Abstract
Extensive research has identified various social-cognitive vulnerabilities for internalizing disorders. However, few studies have assessed multiple disorders simultaneously, so it is unclear whether these vulnerabilities are transdiagnostic or specific risk factors. Their unique associations with disorders are also uncertain, given that they correlate strongly with neuroticism and one other. Psychiatric outpatients completed self-report and interview measures of six disorders (depression, generalized anxiety disorder, posttraumatic stress disorder, social anxiety, panic, obsessive-compulsive disorder), and personality (the Big Five, neuroticism facets, and four vulnerabilities: anxiety sensitivity, intolerance of uncertainty, perfectionism, experiential avoidance). All constructs were modeled as latent variables using structural equation modeling. All four vulnerabilities were closely associated with neuroticism, loading on its anxiety facet in factor analyses. Furthermore, after accounting for the contribution of neuroticism facets, intolerance of uncertainty and experiential avoidance were not uniquely associated with any disorders, and perfectionism was only related to obsessive-compulsive disorder. However, anxiety sensitivity accounted for substantial unique variance in several disorders (i.e., depression, social anxiety, posttraumatic stress disorder, and panic). We discuss theoretical and clinical implications of these results.
Keywords
There is strong interest in a number of relatively stable social-cognitive vulnerabilities (also called clinical traits) that are associated with psychopathology. These vulnerabilities (e.g., rumination, anxiety sensitivity, experiential avoidance, fear of negative evaluation) describe individual differences in thoughts, emotional experiences, and behaviors that are hypothesized to be related to the onset and/or maintenance of internalizing symptoms, such as anxiety and depression. Numerous researchers have suggested that they may represent more proximal contributors to the development of a specific emotional disorder, whereas broader traits like neuroticism may be more distally related (e.g., Hong, 2013; Nolen-Hoeksema & Watkins, 2011). As such, focusing on these relatively narrow traits can improve our understanding of how specific manifestations of psychopathology develop, as well as suggesting possible sources of comorbidity. Furthermore, some of these vulnerabilities are malleable and are the target of recently developed interventions (e.g., Boswell et al., 2013; Hayes, Strosahl, & Wilson, 2011; Mahoney & McEvoy, 2012; Wald, Taylor, Chiri, & Sica, 2010). We focus here on four social-cognitive traits—anxiety sensitivity, perfectionism, intolerance of uncertainty, and experiential avoidance—that have been extensively researched with regard to depression and anxiety and that appear to be transdiagnostic in nature.
Social-Cognitive Vulnerabilities
Definitions and Psychopathological Correlates
Anxiety sensitivity describes individual differences in the fear of anxiety symptoms, due to a belief that the symptoms are likely to have harmful consequences (Reiss & McNally, 1985). This vulnerability consists of three related components: fear of physical sensations of anxiety (physical concerns), fear of mental incapacitation or cognitive dyscontrol (cognitive concerns), and fear of public observation of anxiety (social concerns; e.g., Naragon-Gainey, 2010). Perfectionism is characterized by the setting of excessively high standards of performance. However, it is a multidimensional construct, and only maladaptive perfectionism—a tendency toward negative emotional responses and self-criticism that follow the failure to meet one’s high standards—is related to psychopathology (Lo & Abbott, 2013). Intolerance of uncertainty denotes a tendency to respond with discomfort and anxiety to unclear situations (Freeston, Rheaume, Letarte, Dugas, & Ladouceur, 1994). The construct has two components: prospective intolerance of uncertainty (i.e., a desire for and attempts to obtain predictability) and inhibitory intolerance of uncertainty (i.e., paralysis and lack of action in the face of uncertainty; e.g., Birrell, Meares, Wilkinson, & Freeston, 2011; Hong & Lee, 2015). Last, experiential avoidance is an unwillingness to remain in contact with distressing emotions, thoughts, and sensations (Hayes, Wilson, Gifford, Follette, & Strosahl, 1996). Although experiential avoidance may decrease distress in the short term, it has the paradoxical effect of increasing these same unwanted thoughts and emotions over time, potentially leading to psychopathology (e.g., Hayes et al., 1996; Hayes et al., 2011).
Intolerance of uncertainty and anxiety sensitivity were originally hypothesized to be specific vulnerabilities for a single disorder (i.e., generalized anxiety disorder [GAD] and panic disorder, respectively; Dugas, Gagnon, Ladouceur, & Freeston, 1998; Reiss & McNally, 1985), whereas perfectionism and experiential avoidance were thought to be more broadly relevant to psychopathology (Egan, Wade, & Shafran, 2011; Hayes et al., 1996; Hayes et al., 2011). However, accumulating evidence suggests that, although effect sizes vary across disorders, each of these vulnerabilities is substantially related to most internalizing symptoms, including social anxiety, depression, panic, obsessive-compulsive disorder (OCD), posttraumatic stress disorder (PTSD), and GAD (e.g., Egan et al., 2011; Fledderus, Oude Voshaar, ten Klooster, & Bohlmeijer, 2012; Gamez, Chmielewski, Kotov, Ruggero, & Watson, 2011; Gentes & Ruscio, 2011; Naragon-Gainey, 2010; Taylor et al., 2007).
Associations Among Social-Cognitive Vulnerabilities
Given that these four social-cognitive vulnerabilities are all linked to internalizing disorders and they have some conceptual similarities (e.g., negative perceptions, rigid cognitive styles, avoidance), it is not surprising that they correlate moderately with one another (e.g., Brown & Naragon-Gainey, 2013; Hong, 2013; Hong & Cheung, 2015). In fact, a recent meta-analysis found that six social-cognitive vulnerabilities (including anxiety sensitivity and intolerance of uncertainty) all loaded on a single factor, suggesting a shared etiological basis (Hong & Cheung, 2015). As such, it is important to consider whether, in relation to psychopathology, they are (a) somewhat distinct constructs that provide useful incremental information or (b) largely redundant with one another.
Most prior research in this area has focused on experiential avoidance and anxiety sensitivity. One factor analytic study found that experiential avoidance and anxiety sensitivity were best modeled as two related dimensions, rather than as a single construct (Carleton, Norton, & Asmundson, 2007). Although they are clearly related, the directionality of their association is ambiguous: Some evidence indicates that experiential avoidance accounts for the association between anxiety sensitivity and internalizing disorders (Gratz, Tull, & Gunderson, 2008; Kampfe et al., 2012), whereas other results suggest that anxiety sensitivity is responsible for the association between experiential avoidance and these disorders (Berman, Wheaton, McGrath, & Abramowitz, 2010; Wheaton, Berman, & Abramowitz, 2010). Furthermore, there is evidence that it is the interactive combination of high anxiety sensitivity and high experiential avoidance that best predicts psychopathology (e.g., Bardeen, Fergus, & Orcutt, 2013, 2014; Kashdan, Zvolensky, & McLeish, 2008).
