Multiple disorders tend to co-occur within the same individual, such that comorbidity is a pervasive feature of psychopathology (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017; Watson, Reference Watson2005, Reference Watson2009). This has led to intense interest in transdiagnostic processes that (a) are common to different disorders and (b) help to explain the substantial associations between them. This chapter focuses on individual differences in neuroticism and its core components. Neuroticism is the most widely studied and best-established transdiagnostic personality factor in psychopathology. It has been described as “an almost ubiquitously elevated trait within clinical populations” (Widiger & Costa, Reference Widiger and Costa1994, p. 81). Elevated levels of neuroticism have been reported in virtually all major forms of psychopathology, including the psychotic disorders, bipolar disorders, depressive disorders, anxiety disorders, obsessive-compulsive and related disorders, trauma- and stressor-related disorders, dissociative disorders, somatic symptom disorders, substance use disorders, feeding and eating disorders, and personality and conduct disorders (e.g., Koffel & Watson, Reference Watson2009; Kotov, Gamez, Schmidt, & Watson, Reference Kotov, Gamez, Schmidt and Watson2010; Samuel & Widiger, Reference Samuel and Widiger2008).
In this chapter, I examine associations between neuroticism and anxiety and related disorders. Specifically, I consider symptoms and diagnoses that fell within the anxiety disorders in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (American Psychiatric Association, 2000). In the revised framework of DSM-5 (American Psychiatric Association, 2013), these conditions now fall into three adjacent diagnostic classes: anxiety disorders, obsessive-compulsive and related disorders, and trauma- and stressor-related disorders.
I begin by providing a brief review of the basic features of neuroticism. I then examine how broad individual differences in neuroticism are linked to anxiety-related psychopathology, looking first at its associations with diagnoses, and then considering relations with the specific symptom dimensions subsumed within various disorders. Next, I explicate the nature of these associations further by examining links between anxiety-related psychopathology and the core component traits – or facets – within neuroticism. In the final sections of the chapter, I (a) highlight the complex causal mechanisms underlying these relations, (b) examine associations between neuroticism and other transdiagnostic traits that have been linked to the anxiety and related disorders (e.g., anxiety sensitivity, experiential avoidance), and (c) discuss key methodological issues and directions for future research.
Basic Features of Neuroticism
Relations with Negative Affect
Neuroticism reflects broad individual differences in subjective distress and dissatisfaction. In particular, it has been shown to be the strongest and broadest predictor of negative emotional experience. For example, Watson, Wiese, Vaidya, and Tellegen (Reference Watson, Wiese, Vaidya and Tellegen1999) reported a 0.58 correlation between neuroticism and the trait Negative Affect scale from the Positive and Negative Affect Schedule (PANAS) (Watson, Clark, & Tellegen, Reference Watson, Clark and Tellegen1988). Similarly, Beer, Watson, and McDade-Montez (Reference Beer, Watson and McDade-Montez2013) found that neuroticism was strongly correlated with scores on the PANAS Negative Affect scale in both self-ratings (weighted mean r = 0.59) and other-ratings (weighted mean r = 0.69). Watson (Reference Watson2000a) examined relations between neuroticism and four specific types of negative affectivity, which were assessed using the Expanded Form of the PANAS (PANAS-X) (Watson & Clark, Reference Watson and Clark1999): Fear (e.g., afraid, nervous), Sadness (e.g., blue, lonely), Guilt (e.g., ashamed, dissatisfied with self), and Hostility (e.g., angry, disgusted). Watson (Reference Watson2000a) reported that neuroticism was strongly related to individual differences in Sadness (r = 0.59), Guilt (r = 0.57), and Fear (r = 0.52), and more moderately linked to Hostility (r = 0.41). In light of these associations, neuroticism often is called negative emotionality or negative affectivity in the personality literature (e.g., Soto & John, Reference Soto and John2017; Watson, Reference Watson2000a).
Genetic and Biological Basis
Heritability of neuroticism.
All major personality traits have a strong genetic basis, with heritability estimates from twin studies generally falling in the 0.40 to 0.60 range. In a review of the behavior genetics literature that drew primarily from twin studies, Bouchard (Reference Bouchard2004) reported a 0.48 heritability estimate for neuroticism, noting that it displays evidence of both additive and nonadditive genetic effects. In a subsequent meta-analysis that included a broader range of both twin and adoption studies, Vukasović and Bratko (Reference Vukasović and Bratko2015) reported lower – but still substantial – heritability estimates ranging from 0.37 to 0.47 across various models and measures (see their Table 4.4).
Biological basis of neuroticism.
In terms of its underlying biology, neuroticism reflects individual differences in the withdrawal-oriented behavioral inhibition system (BIS) (Fowles, Reference Fowles and Spaulding1994; Gray, Reference Gray1982, Reference Gray1987; Watson et al., Reference Watson, Wiese, Vaidya and Tellegen1999). The essential purpose of the BIS is to inhibit behavior that might lead to pain, punishment, or some other aversive consequence. Gray (Reference Gray1987) called the BIS a “stop, look, and listen system” to emphasize how it redirects attention toward the environment. According to Gray, BIS activity focuses maximum attention on analyzing environmental stimuli, especially novel stimuli that potentially signal danger. The BIS also has a strongly anticipatory quality: it promotes a vigilant scanning of the environment for potential threats and motivates the organism to proceed cautiously until safety is indicated. Finally, elevated BIS activity is associated with a heightened sensitivity to punishment (e.g., Aluja & Blanch, Reference Aluja and Blanch2011; Corr & Cooper, Reference Corr and Cooper2016; Torrubia, Ávila, Moltó, & Caseras, Reference Torrubia, Ávila, Moltó and Caseras2001). Extensive evidence has established strong associations between neuroticism and various indicators of the BIS and punishment sensitivity (Aluja & Blanch, Reference Aluja and Blanch2011; Carver & White, Reference Carver and White1994; Corr & Cooper, Reference Corr and Cooper2016; Torrubia et al., Reference Torrubia, Ávila, Moltó and Caseras2001; Watson et al., Reference Watson, Wiese, Vaidya and Tellegen1999).
Given this link to the BIS – which is an ancient biobehavioral system – it is interesting to note that reliable individual differences in neuroticism also have been identified in numerous other species. For example, in an early review of the animal personality literature, Gosling and John (Reference Gosling and John1999) identified neuroticism-like factors in a broad range of species, including chimpanzees, gorillas, rhesus and vervet monkeys, hyenas, dogs, cats, rats, guppies, and octopi. Gartner, Powell, and Weiss (Reference Gartner, Powell and Weiss2014) subsequently found robust individual differences in neuroticism among snow leopards, clouded leopards, and lions.
Stability versus Change
Rank-order stability.
Similar to other personality traits, neuroticism shows only modest to moderate rank-order stability during childhood and adolescence. Starting in the 20s, however, it begins to display strong levels of stability, eventually reaching correlations > 0.70 over periods of six to seven years (Roberts & DelVecchio, Reference Roberts and DelVecchio2000). Collapsed across all age ranges, Roberts and DelVecchio reported an overall stability coefficient of 0.50 for neuroticism, a value very similar to those obtained for other general traits (see their Table 4.5).
Mean-level change.
Average levels of neuroticism vary systematically over the life span. Specifically, neuroticism scores tend to increase during adolescence (especially in females), and then show a consistent decline during early and middle adulthood (Roberts, Walton, & Viechtbauer, Reference Roberts, Walton and Viechtbauer2006; Soto, John, Gosling, & Potter, Reference Soto, John, Gosling and Potter2011). It is noteworthy that this decline in neuroticism during adulthood has been shown to be quite robust cross-culturally, although the timing of these age-related changes differs substantially across countries (Bleidorn, Reference Bleidorn2015).
Cohort effects.
Average levels of neuroticism have increased over time. From the 1950s through the early 1990s, there is evidence of birth cohort increases in neuroticism that is consistent across samples and nationalities (Twenge, Gentile, & Campbell, Reference Twenge, Gentile, Campbell, Mikulincer, Shaver, Cooper and Larsen2015). Since the early 1990s, however, trait levels have remained relatively stable or even declined slightly. Nevertheless, Twenge et al. (Reference Twenge, Gentile, Campbell, Mikulincer, Shaver, Cooper and Larsen2015) conclude their review by stating that neuroticism remains “at historically high levels” (p. 543).
The Hierarchical Structure of Personality
Personality traits are ordered hierarchically at different levels of abstraction or breadth (Markon, Krueger, & Watson, Reference Watson2005; Watson, Nus, & Wu, Reference Watson, Nus and Wu2017). Neuroticism is a key component in all of the major structural models of personality, including both the prominent Big Three and Big Five schemes (Markon et al., Reference Markon, Krueger and Watson2005). It is one of the “Big” traits in these models, which means that it is a broad, higher-order dimension that can be decomposed into more specific lower-order facets.
