Anxiety sensitivity: Relations to psychopathy, DSM-IV personality disorder features, and personality traits

Abstract

Relatively few data are available concerning the relations between anxiety sensitivity (AS) and both abnormal and normal personality traits. In particular, little is known about the associations between AS and personality disorders, although Shostak and Peterson [Behav. Res. Ther. 28 (1990) 513.] hypothesized that AS would be negatively correlated with antisocial personality disorder (ASPD) and perhaps related conditions (e.g., psychopathy). We examined the relations between AS, as assessed by the AS Index (ASI), and measures of psychopathy/ASPD, personality disorder features, and personality traits in a sample of 104 undergraduates. The ASI was not significantly associated with global measures of psychopathy or ASPD, although it was negatively correlated in some cases with the core affective deficits of psychopathy. In addition, the ASI was positively correlated with features of several Clusters B (e.g., borderline) and C (e.g., dependent) personality disorders and with features of passive–aggressive personality disorder. In addition, the ASI was positively associated with measures of several normal-range personality traits, including trait anxiety, alienation, well being, Negative Emotionality, and Constraint. Some, although not all, of the abnormal and normal personality correlates of the ASI were attributable to the variance shared by the ASI with trait anxiety measures. Implications and limitations of the present findings for the correlates and etiology of AS are outlined.

Introduction

The construct of anxiety sensitivity (AS) is hypothesized to reflect relatively stable individual differences in the fear of anxiety (Reiss, 1991). Individuals with elevated levels of AS are posited to harbor beliefs that anxiety symptoms (e.g., a rapid heart beat) have adverse consequences (e.g., a heart attack). Such individuals have been found to exhibit elevated rates of a number of anxiety disorders, particularly panic disorder McNally, 1990, McNally, 1996, and to be susceptible to marked anxiety reactions following biological challenge procedures (e.g., hyperventilation and carbon dioxide inhalation; see McNally, 1996, for a review). In addition, evidence from several longitudinal investigations (e.g., Maller & Reiss, 1992, Schmidt et al., 1997) suggests that AS is a predictor of subsequent panic attacks in nonclinical samples. In the adult literature, AS has generally been assessed by the AS Index (ASI), a self-report instrument measuring the extent to which individuals report being frightened by their own anxiety sensations (Reiss, Peterson, Gursky, & McNally, 1986).

Although the relation of AS to anxiety disorders has received increasing attention in recent years (see Taylor, 1999 for a review), relatively little is known concerning the associations between AS and either personality disorders or normal-range personality traits. As a consequence, little information is available concerning the links between AS and the broader personality domain. Data regarding the personality correlates of AS are important for at least three reasons (see also Lilienfeld, 1999).

First, such information may lead to a better understanding of how AS maps onto the factor space defined by higher-order and lower-order personality dimensions. Because at least some higher-order personality dimensions can be conceptualized as “source traits” (Cattell, 1950), i.e., broad underlying traits that give rise to narrower and more specific “surface traits” or lower-order dimensions, a better understanding of the relation of AS to higher-order dimensions may provide important clues regarding the etiology of AS.

Second, higher-order and lower-order personality dimensions can sometimes provide competing explanations for hypotheses in the personality domain (Watson & Clark, 1992). If an investigator proposes a hypothesis concerning the relation between a lower-order dimension and external criteria but neglects to include a measure of the higher-order dimension on which this lower-order dimension loads, the investigator may mistakenly conclude that this hypothesis has been corroborated. But in fact, the observed relation may be attributable to the influence of the unmeasured higher-order dimension (e.g., see Watson & Pennebaker, 1989 for an illustration of how the association between stressful life events and self-reported health complaints appears to be mediated by the higher-order dimension of Negative Emotionality). Consequently, an examination of higher-order dimensions can provide valuable information concerning whether the personality correlates of AS are specific to AS per se.

Third, data on the relation between AS and personality traits can provide helpful information concerning potential personality risk factors for AS. Although such data are correlational and therefore do not permit definitive causal inferences regarding the association between personality traits and AS, they can be useful for generating hypotheses to be tested in longitudinal studies, which may be better suited for drawing cause-and-effect inferences.

