Abstract
The Personality Inventory for DSM-5 (PID-5) assesses traits relevant for diagnosing personality disorder in the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5). We examined the PID-5 in relation to the Big-Three and Big-Five personality traits in outpatient and community adult samples. Domain-level analyses revealed that PID-5 Negative Affectivity correlated strongly with Neuroticism, and PID-5 Antagonism and Disinhibition correlated strongly negatively with Agreeableness and Conscientiousness, respectively; Antagonism and Disinhibition also were both linked strongly to Big-Three trait Disinhibition. PID-5 Detachment related strongly to personality, including Extraversion/Positive Temperament, but did not show its expected specificity to this factor. Finally, PID-5 Psychoticism correlated only modestly with Openness. Facet-level analyses indicated that some PID-5 scales demonstrated replicable deviations from their DSM-5 model placements. We discuss implications of these data for the DSM-5 model of personality disorder, and for integrating it with well-established structures of normal personality.
Personality pathology in the fourth edition (text revision) of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR; American Psychiatric Association, 2000) is based on a model of categorical personality disorder (PD) diagnoses (e.g., schizotypal PD, borderline PD, dependent PD). The problems with this system—including low reliability, poor coverage, artificial diagnostic thresholds, substantial diagnostic comorbidity, and within-category heterogeneity—are well known and have been reviewed extensively (e.g., Clark, 2005, 2007; Krueger & Eaton, 2010; Livesley, 2003; Widiger & Samuel, 2005). Widespread dissatisfaction with this categorical scheme created substantial pressure for a significant change in the DSM-5.
The most frequently mentioned alternative is a hierarchical trait-based dimensional scheme. The American Psychiatric Association—with the support of the World Health Organization, and with funding from the National Institute of Mental Health, the National Institute on Alcohol Abuse and Alcoholism, and the National Institute on Drug Abuse—held a DSM-5 planning conference titled “Dimensional Models of Personality Disorder: Etiology, Pathology, Phenomenology, and Treatment” in December 2004 (Widiger & Simonsen, 2005). The consensus recommendation of the attendees was that the current categorical system be replaced by a hierarchical dimensional scheme in DSM-5. More specifically, the consensus converged on a multilevel Big-Four model; the higher order factors at the apex of this hierarchy essentially are maladaptive variants of the personality traits of the five-factor model (FFM), minus openness (Krueger, Derringer, Markon, Watson, & Skodol, 2012; Widiger & Simonsen, 2005; Watson, Clark, & Chmielewski, 2008). Widiger and Simonsen (2005) subsequently reviewed evidence indicating that a model based on four bipolar dimensions—negative affect versus emotional stability, introversion versus extraversion, antagonism versus compliance, and constraint versus impulsivity—could provide the basis for a trait dimensional scheme in DSM-5.
The DSM-5 Trait Dimensional Scheme
Development of the DSM-5 Trait Dimensional Model
In 2007, the DSM-5 Personality and Personality Disorders Work Group began the process of developing this trait dimensional scheme. In doing so, they recognized the need to modify mainstream models of personality—such as the FFM—which were not designed specifically to capture the pathological range of these dimensions (Krueger et al., 2012). In particular, to maximize clinical utility, they emphasized the need to focus on the poles of these dimensions that are associated with clinically significant personality pathology. In this regard, in a meta-analytic review of the literature, Samuel and Widiger (2008) found that DSM-IV PDs were associated with negative affectivity (i.e., FFM neuroticism), introversion (i.e., low FFM extraversion), antagonism (i.e., low FFM agreeableness), and impulsivity (i.e., low FFM conscientiousness). Consequently, with a few exceptions (see Krueger et al., 2012), the Work Group developed dimensions that were largely unipolar in nature, essentially tapping the presence versus absence of personality pathology. For instance, the Antagonism 1 domain focuses on clearly maladaptive characteristics such as callousness, deceitfulness, manipulativeness, and hostility, and makes no systematic attempt to assess adaptive qualities such as warmth, generosity, and forgiveness, which might form the opposite end of a bipolar dimension. Similarly, the Detachment domain emphasizes maladaptive traits such as social withdrawal and restricted affectivity, but does not also assess desirable qualities such as sociability and assertiveness, which might lie on the other end of the dimension if it were represented as bipolar.
In addition, the Work Group recognized that the original four-domain model failed to capture characteristics related primarily to the “odd or eccentric” PDs (schizoid PD, schizotypal PD, and paranoid PD; see Krueger et al., 2011; Krueger et al., 2012; Tackett, Silberschmidt, Krueger, & Sponheim, 2008; Watson et al., 2008). They therefore added a fifth higher order domain of Psychoticism. Consequently, the current DSM-5 trait scheme consists of five higher order domains labeled Negative Affectivity, Detachment, Antagonism, Disinhibition, and Psychoticism (American Psychiatric Association, 2013). Although this revised and expanded model clearly overlaps substantially with the FFM, it actually bears a stronger resemblance to the Personality Psychopathology–Five model (PSY-5) of Harkness and McNulty (1994; for discussions, see Harkness, Finn, McNulty, & Shields, 2012; Krueger et al., 2011; Krueger et al., 2012).
Current DSM-5 PD Model
This five-domain trait scheme currently forms an integral part of the DSM-5 Section 3 system for PD diagnosis. That system also incorporates a set of six specific PDs: antisocial, avoidant, borderline, narcissistic, obsessive–compulsive, and schizotypal. In contrast to DSM-IV (reproduced in DSM-5 Section 2), however, these PDs are diagnosed on the basis of specific trait characteristics in the system. For example, antisocial PD is characterized by pathological traits (six or seven are required) from the domains of Antagonism (manipulativeness, deceitfulness, callousness, and hostility) and Disinhibition (irresponsibility, impulsivity, and risk taking). Similarly, avoidant PD is diagnosed using traits from Negative Affectivity (anxiousness) and Detachment (two or three of withdrawal, intimacy avoidance, and anhedonia). Finally, patients with personality pathology who do not fit any of these specific types are classified using the Personality Disorder–Trait Specified (PD-TS) diagnosis, which involves identifying particular pathological characteristics from any of the five domains (for more information about the PD-TS diagnosis, see American Psychiatric Association, 2013).