We currently lack consensus regarding the specific facets that fall within the higher-order domain of neuroticism. For instance, the NEO Personality Inventory-3 (NEO-PI-3) (McCrae, Costa, & Martin, Reference McCrae, Costa and Martin2005) contains six neuroticism facets; in contrast, the Faceted Inventory of the Five-Factor Model (FI-FFM) (Watson et al., Reference Watson, Nus and Wu2017) has only five neuroticism facets, whereas the Big Five Inventory-2 (BFI-2) (Soto & John, Reference Soto and John2017) includes only three. To create a consensually defined facet structure of neuroticism, Naragon-Gainey and Watson (Reference Naragon-Gainey and Watson2014) conducted a structural analysis using specific trait scales from five hierarchical instruments: the Revised NEO Personality Inventory (NEO PI-R) (Costa & McCrae, Reference Costa and McCrae1992), the Multidimensional Personality Questionnaire (MPQ) (Tellegen & Waller, Reference Tellegen, Waller, Boyle, Matthews and Saklofske2008), the Sixteen Personality Factor Questionnaire (16PF) (Cattell, Eber, & Tatsuoka, Reference Cattell, Eber and Tatsuoka1970), the Hogan Personality Inventory (HPI) (Hogan & Hogan, Reference Hogan and Hogan1992), and the Jackson Personality Inventory-Revised (JPI-R) (Jackson, Reference Jackson1994). These analyses revealed three consensual facets that represent specific types of negative affectivity: Anxiety (defined by scales such as NEO PI-R Anxiety, 16 PF Apprehension, JPI-R Anxiety, and MPQ Stress Reaction), Anger (marked on one end by NEO PI-R Angry Hostility, and on the other by scales such as HPI Even-Temperedness), and Depression (defined on one end by NEO PI-R Depression, and on the other by scales such as HPI No Depression and HPI No Guilt). I use this consensual three-facet structure to model relations with psychopathology at the lower-order level.
Methods
Overview
I examine four types of data to explicate the associations between neuroticism and anxiety-related psychopathology: (a) domain scores and anxiety diagnoses, (b) domain scores and anxiety symptoms, (c) facet scores and anxiety diagnoses, and (d) facet scores and anxiety symptoms. In the initial set of analyses, I summarize the meta-analytic findings of Kotov et al. (Reference Kotov, Gamez, Schmidt and Watson2010). For the other three sets of analyses, I use data collected from three samples. In the following sections, I describe these participants and the measures that are examined in subsequent analyses.
Participants and Procedure
Sample 1.
This sample (N = 669) included 296 psychiatric outpatients and 373 college students (for more details, see Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014; Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012). These participants completed multiple self-report measures of various types of anxiety symptoms (e.g., several social anxiety scales). I first standardized the scores on these highly correlated indicators – so they would be equally weighted – and then averaged them to produce an overall composite score for each type of symptom; these self-report composites are described later. To eliminate mean-level differences across populations, I standardized all of the symptom and trait scores on a within-population basis (i.e., the patient and student responses were standardized separately) so that they could be combined in a single overall analysis.
These participants also were interviewed using the Clinician Rating version of the Inventory of Depression and Anxiety Symptoms (IDAS-CR) (Koffel, Reference Koffel2011; Watson, Reference Watson2009; Watson et al., Reference Watson, O’Hara, Chmielewski, McDade-Montez, Koffel, Naragon and Stuart2008, Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012) and a companion instrument, the Personality, Cognitions, Consciousness, and Perceptions Interview (PCCP) (Koffel, Reference Koffel2011; Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012). The IDAS-CR and PCCP both consist of a series of single-item ratings representing specific types of symptoms. Each rating is made on a three-point scale (absent, subthreshold, present). To rate each symptom, the clinicians asked a standard initial probe question, as well as several standard follow up questions.
IDAS-CR ratings were available for six specific anxiety symptoms: the assessed variables included Generalized Anxiety, Social Anxiety, and Panic, as well as three types of posttraumatic stress disorder (PTSD) symptoms (Hyperarousal, Intrusions, Avoidance), which also were summed to create an overall PTSD composite. In addition, the PCCP was used to assess five specific types of obsessive-compulsive disorder (OCD) symptoms: Checking, Cleaning, Ordering, Hoarding, and Obsessions; these ratings also were summed to create a general OCD composite. Across the two instruments, I report data on 13 interview-based symptom ratings; this includes five overall scores (Generalized Anxiety, Social Anxiety, Panic, PTSD, OCD), three specific types of PTSD symptoms, and five types of OCD symptoms.
IDAS-CR/PCCP data were available on 625 participants (252 patients, 373 students). As before, I standardized the scores on a within-population basis and then examined them in a single overall analysis.
Finally, the outpatients (N = 252) also were interviewed using the Structured Clinical Interview for DSM-IV (SCID-IV) (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams1997). I report associations between specific neuroticism facets and five DSM-IV anxiety disorder diagnoses: generalized anxiety disorder (GAD), PTSD, panic disorder, social phobia, and OCD.
Sample 2.
The participants were 406 community adults who completed an extensive battery of self-report symptom measures (see Watson et al., Reference Watson, Nus and Wu2017; Watson, Stasik, Ellickson-Larew, & Stanton, Reference Watson, Stasik, Ellickson-Larew and Stanton2015). Although this sample was not fully clinical in nature, roughly half of the participants (46.3%) were either currently receiving treatment for mental health issues or had received it in the past. Consequently, these participants exhibited a relatively high level of psychopathology. As in Sample 1, multiple measures were available for most types of symptoms; these indicators were standardized and then combined to yield an overall composite for each symptom dimension.
In addition, the participants were interviewed using an adapted version (with the authorization of the author) of the Mini-International Neuropsychiatric Interview (M.I.N.I) (Sheehan et al., Reference Sheehan, Lecrubier, Sheehan, Amorim, Janavas, Weiller and Dunbar1998) that incorporated diagnostic changes for DSM-5 (see Watson et al., Reference Watson, Stasik, Ellickson-Larew and Stanton2015). I examine associations between neuroticism facets and six DSM-5 diagnoses: GAD, PTSD, panic disorder, agoraphobia, social phobia, and OCD.1
Sample 3.
This sample consisted of 268 outpatients (see Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014; Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012) who were interviewed using the revised version of the Interview for Mood and Anxiety Symptoms (IMAS) (Gamez, Kotov, & Watson, Reference Gamez, Kotov and Watson2010; Waszczuk, Kotov, Ruggero, Gamez, & Watson, Reference Waszczuk, Kotov, Ruggero, Gamez and Watson2017). The IMAS provides multi-item measures of current (i.e., past month) symptoms; each item is scored on a three-point rating scale (absent, subthreshold, above threshold). Its items were designed to cover all DSM-IV Mood and Anxiety Disorder symptom criteria.
Based on data from prior studies (Watson et al., Reference Watson, O’Hara, Simms, Kotov, Chmielewski, McDade-Montez and Stuart2007; Gamez et al., Reference Gamez, Kotov and Watson2010), the IMAS was revised to improve its symptom coverage. I examine 18 anxiety symptom scores derived from this revised version (see Waszczuk et al., Reference Waszczuk, Kotov, Ruggero, Gamez and Watson2017): GAD (12 items), Social Phobia (15 items), Panic (23 items), four types of PTSD symptoms (overall PTSD [25 items], Hyperarousal [6 items], Intrusions [4 items], Avoidance [3 items]), Agoraphobia (15 items), Claustrophobia [6 items], four indicators of specific phobia (overall Specific Phobia [13 items], Animal Phobia [3 items], Situational Phobia [3 items], Blood-Injection-Injury Phobia [4 items]), and five measures of OCD symptoms (overall OCD [18 items], Obsessions [6 items], Checking [4 items], Ordering [6 items], Cleaning [5 items]).
Neuroticism Measures
Sample 1.
The Sample 1 participants completed three overall measures of neuroticism: (a) an eight-item version of the Big Five Inventory (BFI) (John & Srivastava, Reference John, Srivastava, Pervin and John1999) Neuroticism scale; (b) the sum of the Anxiety, Depression, and Angry Hostility facet scales from the NEO PI-R (Costa & McCrae; Reference Costa and McCrae1992); and (c) the sum of the Anxiety, Depression, and Anger Proneness facet scales from the FI-FFM (Watson et al., Reference Watson, Nus and Wu2017). Correlations between these measures ranged from 0.79 to 0.90, with a mean coefficient of 0.85. These indicators were standardized and then averaged to form an overall domain score for neuroticism.
