, in turn, more likely to experience more depressive or social anxiety symptoms. The network model revealed that individuals who are inflexible in revising initial negative interpretations also tend to dampen positive emotions within close relationships, potentially elevating their levels of social anxiety. These results indicate that inflexible negative interpretations may set the stage for a stressful interpersonal environment that dampens positive emotions and enhances negative emotions, conducive to the formation and maintenance of depression and social anxiety (Bastin et al., 2018; Hames et al., 2013). This extends previous research linking inflexible negative interpretations to intrapersonal dampening of positive emotions (Everaert et al., 2020) by generalizing this association to interpersonal dampening behavior.
The findings concerning interpretation bias are consistent with the stress generation hypotheses for depression (Hammen, 2006). Individuals who infer more negative interpretations of ambiguous social situations may experience more negative social interactions, anxiously anticipate rejection, and tend to seek feedback that confirms their pre-existing negative self-beliefs or ask for reassurance about one’s self-worth, which in turn, might encourage depression symptom levels. Negative interpretation bias was also directly and indirectly related to symptoms of social anxiety via co-rumination. People who tend to interpret social situations more negatively may excessively discuss personal problems during social interactions, thereby increasing symptoms of social anxiety. This result extends previous work linking negative interpretation bias to intrapersonal forms of repetitive negative thinking (Everaert et al., 2020), and suggests that its link with co-rumination may be particularly relevant to explain social anxiety (when controlling for depression symptoms and other nodes within the network; see also: Spendelow et al., 2017).
Positive interpretation bias was indirectly associated with symptoms of depression and social anxiety, largely via its negative relations with rejection sensitivity. This finding suggests that less positive interpretation bias may shape rejection sensitivity, indicating positive bias may protect against anxious anticipation of rejection and quiet worries about whether others will be accepting and supportive (in line with Bronstein et al., 2022). Given that rejection sensitivity likely contributes to interpersonal conflict (Downey & Feldman, 1996), these results imply that less positive interpretations observed in people with depression or social anxiety generate social stress and thereby perpetuate symptoms (Hammen, 2006). This finding extends past research suggesting that rejection sensitivity promotes biased interpretation of ambiguity (Normansell & Wisco, 2017). Future research should examine potential reciprocal relations using longitudinal designs.
Contrary to this study’s hypothesis, the findings suggest that positive interpretation bias is positively related to interpersonal emotion regulation strategies such as excessive reassurance-seeking, co-dampening, and co-rumination. One possible explanation is that a more positive interpretation bias results from the social benefits offered by engaging in interpersonal emotion regulation such as increased social support provision and feelings of understanding as well as social connection (Rose, 2021). However, individuals with a positive interpretation bias may also experience disappointment or emotional distress upon realizing their overly positive interpretations are inaccurate. Consequently, they may engage in maladaptive emotion regulation strategies (e.g., excessively asking for reassurance) to cope with their uncertainty. Future studies should further explore when and why interpersonal emotion regulation shapes and is shaped by both positive and negative interpretation biases over time.
Future research should also continue to investigate the impact of interpretation processes on social relationships. Results suggested that pathways connecting interpretation processes and depression/social anxiety do not involve social integration metrics. This is inconsistent with theories positing that behaviors linked to interpretation processes in this study (e.g., excessive reassurance-seeking) undermine social relationships and may cause attrition from social networks (Hames et al., 2013). One potential reason for this result is the cross-sectional nature of our study. Distorted interpretations and problematic behavior may fluctuate over time, and it is possible that only sustained interpretation bias, inflexibility, and/or problematic behavior may be sufficient to cause ruptures in people’s social networks. Another reason for this result might be that interpretation bias/inflexibility worsens the quality of close relationships (e.g., less closeness, intimacy, support) rather than, or prior to, reducing social networks.
Another direction for future research is to examine measures of social networks that account for various types of social interactions. Research suggests that individuals can have different social functions across relationships, such as sharing negative news or having fun (Morelli et al., 2017). It is possible that the social network metrics used here-in may not have captured the variation in functional roles people experience across these networks. Furthermore, individuals can have different emotion-regulation needs for specific relationships (e.g., going to one’s partner for regulating fear vs. going to one’s best friend for regulating sadness) (Cheung et al., 2015). Taking this complexity into account may enable a better understanding of how interpretation processes are linked to social integration.
Limitations
This study has several limitations. First, the cross-sectional nature of this study precludes claims about causality and within-person processes. Intensive longitudinal research is needed to probe temporal relations indicative of causality and to examine whether between-person correlations adequately describe within-person dynamics over time.
Second, although this study examined partial correlations to identify possible relations between cognitive processes, socio-affective processes, and psychopathology, it is still possible that currently unmeasured variables explain the observed relations. For future research, it is important to investigate other potential mechanisms that might also play a role. For example, future studies could examine how negative interpretations contribute to social anxiety through safety-seeking behaviors (e.g., avoidance of social situations; Ginat-Frolich et al., 2024). These behaviors may prevent the disconfirmation of negative beliefs, thereby reinforcing the individual’s distorted perceptions. Moreover, this study is limited by its focus on process-based risk factors of depression and social anxiety. These process-based risk factors likely shape and be shaped by self-view and belief factors (e.g., dysfunctional attitudes, negative self-imagery; Beck & Haigh, 2014; Hirsch et al., 2006). Future studies could integrate both process-based risk factors and beliefs variables to model their interplay comprehensively.
Third, another limitation stems from the use of self-report measures. Individuals’ subjective experiences of socio-affective functioning may be colored by inaccurate interpretation processes, resulting in relations between the study measures that do not reflect a causal effect of interpretation on objective functioning. Future research should make use of alternate reporters (e.g., partners/friends) and measures of socio-affective processes that may be less susceptible to biases (e.g., experience sampling). Additionally, the reliance on self-report measures raises the potential issue of common method variance, which could have inflated the magnitude of the observed correlations (Podsakoff et al., 2003). Though this issue is partly mitigated by statistically controlling for shared variance (e.g., partial correlations in our network model) and using distinct measures (e.g., self-report questionnaires, cognitive task, self-report social network characteristics), future research should consider adopting a multi-method approach to further reduce the potential impact of common method variance.
Fourth, this study invited individuals who were “feeling blue, down, or anxious lately” to ensure sufficient variability in depression and anxiety symptom severity. While oversampling for elevated symptom levels is needed to adequately study risk mechanisms along the continuum of symptom severity, it may limit the generalizability of the findings. Further research in specific (non-)clinical populations is needed to determine the generalizability of the observed pathways and parameters estimates.
Finally, the data for this study was collected in a predominantly Western and well-educated crowdsourced sample. This further limits the study’s generalizability. Future studies should aim to replicate this study’s findings in more diverse and representative populations to enhance the generalizability of the study.
Conclusion
This study extends previous research by identifying potential pathways linking biased and inflexible interpretations to socio-affective functioning in ways that contribute to depression and/or social anxiety symptom severity. This pattern of results corroborates cognitive models positing that individuals with depression or social anxiety exhibit overly pessimistic perceptions of their social worlds, thereby setting the stage for a stressful environment of dampening positive and enhancing negative emotions that increase symptom severity. This study also suggests exciting new advances toward more integrative socio-cognitive-affective theories of psychopathology.
Acknowledgements
This work was supported by research grants (1202119 N, 1202122 N) from the Research Foundation – Flanders awarded to Jonas Everaert.
Declarations
Conflict of Interest
The authors have no conflicts of interest to declare.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.