The limited literature that has examined other combinations of social-cognitive vulnerabilities found that anxiety sensitivity and intolerance of uncertainty both were incrementally associated with OCD, GAD, and social anxiety (Norr et al., 2013). In contrast, intolerance of uncertainty was associated with social anxiety after accounting for shared variance with perfectionism, but the reverse was not true (Whiting et al., 2014).
Social-Cognitive Vulnerabilities and Personality
Associations With the Big Five and Facets
In most studies, a single social-cognitive vulnerability has been examined in relation to a single disorder or symptom, making it difficult to consolidate the literature across traits and symptoms. One way to synthesize the research on these various social-cognitive vulnerabilities is to place them within the well-studied framework of normal personality, such as the Big Five, which have robust associations with depression and anxiety (e.g., Kotov, Gamez, Schmidt, & Watson, 2010). All four social-cognitive vulnerabilities are primarily associated with neuroticism, with moderate to strong effect sizes (generally, rs = .30-.70; e.g., Berenbaum, Bredemeier, & Thompson, 2008; Clara, Cox, & Enns, 2007; Cox, Borger, Taylor, Fuentes, & Ross, 1999; Fergus & Rowatt, 2014; Gamez et al., 2011; Lilienfeld, 1997; Rice, Ashby, & Slaney, 2007). In many studies, these correlations are strong enough that they call into question whether the social-cognitive vulnerabilities are meaningfully and empirically distinct from neuroticism.
With regard to the other Big Five traits, maladaptive perfectionism has the broadest secondary associations, showing significant links to extraversion, conscientiousness, and agreeableness (Dunkley, Blankstein, & Berg, 2012; Hill, McIntire, & Bacharach, 1997; Rice et al., 2007), whereas anxiety sensitivity and intolerance of uncertainty are secondarily associated with low extraversion only (Berenbaum et al., 2008; Cox et al., 1999; Fergus & Rowatt, 2014; Hong & Lee, 2015). Although few studies have examined the personality correlates of experiential avoidance, it appears to be related substantially only to neuroticism (Gamez et al., 2011). For all the social-cognitive vulnerabilities, it is noteworthy that associations with extraversion, conscientiousness, agreeableness, or openness were consistently small in magnitude (i.e., |.10| to |.25|), suggesting that neuroticism is the only Big Five domain that is closely associated with them.
Each of the Big Five domains can be broken down into numerous lower order components or facets, which provide relatively narrow bandwidth descriptions of individual differences (e.g., Costa & McCrae, 1992; Markon, Krueger, & Watson, 2005). Most relevant to social-cognitive vulnerabilities, neuroticism includes separable components such as sadness, anxiety, anger, and mistrust (e.g., Costa & McCrae, 1992; John & Srivastava, 1999; Naragon-Gainey & Watson, 2014). Few studies have examined how these vulnerabilities are associated with facets of neuroticism, but there is some evidence of specific associations. In particular, intolerance of uncertainty (and primarily, the inhibitory component) is most closely associated with the anxiety facet (Fergus & Rowatt, 2014; Hong & Lee, 2015). Maladaptive perfectionism is broadly related to neuroticism facets, but shows the strongest unique association with sadness (Hill et al., 1997), whereas anxiety sensitivity has broad moderate correlations with multiple facets, including anxiety, sadness, self-consciousness, stress reactivity, and alienation (Cox et al., 1999; Lilienfeld, 1997). To date, no studies have examined the facet-level placement of experiential avoidance. In addition, multivariate analyses that account for the shared variance among the facets are needed to better understand specific facet-level associations.
Evidence for Unique Associations With Psychopathology
The strong associations of the social-cognitive vulnerabilities with neuroticism prompt an important question: To what extent are these vulnerabilities informative vis-à-vis disorders beyond their shared variance with neuroticism and its facets? If they provide little incremental information and are essentially isomorphic with neuroticism or its facets, then it is more parsimonious and informative to focus solely on neuroticism in research and treatment. On the other hand, if these vulnerabilities contribute relevant additional information to our understanding of psychopathology after accounting for shared variance with neuroticism, then it is important to include them in models of psychopathology and its treatment.
The current evidence suggests that these vulnerabilities account for additional variance beyond neuroticism for some disorders but not for others. The associations of maladaptive perfectionism with depression and social anxiety appear to be relatively independent of neuroticism, with multiple studies finding evidence of unique associations (e.g., Dunkley et al., 2012; Juster et al., 1996; Rice et al., 2007). However, the incremental validity of maladaptive perfectionism has not been examined with regard to other anxiety disorders. Anxiety sensitivity accounts for variance beyond the Big Five for most mood and anxiety disorders, including panic disorder (e.g., Hong, 2013; Kotov, Watson, Robles, & Schmidt, 2007), depression (Reardon & Williams, 2007), social anxiety disorder (Hong, 2013; Kotov et al., 2007; Norr et al., 2013), and PTSD (Collimore, McCabe, Carleton, & Asmundson, 2008; Vujanovic, Zvolensky, & Bernstein, 2008); evidence is more mixed for GAD and OCD (Kotov et al., 2007; Norr et al., 2013). Intolerance of uncertainty typically provides incremental information beyond neuroticism (e.g., Fetzner, Horswill, Boelen, & Carleton, 2013; Hong, 2013; McEvoy & Mahoney, 2011, 2012; Norr et al., 2013; Sarawgi, Oglesby, & Cougle, 2013; Whiting et al., 2014). However, several studies failed to find evidence of incremental validity for certain disorders, perhaps because they included additional covariates (Boswell et al., 2013; Brown & Naragon-Gainey, 2013; Fergus & Wu, 2011). Finally, there is limited evidence that experiential avoidance accounts for unique variance in PTSD (Meyer, Morissette, Kimbrel, Kruse, & Gulliver, 2013), depression, and OCD (Gamez et al., 2011; Gamez et al., 2014). Overall, this literature is quite mixed and limited by the fact that traits and disorders were generally studied in a piecemeal fashion, with little information about some disorders.