Several investigators have recently begun to examine the relations between AS and normal-range personality traits (see Lilienfeld, 1999 for a review). All of the studies conducted thus far have relied exclusively on self-report indices of personality. At the lower-order level, the most consistent finding is that AS is moderately associated with trait anxiety, i.e., a propensity to respond anxiously to stressors Borger et al., 1996, Lilienfeld et al., 1993. This moderate correlation led Lilienfeld, Jacob, and Turner (1989) to suggest that a number of findings previously attributed to AS per se, such as the association between AS and panic disorder (Taylor, Koch, & McNally, 1992), might be due to trait anxiety. Nevertheless, there is compelling evidence that the ASI possesses incremental validity (Meehl, 1959) above and beyond measures of trait anxiety in the prediction of a number of clinically relevant phenomena, including panic disorder, panic attacks, and anxiety responses to hyperventilation (e.g., Brown & Cash, 1990, McNally, 1989, Rapee & Medoro, 1994, Schmidt et al., 1997). Thus, the ASI possesses reliable variance that is not shared with trait anxiety, although the nature and correlates of this unique variance are unclear. Some researchers (e.g., Lilienfeld et al., 1993) have conjectured that AS is a lower-order trait nested hierarchically within a higher-order trait anxiety dimension, although this possibility has received relatively little systematic examination. If so, one would expect AS and trait anxiety measures to be moderately correlated but AS measures to possess both unique variance and unique psychological correlates above and beyond trait anxiety measures. Because of the moderate covariation between AS and trait anxiety, it is important for researchers to examine the extent to which the correlates of AS are attributable to AS per se as opposed to trait anxiety.

A recent investigation of 220 undergraduates Lilienfeld, 1997, Lilienfeld, 1999 provides further information regarding the lower-order and higher-order personality correlates of AS. Lilienfeld found that several measures of AS, including the ASI, were significantly, although modestly, correlated with the lower-order dimension of absorption. In addition, absorption was significantly and positively correlated with panic attack history. Absorption is a tendency to become immersed in sensory or imaginative experiences and has been found to be associated with hypnotic susceptibility (e.g., Council et al., 1986, Tellegen & Atkinson, 1974). Lilienfeld (1997) hypothesized that elevated absorption levels could make individuals more attuned to unpleasant and potentially frightening internal sensations and thereby heighten their risk for anxiety disorders (e.g., panic disorder) associated with a hypersensitivity to interoceptive cues.

Lilienfeld, 1997, Lilienfeld, 1999 also found that AS measures were positively correlated with the lower-order trait of alienation, which is a propensity to mistrust others and to perceive malevolent intent in others' actions (Tellegen, 1978/1982). Partial correlation analyses controlling for trait anxiety, however, suggested that these correlations were largely attributable to the variance shared by trait anxiety and alienation.

At the higher-order level, AS has been found to be associated with the dimension of Negative Emotionality Arrindell, 1993, Lilienfeld, 1997. Negative Emotionality is related to, although broader than, the Eysenck's (1975) construct of neuroticism and is a propensity to experience negative affects of many kinds (e.g., anxiety, guilt, anger, and mistrust; Tellegen & Waller, 1994). Both trait anxiety and alienation are lower-order markers of Negative Emotionality (Tellegen, 1978/1982).

In analyses of the undergraduate data set mentioned earlier, Lilienfeld (1999) found that the ASI, but not other AS measures, was significantly correlated with the higher-order dimension of Constraint, although this correlation was weak in magnitude (r=.18). Constraint is a fearfulness or response inhibition dimension that is largely orthogonal to Negative Emotionality (Tellegen & Waller, 1994). Lilienfeld et al. (1993) conjectured that because Constraint is hypothetically associated with a sensitivity to threat cues (Tellegen, 1978/1982), individuals with high levels of this dimension might be particularly susceptible to developing fears of their own anxiety symptoms. Nevertheless, the low correlations between AS indices and Constraint reported by Lilienfeld call this hypothesis into question.