Personality Inventory for DSM-5
The DSM-5 Work Group also decided to create a self-report assessment instrument to test and refine this trait-dimensional model. This eventually led to the development of the Personality Inventory for DSM-5 (PID-5; Krueger et al., 2012), which was created by several Work Group members and consultants. The PID-5 assesses the 25 specific facets included in the DSM-5 Section 3 system; these then can be combined to create the overall domain scores in this model. For example, scores on six facets—Anhedonia, Depressivity, Intimacy avoidance, Restricted affectivity, Suspiciousness, and Withdrawal—can be aggregated to create an overall Detachment domain score. Use of the PID-5 facilitates analyses that compare and contrast the DSM-5 trait scheme with other personality models.
The Current Study
One clearly would expect this DSM-5 model to be related strongly and systematically to general traits of personality, such as those of the FFM. More specifically, in terms of the Big Five, one would expect (a) Negative Affectivity to be strongly related to neuroticism, (b) Detachment to be inversely related to extraversion, (c) Antagonism to be negatively associated with agreeableness, and (d) Disinhibition to be inversely linked to conscientiousness. As noted earlier, however, in developing the DSM-5 trait model, the Work Group emphasized assessing the pathological range within these dimensions—as opposed to the largely adaptive range captured by most standard measures of the FFM—which led to the creation of largely unipolar constructs (e.g., the low end of extraversion). It seems reasonable to consider that creating pathological variants of these traits may have altered at least some of them in a significant way compared with their normal-range FFM counterparts. This raises some key questions, such as (a) how strong are these associations and (b) how closely do the DSM-5 domains parallel the normal personality dimensions? Consequently, the basic goal of this study was to examine the DSM-5 trait model—as operationalized by the PID-5—in relation to the higher order dimensions of the Big-Three and Big-Five models of personality (Markon, Krueger, & Watson, 2005; Watson, Clark, & Harkness, 1994; Watson et al., 2008).
We examined these associations in two groups of participants. First, outpatients were assessed on the PID-5 and the Schedule for Nonadaptive and Adaptive Personality–Second Edition (SNAP-2; Clark, Simms, Wu, & Casillas, in press), which contains scales assessing the core of the higher order Big-Three traits. Based on previous research (e.g., Markon et al., 2005; Watson et al., 1994), we expected (a) SNAP-2 Negative Temperament to be strongly related to PID-5 Negative Affectivity, (b) SNAP-2 Positive Temperament to be negatively associated with PID-5 Detachment, and (c) SNAP-2 Disinhibition—which, in FFM terms, is a blend of low conscientiousness and low agreeableness (particularly the former; see, Markon et al., 2005; Watson et al., 1994)—to be related to both PID-5 Disinhibition and PID-5 Antagonism. We made no formal prediction regarding PID-5 Psychoticism, which we did not expect to map specifically onto any particular Big-Three domain, although it should relate to SNAP-2 Eccentric Perceptions, a lower order scale associated with the higher order negative affectivity domain.
Second, a community adult sample completed the PID-5, the SNAP-2, and two different measures of the FFM: the Big Five Inventory (BFI; John & Srivastava, 1999) and the Faceted Inventory of the Five-Factor Model (FI-FFM; Simms, 2009; see also Naragon-Gainey, Watson, & Markon, 2009). This expanded set of general personality scales allowed us to model the Big Five as latent factors. For reasons discussed earlier, we predicted that (a) PID-5 Negative Affectivity would be strongly related to neuroticism, (b) PID-5 Detachment would be inversely related to extraversion, (c) PID-5 Antagonism would be negatively associated with agreeableness, and (d) PID-5 Disinhibition would be inversely linked to conscientiousness. In addition, we expected that PID-5 Psychoticism would be related to individual differences in openness, albeit only moderately, because previous research has indicated that some aspects of openness (e.g., fantasy, imagination) are much more highly related to psychoticism than are others (e.g., intellect, values; see DeYoung, Grazioplene, & Peterson, 2012; Watson et al., 2008), and our openness indicators did not focus on the types of content that show particularly strong associations with psychoticism. 2
In addition, the SNAP-2 (Eccentric Perceptions) and the FI-FFM (Unusual Experiences, Eccentric Beliefs) both contain content assessing individual differences in odd or peculiar characteristics (e.g., Simms, 2009). We expected these scales to form a sixth personality factor (see, Watson et al., 2008) that would be highly correlated with PID-5 Psychoticism.
Method
Participants and Procedure
All study procedures were approved by the Institutional Review Board of the University of Notre Dame. Patients gave written informed consent before participation, and community adults gave informed consent as part of the online protocol.
Outpatient Sample
Outpatients living in the greater South Bend metropolitan area were recruited from the Oaklawn Psychiatric Center, and from listservs, newsletters, and mass e-mails sent to University of Notre Dame staff, faculty, and graduate students. All participants were run in small group sessions at the research facility of the Center for the Advanced Measurement of Personality and Psychopathology. Patients were compensated for their participation in each session.