Current Study
The current study has two primary goals: (a) To clarify the nature and strength of the associations of maladaptive perfectionism, anxiety sensitivity, experiential avoidance, and intolerance of uncertainty with the Big Five and with facets of neuroticism and (b) To examine the unique contributions of these social-cognitive vulnerabilities beyond one another and neuroticism in relation to symptoms of six internalizing disorders (i.e., depression, GAD, PTSD, OCD, social anxiety, and panic). We focus on neuroticism facets (rather than those of other Big Five traits) because of their specific and strong associations with the social-cognitive vulnerabilities, and these six common internalizing disorders were selected because of their empirical and theoretical relevance to the vulnerabilities. Overall, we seek to identify traits that are most specifically and robustly associated with each disorder, yielding more informative and parsimonious etiological models.
The four social-cognitive vulnerabilities generally have been examined in nonclinical samples; here we use a diagnostically heterogeneous clinical sample because synergistic effects among these vulnerabilities (e.g., Bardeen et al., 2013, 2014) suggest that they may be more highly correlated with one another and relate differently to psychopathology in a currently distressed clinical sample. All disorders were assessed dimensionally with multiple self-report and interview measures, and structural equation modeling was used to remove the influence of method effects and measure-specific error from all variables, consistent with recent recommendations for examining incremental validity (Westfall & Yarkoni, 2016).
Method
Participants
The sample consisted of 299 psychiatric outpatients recruited from numerous treatment centers and via community postings. Eligibility criteria included current treatment for a psychological concern, 18 years of age or older, and fluency in English. Individuals who had been diagnosed with dementia or mental retardation, or who were currently psychotic, were excluded. Most of the sample completed interview measures in addition to self-report scales (n = 255), but the remaining 44 participants did not want to or were unable to come to the lab, so they opted to complete self-report measures from home and were not interviewed. 1 Three participants who omitted more than 40% of the items overall were removed from all analyses, resulting in a final sample size of 296 (n = 252 with both self-report and interview measures).
Most participants were female (73.9%) and age ranged from 18 to 73 years (M = 36.73 years, SD = 12.19); five people did not report their age. Eighty-nine percent identified as Caucasian, 4.7% as multiracial, 3.0% as Black or African American, 1.3% as Asian, and 2.0% did not report their ethnicity. This sample was relatively well educated, as the modal highest level of education was a bachelor’s degree (35.8%). A substantial portion of participants reported that their highest level of education was some college/2-year degree (n = 29.4%), or a graduate degree (n = 27.0%). Almost 8% did not receive any educational training beyond high school. The majority of the sample reported currently receiving therapy (75.3%) and psychotropic medications (86.1%). Based on the Structured Clinical Interview for DSM-IV (SCID-IV), the most prevalent diagnoses were GAD (37.3%), major depressive disorder (34.9%), social anxiety disorder (28.2%), panic disorder with or without agoraphobia (16.7%), PTSD (13.5%), and OCD (8.3%).
Procedure
Participants came to the laboratory in small groups to complete questionnaires and a brief clinical interview conducted individually in a private room. Total completion time was roughly 1.5 to 2 hours, and participants received $30 compensation. Participants who could not or did not want to come to our laboratory were offered the option of completing the self-report questionnaire at their homes (either online or with a mail-in packet) and were compensated $20.
All interviewers (graduate students and advanced undergraduate research assistants) underwent extensive training, consisting of weekly training sessions for a month prior to the study and competency assessments. Interviews in the study were audiotaped and interrater reliability was assessed periodically by having a second interviewer code approximately 20% of the recorded interviews (n = 51), selected at random.
Measures
The indicators for each latent variable are listed in the online supplement (Table S1 for the personality measures and Table S2 for the psychopathology measures; available at http://asm.sagepub.com/content/by/supplemental-data). These tables also provide descriptive statistics (means and standard deviations) and reliability coefficients (coefficient alpha for the self-report scales, intraclass correlations [ICC] for the interview ratings) in the current sample, as well as the number of items for all measures. With regard to reliability, all self-report measures had acceptable internal consistency in this sample (αs = .76-.96), and all interview measures had good interrater reliability (ICCs for dimensional ratings = .71-1.00; κs for diagnoses = .83-1.00).
Social-Cognitive Vulnerabilities
Anxiety Sensitivity
The Anxiety Sensitivity Index–3 (ASI-3; Taylor et al., 2007) was created to provide better coverage of the three anxiety sensitivity facets (i.e., physical, cognitive, and social concerns) than the original ASI. It consists of 18 statements (six for each facet) that are rated on a 5-point Likert-type scale. The ASI-3 has good factorial validity across multiple samples, and the scales distinguish among diagnostic groups according to expectations (Taylor et al., 2007). The three ASI-3 subscales served as indicators for the anxiety sensitivity latent variable.
Perfectionism
Two subscales from the Frost Multidimensional Perfectionism Scale (FMPS; Frost, Marten, Lahart, & Rosenblate, 1990)—which show the strongest links to internalizing psychopathology—were used to assess different types of maladaptive perfectionism: Concern over Mistakes (COM; self-critical and catastrophic beliefs regarding making mistakes) and Doubts about Actions (DAA; doubts about one’s ability to do things “right”). Each statement is rated on a 5-point Likert-type scale. COM and DAA have acceptable internal consistency as well as good discriminant validity and convergent validity (Frost et al., 1990). These two scales served as indicators for the perfectionism latent variable.
Intolerance of Uncertainty
The Intolerance of Uncertainty Scale–12 (IUS-12; Carleton et al., 2007) is a 12-item short form of the original Intolerance of Uncertainty Scale (Freeston et al., 1994). It was created to reduce the item redundancy in—and improve the factor structure of—the original measure. The IUS-12 has a two-factor structure, consisting of Prospective Anxiety and Inhibitory Anxiety. The IUS-12 shows the expected patterns of convergent and discriminant validity with measures of depression and anxiety (Carleton et al., 2007). The intolerance of uncertainty latent variable was marked by the Prospective Anxiety and Inhibitory Anxiety subscales.
Experiential Avoidance
The Multidimensional Experiential Avoidance Questionnaire (MEAQ; Gamez et al., 2011) contains six scales measuring different aspects of experiential avoidance using a 6-point Likert-type scale. For this study, the Distress Aversion Scale (negative attitudes toward and attempts to avoid emotional distress) was used, as analyses indicate that it is most central to the construct (r = .60 with other scales of experiential avoidance; Gamez et al., 2011). The MEAQ has good convergent and discriminant validity (Gamez et al., 2011). Because this scale has been shown to be unidimensional (Gamez et al., 2011), we created three parcels using a random number generator that served as indicators for the experiential avoidance latent variable.
Domain-Level Personality Traits
Indicators of the Big Five were taken from the Big Five Inventory (BFI; John & Srivastava, 1999) and the Positive and Negative Affect Schedule–Expanded Form (PANAS-X; Watson & Clark, 1999).