I also used the individual NEO PI-R and FI-FFM scales to create three standardized facet scores. Thus, NEO PI-R Anxiety and FI-FFM Anxiety (r = 0.81) were averaged to assess Anxiety; NEO PI-R Depression and FI-FFM Depression (r = 0.85) were used to model Depression; and NEO PI-R Angry Hostility and FI-FFM Anger Proneness (r = 0.84) were combined to measure Anger.
Sample 2.
These participants completed the full NEO-PI-3 (McCrae et al., Reference McCrae, Costa and Martin2005) and FI-FFM (Watson et al. Reference Watson, Nus and Wu2017). The Neuroticism domain scores from these instruments correlated 0.84 with each other; they were standardized and then combined into a composite. In addition, I again created standardized facet scores by averaging NEO-PI-3 Anxiety and FI-FFM Anxiety (r = 0.81), NEO-PI-3 Depression and FI-FFM Depression (r = 0.81), and NEO-PI-3 Angry Hostility and FI-FFM Anger Proneness (r = 0.79).
Sample 3.
These participants completed a nine-item version of the BFI Neuroticism scale (John & Srivastava, Reference John, Srivastava, Pervin and John1999).
Self-Rated Symptoms
GAD symptoms.
The participants in Samples 1 and 2 completed three GAD symptom measures that were combined into composites: (1) the 7-item Anxious Mood scale from the Inventory of Depression and Anxiety Symptoms (IDAS) (Watson et al., Reference Watson, O’Hara, Simms, Kotov, Chmielewski, McDade-Montez and Stuart2007); (2) the 9-item Generalized Anxiety Disorder Questionnaire-IV (GADQ-IV) (Newman et al., Reference Newman, Zuellig, Kachin, Constantino, Przeworski, Erickson and Cashman-McGrath2002); and (3) the 10-item Worry Domains Questionnaire-Short Form (Stöber & Joormann, Reference Stöber and Joormann2001).
PTSD symptoms.
The Sample 1 participants were assessed on two PTSD symptom measures: (1) the 17-item PTSD Checklist-Civilian Version (PCL) (Weathers, Litz, Herman, Huska, & Keane, Reference Weathers, Litz, Herman, Huska and Keane1993); and (2) an aggregate score based on the Traumatic Intrusions (4 items) and Traumatic Avoidance (4 items) scales of the Expanded Version of the IDAS (IDAS-II) (Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012). In addition to creating an overall PTSD score, I also examined four specific types of PTSD symptoms (for discussions of PTSD symptom structure, see Watson, Reference Watson2009; Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012; Yufik & Simms, Reference Yufik and Simms2010). The Sample 1 battery included two measures of Intrusions (PCL Intrusions, IDAS-II Traumatic Intrusions), two indicators of Avoidance (PCL Avoidance, IDAS-II Traumatic Avoidance), and single markers of both Dysphoria (PCL Dysphoria) and Hyperarousal (PCL Hyperarousal).
The Sample 2 participants completed (1) the five intrusions items and two avoidance items from the PCL and (2) the IDAS-II Traumatic Intrusions and Traumatic Avoidance scales. These measures were used to create three composites: overall PTSD, Intrusions, and Avoidance.
Social anxiety.
The Sample 1 participants completed three indicators of social anxiety: (1) the 5-item Social Phobia scale from the Fear Questionnaire (FQ) (Marks & Mathews, Reference Marks and Mathews1979); (2) the 10-item Social Phobia scale from the Albany Panic and Phobia Questionnaire (APPQ) (Rapee, Craske, & Barlow, Reference Rapee, Craske and Barlow1994/1995); and (3) the 6-item IDAS-II Social Anxiety scale (Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012). The Sample 2 participants were assessed on these same three measures, plus a factor analytically derived 10-item Social Anxiety scale from the Schizotypal Personality Questionnaire (Chmielewski & Watson, Reference Chmielewski and Watson2008).
Panic.
The Sample 1 battery contained two measures of panic symptoms: (1) the 17-item Anxious Arousal scale from the Mood and Anxiety Symptom Questionnaire (MASQ) (Watson et al., Reference Watson, Weber, Assenheimer, Clark, Strauss and McCormick1995) and (2) the 13-item Panic Attack Symptom Questionnaire (PASQ) (Watson, Reference Watson2000b). The Sample 2 participants completed three indicators of panic: (1) a reduced, 9-item version of the MASQ Anxious Arousal scale (Watson et al., Reference Watson, Weber, Assenheimer, Clark, Strauss and McCormick1995), (2) an abbreviated, 6-item version of the PASQ (Watson, Reference Watson2000b), and (3) the 8-item IDAS-II Panic scale (Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012).
OCD.
Participants in both samples completed three OCD measures: (1) the 18-item Obsessive-Compulsive Inventory-Revised (OCI-R) (Foa et al., Reference Foa, Huppert, Leiberg, Langner, Kichic, Hajcak and Salkovskis2002); (2) a total score based on the Obsessive Checking (14 items), Obsessive Cleanliness (12 items), Compulsive Rituals (8 items), and Hoarding (5 items) scales from the Schedule of Compulsions, Obsessions, and Pathological Impulses (SCOPI) (Watson & Wu, Reference Watson and Wu2005); and (3) a combined score based on the IDAS-II Checking (3 items), Ordering (5 items), and Cleaning (7 items) scales. In addition to creating an overall OCD score, I constructed five specific OCD symptom composites in each sample (for reviews of OCD symptom structure, see Watson, Reference Watson2009; Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012): Obsessions (OCI-R Obsessions), Checking (OCI-R Checking, SCOPI Obsessive Checking, IDAS-II Checking), Ordering (OCI-R Ordering, SCOPI Compulsive Rituals, IDAS-II Ordering), Cleaning (OCI-R Washing, SCOPI Obsessive Cleanliness, IDAS-II Cleaning), and Hoarding (OCI-R Hoarding, SCOPI Hoarding).
Agoraphobia.
The Sample 2 battery included two measures of agoraphobia: (1) the nine-item APPQ Agoraphobia scale and the five-item FQ Agoraphobia scale.
Claustrophobia.
The Sample 2 participants completed the five-item IDAS-II Claustrophobia scale, which assesses experienced fear and behavioral avoidance related to crowded public settings (e.g., afraid of getting trapped in a crowd); small, tight spaces (e.g., avoided small spaces); and dark, confined places (e.g., afraid of tunnels).
Specific phobia.
Participants in Sample 2 were assessed on two types of specific phobia. First, they completed three measures of blood-injection-injury (BII) phobia: (1) the 5-item FQ Blood-Injection scale; (2) the 10-item Blood-Injection scale from the Phobic Stimuli Response Scales (PSRS) (Cutshall & Watson, Reference Cutshall and Watson2004); and (3) a 5-item BII scale created from the Fear Survey Schedule (FSS) (Wolpe & Lang, Reference Wolpe and Lang1964). Second, they were assessed on three measures of animal phobia: (1) the 8-item PSRS Animal fear scale (Cutshall & Watson, Reference Cutshall and Watson2004); (2) a 5-item Animal Phobia scale constructed from the FSS (Wolpe & Lang, Reference Wolpe and Lang1964); and the provisional 9-item Animal Phobia scale of the IDAS-II (Watson et al., Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012). These scores were used to create three composite measures: Blood-Injection Phobia, Animal Phobia, and overall Specific Phobia.
Neuroticism and Anxiety Disorder Diagnoses
Kotov et al. (Reference Kotov, Gamez, Schmidt and Watson2010) conducted a meta-analysis in which they compared the mean trait scores of individuals with and without 14 DSM-IV diagnoses (e.g., the mean neuroticism scores of individuals with and without GAD). Specifically, they examined three DSM-IV mood disorder (major depression, dysthymic disorder, unspecified unipolar depression), seven anxiety disorder (GAD, PTSD, panic disorder, agoraphobia, social phobia, specific phobia, OCD), and four substance use disorder (overall substance use, primarily alcohol, primarily drugs, mixed use) diagnoses. The mean differences between diagnostic cases and non-cases were divided by their pooled standard deviations to yield effect sizes expressed as Cohen’s d (Cohen, Reference Cohen1992; see their Table 4.4); positive d values indicate that the diagnostic cases scored higher on the trait.