We report analyses based on the 202 patients with complete PID-5 and SNAP-2 data. Responses to these questionnaires were collected in two separate sessions conducted an average of approximately 1 month apart (M interval = 29.5 days); depending on how they were scheduled, participants could complete either session first. The sample consisted of 84 men and 118 women (58.4%), with a mean age of 43.5 years (range = 19-68 years). The sample was 63.9% White, 26.1% African American, and 10.0% other races/ ethnicities. Current diagnoses were obtained using the Mini-International Psychiatric Interview 6.0 (Sheehan et al., 1998, modified with permission to obtain DSM-5 diagnoses). The most common current diagnoses were depressive disorders (85 patients, or 42.1% of the sample), anxiety disorders 3 (99 patients; 49.0%), and substance use disorders (62 patients; 30.7%). Overall, 144 patients (71.3% of the sample) had a current mood, anxiety, and/or substance use disorder diagnosis.
Community Adult Sample
Adults living in the greater South Bend metropolitan area were recruited for two online data sessions that were completed an average of approximately 1 week apart (M interval = 6.2 days). Participants were compensated for each session.
In the first session, 388 individuals completed both the PID-5 and the FI-FFM. However, the responses of 13 participants were dropped due to validity concerns, based on scores on the SNAP-2 validity scales and on individual validity items that were scattered throughout the protocol to detect careless, inattentive responding. Of the 375 participants with valid Session 1 data, 335 (89.3%) completed the SNAP-2 and the BFI in the second session. This sample consisted of 107 men and 228 women (68.1%), with a mean age of 36.6 years (range = 20-83 years). The sample was 87.2% White, 6.3% African American, and 6.5% other.
Measures
PID-5
The PID-5 (Krueger et al., 2012; see also, Hopwood, Thomas, Markon, Wright, & Krueger, 2012; Wright et al., 2012) was used to assess the DSM-5 traits. The instrument consists of sentences that are rated on a 4-point scale ranging from 0 (very false or often false) to 3 (very true or often true). The patients completed the full, 220-item version of the PID-5; three suicidality items were dropped in the community adult sample, however, yielding a reduced 11-item version of the Depressivity scale. We used this reduced version of the scale in the analyses reported in Table 1; the complete, 14-item version of Depressivity was used in all other analyses of the patient data.
Descriptive Statistics for the PID-5 Facet Scales in the Patient and Community Samples.
Note. N = 202 (patients), 335 (community). PID-5 = Personality Inventory for DSM-5; d = Cohen’s d; NA = Negative Affectivity; DET = Detachment; ANT = Antagonism; DIS = Disinhibition; COMP = Compulsivity; PSY = Psychoticism; Unusual bel & exper = Unusual beliefs and experiences; Cog & percept dysreg = Cognitive and perceptual dysregulation.
Differences are significant (p < .05) for all but six scales (d < .19), italicized.
Results are based on an 11-item version of the scale; see text for details.
The PID-5 includes 25 primary facet scales, varying in length from 4 to 14 items; coefficient alphas for these scales (see Table 1) ranged from .73 to .95 (median = .87) in the patients and from .77 to .95 (median = .87) in the community sample. As noted earlier, scores on these facet scales also can be combined to assess the five higher order DSM-5 trait domains. Because we were interested in seeing how the DSM-5 model related to broad, general personality traits, we used the higher order structure that is articulated in the DSM-5 to score these domains. Of note, four facets are listed in more than one domain in the DSM-5 scheme; therefore, we scored these facets in both domains: (a) hostility is listed as a facet of both high Negative Affectivity and high Antagonism; (b) depressivity and (c) suspiciousness are both listed as facets of both high Negative Affectivity and high Detachment; and (d) restricted affectivity is listed as a facet of both high Detachment and low Negative Affectivity. In addition (lack of) rigid perfectionism is listed as a facet of Disinhibition, so we reverse-keyed this PID-5 facet and then combined it with the other PID-5 Disinhibition facets to create an overall Disinhibition domain score. We first standardized the facet scales so that each was weighted equally in the final composite and then averaged scores on the facets comprising each domain.
SNAP-2
Participants in both samples completed the SNAP-2. The original SNAP (Clark, 1993) is a widely used, 375-item true–false questionnaire that assesses the Big-Three personality dimensions using 15 trait and temperament scales. The revised SNAP-2 (Clark et al., in press) contains an identical set of trait and temperament scales. 4 We report results in both samples based on the Big-Three temperament scales of the SNAP-2: Negative Temperament (28 items), Positive Temperament (27 items), and Disinhibition (35 items). As noted earlier, we also use the 15-item SNAP-2 Eccentric Perceptions scale—which correlates strongly with markers of psychoticism (Watson et al., 2008)—to model a Peculiarity factor in the community participants. Coefficient alphas for these scales (see Table 2) ranged from .85 to .93 in the patient sample (median = .88), and from .81 to .93 in the community sample (median = .88).
Descriptive Statistics (Based on T Scores) for the SNAP-2 Scales in the Patient and Community Samples.
Note. N = 202 (patients), 335 (community). SNAP-2 = Schedule for Nonadaptive and Adaptive Personality–Second Edition.
All differences are significant at p < .01.
BFI
The community participants also completed the 44-item version of the BFI (John & Srivastava, 1999), which contains 8-item scales assessing Neuroticism and Extraversion, 9-item measures of Agreeableness and Conscientiousness, and a 10-item Openness scale. The instrument begins with the general stem “I see myself as someone who . . .,” followed by a series of short phrases; responses are made on a 5-point Likert-type scale (1 = disagree strongly, 3 = neither agree or disagree, 5 = agree strongly). Coefficient alphas in this sample ranged from .81 to .87 (median = .83).
FI-FFM
Finally, the community participants completed the FI-FFM (Simms, 2009), a factor analytically derived, 247-item self-report inventory that assesses specific lower order traits within the framework of the FFM. The items are sentences that are rated on a 5-point Likert-type scale ranging from strongly disagree to strongly agree.