The BFI consists of 44 short phrases that are rated on a 5-point Likert-type scale and that correspond to one of the “Big Five” domains. The BFI scales have good internal consistency and strong convergent and discriminant validity with other measures of the Big Five, including peer ratings. In addition, the BFI has demonstrated good retest reliability (mean r = .85 after a 3-month interval; John & Srivastava, 1999). For the current study, parcels of the Big Five domains were created using a random number generator to ensure that each trait had three indicators in factor analyses (all parcel alphas were ≥.60).
The PANAS-X is a self-report measure of specific types of affect; each mood term is rated on 5-point intensity scale anchors. Our participants rated how they “generally” felt using the trait version of the instrument. These scales have demonstrated good convergent and discriminant validity with other measures of affectivity, and adequate retest reliability after 2 months (r = .59; Watson & Clark, 1999). The Positive Affect scale was used as an indicator of extraversion in this study, whereas Negative Affect scale was used as an indicator of neuroticism.
Facet-Level N/NE Traits
Based on a review of major personality inventories (including measures of both abnormal and normal personality), a six-faceted structure of neuroticism was modeled, consisting of Anxiety, Sadness, Angry Hostility, Mistrust, Dependency, and Stress Vulnerability.
Anxiety
Indicators of the Anxiety factor included (a) the Anxiety scale from the Faceted Inventory of the five-factor model (FI-FFM; Simms, 2009). The FI-FFM is a factor analytically derived measure designed to assess higher and lower order personality traits within the framework of the five-factor model; (b) the Anxiety facet from the HEXACO Personality Inventory (HEXACO PI; Lee & Ashton, 2004), a measure that assesses six domains of personality and their facets; (c) the Anxiety scale from the Revised NEO Personality Inventory (NEO PI-R; Costa & McCrae, 1992). The NEO PI-R measures personality domains and facets in the five-factor model; and (d) the Fear Scale from the PANAS-X.
Sadness
The Sadness factor was marked by (a) the Depression facet from the NEO-PI-R, (b) the Depression facet from the FI-FFM, and (c) the Sadness scale from the PANAS-X.
Angry Hostility
Indicators for Angry Hostility were (a) Anger Proneness from the FI-FFM, (b) Angry Hostility from the NEO-PI-R, and (c) the Hostility scale from the PANAS-X.
Mistrust
Mistrust was assessed with (a) the Distrust scale from the International Personality Item Pool 16PF (IPIP-16PF; Goldberg, 2009), which is modeled after the 16PF Vigilance scale; (b) FI-FFM Trust versus Cynicism, and (c) the Mistrust scale from the Schedule for Nonadaptive and Adaptive Personality (SNAP; Clark, 1993). The SNAP was designed to assess personality pathology; each item is answered using a true–false format.
Dependency
Markers for the Dependency factor were (a) the Submissive scale from the Three Vector Dependency Inventory (3VDI; Pincus & Wilson, 2001), which is based on the interpersonal circumplex. The Submissive scale measures the tendency to yield to others and is most closely related to the passive dependency that is linked to internalizing psychopathology; (b) the Lack of Social Self-Confidence scale from the Interpersonal Dependency Inventory (IDI; Hirschfeld et al., 1977); and (c) the Dependency scale from the SNAP.
Stress Vulnerability
Stress Vulnerability had the following indicators: (a) the Vulnerability scale from the NEO PI-R and (b) the Stress Reaction scale from the Multidimensional Personality Questionnaire–Brief Form (Patrick, Curtin, & Tellegen, 2002).
Internalizing Symptoms
Symptoms of six internalizing disorders were each modeled with multiple self-report and interview measures.
Depression
The indicators for Depression included (a) the Dysphoria scale from the Inventory of Depression and Anxiety Symptoms (IDAS; Watson et al., 2007, 2008). The IDAS is a factor analytically derived, multidimensional inventory that uses a 5-point Likert-type scale to assess symptoms over the past 2 weeks. The Dysphoria scale contains 10 items, but one of them is identical to an item in the Anxious Mood scale, so it was removed from Dysphoria to maintain scale independence; (b) the Personal Health Questionnaire–9 (PHQ-9; Kroenke, Spitzer, & Williams, 2001) which is based on the DSM-IV diagnostic criteria for depression; (c) the Anhedonic Depression—Loss of Interest scale from the Mood and Anxiety Symptom Questionnaire (MASQ; Watson & Clark, 1991). The MASQ was developed as a test of the tripartite model, and this scale was designed to measure relatively specific symptoms of depression; (d) the low mood screener from the Structured Clinical Interview for DSM-IV (SCID-IV; First, Spitzer, Gibbon, & Williams, 1997). (Note that we use the trichotomous screeners rather than diagnoses as indicators for all the symptom factors because of their greater information and variability.); (e) the anhedonia screener from the SCID-IV; and (f) the Dysphoria rating from the Clinician Rating Version of the Inventory of Depression and Anxiety Symptoms (IDAS-CR; Watson et al., 2008).
The IDAS-CR is a semistructured interview version of the IDAS, wherein a series of ratings are made that correspond to each of the 11 nonoverlapping IDAS scales. A 3-point scale is used to rate each symptom dimension over the past 2 weeks as absent, subthreshold, or present, based on responses to a standard initial probe as well as three-to-five follow-up questions (see, Watson et al., 2008).
Generalized Anxiety Disorder
Indicators included (a) the IDAS Anxious Mood scale; (b) Generalized Anxiety Disorder Questionnaire–IV (GAD-Q-IV; Newman et al., 2002). This scale assesses the nine GAD criteria from DSM-IV, and it was scored dimensionally in the current study; (c) Worry Domains Questionnaire–Short Form (WDQ-SF; Stöber & Joormann, 2001). The WDQ-SF samples two items from each of five domains of worry (e.g., relationships, work, finances) assessed in the original WDQ (Tallis, Eysenck, & Mathews, 1992); (d) the GAD screener from the SCID-IV; and (e) the Generalized Anxiety rating from the IDAS-CR.
Posttraumatic Stress Disorder
The PTSD factor was assessed with (a) the sum of the IDAS-II Traumatic Intrusions and Traumatic Avoidance scales (Watson et al., 2012); (b) the PTSD Checklist–Civilian Version (PCL-C; Weathers, Litz, Herman, Huska, & Keane, 1993), which assesses symptoms over the past month as described in the DSM-IV criteria; (c) the sum of the Intrusions, Avoidance, Dysphoria, and Hyperarousal scales from the Iowa Traumatic Response Inventory (ITRI; Gootzeit, Markon, & Watson, 2015); (d) the PTSD screener from the SCID-IV; and (e) the sum of the IDAS-CR Traumatic Intrusions, Traumatic Avoidance, and Traumatic Hyperarousal ratings.