Kotov et al. (Reference Kotov, Gamez, Schmidt and Watson2010) examined relations with six broad traits: the higher-order Big Five domains, plus disinhibition from the Big Three model. Neuroticism easily was the strongest predictor of psychopathology in these analyses. It was significantly – and substantially – related to every diagnosis. In fact, it produced large effect sizes (Cohen, Reference Cohen1992) with 13 of the 14 analyzed disorders (the exception was alcohol use) and displayed very little diagnostic specificity overall (ds ranged from 0.77 to 2.25, mean d = 1.65). Table 4.1 presents Kotov et al.’s (Reference Kotov, Gamez, Schmidt and Watson2010) meta-analytic results for the seven anxiety disorder diagnoses. It is noteworthy that six of the d values exceeded 1.60; the d for specific phobia was substantially smaller (0.92), but still represented a large effect size. These findings clearly establish neuroticism as a strong transdiagnostic factor in the DSM-IV anxiety disorders (as well as in the depressive and substance use disorders).
Table 4.1 Meta-analytic associations between neuroticism and anxiety disorder diagnoses
| Diagnosis | K | NC | NNC | d |
|---|---|---|---|---|
| Posttraumatic stress disorder | 16 | 1,714 | 22,174 | 2.25 |
| Obsessive-compulsive disorder | 18 | 905 | 25,152 | 2.07 |
| Generalized anxiety disorder | 14 | 1,674 | 44,570 | 1.96 |
| Panic disorder | 24 | 2,556 | 32,227 | 1.92 |
| Social phobia | 18 | 3,188 | 36,165 | 1.63 |
| Agoraphobia | 15 | 1,451 | 24,902 | 1.61 |
| Specific phobia | 10 | 2,800 | 21,367 | 0.92 |
| Mean effect size | 1.77 |
Note. K = number of studies. NC = number of cases (patients). NNC = number of non-cases.
Neuroticism and Anxiety Symptoms
The Problem of Heterogeneity
Disorder-based analyses are limited in several important ways (see Watson, Reference Watson2005, Reference Watson2009). One particularly important problem is that diagnosis-based analyses fail to capture the marked heterogeneity of many DSM disorders. As already noted, structural analyses have identified replicable symptom dimensions within several of the DSM-IV anxiety disorders, including PTSD, OCD, and specific phobia (Watson, Reference Watson2005, Reference Watson2009; Watson et al., Reference Watson, O’Hara, Simms, Kotov, Chmielewski, McDade-Montez and Stuart2007, Reference Watson, O’Hara, Naragon-Gainey, Koffel, Chmielewski, Kotov and Ruggero2012; Watson & Wu, Reference Watson and Wu2005); moreover, these symptom dimensions are highly distinctive and often correlate quite differently with other variables. Consequently, symptom-based analyses provide a clearer, more nuanced view of the associations between neuroticism and psychopathology; they also show much greater evidence of specificity.
Associations with Anxiety Symptoms
Overall symptom composites.
Neuroticism shows substantial specificity at the symptom level (Watson, Gamez, & Simms, Reference Watson, Gamez and Simms2005; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014). It correlates most strongly with indicators of subjective distress (e.g., the anxious, apprehensive mood that is the core feature of GAD), more moderately with symptoms associated with more limited forms of distress (e.g., panic, social anxiety), and weakly with symptoms characterized primarily by behavioral avoidance of specific stimuli (e.g., phobias).
Table 4.2 presents correlations between neuroticism domain scores and self-rated (Samples 1 and 2) and interview-based (IDAS-CR in Sample 1, IMAS in Sample 3) anxiety symptoms in the three current samples. Consistent with previous studies (Watson et al., Reference Watson, Gamez and Simms2005; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014), neuroticism was significantly related to every type of anxiety symptom, which again establishes it as a strong transdiagnostic factor in the DSM-IV anxiety disorders. Nevertheless, the magnitude of these relations varied dramatically, ranging from a high of 0.81 (self-rated GAD symptoms in Sample 1) to a low of only 0.15 (IMAS Blood-Injection Phobia in Sample 3). Replicating previous findings (Watson et al., Reference Watson, Gamez and Simms2005; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014), neuroticism was most strongly linked to GAD symptoms: across the four sets of analyses, these correlations ranged from 0.47 to 0.81, with a weighted mean value (after r-to-z transformation) of 0.70. Neuroticism also was strongly related to overall symptoms of PTSD (weighted mean r = 0.59), Panic (r = 0.55), and Social Anxiety (r = 0.53). It was more moderately related to Agoraphobia (r = 0.45), OCD (r = 0.42), and Claustrophobia (r = 0.38), and only modestly linked to Specific Phobia (r = 0.28). These data again demonstrate that neuroticism shows substantial specificity at the symptom level: it is most strongly related to core indicators of subjective distress/negative mood and exhibits more moderate associations with other types of anxiety symptoms.
Table 4.2 Correlations between neuroticism and anxiety symptoms
| Symptom Score | Self-Report | Interview | ||
|---|---|---|---|---|
| Sample 1 | Sample 2 | IDAS-CR | IMAS | |
| Generalized anxiety | 0.81 | 0.74 | 0.47 | 0.73 |
| PTSD | 0.62 | 0.60 | 0.42 | 0.71 |
| Dysphoria | 0.71 | .— | 0.41 | .— |
| Hyperarousal | 0.55 | .— | 0.29 | 0.58 |
| Intrusions | 0.50 | 0.60 | 0.30 | 0.39 |
| Avoidance | 0.47 | 0.52 | 0.31 | 0.42 |
| Social anxiety | 0.57 | 0.64 | 0.43 | 0.47 |
| Panic | 0.57 | 0.64 | 0.34 | 0.64 |
| OCD | 0.42 | 0.45 | 0.40 | 0.40 |
| Obsessions | 0.53 | 0.55 | 0.41 | 0.43 |
| Checking | 0.43 | 0.44 | 0.24 | 0.33 |
| Ordering | 0.34 | 0.34 | 0.21 | 0.30 |
| Hoarding | 0.29 | 0.35 | 0.21 | .— |
| Cleaning | 0.26 | 0.25 | 0.11 | 0.36 |
| Agoraphobia | .— | 0.44 | .— | 0.46 |
| Claustrophobia | .— | 0.39 | .— | 0.36 |
| Specific Phobia | .— | 0.26 | .— | 0.31 |
| Situational | .— | .— | .— | 0.30 |
| Blood-Injection | .— | 0.24 | .— | 0.15 |
| Animal | .— | 0.22 | .— | 0.23 |
Note. N = 669 (Sample 1), 406 (Sample 2), 625 (IDAS-CR), 268 (IMAS). Correlations ≥ 0.40 are highlighted. IDAS-CR = Clinician Rating Version of the IDAS. IMAS = Interview of Mood and Anxiety Symptoms. PTSD = posttraumatic stress disorder. OCD = obsessive-compulsive disorder.
PTSD symptoms. PTSD symptoms differ markedly in their specificity and in their associations with neuroticism/negative affectivity. In particular, Dysphoria symptoms (e.g., feeling numb, feeling irritable, difficulty concentrating) show much greater non-specificity, correlating more strongly with indicators of depression (Gootzeit & Markon, Reference Gootzeit and Markon2011; Watson, Reference Watson2009), anxiety (Gootzeit & Markon, Reference Gootzeit and Markon2011), and neuroticism (Watson et al., Reference Watson, Gamez and Simms2005; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014) than other types of PTSD symptoms. Table 4.2 demonstrates that the current participants exhibited the same basic pattern. Thus, neuroticism correlated more strongly with Dysphoria (weighted r = 0.58) than with Hyperarousal (r = 0.46), Intrusions (r = 0.45), and Avoidance (r = 0.43). Follow-up significance tests (Kenny, Reference Kenny1987) indicated that neuroticism correlated significantly more strongly with Dysphoria than with other types of PTSD symptoms in all six comparisons (all ps < 0.05, two-tailed). These results illustrate the heterogeneity of the PTSD diagnosis and establish the importance of examining specific types of symptoms separately.
OCD symptoms.
OCD symptoms also display different levels of specificity, although the observed pattern varies somewhat across instruments (Watson, Reference Watson2009; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014). The most robust finding is that obsessive intrusions correlate more strongly with indicators of depression, anxiety, and neuroticism than other types of OCD symptoms. Consistent with these previous data, Table 4.2 establishes that neuroticism was more strongly related to Obsessions (weighted mean r = 0.48) than to Checking (r = 0.36), Ordering (r = 0.30), Hoarding (0.28), and Cleaning (0.25). Follow-up tests indicated that neuroticism correlated significantly more strongly (p < 0.01, 2-tailed) with Obsessions than with other types of OCD symptoms in 13 of 15 comparisons; the only exceptions were that it did not correlate more strongly with IMAS Obsessions (r = 0.43) than with IMAS Checking (r = 0.33, z = 1.53) and IMAS Cleaning (r = 0.36, z = 1.14) in Sample 3. Paralleling the PTSD analyses, these findings demonstrate the value of examining specific types of OCD symptoms separately.