The full FI-FFM consists of 26 specific trait scales. However, based on an extensive series of results—including factor analyses of both self-report and informant data, as well as convergent/discriminant validity analyses in relation to the BFI and the Revised NEO Personality Inventory (NEO PI-R; Costa & McCrae, 1992)—Simms (2009) concluded that only 22 scales were clear markers of the higher order FFM domains. Specifically, the FI-FFM contains five neuroticism facets (Anxiety, Depression, Anger Proneness, Somatic Complaints, and Envy), five markers of extraversion (Positive Temperament, Sociability, Ascendance, Venturesomeness, and Frankness), four indicators of Agreeableness (Empathy, Trust, Straightforwardness, and Modesty), five facets of conscientiousness (Self-Discipline, Dutifulness, Deliberation, Achievement Striving, and Order), and three indicators of Openness (Intellectance, Novel Experience Seeking, and Nontraditionalism). We use these 22 facet scales to model the FFM domains in the community sample. In addition, we used two other FI-FFM scales—Unusual Experiences (e.g., “I often feel that things are not real,” “Sometimes I can’t tell if I’m awake or dreaming”) and Eccentric Beliefs (e.g., “I believe the earth has been visited by beings from another planet,” “I believe in the existence of ghosts or spirits”) to model a Peculiarity factor in these participants.
Simms (2009) presented extensive reliability (including both internal consistency and 2-week dependability coefficients) and validity data on the FI-FFM scales. In addition, Naragon-Gainey et al. (2009) showed that the FI-FFM extraversion facets displayed a clear convergent/discriminant pattern in relation to other domain markers—including facet scales from the NEO PI-R—in both student and patient samples. In the current sample, coefficient alphas ranged from .80 to .93, with a median value of .86.
Results
Descriptive Statistics
PID-5 Facet Scales
Table 1 presents means and standard deviations for the PID-5 facet scales in the patient and community adult samples. As would be expected, the patients generally reported higher levels of personality pathology, with 19 of the 25 individual comparisons (76%) reaching statistical significance (p < .05, two-tailed). To quantify these group differences, we computed effect sizes using Cohen’s d; d values from |.20| to |.49| represent small effect sizes, those ranging from |.50| to |.79| reflect medium effect sizes, and those ≥|.80| indicate large effect sizes (Cohen, 1992). There were 12 medium effects (ds ranged from .55 to .71) and five small effects (ds ranged from .40 to .49); the other two significant differences fell just below the cutoff for a small effect size (d = .19). In general, markers of the Detachment and Negative Affectivity domains tended to show the largest group differences—two thirds of the medium effect sizes were from these domains (i.e., Suspiciousness, Depressivity, Emotional lability, Separation insecurity, Anhedonia, Anxiousness, Intimacy avoidance, and Withdrawal). Conversely, indicators of Antagonism generally displayed weaker differences—only Hostility (d = .44) and Callousness (d = .40) had small effect sizes, and the rest (i.e., Deceitfulness, Grandiosity, Attention seeking, and Manipulativeness) were <.20. Finally, markers of Psychoticism all had medium effect size differences (range = .47-.67), whereas those for Disinhibition covered a broad range from .09 to .66.
To explicate further the level of pathology reported by our patient sample, we compared their responses with the help-seeking sample presented in Krueger et al. (2012; Round 2 sample in their table 2). Our patient participants had higher mean scores on 23 of the 25 PID-5 facet scales (the two exceptions were Restricted affectivity and Grandiosity). Once again, we quantified these group differences via Cohen’s d, using the scale standard deviations reported in Krueger et al. (2012) as the denominator. These comparisons yielded one large effect (Depressivity) and 10 medium effects (Emotional lability, Distractibility, Separation insecurity, Anxiousness, Irresponsibility, Cognitive and perceptual dysregulation, Impulsivity, Anhedonia, Suspiciousness, and Perseveration). Taken together, these analyses indicate that our patient participants reported substantial levels of personality pathology, with particular elevations on markers of Negative Affectivity, Detachment, or both (e.g., Depressivity, Emotional lability, Suspiciousness).
SNAP-2 Scales
Table 2 presents parallel findings for the SNAP-2 scales; the descriptive statistics are based on T scores derived using SNAP-2 norms (Clark et al., in press). Compared with scale norms, the patients reported substantially elevated levels of Negative Temperament (mean T = 60.8) and Eccentric Perceptions (M = 58.4), as well as reduced levels of Positive Temperament (M = 44.0); in contrast, the scale scores for the community respondents were typical of the general population (mean Ts ranged from 46.8 to 52.8).
As shown in Table 2, the samples differed significantly on all four SNAP-2 scales. The effect size difference was large for Eccentric Perceptions (d = .93), medium for Negative Temperament (d = .69), and small for both Disinhibition (d = .36) and Positive Temperament (d = −.25).
Patient Sample Analyses
Domain-Level Analyses
Table 3 presents correlations between the SNAP-2 temperament scales and the PID-5 domain scores. These results are largely consistent with our predictions. As expected, SNAP-2 Negative Temperament correlated highly with PID-5 Negative Affectivity (r = .75), and SNAP-2 Disinhibition was strongly associated with both Disinhibition (r = .68) and Antagonism (r = .58). Contrary to hypothesis, however, PID-5 Detachment was linked substantially to Negative Temperament (r = .56) as well as Positive Temperament (r = −.49), rather than showing specificity to the latter, perhaps because two of the six Detachment facet scales (Suspiciousness and Depressivity) also are scored on Negative Affectivity. Finally, PID-5 Psychoticism correlated moderately with both Negative Temperament (r = .45) and Disinhibition (r = .32).