Social Anxiety
The Social Anxiety indicators were (a) the IDAS-II Social Anxiety scale; (b) Social Phobia from the Albany Panic and Phobia Questionnaire (APPQ; Rapee, Craske, & Barlow, 1994/1995). Respondents use a 9-point rating scale to describe how much fear they would experience if they encountered various social situations; (c) the Fear Questionnaire (FQ; Marks & Mathews, 1979) assesses avoidance of social situations, rated on a 5-point Likert-type scale; (d) the social phobia screener from SCID-IV; and (e) the IDAS-CR Social Anxiety rating.
Panic
The Panic latent variable consisted of (a) MASQ Anxious Arousal; (b) the Panic Attack Symptom Questionnaire (PASQ; Watson, 2000), which assesses the symptom criteria for a panic attack in the DSM-IV. Ratings are made on a 5-point intensity scale for symptoms that occurred over the past month; (c) the panic disorder screener from SCID-IV; and (d) the IDAS-CR Panic rating.
Obsessive-Compulsive Disorder
Indicators for OCD included (a) the sum of the Cleaning, Checking, and Ordering scales from the IDAS-II (Watson et al., 2012); (b) the Obsessive-Compulsive Inventory–Revised (OCI-R; Foa et al., 2002) total score, which is the sum of six subscales (Checking, Washing, Ordering, Hoarding, Obsessing, Neutralizing); (c) the sum of four scales (Obsessive Checking, Obsessive Cleanliness, Compulsive Rituals, and Hoarding) from the Schedule of Compulsions, Obsessions, and Pathological Impulses (SCOPI; Watson & Wu, 2005). The SCOPI is a multidimensional measure of OCD symptoms that was created using factor analysis; (d) the obsessions screener; and (e) the compulsions screener from the SCID-IV; (f) the sum of five OCD ratings from the Personality, Cognitions, Consciousness, and Perceptions Interview (PCCP; Chmielewski & Watson, 2007), consisting of Checking/Doubting, Cleaning/Washing, Intrusive Thoughts/Obsessions, Ordering/Rituals, and Hoarding. As with the IDAS-CR, the PCCP uses a 3-point rating system (absent, subthreshold, present) for each symptom, using a standard initial probe and three to five follow-up questions.
Data Analyses
Confirmatory factor analyses and structural regression models were run in MPlus 7.31, and other analyses were conducted in SAS 9.4. Multiple imputation was used to impute missing data in SAS. To examine how the social-cognitive vulnerabilities fit into the larger personality hierarchy, we conducted an exploratory factor analysis of the four vulnerabilities with Big Five domains and trait affect, using a promax rotation (power = 3). As noted earlier, parcels were created from the five BFI scales using a random number generator to assign items, with the condition of at least three items per parcel (see Table 2 for items assigned to each parcel); the PANAS-X NA and PANAS-X PA scales were also included in the factor analysis. Given our hypothesis that the social-cognitive vulnerabilities would be most strongly associated with neuroticism, a second exploratory factor analysis was performed on the neuroticism facet scales and the social-cognitive vulnerabilities to locate them more precisely in this lower level of the personality hierarchy.
Prior to testing the unique and incremental associations of the social-cognitive vulnerabilities in relation to internalizing symptoms, we first conducted separate confirmatory factor analyses on all latent constructs. Although indicators were carefully selected a priori to be valid and reliable markers of their latent variables, measure properties are sample-specific. Thus, it is important to determine empirically whether, in this sample, (a) the indicators are strong markers of each construct as expected and (b) the data are a good fit to the model (see Brown, 2015). Within each symptom construct, the error terms of the interview measures were allowed to covary to account for their shared method variance whenever possible (i.e., if doing so did not result in an improper solution). The metric of each latent variable was set with marker indicators, using the first measure listed for each latent variable in the “Measures” section. We used robust weighted least squares estimators (referred to as WLSMV estimators in MPlus) because they account for the categorical nature of the interview indicators, while still providing standard fit indices to evaluate goodness of fit. Maximum likelihood estimators with robust standard errors (MLR in MPlus) were used to evaluate the models with neuroticism and its facets and the social-cognitive vulnerabilities, as these models used only continuous indicators.
We focus on several complementary fit indices to evaluate the CFAs: the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean squared residual (SRMR; for continuous indicators only) or weighted root mean squared residual (WRMR; for categorical indicators only). Interpretation of these indices is based on the guidelines set forth by Hu and Bentler (1999) and Browne and Cudeck (1993). Hu and Bentler suggested that CFI should be “close to” .95 or greater for good fit; however, given debate regarding the best cutoff value and difficulties in generalizing cutoff values across data sets and models, slightly lower values are often considered acceptable (Marsh, Hau, & Wen, 2004). Browne and Cudeck (1993) suggested that RMSEA values less than .08 and greater than .10 reflect good fit and poor fit, respectively. SRMR should be less than or equal to .08 (Hu & Bentler, 1999) and WRMR should be less than 1.0 (Yu, 2002). The solutions were also examined for any salient areas of strain that would suggest a theoretically sound modification to improve model fit.
A series of structural regression models were analyzed to evaluate the unique incremental contributions of the social-cognitive vulnerabilities, beyond one another and the neuroticism facets, in accounting for variance in the internalizing symptom factors. The structural models were specified in a hierarchical fashion. In the first series of models, each disorder was separately regressed on the neuroticism facet latent variables. The four latent social-cognitive vulnerabilities were added in the second series of structural models to test their incremental and unique contributions. 2 We used Hunsley and Meyer’s (2003) guidelines when interpreting the magnitude of incremental associations, where a change in R2 of approximately .03 to .05 or larger is indicative of a “reasonable” incremental contribution. MLR estimators were used for the structural regression analyses, as they can accommodate categorical indicators and are preferable to WLSMV because they are more efficient. However, MLR does not allow for the calculation of standard fit indices when categorical indicators (i.e., the interview measures) are included (Muthén & Muthén, 1998-2010).
Results
Correlations among the social-cognitive vulnerabilities are shown in Table 1. Excluding part–whole correlations between subscales and total scale scores, most correlations fell between .35 and .60, indicating moderate to strong associations that one might expect for related but distinguishable traits.
Correlations Among Social-Cognitive Vulnerabilities.