Specific phobia symptoms.
All of the specific phobia measures displayed relatively weak associations with neuroticism. In fact, the strongest individual coefficient was only 0.30 (with IMAS Situational Phobia in Sample 3). These results are consistent with previous findings indicating that neuroticism is weakly linked to symptoms characterized primarily by behavioral avoidance (Watson et al., Reference Watson, Gamez and Simms2005; Watson & Naragon-Gainey, Reference Watson and Naragon-Gainey2014).
Neuroticism Facets and Anxiety Disorder Diagnoses
PANAS-X Analyses
I now examine how specific types of neuroticism/negative affectivity relate to anxiety disorder diagnoses. Watson, Clark, and Stasik (Reference Watson, Clark and Stasik2011) reported associations between the specific affect scales of the PANAS-X and five DSM-IV anxiety disorder diagnoses – GAD, PTSD, panic disorder, social phobia, and OCD. Data were obtained from two samples of outpatients who completed the PANAS-X using either general trait instructions (N = 331) or “past week” instructions (N = 253).
Watson et al. (Reference Watson, Clark and Stasik2011) conducted logistic regression analyses to identify the unique, incremental predictive power of the individual affect scales. PANAS-X Fear clearly emerged as the strongest individual predictor of the anxiety disorders. It produced significant positive effects in 8 of 10 analyses (80%), showing robust associations with GAD, PTSD, and panic disorder that replicated across samples. In contrast, the Sadness and Guilt scales had consistent associations with major depression, but failed to contribute significantly to any of the anxiety disorder diagnoses. Finally, PANAS-X Hostility had a significant positive effect in only one analysis (involving GAD). These results suggest that individual differences in fear and nervousness primarily are responsible for observed associations between neuroticism and DSM-IV anxiety diagnoses.
Neuroticism Facet Analyses
Bivariate analyses.
We can test the generalizability of these findings across measures (viz., the PANAS-X scales vs. the neuroticism facet scales from the NEO-PI-3 and FI-FFM) and diagnostic systems (i.e., DSM-IV vs. DSM-5). Table 4.3 presents polyserial correlations between the neuroticism facet scales and anxiety disorder diagnoses in Samples 1 and 2, as well as weighted mean coefficients computed across both samples. Polyserial correlations estimate the linear association between two normally distributed latent continuous variables when one of the observed variables is ordinal and the other is continuous (Flora & Curran, Reference Flora and Curran2004; Olsson, Drasgow, & Dorans, Reference Olsson, Drasgow and Dorans1982). They retain the relative rank order information provided by Pearson correlations but are unaffected by differences in prevalence rates, thereby facilitating comparisons across dichotomous diagnoses. The interview variables were scored as 0 = absent, 1 = present, so that positive correlations indicate that higher scores on a facet were associated with an increased likelihood of receiving that rating.
Table 4.3 Polyserial correlations between neuroticism facet scales and anxiety disorder diagnoses
| Facet/Sample | GAD | PTSD | Panic | Agora | Social Phobia | OCD |
|---|---|---|---|---|---|---|
| Anxiety | ||||||
| Sample 1 | 0.55 | 0.42 | 0.52 | .— | 0.35 | 0.39 |
| Sample 2 | 0.57 | 0.54 | 0.50 | 0.43 | 0.53 | 0.61 |
| Weighted M | 0.56 | 0.49 | 0.51 | 0.43 | 0.46 | 0.52 |
| Depression | ||||||
| Sample 1 | 0.44 | 0.44 | 0.31 | .— | 0.36 | 0.22 |
| Sample 2 | 0.60 | 0.53 | 0.40 | 0.37 | 0.49 | 0.48 |
| Weighted M | 0.54 | 0.50 | 0.36 | 0.37 | 0.44 | 0.38 |
| Anger | ||||||
| Sample 1 | 0.28 | 0.29 | 0.21 | .— | 0.18 | 0.25 |
| Sample 2 | 0.42 | 0.46 | 0.36 | 0.36 | 0.27 | 0.40 |
| Weighted M | 0.36 | 0.39 | 0.31 | 0.36 | 0.24 | 0.35 |
Note. N = 252 (Sample 1), 402 (Sample 2). Correlations ≥ 0.40 are highlighted.
Consistent with the findings of Watson et al. (Reference Watson, Clark and Stasik2011), the Anxiety facet displayed the strongest and broadest associations with anxiety disorder diagnoses in these data. Across the six assessed diagnoses, it had weighted mean correlations ranging from 0.43 (agoraphobia) to 0.56 (GAD), with a grand mean value of 0.50. The Depression facet also showed moderate to strong associations with these disorders. It had weighted mean correlations ranging from 0.36 (panic disorder) to 0.54 (GAD), with a grand mean value of 0.43. Finally, the Anger facet clearly had the weakest links to anxiety diagnoses: It had weighted mean correlations ranging from only 0.24 (social phobia) to 0.39 (PTSD), with a grand mean value of 0.33.
Multivariate analyses.
To determine the unique incremental information provided by each trait, I conducted logistic regression analyses in which the three facets were included as predictors of each diagnosis. Table 4.4 presents the odds ratios (ORs) from these logistic regressions. Note that an OR significantly greater than 1.00 indicates that the facet was associated with an increased likelihood of receiving that diagnosis (i.e., greater psychopathology). In contrast, an OR significantly less than 1.00 indicates that higher scores on the trait were associated with a reduced likelihood of receiving that diagnosis (i.e., lower levels of psychopathology). The latter represent suppressor effects (Watson, Clark, Chmielewski, & Kotov, Reference Watson, Clark, Chmielewski and Kotov2013), wherein positive bivariate relations (see Table 4.3) are transformed into significant negative associations at the multivariate level.
Table 4.4 Odds ratios from logistic regression analyses predicting DSM anxiety disorder diagnoses from the neuroticism facet scales
| Facet/Sample | GAD | PTSD | Panic | Agora | Social Phobia | OCD |
|---|---|---|---|---|---|---|
| Anxiety | ||||||
| Sample 1 | 2.86 | 1.57 | 3.45 | .— | 1.48 | 2.35 |
| Sample 2 | 1.96 | 2.04 | 2.81 | 2.34 | 3.33 | 3.68 |
| Depression | ||||||
| Sample 1 | 1.28 | 1.77 | 0.92 | .— | 1.59 | 0.79 |
| Sample 2 | 2.64 | 1.67 | 1.10 | 1.03 | 2.05 | 1.19 |
| Anger | ||||||
| Sample 1 | 0.89 | 1.13 | 0.93 | .— | 0.91 | 1.33 |
| Sample 2 | 0.89 | 1.43 | 1.16 | 1.34 | 0.57 | 1.22 |
Note. N = 252 (Sample 1), 402 (Sample 2). Significant effects (p < 0.05) are highlighted.
Replicating the findings of Watson et al. (Reference Watson, Clark and Stasik2011), the Anxiety facet clearly was the strongest and broadest predictor of anxiety disorder diagnoses in these samples. It produced significant positive effects in 8 of 11 analyses (72.7%) and showed consistent associations with three diagnoses – GAD, panic disorder, and OCD – that replicated across samples. It also had two additional associations that approached significance: PTSD in Sample 2 (OR = 2.04, p < 0.06) and social phobia in Sample 1 (OR = 1.48, p < 0.08). In contrast, the Depression facet contributed positively in only three cases (27.3%), although it was a consistent predictor of social phobia. Finally, the Anger facet yielded only a single suppressor effect (OR = 0.57 with social phobia in Sample 2).
Summary
Taken together with the findings of Watson et al. (Reference Watson, Clark and Stasik2011), the current results establish that individual differences in fear/anxiety have the strongest and broadest links to anxiety disorder diagnoses. Overall, the fear/anxiety scales produced significant positive associations in 16 of 21 analyses (76.2%). Moreover, these scales had highly robust associations with both (a) GAD and (b) panic disorder that replicated across four different samples. In marked contrast, indicators of sadness/depression displayed a significant positive effect in only three (14.3%) analyses, whereas the anger/hostility scales contributed positively in only one case (4.8%); furthermore, neither facet produced an association that consistently replicated across samples. Consequently, it seems reasonable to conclude that individual differences in fear and nervousness primarily are responsible for observed associations between neuroticism and DSM-IV/DSM-5 anxiety disorder diagnoses.
Neuroticism Facets and Anxiety Symptoms
Overview
Finally, I examine relations between these three neuroticism facets and anxiety symptoms. These associations are presented in Tables 4.5 (self-rated symptoms in Sample 1), 4.6 (self-rated symptoms in Sample 2), and 4.7 (IDAS-CR ratings in Sample 1). Each table reports both (a) simple Pearson correlations and (b) standardized β weights from multiple regression analyses in which the three facets were entered as joint predictors of each symptom dimension. Overall, 45 sets of analyses are reported across the three tables.