Correlations Between the SNAP-2 Big-Three Scales and PID-5 Domain Scores in the Patient Sample.
Note. N = 202. Correlations ≥|.40| are in boldface. Correlations ≥|.18| are significant at p < .05. SNAP-2 = Schedule of Nonadaptive and Adaptive Personality–Second Edition; PID-5 = Personality Inventory for DSM-5; NegTemp = Negative Temperament; PosTemp = Positive Temperament; Disinhib = Disinhibition.
Next, we conducted multiple regression analyses to determine how well each of the PID-5 domain scores was predicted by standard measures of the Big Three; the overall Rs from these analyses are displayed in the final column of Table 3. The SNAP-2 scales jointly were strong predictors of four of the five domain scores, with Rs ranging from .63 (Detachment) to .75 (Negative Affectivity); in contrast, Psychoticism was only moderately associated (R = .48) with the Big-Three traits.
These data establish strong and systematic links between general personality traits and the PID-5 domain scores. It is particularly noteworthy that three of the PID-5 domains—Negative Affectivity, Disinhibition, and Antagonism—map clearly onto two of the Big-Three traits. Again, the one unexpected finding is that Detachment is a blend of high Negative Temperament and low Positive Temperament, rather than being specifically linked to the latter but, as mentioned previously, this may be due, at least in part, to the two shared facets that mark both Detachment and Negative Affectivity.
Facet-Level Analyses
Correlations between the SNAP-2 temperament scales and the PID-5 facet scales are presented in Table 4. These data clearly establish that the PID-5 is rich in content related to both Negative Temperament and Disinhibition. Fourteen PID-5 scales had correlations of .40 or greater with Negative Temperament, and eight of these (e.g., Anxiousness, Emotional lability, Depressivity, Hostility) had correlations greater than .50. Similarly, nine scales had correlations of .40 or greater with Disinhibition, and five of these coefficients (e.g., Deceitfulness, Impulsivity, Risk taking) were at least .50. In marked contrast, only three scales (Anhedonia, Depressivity, Withdrawal) had correlations exceeding −.40 with Positive Temperament, and only that for Anhedonia exceeded −.50. Taken together with the results of Table 3, these data indicate that PID-5 content related to positive affectivity is somewhat limited.
Correlations Between the SNAP-2 Big-Three Scales and PID-5 Facet Scales in the Patient Sample.
Note. N = 202. Correlations ≥|.40| are in boldface. Correlations ≥|.15| are significant, p < .05. SNAP-2 = Schedule of Nonadaptive and Adaptive Personality–Second Edition; PID-5 = Personality Inventory for DSM-5; NegTemp = Negative Temperament; PosTemp = Positive Temperament; Disinhib = Disinhibition; NA = Negative Affectivity; DET = Detachment; ANT = Antagonism; DIS = Disinhibition; COMP = Compulsivity; PSY = Psychoticism; Cog & percept dysreg = Cognitive and perceptual dysregulation; Unusual beliefs & exper = Unusual beliefs and experiences.
It is also interesting to consider these relations in the context of the placements of these trait facets within the DSM-5 model. Although most scales behaved as expected, there were noteworthy anomalies. First, several scales that are putative markers of other domains—such as Anhedonia (r = .55), Distractibility (r = .49), and Eccentricity (r = .47)—had relatively strong correlations with Negative Temperament. Second, a number of scales—including Restricted affectivity (an intended marker of both high Detachment and low Negative Affectivity), Intimacy avoidance (Detachment), Grandiosity (Antagonism), and Submissiveness (Negative Affectivity)—correlated weakly with all three SNAP-2 scales (all rs < |.30|). Third, Rigid perfectionism—intended to be an indicator of low Disinhibition—correlated only .01 with SNAP-2 Disinhibition and instead correlated moderately (r = .34) with Negative Temperament. We return to this issue of DSM-5 placements after presenting results for the community sample.
We again conducted multiple regression analyses to determine how well each of the PID-5 facet scales was predicted by the Big-Three traits; the overall Rs from these analyses are displayed in the final column of Table 4. Eighteen scales (72%) had Rs of .50 or greater, and 23 scales (92%) had Rs of .30 or greater. Thus, the large majority of the PID-5 facets share substantial variance with these higher order personality traits. Anxiousness (R = .72), Anhedonia (R = .72), Depressivity (R = .66) and Emotional lability (R = .66) displayed the strongest overall associations in these analyses, whereas Intimacy avoidance (R = .27) and Restricted affectivity (R = .17) were only weakly related to the Big Three.
Community Sample Analyses
Factor Analysis of the FI-FFM Facet Scales
As noted earlier, the FI-FFM contains five facet scales each for Neuroticism, Extraversion, and Conscientiousness; four facets for Agreeableness; and three facets for Openness. Because the FI-FFM is a new and relatively unknown instrument, we were interested in whether its factor structure would replicate in a new sample. Accordingly, we conducted a principal factor analysis of the 22 FI-FFM facet scales in the community sample (N = 375 for these analyses), extracted five factors, and rotated them using varimax. As shown in Table 5, all 22 scales marked the intended factor. That is, they all had their highest loading (and all ≥.55) on their target factor. Thus, we summed the facet scale scores to create overall Big-Five domain scores, which we used in subsequent analyses.
Varimax Loadings of the FI-FFM Scales in the Community Sample.
Note. N = 375. Loadings ≥|.40| are in boldface. FI-FFM = Faceted Inventory of the Five-Factor Model; Neur = Neuroticism; Agree = Agreeableness; Con = Conscientiousness; Extra = Extraversion; Open = Openness.