Note. N = 296. All correlations are significant at p < .01. Correlations greater than or equal to |.50| are shown in boldface. AS = Anxiety Sensitivity; Perf. = Perfectionism; COM = Concern Over Mistakes; DAA = Doubts About Actions; IU = Intolerance of Uncertainty; Exp. = Experiential.
Associations of Social-Cognitive Vulnerabilities Traits With Personality Traits
Higher Order Traits
The parcels formed from the BFI scales, as well as PANAS-X NA, PANAS-X PA, and the four social-cognitive vulnerabilities (total scores), were submitted to an exploratory factor analysis with a promax rotation. Five factors were extracted, each of which had at least three markers. These factors clearly corresponded to the Big Five domains (see Table 2; rs among factors = |.10| to |.36|). All four social-cognitive vulnerabilities loaded primarily and strongly on the Neuroticism factor (loadings = .63-.78); in particular, the loading for Intolerance of Uncertainty was as strong or stronger than those of the Neuroticism indicators. Secondary loadings were generally minimal for the vulnerabilities, with the exception of Experiential Avoidance’s small to moderate negative loading on Openness (−.25). Thus, all the social-cognitive vulnerabilities were primarily and specifically associated with neuroticism.
Standardized Factor Loadings from the Exploratory Factor Analysis (Promax Rotation) of Big Five Domains and Social-Cognitive Vulnerabilities.
Note. N = 296. Loadings greater than or equal to |.20| are shown in boldface. Social-cognitive vulnerabilities are italicized. N = neuroticism; E = extraversion; C = conscientiousness; A = agreeableness; O = openness. BFI = Big Five Inventory. Numbered BFI scales indicate parcels formed from the BFI items, with the following item numbers assigned to each parcel: N1 = 4, 14, 24, 39; N2 = 9, 19, 29, 34; E1 = 1, 11, 21, 36; E2 = 6, 16, 26, 31; C1 = 3, 13, 28; C2 = 8, 33, 38; C3 = 18, 23, 43; A1 = 7, 12, 32; A2 = 17, 27, 42; A3 = 2, 22, 37; O1 = 5, 20, 35, 40; O2 = 10, 15, 25; O3 = 30, 41, 44.
Neuroticism Facets
Prior to examining how the social-cognitive vulnerabilities fit into a facet-level model of neuroticism (consisting of Sadness, Anxiety, Angry Hostility, Dependency, Mistrust, and Stress Vulnerability), we first conducted a CFA of the neuroticism facets to evaluate the acceptability of the model. The error terms of the three PANAS-X scales were allowed to covary to capture their shared method variance (i.e., these scales all use single adjectives as items, whereas all other scales use full sentences). This model did not yield a proper solution, however, because Stress Vulnerability exhibited a linear dependency with other factors (viz., rs with Sadness, Anxiety, and Dependency = .89-.95), resulting in a nonpositive definite matrix. 3 Therefore, Stress Vulnerability was removed from all subsequent analyses. The revised model with the remaining five neuroticism facets converged on an admissible solution and was an acceptable fit to the data: χ2(91) = 251.06, p > .05, CFI = .96, RMSEA = .07, and SRMR = .05. Each scale was a good indicator of its factor, with standardized factor loadings ranging from .66 to .96 (p < .001; see Table S3 of the online supplement, available at http://asm.sagepub.com/content/by/supplemental-data). Correlations among the factors ranged from .34 to .72 (mean r = .56). Given the soundness of this model, it was used in subsequent structural regression models.
Next, these scales and the four social-cognitive vulnerability total scale scores were submitted to an exploratory factor analysis with a promax rotation to locate the vulnerabilities within the structure of the five neuroticism facets. Standardized factor loadings are shown in Table 3. When five factors (one for each of the five neuroticism facets) were extracted, all four of the social-cognitive vulnerabilities loaded most strongly on the Anxiety factor, although the magnitudes varied widely. Anxiety Sensitivity loaded very strongly on Anxiety (.76), with minimal secondary loadings (|.02| to |.20|; mean = |.11|) that were similar in magnitude to those of the scales specifically selected as indicators of Anxiety. Intolerance of Uncertainty and Maladaptive Perfectionism had weaker primary loadings on Anxiety (.63 and .47, respectively), and slightly stronger secondary loadings than Anxiety Sensitivity (|.03| to |.23|; mean = |.14|). Finally, Experiential Avoidance loaded more weakly on the Anxiety facet (.32) and had two nonzero secondary loadings (.19 on Dependency and .24 on Trust); thus, although it was primarily associated with the Anxiety facet, it may be more closely related to higher order neuroticism.
Standardized Factor Loadings from the Exploratory Factor Analysis (Promax Rotation) of Neuroticism Facet Scales and Social-Cognitive Vulnerabilities.
Note. N = 296. Loadings greater than or equal to |.20| are shown in boldface. Social-cognitive vulnerabilities are italicized. NEO PI-R = Revised NEO Personality Inventory; HEXACO PI = HEXACO Personality Inventory; PANAS-X = Positive and Negative Affect Schedule–Expanded Form; FI-FFM = Faceted Inventory of the Five-Factor Model; 3VDI = 3-Vector Dependency Inventory; IDI = Interpersonal Dependency Inventory; SNAP = Schedule for Nonadaptive and Adaptive Personality; IPIP 16PF = International Personality Item Pool version of the 16 Personality Factor Questionnaire.
Unique Contributions of Social-Cognitive Vulnerabilities to Symptoms
All disorders and social-cognitive vulnerabilities were modeled as latent variables (see the “Measures” section and online Table S2 for the indicators of each construct), so confirmatory factor analyses of each construct were examined prior to conducting the structural regression analyses. Fit indices for each CFA are reported in Table 4. Generally, the latent variable models demonstrated an excellent fit to the data, though RMSEA values were somewhat high for PTSD and social anxiety (see the Table 4 note for details). Zero-order correlations among the internalizing symptom factors were strong, indicating substantial comorbidity across symptoms (rs = .48-.88; see online Table S4). Most correlations between the internalizing symptom factors and the social-cognitive vulnerability factors were moderate to strong, ranging from .31 to .78 (Table 5). Overall, these associations did not display clear evidence of specificity or differential patterns of associations across disorders. Correlations between experiential avoidance and disorders tended to be weaker than the correlations between the other social-cognitive vulnerabilities and disorders.
Fit Indices for Confirmatory Factor Analyses (CFAs) of Internalizing Symptoms and Social-Cognitive Vulnerabilities.