Table 4.5 Associations between neuroticism facets and self-rated anxiety symptoms (Sample 1)
| Symptom Score | Correlations | Regressions | ||||
|---|---|---|---|---|---|---|
| Anxiety | Depression | Anger | Anxiety | Depression | Anger | |
| Generalized anxiety | 0.77 | 0.76 | 0.52 | 0.47* | 0.44* | 0.01 |
| PTSD | 0.53 | 0.63 | 0.43 | 0.18* | 0.47* | 0.06 |
| Dysphoria | 0.56 | 0.74 | 0.51 | 0.07 | 0.64* | 0.10* |
| Hyperarousal | 0.51 | 0.48 | 0.42 | 0.31* | 0.19* | 0.14* |
| Intrusions | 0.45 | 0.50 | 0.35 | 0.20* | 0.34* | 0.04 |
| Avoidance | 0.41 | 0.49 | 0.31 | 0.14* | 0.40* | −0.00 |
| Social anxiety | 0.53 | 0.56 | 0.35 | 0.30* | 0.38* | −0.04 |
| Panic | 0.53 | 0.53 | 0.39 | 0.31* | 0.29* | 0.05 |
| OCD | 0.39 | 0.34 | 0.33 | 0.26* | 0.09 | 0.13* |
| Obsessions | 0.47 | 0.51 | 0.36 | 0.22* | 0.34* | 0.04 |
| Checking | 0.44 | 0.32 | 0.33 | 0.37* | −0.01 | 0.12* |
| Ordering | 0.33 | 0.24 | 0.29 | 0.26* | −0.02 | 0.16* |
| Hoarding | 0.24 | 0.26 | 0.27 | 0.08 | 0.12* | 0.16* |
| Cleaning | 0.29 | 0.16 | 0.21 | 0.30* | −0.10 | 0.11* |
Note. N = 669. Correlations ≥ 0.40 are highlighted. Depress = Depression. PTSD = posttraumatic stress disorder. OCD = obsessive-compulsive disorder.
* Standardized β weight is significant at p < 0.05.
Table 4.6 Associations between neuroticism facets and self-rated anxiety symptoms (Sample 2)
| Symptom Score | Correlations | Regressions | ||||
|---|---|---|---|---|---|---|
| Anxiety | Depression | Anger | Anxiety | Depression | Anger | |
| Generalized anxiety | 0.71 | 0.73 | 0.53 | 0.39* | 0.45* | −0.02 |
| PTSD | 0.57 | 0.56 | 0.49 | 0.29* | 0.26* | 0.13* |
| Intrusions | 0.55 | 0.57 | 0.48 | 0.24* | 0.30* | 0.12* |
| Avoidance | 0.51 | 0.48 | 0.43 | 0.31* | 0.17* | 0.11 |
| Social anxiety | 0.59 | 0.61 | 0.42 | 0.33* | 0.42* | −0.08 |
| Panic | 0.56 | 0.60 | 0.52 | 0.20* | 0.34* | 0.17* |
| OCD | 0.41 | 0.38 | 0.42 | 0.20* | 0.07 | 0.24* |
| Obsessions | 0.48 | 0.51 | 0.47 | 0.17* | 0.26* | 0.18* |
| Checking | 0.43 | 0.39 | 0.36 | 0.27* | 0.12 | 0.11 |
| Ordering | 0.31 | 0.28 | 0.33 | 0.16* | 0.01 | 0.22* |
| Hoarding | 0.29 | 0.31 | 0.31 | 0.07 | 0.15 | 0.17* |
| Cleaning | 0.24 | 0.19 | 0.28 | 0.14 | −0.08 | 0.25* |
| Agoraphobia | 0.45 | 0.35 | 0.31 | 0.44* | −0.00 | 0.03 |
| Claustrophobia | 0.32 | 0.33 | 0.39 | 0.07 | 0.09 | 0.28* |
| Specific phobia | 0.27 | 0.19 | 0.23 | 0.24* | −0.07 | 0.12 |
| Blood-injection | 0.25 | 0.19 | 0.18 | 0.23* | −0.01 | 0.04 |
| Animal | 0.23 | 0.16 | 0.23 | 0.21* | −0.11 | 0.17* |
Note. N = 406. Correlations ≥ 0.40 are highlighted. Depress = Depression. PTSD = posttraumatic stress disorder. OCD = obsessive-compulsive disorder.
* Standardized β weight is significant at p < 0.05
Table 4.7 Associations between neuroticism facets and clinician-rated anxiety symptoms (Sample 1)
| Symptom Score | Correlations | Regressions | ||||
|---|---|---|---|---|---|---|
| Anxiety | Depress | Anger | Anxiety | Depress | Anger | |
| Generalized anxiety | 0.45 | 0.40 | 0.32 | 0.31* | 0.15* | 0.05 |
| PTSD | 0.36 | 0.44 | 0.29 | 0.11* | 0.36* | 0.02 |
| Dysphoria | 0.32 | 0.48 | 0.26 | −0.00 | 0.51* | −0.04 |
| Hyperarousal | 0.26 | 0.27 | 0.23 | 0.12* | 0.15* | 0.07 |
| Intrusions | 0.27 | 0.30 | 0.20 | 0.11* | 0.23* | 0.00 |
| Avoidance | 0.29 | 0.32 | 0.24 | 0.12* | 0.21* | 0.05 |
| Social anxiety | 0.36 | 0.43 | 0.29 | 0.11* | 0.33* | 0.03 |
| Panic | 0.34 | 0.31 | 0.21 | 0.24* | 0.14* | −0.01 |
| OCD | 0.36 | 0.36 | 0.33 | 0.18* | 0.15* | 0.15* |
| Obsessions | 0.35 | 0.44 | 0.25 | 0.09 | 0.40* | −0.03 |
| Checking | 0.24 | 0.17 | 0.19 | 0.21* | −0.03 | 0.09 |
| Ordering | 0.20 | 0.14 | 0.23 | 0.14* | −0.07 | 0.19* |
| Hoarding | 0.14 | 0.21 | 0.23 | −0.05 | 0.14* | 0.18* |
| Cleaning | 0.14 | 0.07 | 0.08 | 0.17* | −0.06 | 0.02 |
Note. N = 625. Correlations ≥ 0.40 are highlighted. Depress = Depression. PTSD = posttraumatic stress disorder. OCD = obsessive-compulsive disorder.
* Standardized β weight is significant at p < 0.05.
Anxiety
Bivariate relations.
Consistent with the diagnostic data, the Anxiety facet was the strongest and broadest predictor of these symptom dimensions. Across the three sets of analyses, it correlated most strongly with GAD symptoms (weighted mean r = 0.66), but also displayed substantial associations with Social Anxiety (r = 0.49), overall PTSD (r = 0.48), Panic (r = 0.47), Dysphoria (r = 0.45), Agoraphobia (r = 0.45), Obsessions (r = 0.43), Intrusions (r = 0.41), and Hyperarousal (r = 0.40). Its weakest associations were with various indicators of specific phobia (mean r = 0.27 with overall Specific Phobia, 0.25 with Blood-Injection Phobia, and 0.23 with Animal Phobia) and with certain types of OCD symptoms (r = 0.28 with Ordering, 0.22 with Hoarding, 0.22 with Cleaning).
Multivariate relations.
At the multivariate level, Anxiety made a significant incremental contribution in 37 analyses (82.2%). It is noteworthy that it had robust associations with several types of symptoms that replicated across multiple samples; these included GAD, overall PTSD, Hyperarousal, Intrusions, Avoidance, Social Anxiety, Panic, overall OCD, Checking, and Ordering. In contrast, it failed to predict Dysphoria and Hoarding in any analysis. Overall, however, Anxiety emerged as a consistent predictor across a very broad range of symptoms.
Depression
Bivariate relations.
The Depression facet was a stronger predictor of anxiety symptoms than anxiety diagnoses. It showed several substantial associations at the bivariate level. Across the three sets of analyses, its strongest links were with GAD symptoms (weighted mean r = 0.65), Dysphoria (r = 0.63), overall PTSD (r = 0.55), Social Anxiety (r = 0.53), Obsessions (r = 0.49), Panic (r = 0.47), Intrusions (r = 0.45), and Avoidance (r = 0.43). Similar to Anxiety, Depression had its weakest associations with measures of specific phobia (mean r = 0.19 with overall Specific Phobia, 0.19 with Blood-Injection Phobia, and 0.16 with Animal Phobia) and with most types of OCD symptoms (r = 0.28 with Checking, 0.22 with Ordering, 0.25 with Hoarding, 0.13 with Cleaning).