Factor Analysis of the Higher Order Personality Scales
Next, we subjected the FI-FFM Big-Five domain scores, the BFI scales, the SNAP-2 Big-Three scales, and our three hypothesized Peculiarity markers (SNAP-2 Eccentric Perceptions; FI-FFM Unusual Experiences and Eccentric Beliefs) to a principal factor analysis with varimax rotation. As expected, these variables defined six clear factors that obviously can be labeled Neuroticism, Extraversion, Conscientiousness, Peculiarity, Agreeableness, and Openness, respectively (see, Table 6). We computed regression-based factor scores to model these dimensions in subsequent analyses.
Varimax Loadings of the General Personality Scales in the Community Sample.
Note. N = 335. Loadings ≥|.40| are in boldface. SNAP-2 = Schedule of Nonadaptive and Adaptive Personality–Second Edition; BFI = Big-Five Inventory; FI-FFM = Faceted Inventory of the Five-Factor Model.
Domain-Level Analyses
Table 7 presents correlations between these factor scores and the PID-5 domain scales in the community sample. These results provide partial support for our hypotheses, and are highly consistent with the domain-level findings from the patient sample. As expected (a) Neuroticism correlated strongly with PID-5 Negative Affectivity (r = .76), (b) Conscientiousness had a strong negative association with Disinhibition (r = −.74), (c) Agreeableness was inversely related to Antagonism (r = −.72), and (d) Peculiarity was strongly linked to Psychoticism (r = .71). Contrary to hypothesis—but consistent with the patient results—PID-5 Detachment correlated with—but did not show specificity to—Extraversion (r = −.47), as it also was substantially linked to both Neuroticism (r = .47) and Agreeableness (r = −.41). Finally, Openness related weakly to all five PID-5 domain scales, with correlations ranging from only −.09 to .15.
Correlations Between Big-Five Factor Scores and PID-5 Domain Scores in the Community Sample.
Note. N = 335. Correlations ≥|.40| are in boldface; those with strong convergent discriminant relations are underlined. Correlations ≥|.11| are significant, p < .05. PID-5 = Personality Inventory for DSM-5; N = Neuroticism; E = Extraversion; C = Conscientiousness; P = Peculiarity; A = Agreeableness; O = Openness.
Next, we conducted multiple regression analyses to determine how well the PID-5 domain scores could be predicted by the combined power of these basic personality factors; the overall Rs from these analyses are displayed in the final column of Table 7. The personality factors jointly were very strong predictors of all five domain scores, with Rs ranging from .80 (Detachment) to .87 (Negative Affectivity). Overall, we see very strong overlap between the DSM-5 trait model and basic factors of normal personality.
Thus, we see strong, clear parallels between three of the PID-5 domains and the Big-Five traits (neuroticism, agreeableness, conscientiousness) that define the alpha/stability factor in “Big-Two” models of personality (DeYoung, 2006; Digman, 1997; Markon et al., 2005). In contrast, although the two remaining domain scores clearly are strongly related to basic dimensions of personality, they do not map directly onto the traits (extraversion, openness) that define beta/plasticity in the Big Two.
Facet-Level Analyses
Table 8 reports parallel results for the PID-5 facet scales. Consistent with the patient findings, these data demonstrate that the PID-5 is particularly rich in content related to Neuroticism and (low) Agreeableness. Ten PID-5 scales had correlations of .40 or greater with Neuroticism, with six scales (e.g., Anxiousness, Emotional lability, Depressivity, Perseveration) correlating >.50. Similarly, nine scales had correlations ≥|.40| with Agreeableness, and four of these coefficients (Callousness, Deceitfulness, Hostility, Manipulativeness) were at least |.50|. Three scales (Irresponsibility, Impulsivity, and Distractibility) correlated ≥|.40| with Conscientiousness; all these associations exceeded |.50|. All three PID-5 Psychoticism scales had correlations exceeding |.40| with Peculiarity; those for Cognitive and perceptual dysregulation (r = .72) and Unusual beliefs and experiences (r = .72) exceeded |.70|. Corroborating the Table 4 findings, four scales had correlations ≥|.40| with Extraversion, but only those for Withdrawal and Anhedonia were ≥|.50|. Finally, no PID-5 scales correlated strongly with Openness; indeed, only two coefficients were ≥.20 (r = .27 for Risk taking, r = .20 for Unusual beliefs and experiences).
Correlations Between Big-Five Factor Scores and PID-5 Facet Scales in the Community Sample.
Note. N = 335. Correlations ≥|.40| are in boldface; those with strong convergent/discriminant patterns are underlined. Correlations ≥|.12| are significant, p < .05. PID-5 = Personality Inventory for DSM-5; N = Neuroticism; E = Extraversion; C = Conscientiousness; P = Peculiarity; A = Agreeableness; O = Openness; NA = Negative Affectivity; DET = Detachment; ANT = Antagonism; DIS = Disinhibition; COMP = Compulsivity; PSY = Psychoticism; Cog & percept dysreg = Cognitive and perceptual dysregulation; Unsual bel & exper = Unusual beliefs and experiences.
As in the patient sample, although most scales behaved as expected, there were some notable deviations from the DSM-5 model. First, consistent with the Table 4 results, Anhedonia (r = .53) and Distractibility (r = .41) again had relatively strong associations with Neuroticism; in fact, Anhedonia had its highest correlation with that factor. Second, similar to its pattern in the patient data, Intimacy avoidance (a putative facet of Detachment) correlated weakly with all six factor scores (all rs < |.30|). Third, Rigid perfectionism—intended to be an indicator of low Disinhibition—correlated only .28 with Conscientiousness and was more strongly linked to Neuroticism (r = .41). Fourth, Attention seeking not only displayed its hypothesized negative association with Agreeableness (r = −.42) but also had a link of comparable magnitude with Extraversion (r = .44). Fifth, Submissiveness again correlated relatively weakly with Neuroticism (r = .31) and also had an association of similar magnitude with Extraversion (r = −.29). Finally, Restricted affectivity failed to correlate strongly with either Neuroticism (r = −.04) or Extraversion (r = −.23), and actually was most strongly related to Agreeableness (r = −.38).