Note. N = 252 for symptom CFAs, 296 for social-cognitive vulnerabilities CFA. df = degrees of freedom; CFI = comparative fit index; RMSEA = root mean square error of approximation; WRMR = weighted root mean squared residual; SRMR = standardized room mean squared residual; PTSD = posttraumatic stress disorder; GAD = generalized anxiety disorder; OCD = obsessive-compulsive disorder. Because disorder CFAs included categorical indicators, WLSMV estimators were used and WRMR is reported, whereas MLR estimators were used for the social-cognitive vulnerabilities CFA and SRMR is reported (see text for detail). Note that RMSEA values exceeded typical cutoffs consistent with good fit for Social Anxiety and PTSD (.13 and .16, respectively). Because (a) all other fit indices indicated good fit in these three models, (b) no salient areas of strain were identified in the models, and (c) the models are near saturation (df = 4), the high RMSEA value likely reflects the fact that this index strongly penalizes greater model saturation, as opposed to indicating a problem with model fit per se (Brown, 2015).
Correlations Between Social-Cognitive Vulnerability Factors and Internalizing Symptom Factors.
Note. N = 252. PTSD = posttraumatic stress disorder; OCD = obsessive-compulsive disorder; GAD = generalized anxiety disorder. All correlations are significant at p < .001. Correlations greater than or equal to |.60| are shown in boldface.
We next examined six separate structural regression models with each internalizing symptom as the outcome, first with neuroticism facet latent variables only, and then with both the neuroticism facets and the social-cognitive vulnerability latent variables, to assess the incremental contributions of each vulnerability. Results are shown in Table 6; because of the number of tests conducted and to minimize Type I error, we focus on parameter estimates significant at p < .01 when interpreting these results. We note that some standard errors of parameter estimates were large, particularly those of perfectionism, suggesting that these latent variables were not measured with optimal precision.
Structural Regressions of Social-Cognitive Vulnerabilities and Neuroticism Facets on Internalizing Symptoms.
Note. N = 252. Standardized parameter estimates are shown with standard errors in parentheses. Intoler. = Intolerance; Exp. = Experiential; PTSD = posttraumatic stress disorder; OCD = obsessive-compulsive disorder; GAD = generalized anxiety disorder.
p < .05. **p < .01. ***p < .001.
In the first series of models, the neuroticism facets accounted for significant variance (p < .001) in each of the six internalizing symptoms, but the magnitudes varied. Specifically, the facets accounted for almost all the variance in GAD (R2 = .90), a large proportion of variance in depression and social anxiety (R2 = .72 and .68, respectively), and moderate variance in the other symptoms (R2 = .43-.53).
The social-cognitive vulnerability factors were added as predictors in the second series of models. They did not account for additional variance in GAD (ΔR2 = .007), but they did contribute significant additional variance for all other symptoms, with effect sizes that exceeded Hunsley and Meyer’s (2003) suggested threshold for a substantial incremental contribution (ΔR2 = .091-.206, p < .001). Anxiety sensitivity was significantly associated with panic (β = .79, p < .01), depression (β = .50, p < .01), PTSD (β = .64, p < .01), and social anxiety (β = .48, p < .01), whereas perfectionism was significantly associated with OCD (β = .98, p < .01). There was also a weaker association between intolerance of uncertainty and depression (β = .29, p < .05), and a weak inverse suppressor association between experiential avoidance and panic (β = −.16, p < .05).
For those social-cognitive vulnerabilities that accounted for significant (p < .01) incremental variance in a symptom, we conducted post hoc analyses to determine whether specific subscales or components of these traits were primarily responsible for the observed associations. Thus, we ran five series of structural regression models, examining the unique associations of the anxiety sensitivity subscales in relation to panic, depression, PTSD, and social anxiety, and the unique associations of the perfectionism subscales in relation to OCD. Parallel to the analyses shown in Table 6, after entering the neuroticism facet latent variables in Regression Model 1, we added the subscales of the associated social-cognitive vulnerabilities in Regression Model 2.
4
Associations with the three anxiety sensitivity subscales varied by disorder. Namely, both the cognitive concerns and physical concerns components were uniquely associated with PTSD (cognitive β = .31, p < .001; physical β = .21, p < .05; also a suppressor effect with social concerns: β = −.22, p < .05) and with depression (cognitive β
Discussion
The aim of this study was to examine how social-cognitive vulnerabilities fit into the broader, well-established structure of normal personality, and to determine their unique contributions (beyond one another and beyond neuroticism facets) in accounting for variance in multiple internalizing disorders in a clinical sample. With regard to the first aim, domain-level structural analyses revealed that the four social-cognitive vulnerabilities were not clearly distinguishable from indicators of neuroticism: All loaded strongly on neuroticism, with very few significant secondary loadings. Furthermore, this study was the first to explore how these vulnerabilities relate to a multi-inventory model of facets of neuroticism, wherein all vulnerabilities loaded primarily and specifically on the anxiety facet. The anxiety facet scales focus primarily on worry and tension, similar to the “anxious apprehension” described in the clinical literature and most closely associated with GAD. Anxious apprehension also is a core, central aspect of neuroticism. Intolerance of uncertainty and anxiety sensitivity clearly have a close conceptual relationship with anxiety/anxious apprehension, but it is noteworthy that perfectionism and experiential avoidance—which arguably have less conceptual overlap with anxiety— are also most closely related to this particular facet. Overall, these analyses suggest that—at least structurally—the social-cognitive vulnerabilities may be little more than specific manifestations of neuroticism.
Given these results, it is perhaps not surprising that the social-cognitive vulnerabilities generally did not demonstrate unique associations with the six internalizing disorders, after accounting for shared variance with neuroticism facets. Specifically, experiential avoidance and intolerance of uncertainty were not significantly associated (p < .01) with any of the disorders in these analyses (although there was a weak association between depression and intolerance of uncertainty at p < .05), and perfectionism was only significantly associated with OCD. However, anxiety sensitivity was the exception to this pattern. Despite loading very strongly on neuroticism and its anxiety facet in exploratory structural analyses, anxiety sensitivity uniquely contributed between 9% and 20% of the variance in multiple disorders (i.e., depression, social anxiety, PTSD, and panic).
Anxiety sensitivity may be broadly and uniquely associated with internalizing psychopathology for several reasons. First, it may reflect perceptions of impaired psychosocial functioning (e.g., starting to “go crazy,” experiencing abnormal physical or social functioning), and such perceived impairments are often distressing and are associated with many disorders (American Psychiatric Association, 2013). In addition, anxiety sensitivity may be indicative of a maladaptive cycle (similar to experiential avoidance) in which attempts to suppress or control one’s cognitions and physical sensations actually serve to increase the frequency and salience of these unwanted internal experiences, leading to greater distress and anxiety (e.g., Hayes et al., 2011). It is noteworthy that although anxiety sensitivity was similarly associated with numerous disorders, greater differentiability was apparent when examining the subscales (physical, cognitive, and social concerns), consistent with prior literature illustrating the utility of considering these components separately (e.g., Kemper, Lutz, Bahr, Ruddel, & Hock, 2012; Naragon-Gainey, 2010).