Multivariate relations.
Depression contributed significantly in 28 regression analyses (62.2%). It had consistent associations with GAD, overall PTSD, Dysphoria, Hyperarousal, Intrusions, Avoidance, Social Anxiety, Panic, and Obsessions that replicated across analyses. In contrast, it failed to add significant incremental variance to the OCD symptoms of Checking, Ordering, and Cleaning in any analysis.
Anger
Bivariate relations.
Consistent with the diagnostic data, the Anger facet clearly showed the weakest overall links to anxiety symptoms. In fact, across the three sets of analyses, it produced only two weighted mean correlations ≥ 0.40: 0.46 with GAD and 0.40 with overall PTSD. Similar to the other facets, its weakest associations were with specific phobia (mean r = 0.23 with overall Specific Phobia, 0.18 with Blood-Injection Phobia, and 0.23 with Animal Phobia) and with most types of OCD symptoms (r = 0.29 with Checking, 0.28 with Ordering, 0.26 with Hoarding, 0.18 with Cleaning).
Multivariate relations.
Anger produced 20 significant regression effects (44.4%). Most of its significant links were with OCD-related symptoms. In particular, it displayed replicable associations with overall OCD, Ordering, and Hoarding. In contrast, it failed to predict GAD, Avoidance, and Social Anxiety in any analysis.
General Summary of Results
Diagnostic Analyses
Among the “big” traits of personality, neuroticism has the strongest and broadest associations with internalizing psychopathology. The meta-analytic results of Kotov et al. (Reference Kotov, Gamez, Schmidt and Watson2010) establish neuroticism as a strong transdiagnostic factor in the DSM-IV anxiety disorders, with ds ranging from 0.92 (specific phobia) to 2.25 (PTSD), and an overall mean d of 1.77 (see Table 4.1). Subsequent facet-level analyses strongly suggest that individual differences in fear and nervousness (i.e., the Anxiety facet) primarily are responsible for these associations.
Symptom Analyses
Although neuroticism is significantly related to every type of anxiety symptom, it shows much greater specificity in these analyses. It correlates most strongly with the apprehensive mood and worry that is the core feature of GAD, but also is substantially associated with symptoms of PTSD (particularly those related to Dysphoria), panic, and social anxiety. It is more moderately related to agoraphobia, OCD (correlating more strongly with obsessive intrusions than with other symptoms), and claustrophobia, and only modestly associated with various types of specific phobia. Lower-order analyses revealed that the Anxiety and Depression facets had the strongest associations with anxiety symptoms, although Anger also contributed significantly to the prediction of OCD-related symptoms. The Anxiety facet was substantially related to most types of symptoms, with particularly strong links to GAD, social anxiety, PTSD, panic, and agoraphobia. The Depression facet had noteworthy associations with GAD symptoms, Dysphoria symptoms within PTSD, social anxiety, and obsessive intrusions.
Underlying Causal Mechanisms
Overview
Why is neuroticism so broadly linked to anxiety symptoms and diagnoses? A thorough review of this complex topic is beyond the scope of this chapter, as it is clear that multiple mechanisms are involved. In the following sections, I discuss the major models that have been proposed and review evidence that supports them.
Six classic models have been proposed to explain the associations between personality and psychopathology (Clark, Watson, & Mineka, Reference Clark, Watson and Mineka1994; Klein, Kotov, & Bufferd, Reference Klein, Kotov and Bufferd2011; Ormel et al., Reference Ormel, Jeronimus, Kotov, Riese, Bos, Hankin and Oldehinkel2013). They can be further organized into three broad types. The first two models emphasize the causal role of traits in the development and/or course of psychopathology. Thus, the vulnerability or predisposition model posits that traits such as neuroticism play a significant role in the subsequent onset of psychopathology, whereas the pathoplasty model argues that personality influences the course, chronicity, and/or expression of disorders. The next two models reverse these causal roles by positing that prominent episodes of psychopathology affect scores on neuroticism measures. Specifically, the scar model asserts that significant psychopathology produces permanent increases in neuroticism (i.e., trait scores continue to remain elevated even after symptom levels fade), whereas a complication, concomitants, or state model proposes that this increase is only temporary (i.e., neuroticism scores return to premorbid baseline levels once the episode has ended). The final two models argue that neuroticism and internalizing psychopathology reflect common underlying processes. The common cause model proposes that traits and psychopathology reflect shared etiological factors (such as genetic diatheses), whereas the spectrum or continuum model argues that they represent different locations on the same underlying dimension.
Neuroticism as a Cause
We now have ample evidence establishing that neuroticism is a prospective predictor of psychopathology; moreover, it is significantly related to the subsequent development of internalizing problems, even after controlling for baseline symptoms and psychiatric history (for a meta-analytic review, see Jeronimus, Kotov, Riese, & Ormel, Reference Jeronimus, Kotov, Riese and Ormel2016). At the symptom level, for example, Levenson, Aldwin, & Bossé (Reference Levenson, Aldwin, Bossé and Spiro1988) reported that neuroticism was substantially related to levels of internalizing psychopathology (including anxiety) assessed 10 years later. Watson and Walker (Reference Watson and Walker1996) found that negative emotionality scores predicted anxious mood and panic symptoms assessed six to seven years later. Kopala-Sibley et al. (Reference Kopala-Sibley, Danzig, Kotov, Bromet, Carlson, Olino and Klein2016) reported that individual differences in fearfulness at age three predicted anxiety symptoms at age 11, even after controlling for symptom levels at age nine.
Neuroticism also has been shown to predict the future onset of diagnoses. Krueger (Reference Krueger1999) found that elevated levels of negative emotionality assessed at age 18 were associated with an increased likelihood of anxiety disorder diagnoses at age 21. Zinbarg et al. (Reference Zinbarg, Mineka, Bobova, Craske, Vrshek-Schallhorn, Griffith and Anand2016) reported that neuroticism (assessed when participants were high school juniors) predicted first onsets of anxiety disorders over a subsequent three-year period. Conway, Craske, Zinbarg, and Mineka (Reference Conway, Craske, Zinbarg and Mineka2016) found that individual differences in negative emotionality predicted both first onsets and recurrences of internalizing disorders across a four-year follow-up period. Goldstein, Kotov, Perlman, Watson, and Klein (Reference Goldstein, Kotov, Perlman, Watson and Klein2017) examined both neuroticism domain and facet scores in adolescent females; these trait scores (assessed when the participants were 13 to 15 years old) were used to predict first onset of disorders over an 18-month follow-up period. Neuroticism domain scores predicted first onsets of GAD, social phobia, and specific phobia in these data. At the lower-order level, the Anxiety facet predicted first onsets of all three anxiety disorders. In contrast, the Depression facet predicted the first onsets of GAD and specific phobia, but not social phobia, whereas the Anger facet was significantly linked only to the initial onset of specific phobia.
Neuroticism as an Effect
The scar model has received mixed support to date, with studies finding evidence both consistent (e.g., Kendler, Neale, Kessler, Heath, & Eaves, Reference Kendler, Neale, Kessler, Heath and Eaves1993; Rohde, Lewinsohn, & Seeley, Reference Rohde, Lewinsohn and Seeley1994) and inconsistent (e.g., Ormel, Oldehinkel, & Vollebergh, Reference Ormel, Oldehinkel and Vollebergh2004; Shea et al., Reference Shea, Leon, Mueller, Solomon, Warshaw and Keller1996) with it. Moreover, the supportive evidence essentially is limited to major depression, such that there currently is very little evidence indicating that anxiety-related psychopathology can produce permanent increases in neuroticism (see Klein et al., Reference Klein, Kotov and Bufferd2011; Ormel et al., Reference Ormel, Jeronimus, Kotov, Riese, Bos, Hankin and Oldehinkel2013).
In contrast, the state model has received much stronger support (see Klein et al., Reference Klein, Kotov and Bufferd2011; Ormel et al., Reference Ormel, Jeronimus, Kotov, Riese, Bos, Hankin and Oldehinkel2013), as multiple studies have reported that neuroticism scores show transient elevations during episodes of internalizing psychopathology (e.g., De Fruyt, van Leeuwen, Bagby, Rolland, & Rouillon, Reference De Fruyt, van Leeuwen, Bagby, Rolland and Rouillon2006; Morey et al., Reference Morey, Shea, Markowitz, Stout, Hopwood, Gunderson and Skodol2010; Ormel, Oldehinkel et al., Reference Ormel, Oldehinkel and Vollebergh2004). Again, the supportive evidence is confined largely to episodes of major depression, although Spinhoven, Penelo, de Rooij, Penninx, and Ormel (Reference Spinhoven, Penelo, de Rooij, Penninx and Ormel2014) found that neuroticism scores fluctuated concomitantly with both distress disorders (which included GAD) and fear disorders (which included panic disorder, agoraphobia, and social phobia). Furthermore, the state model is consistent with extensive data demonstrating that general trait measures – including neuroticism – are influenced by transient error, such that they reflect the respondents’ current affective state to a significant extent (Chmielewski & Watson, Reference Watson2009; Gnambs, Reference Gnambs2014). Finally, in a recent meta-analysis, Roberts et al. (Reference Roberts, Luo, Briley, Chow, Su and Hill2017) reported that clinical interventions were associated with marked reductions in neuroticism trait scores. Thus, it now is clearly established that trait neuroticism scores are sensitive to respondents’ current clinical and affective state.