We again conducted multiple regression analyses to determine how well the PID-5 facet scales were predicted by the combined contributions of these basic trait scores; the overall Rs from these analyses are displayed in the final column of Table 8. Twenty-two scales (88%) had Rs ≥ .50, and 19 scales (76%) had Rs ≥ .60. Thus, the large majority of the PID-5 facets share substantial variance with these higher order personality traits. Anxiousness (R = .84), Cognitive and perceptual dysregulation (R = .82), Hostility (R = .81), and Depressivity (R = .81) showed the strongest overall associations in these analyses; in contrast, Restricted affectivity (R = .47), Submissiveness (R = .46), and Intimacy avoidance (R = .42) displayed the weakest associations with these traits.
Discussion
Integrating Normal and Pathological Personality
A Common “Big Three”
The basic goal of this study was to examine relations between the DSM-5 trait model—as operationalized by the PID-5 (Krueger et al., 2012)—and the higher order dimensions contained in the prominent Big-Three and Big-Five models of personality. We found highly consistent results across our two samples. First, in our patient sample, we related the PID-5 to the Big-Three scales of the SNAP-2. PID-5 Negative Affectivity was highly associated with SNAP-2 Negative Temperament (r = .75), whereas Disinhibition (r = .68) and Antagonism (r = .58) both were strongly related to SNAP-2 Disinhibition, which is a blend of low conscientiousness and low agreeableness (see Markon et al., 2005; Watson et al., 1994). Next, in our community adult sample, we examined correlations between the PID-5 and basic trait factor scores. Here, we found strong associations between Negative Affectivity and Neuroticism (r = .76), Disinhibition and low Conscientiousness (r = −.74), and Antagonism and low Agreeableness (r = −.72). On the basis of these results, it seems reasonable to conclude that these three domains in the DSM-5 trait model represent pathological variants of basic, higher order personality dimensions. These domains therefore represent a “Big Three” that is common to both normal and pathological personality. Of note, these three domains define the alpha/stability factor in Big-Two models of personality (DeYoung, 2006; Digman, 1997; Markon et al., 2005; Watson et al., 2008).
Detachment
The PID-5 Detachment domain score also was strongly linked to basic personality traits in our data (R = .63 and .80 in the patients and community participants, respectively). We predicted that it would show a strong, specific negative association with extraversion but here, the results were more complex and did not clearly support our hypothesis. The PID-5 Detachment score was negatively related to both Positive Temperament (r = −.49) and Extraversion (r = −.47), but these correlations were only moderate in magnitude. Moreover, Detachment actually correlated as strongly with Negative Temperament (r = .56) and Neuroticism (r = .47) in both samples; it also had a moderate negative association with Agreeableness (r = −.41) in the community data. Given this pattern of results, it would not be accurate to describe DSM-5 detachment simply as a pathological variant of (low) extraversion.
These results reflect two basic considerations. First, correlations between PID-5 Detachment and markers of extraversion/positive affectivity were lower than expected, falling only in the −.45 to −.50 range. Related to this, one surprising aspect of our results was the relative paucity of detachment-related content in the PID-5: Only three PID-5 facet scales—Withdrawal (r = −.43 and −.58 in the patient and community data, respectively), Anhedonia (r = −.64 and −.50, respectively), and Depressivity (r = −.48 and −.41, respectively)—had correlations of at least |.40| with extraversion/positive affectivity in both samples. The other three DSM-5 detachment facets—Suspiciousness, Intimacy avoidance, and Restricted affectivity—all had relatively low associations with extraversion/positive affectivity in both samples (rs ranged from −.06 to −.31, mean r = −.20). The results for Intimacy avoidance and Restricted affectivity are particularly surprising, given that Watson et al. (2008) found that scales assessing social aloofness and constricted affect clearly defined the low end of a bipolar Extraversion factor (see their table 5). It is possible that these relatively low correlations reflect the particular way in which these traits have been operationalized in the PID-5 (e.g., Intimacy avoidance focuses almost exclusively on romantic/sexual relationships). This is an important issue for future research.
Second, PID-5 Detachment did not show specificity to extraversion, but instead correlated substantially with other higher order domains, particularly neuroticism. To some extent, this pattern may reflect the fact that the DSM-5 model understandably focuses on the pathological range of this dimension. In this regard, Haigler and Widiger (2001) experimentally altered NEO PI-R items to create more pathological variants of these scales (the EXP-NEOPIR). The resulting EXP-NEOPIR Extraversion scale correlated almost as highly with low NEO PI-R Agreeableness (r = −.44) as it did with NEO PI-R Extraversion (r = .50); its correlation with NEO PI-R Neuroticism remained relatively low, however (r = .21; see Haigler & Widiger, 2001, table 3). Thus, the relatively strong correlation of, for example, Anhedonia, with Negative Affectivity (r = .55 and .53, in the patient and community samples, respectively) may largely reflect its generally psychopathological nature.
Beyond this, our results also likely reflect the manner in which the Detachment domain is specified in the DSM-5 model. In particular, two of its facets—depressivity and suspiciousness—are cross-listed as facets of high Negative Affectivity in the DSM-5 trait model and so, not surprisingly, had relatively strong associations with Neuroticism: Depressivity (r = .60 and .65 in the patient and community data, respectively) and Suspiciousness (r = .51 and .42, respectively). Thus, as currently specified, the DSM-5 model builds in substantial overlap between Detachment and Neuroticism/Negative Affectivity.