A less sanguine interpretation of these results is that anxiety sensitivity (and perfectionism) may have accounted for unique variance in certain disorders because they include content that overlaps with core features of these disorders. An examination of these scales suggests that catastrophic interpretations of physical or cognitive sensations of anxiety share important content with panic disorder (i.e., diagnostic criteria emphasizing feared consequences of panic attacks) and, to a lesser degree, the hyperarousal symptoms of PTSD, whereas the social concerns component of anxiety sensitivity taps into a fear of negative evaluation that is core to social anxiety. Similarly, the Doubt about Actions perfectionism scale is closely linked to certain obsessions (e.g., mental and behavioral rituals involving repeated checking). We cannot rule out the possibility that shared content contributed to these findings (indeed, the scales identified above seemed to drive associations in the post hoc regression analyses), but we believe that our measurement approach likely minimized such effects. Specifically, the social-cognitive vulnerability latent variables were formed only from the shared variance among the subscales for each construct, with the unique variance of each isolated in the error term. As such, the anxiety sensitivity latent variable represents a general fear of anxiety sensations, after removing variance due to content that is specific to physical, cognitive, or social concerns. Similarly, the perfectionism latent variable models the shared variance between the Doubt about Actions and Concern over Mistakes subscales, which likely represents an approach characterized by high standards. The fact that significant associations remained after removing unique subscale variance suggests that these results are at least partially due to associations with broad underlying vulnerabilities, which transcend specific manifestations that may reflect symptoms of specific disorders.
Theoretical and Clinical Implications
Several frameworks have been developed to integrate and understand information regarding vulnerability factors. One such framework is the triple vulnerability model (e.g., Barlow, 2000), which describes three types of vulnerabilities that contribute to the internalizing disorders. Neuroticism and low perceived control are proposed to be two generalized vulnerabilities common to all these disorders, thereby contributing to elevated comorbidity rates among them. In contrast, disorder-specific vulnerabilities reflect a specific focus of distress (e.g., situations, objects, thoughts, somatic sensations) that explain why an individual develops a particular disorder. Thus, neuroticism (and perhaps other broad personality traits) is a distal vulnerability that increases risk for many forms of psychopathology, whereas social-cognitive vulnerabilities may be more proximal characteristics that mediate the association between neuroticism and the development or maintenance of specific internalizing symptoms (e.g., Hong, 2013; Nolen-Hoeksema & Watkins, 2011). Our results suggest that anxiety sensitivity may fall somewhere between a generalized and disorder-specific vulnerability, as it confers specific risk for a subset of the internalizing disorders only. Thus, anxiety sensitivity may be an additional (though more limited) source of comorbidity beyond the well-established effect of neuroticism (e.g., Watson & Naragon-Gainey, 2014).
One may also distinguish broad traits like neuroticism from social-cognitive vulnerabilities on the basis of whether they reflect primary emotional processes or secondary reactions to them. That is, neuroticism may be conceptualized as the typical level of negative emotions experienced by an individual. In contrast, social-cognitive vulnerabilities describe how one interprets and responds to aversive thoughts, sensations, and emotions, or what one does with these unpleasant experiences. These “secondary reactions” to internal experiences are particularly central to acceptance- and mindfulness-based models of psychopathology and therapy, though they are also relevant to cognitive-behavioral therapy (e.g., Hayes et al., 2011).
Despite their conceptual importance to such models, our results suggest that most of the social-cognitive vulnerabilities examined here do not provide much unique information with regard to symptom levels, and so it may be more time-efficient in clinical contexts simply to assess neuroticism. However, it appears that anxiety sensitivity contributes unique, relevant information beyond neuroticism for numerous disorders, even in a clinically distressed sample where the level of neuroticism was relatively high. Thus, particularly for individuals presenting with numerous comorbid internalizing symptoms, it may be helpful to initially assess and then track their anxiety sensitivity. For individuals with markedly high levels at intake, the therapist could consider supplementing treatment with interventions designed specifically to target and modify this vulnerability (e.g., Boswell et al., 2013; Wald et al., 2010). Finally, it is important to note that we only assessed cross-sectional symptoms as outcomes; it may be that some or all these social-cognitive vulnerabilities are informative beyond neuroticism with regard to other outcomes, such as treatment response or maintenance of symptoms over time.
Study Limitations and Future Directions
A number of limitations apply to this study. First, measurement of traits and symptoms was concurrent and cross-sectional, rather than prospective and longitudinal, likely leading to an inflation of correlations and some confounding of these associated constructs (see also the previous discussion of possible content overlap). Longitudinal methodologies are necessary to evaluate the direction of causality and to distinguish personality traits from symptoms more clearly. The current study aimed to minimize measure-specific error by carefully selecting each marker and including multiple indicators of disorders and personality traits. However, we used only one instrument to assess each social-cognitive vulnerability—because there were not three or more measures of most of these constructs—and we did not have multiple measures of the Big Five (or a faceted measure for all domains). Results may have differed if other measures of these constructs had been used. Relatedly, the fact that we did not have multiple measures of the social-cognitive vulnerabilities may have contributed to the relatively low precision observed for the parameter estimates of some of these latent variables. We collected clinical interview data as a complement to self-reported symptoms, but future studies should take this multimethod approach further by also collecting informant data on participants’ personality traits and vulnerabilities. Although many disorders and vulnerabilities were included here, other relevant constructs (e.g., borderline personality disorder, bipolar disorder, health anxiety, rumination, oddity) were omitted due to time constraints and should be examined in the future.
Finally, future work should delineate underlying mediating factors that may account for the associations among neuroticism, social-cognitive vulnerabilities, and disorders. In particular, anxiety sensitivity should be further studied as a potential additional source of comorbidity. In addition, focusing on the interaction of specific vulnerabilities that are most robustly associated with each disorder would provide a more nuanced, contextualized perspective on the circumstances under which a particular type of psychopathology is most likely to arise.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by an American Psychological Association Dissertation Grant Award, Society for a Science of Clinical Psychology Dissertation Grant Award, and University of Iowa Executive Council of Graduate and Professional Students Grant awarded to Kristin Naragon-Gainey, as well as National Institute of Mental Health Grant R01-MH068472 awarded to David Watson.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