Third Variable Models
Several studies have obtained support for the common cause model (see Middeldorp, Cath, van Dyck, & Boomsma, Reference Middeldorp, Cath, van Dyck and Boomsma2005). For instance, Hettema, Prescott, and Kendler (Reference Hettema, Prescott and Kendler2004) found a very strong genetic correlation (r = 0.80) between neuroticism and GAD. Hettema, Neale, Myers, Prescott, and Kendler (Reference Hettema, Neale, Myers, Prescott and Kendler2006) reported genetic correlations ranging from 0.58 (animal phobia) to 0.82 (social phobia) with six different anxiety disorders. In an examination of OCD symptom dimensions, Bergin et al. (Reference Bergin, Verhulst, Aggen, Neale, Kendler, Bienvenu and Hettema2014) found that neuroticism had strong genetic correlations with both aggressive obsessions (r = 0.89) and checking (r = 0.66), and a more moderate association with contamination concerns (r = 0.40). Consequently, it is clear that shared genetic factors are at least partly responsible for observed associations between neuroticism and anxiety-related psychopathology.
With regard to the spectrum model, several authors have highlighted the significant content overlap between neuroticism items and anxiety symptoms (e.g., Ormel, Rosmalen, & Farmer, Reference Ormel, Rosmalen and Farmer2004; Uliaszek et al., Reference Uliaszek, Hauner, Zinbarg, Craske, Mineka, Griffith and Rose2009). GAD provides the most obvious case of content overlap. The DSM-5 diagnostic criteria for GAD include “excessive anxiety and worry (apprehensive expectation), occurring more days than not for at least 6 months” (American Psychiatric Association, 2013, p. 222). Similarly, the FI-FFM Anxiety facet scale contains items such as “I constantly worry about things that might have gone wrong” and “I find myself worrying a lot.” Based on this content overlap, it seems reasonable to argue that GAD and the anxiety facet of neuroticism essentially represent alternative indicators of the same underlying dimension.
Relations with Other Transdiagnostic Traits
Many other transdiagnostic traits have been studied in relation to internalizing psychopathology. It is important to note that neuroticism is strongly linked to many of these so-called clinical traits (see Mahaffey, Watson, Clark, & Kotov, Reference Mahaffey, Watson, Clark and Kotov2016; Naragon-Gainey & Watson, Reference Naragon-Gainey and Watson2018). For example, Stanton, Rozek, Stasik-O’Brien, Ellickson-Larew, and Watson (Reference Stanton, Rozek, Stasik-O’Brien, Ellickson-Larew and Watson2016) reported moderate to strong associations between neuroticism and factors identified within the Difficulties in Emotion Regulation Scale (Gratz & Roemer, Reference Gratz and Roemer2004). Zinbarg et al. (Reference Zinbarg, Mineka, Bobova, Craske, Vrshek-Schallhorn, Griffith and Anand2016) obtained support for a model in which anxiety sensitivity, negative inferential style, sociotropy, and autonomy were specified as distinct facets within the broader neuroticism domain. Mahaffey et al. (Reference Mahaffey, Watson, Clark and Kotov2016) examined six clinical traits – ruminative response style, self-criticism, perfectionism, anxiety sensitivity, fear of negative evaluation, and thought suppression – in student and patient samples. All six clinical traits were strong and clear markers of a neuroticism factor in both samples, with loadings ranging from 0.59 to 0.75 (mean loading = 0.68) in the patients, and from 0.56 to 0.77 (mean loading = 0.69) in the students. Naragon-Gainey and Watson (Reference Naragon-Gainey and Watson2018) found that four clinical traits – anxiety sensitivity, intolerance of uncertainty, maladaptive perfectionism, and experiential avoidance were strongly related to neuroticism. Moreover, structural analyses revealed that all four traits were moderate to strong markers of its Anxiety facet, with loadings ranging from 0.32 to 0.76 (mean loading = 0.55).
This raises the crucial question of whether these clinical traits provide significant incremental information beyond neuroticism – and each other. The answer is a qualified “Yes,” although the results are inconsistent and the incremental gains vary widely in magnitude. For example, Mahaffey et al. (Reference Mahaffey, Watson, Clark and Kotov2016) reported that the six assessed clinical traits jointly provided modest but significant incremental validity beyond a comprehensive set of higher-order and lower-order personality scales, accounting for an additional 2% to 8% of the variance in patients (mean = 6%), and an additional 3% to 7% in students (mean = 4%). Naragon-Gainey and Watson (Reference Naragon-Gainey and Watson2018) examined the incremental validity of four clinical traits vis-à-vis six types of internalizing symptoms: depression, PTSD, GAD, social anxiety, panic, and OCD. The clinical traits jointly contributed an additional 9.1% to 20.6% of the variance (mean = 14.3%) beyond neuroticism in five of the six analyses, but showed no incremental validity in the prediction of GAD. Anxiety sensitivity demonstrated the greatest incremental power, as it contributed significantly to the prediction of panic, depression, PTSD, and social anxiety. Perfectionism added significantly to the prediction of OCD, whereas intolerance of uncertainty contributed unique variance only in the prediction of depression. Finally, experiential avoidance showed only a single suppressor effect, displaying a weak inverse association with panic symptoms.
The key point is that one cannot assume that a given trait has incremental validity in predicting psychopathology; rather, this must be established empirically. These results again highlight the importance of examining multiple traits together in a single integrated framework.
Directions for Future Research
I conclude by discussing three important methodological issues that should direct future work in this area. First, as I noted at the beginning of this chapter, recent interest in transdiagnostic traits has been stimulated by the increasing recognition of the pervasive comorbidity within psychopathology: virtually all symptoms and disorders are positively related to each other, thereby giving rising to an overarching general factor (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Moffitt2014; Lahey et al., Reference Lahey, Applegate, Hakes, Zald, Hariri and Rathouz2012). As I have shown, however, many of these transdiagnostic traits are substantially related to each other. Accordingly, it is crucial to study multiple traits together in joint analyses that allow one to identify their unique incremental effects. For example, Table 4.3 indicates that the Anger facet of neuroticism had mean correlations ranging from 0.24 to 0.39 with various anxiety disorder diagnoses. Viewed in isolation, these results might be taken to suggest that Anger is a significant transdiagnostic factor in these disorders. However, once its shared variance with the Anxiety and Depression facets was removed, Anger failed to show any significant incremental validity across 12 logistic regression analyses (see Table 4.4). These results clearly demonstrate that transdiagnostic traits cannot be studied in isolation.
Second, although I replicated many key findings across methods, measures, and samples, the data I presented are based heavily on self-report, either directly or indirectly through the filter of clinical interviews. As such, these relations are inflated to some extent by mono-method bias (Johnson, Rosen & Djurdjevic, Reference Johnson, Rosen and Djurdjevic2011; Podsakoff, MacKenzie, Lee, & Podsakoff, Reference Podsakoff, MacKenzie, Lee and Podsakoff2003). Future work should examine the generality of these findings across a broader range of methods. It would be particularly valuable to include non-self-report indicators of personality, such as informant ratings. Given that neuroticism-related traits tend to be subjective and internal in nature and, therefore, are low in visibility (Connelly & Ones, Reference Connelly and Ones2010), it will be important to use well-acquainted informants in these studies.
Third, most of the existing evidence is cross-sectional in nature (Kotov et al., Reference Kotov, Gamez, Schmidt and Watson2010). Longitudinal data are needed to clarify the etiological basis of these relations. Longitudinal studies begun relatively early in life – more specifically, before the first onsets of diagnosable disorders – will be particularly helpful in explicating the nature of the relations between traits and anxiety-related psychopathology (see Goldstein et al., Reference Goldstein, Kotov, Perlman, Watson and Klein2017; Zinbarg et al., Reference Zinbarg, Mineka, Bobova, Craske, Vrshek-Schallhorn, Griffith and Anand2016).