The relation between AS measures and extraversion also requires clarification. In a study of 94 psychiatric outpatients, Arrindell (1993) reported no significant associations between AS indices and Extraversion, as measured by the Eysenck Personality Questionnaire (Eysenck & Eysenck, 1975). In contrast, in an investigation of 320 undergraduates, Borger et al. (1996) reported that the ASI was significantly and negatively associated (r=−.26) with Extraversion, as assessed by the Neuroticism-Extraversion-Openness Personality Inventory-Revised (Costa & McCrae, 1992), a measure of the “Big Five” taxonomy of personality traits (see Goldberg, 1993). The reasons for the discrepancy between Arrindell's findings and those of Borger et al. are unclear.

There appears to be no published data concerning the relation between AS and personality disorders. Nevertheless, the results of several investigations provide intriguing suggestions regarding the association between AS and certain personality disorders. On the basis of findings indicating that low AS individuals responded to a challenge procedure (a difficult mental arithmetic task) with normal levels of physiological arousal but with little or no apparent perception of this arousal, Shostak and Peterson (1990) hypothesized that low AS individuals were prone to antisocial personality disorder (ASPD), a condition characterized by a chronic history of illegal, irresponsible, and often criminal behavior (e.g., stealing, vandalism, and physical aggression). According to Shostak and Peterson, “Under some conditions, low (anxiety) sensitivity may relate to ASPDs. If physiological arousal is not viewed as negative and/or is not processed and poor moral development or antisocial value development is present, there will be a failure to inhibit antisocial behavior” (p. 520). Peterson (1991, personal communication) similarly conjectured that psychopathic personality (psychopathy) is associated with abnormally low levels of AS. In addition, Cox, Borger, and Enns (1999), although not explicitly discussing the association between AS and psychopathy, noted that low AS individuals were often observed during interviews to be bored and indifferent. Although such reactions may stem from a variety of sources, it is worth noting that boredom proneness is commonly, although not exclusively, associated with psychopathy Hare, 1991, Lilienfeld & Andrews, 1996.

Although ASPD and psychopathy are overlapping syndromes, the former is operationalized primarily in terms of antisocial and criminal behaviors, whereas the latter is operationalized primarily in terms of personality traits, such as lack of guilt, callousness, failure to form close emotional attachments to others, physical risk taking, boredom proneness, propensity to externalize blame, and superficial charm Cleckley, 1941/1988, Hare et al., 1991, Lilienfeld, 1994. Indeed, factor analyses of global psychopathy measures have typically yielded two moderately correlated dimensions, the first of which (Factor 1) is associated primarily with the core affective traits of psychopathy (e.g., guiltlessness, lack of empathy, and narcissism) and the second of which (Factor 2) is associated primarily with antisocial and illegal behaviors (Harpur, Hare, & Hakstian, 1989).

With respect to features of other personality disorders, Stewart, Knize, and Pihl (1992) found that the ASI was significantly correlated with a measure of interpersonal dependency in a mixed sample of undergraduates and panic disorder patients (see also Borger et al., 1996 for data suggesting an association between the ASI and interpersonal dysfunction). These data suggest that the ASI might correlate positively with features of dependent personality disorder, a condition characterized by intense dependence on others and a willingness to subordinate one's needs to others' demands (American Psychiatric Association, 1994).¹ More broadly, the moderate association between AS measures and both Axis I anxiety disorders and trait anxiety indices Lilienfeld et al., 1989, McNally, 1996 might lead one to expect the former measures to be associated with features of other Cluster C (anxious and fearful) personality disorders in the DSM-IV (APA, 1994), which in addition to dependent personality disorder are avoidant and obsessive-compulsive personality disorders.

The present investigation, which like several other recent investigations of the personality correlates of AS Borger et al., 1996, Lilienfeld, 1997, Lilienfeld, 1999, was conducted with undergraduates, was undertaken to examine in greater detail the relations between the construct of AS, as assessed by the ASI, and indices of personality disorders and personality traits. Although undergraduate samples are characterized by several disadvantages, such as a potentially restricted range on measures of psychopathy and perhaps other personality disorders, these samples have the advantage of being relatively free of severe Axis I disorders (e.g., major depressive disorder, alcohol dependence), which have been found in some studies to distort the reporting of enduring personality traits and personality disorder features (e.g., see Loranger et al., 1991 for data on state-trait artifacts and their detrimental impact on personality disorder assessment). In addition, as there is increasing evidence that psychopathic personality traits can be meaningfully assessed in nonclinical (e.g., undergraduate) samples (Lilienfeld, 1998), the present investigation afforded an opportunity to examine the relations between AS and psychopathic features in a sample free of the problematic effects (e.g., fatigue, resentment, alienation) of incarceration and institutionalization (see also Widom, 1977).

More specifically, the primary goals our study were fivefold. First, we examined Shostak and Peterson's (1990) hypothesis that AS would be negatively associated with psychopathy, ASPD, or both. If low levels of AS were found to be characteristic of one or both of these syndromes, this would suggest that the relation between AS and maladjustment is curvilinear (Cox et al., 1999), with both high and low AS levels being associated with different forms of psychopathology (anxiety disorders and psychopathy/ASPD, respectively). In view of evidence that psychopathy is multidimensional (Lilienfeld & Andrews, 1996), for exploratory purposes, we also examined the relation between AS and specific components of psychopathy, such as physical fearlessness, callousness, and a tendency to externalize blame (see Measures). In addition, for exploratory purposes, we examined the relation between the ASI and the two primary underlying dimensions of psychopathy (Harpur et al., 1989) in order to clarify whether AS was differentially associated with either the core affective deficits of psychopathy (Factor 1) or its concomitant antisocial and criminal behaviors (Factor 2).

Second, we examined the relations between the ASI and features of DSM-IV personality disorders other than ASPD. Because we used undergraduate participants, we focused on dimensional (i.e., continuous) rather than categorical (i.e., diagnostic) measures of these disorders. Based on previous findings, we predicted that the ASI would be positively associated with features of Cluster C personality disorders, including dependent personality disorder, but negligibly associated with features of Clusters A (odd and eccentric) and B (dramatic, emotional, and erratic) personality disorders. Exploratory analyses were conducted to examine the relations between the ASI and the two personality disorders listed in the Appendix of DSM-IV, viz., passive–aggressive (negativistic) and depressive personality disorders (APA, 1994).

Third, we examined the relations between the AS and both lower-order and higher-order dimensions of personality in an effort to replicate the findings of Lilienfeld, 1997, Lilienfeld, 1999. Specifically, based on previous findings, we predicted positive associations between the ASI and trait anxiety, absorption, alienation, Negative Emotionality, and Constraint. In addition, for exploratory purposes, we examined the relation between the ASI and other lower-order and higher-order personality traits.

Fourth, as noted earlier, all of the extant data on the personality correlates of AS measures (e.g., Borger et al., 1996, Lilienfeld, 1997) derive from studies using self-report indices. As a consequence, it is difficult to exclude the possibility that findings on the personality correlates of AS are attributable at least partly to method covariance (Campbell & Fiske, 1959). The present investigation represents an attempt to remedy this shortcoming by supplementing data on self-reported personality traits with corroborative personality ratings from peers. Peer data may be especially crucial in the assessment of psychopathic individuals, who often fail to appreciate the impact of their behavior on others Grove & Tellegen, 1991, Lilienfeld, 1994.

Fifth and finally, we examined the relations between the AS and measures of personality after controlling statistically for measures of trait anxiety. These analyses should shed light on the extent to which both the abnormal and normal personality correlates of AS are attributable to the variance shared by AS and trait anxiety as opposed to the unique variance possessed by AS per se. In addition, these analyses could provide data consistent with a hierarchical model of the relation between AS and trait anxiety (Lilienfeld et al., 1993) by demonstrating that a measure of AS shares unique variance with psychologically meaningful correlates (e.g., measures of personality and personality disorders) above and beyond trait anxiety indices.