It is important to note, moreover, that standard trait models typically do not place either depressivity or suspiciousness within the extraversion domain. For example, the NEO PI-R and the FI-FFM both (a) classify Depressivity as a facet of Neuroticism and (b) place the dimension of Trust versus Mistrust/Suspiciousness within the domain of Agreeableness (for related results, see Table 5). This suggests that these traits potentially are misclassified as facets of detachment in the DSM-5 model.
We tested this possibility by rescoring the PID-5 Detachment domain without these two facets and then recomputing its correlations with our basic trait measures. In comparison with the full Detachment domain score (see Tables 3 and 7), this reduced variable displayed a clearer, more specific pattern vis-à-vis Neuroticism/Negative Affectivity and Extraversion/Positive Affectivity. That is, it had a substantially lower correlation with Neuroticism in both the patient (from .56 to .44) and the community (from .47 to .33) samples. Its correlation with SNAP-2 Positive Temperament decreased slightly from −.49 to −.45 in the patient sample, but its correlation with Extraversion actually increased slightly from −.47 to −.51 in the community data. Although our results clearly need to be replicated and extended in future research, they tentatively suggest that depressivity and suspiciousness are misclassified as Detachment facets.
Psychoticism
As expected, the PID-5 Psychoticism domain score did not correlate strongly with any individual Big-Three or Big-Five dimension. In the patient sample, it was moderately correlated with Negative Temperament (r = .45) and Disinhibition (r = .32). In the community sample, it correlated primarily with Peculiarity (r = .71) but also showed modest associations with low Agreeableness (r = −.35), Neuroticism (r = .32), and low Conscientiousness (r = −.24). Thus, in addition to its strong links to Peculiarity, Psychoticism correlated moderately with the traits that define the alpha/stability factor in Big-Two personality models (DeYoung, 2006; Digman, 1997; Markon et al., 2005; Watson et al., 2008). These results are consistent with the findings of Watson et al. (2008, table 1), who found that scales assessing odd and eccentric characteristics were strong markers of the higher order alpha/stability factor, whereas measures of openness tended to load more strongly on beta/plasticity.
Beyond these individual associations, it is important to note that Psychoticism scores had substantial overall relations with basic dimensions of personality. Analyses of the patient data indicated that the SNAP-2 Big-Three scales jointly accounted for a moderate amount of variance in Psychoticism (R = .48). Our ability to model a Peculiarity factor in the community participants led to a much higher level of prediction of Psychoticism in this sample (R = .84). More generally, our data establish that all five DSM-5 trait domains are substantially related to basic dimensions of personality.
Openness and Psychoticism
We also predicted that Psychoticism would be moderately correlated with Openness in the community data. Although this association was statistically significant, it was lower than expected (r = .15) and actually was weaker than Psychoticism’s link to the various markers of alpha/stability (see Table 7). As discussed earlier, however, the associations between openness and psychoticism are highly complex and depend critically on the specific aspects of the openness domain that are assessed (see DeYoung et al., 2012; Watson et al., 2008). For example, DeYoung et al. (2012) found that indicators of psychoticism were positively associated with traits related to openness to experience per se (e.g., fantasy proneness; appreciation of art, beauty, and aesthetics), but were weakly or negatively related to measures of intellect and intelligence. Similarly, Chmielewski, Bagby, and Markon (2013) conducted analyses that controlled for shared variance due to the higher order trait factor, and found that the unique aspects of openness to experience and intellect were positively and negatively correlated, respectively, with DSM-5 Psychoticism. Thus, depending on how openness is assessed and modeled, it may show positive, negative, or near zero associations with psychoticism. It will be important for future research to explicate the nature of these links more thoroughly.
Implications for the DSM-5 Trait Model
Our analyses are limited in that they are based on a single self-report instrument (the PID-5) and focus almost entirely on a single broad issue (i.e., how the DSM-5 model relates to general, higher order traits of personality). Nevertheless, our data have produced some unexpected findings that merit closer scrutiny in the future. First, as noted, Intimacy avoidance, Restricted affectivity, and Suspiciousness all had relatively weak associations with extraversion/positive affectivity. Moreover, Restricted affectivity—which also is a marker of low negative affectivity in the DSM-5 model—correlated only .15 with Negative Temperament in the patient data, and only −.04 with Neuroticism in the community sample. These results raise questions regarding the placement of these traits within these domains.
Second, Anhedonia (an intended marker of detachment) and Distractibility (a facet of disinhibition) both correlated substantially with neuroticism (for Anhedonia, r = .55 and .53 in the patient and community samples, respectively; for Distractibility, r = .49 and .41, respectively). Conversely, Submissiveness—a facet of Negative Affectivity in the DSM-5 trait model—correlated only .29 with Negative Temperament and .31 with Neuroticism in the patient and community samples, respectively. Interestingly, these were the highest correlations of Submissiveness with any higher order trait, and its multiple Rs with all traits were only .34 and .46, in the two samples, respectively, indicating that Submissiveness assesses specific content not well represented in general trait models. Again, these results raise questions about the optimal specification of the Negative Affectivity domain in the DSM-5 model, although it must be noted that this is not unique to either the DSM-5 or the PID-5. Specifically, submissiveness is one of the two components of dependency (Morgan & Clark, 2010), and dependency scales typically have only moderate correlations with neuroticism/negative affectivity (see, meta-analyses by Saulsman & Page, 2004, and Samuel & Widiger, 2008) as well as relatively large proportions of specific variance (e.g., Markon et al., 2005).
Finally, rigid perfectionism is considered a marker of low Disinhibition in the DSM-5 model. However, the corresponding PID-5 scale correlated only .01 with SNAP-2 Disinhibition in the patient sample, and only .28 with the Conscientiousness factor in the community data. Moreover, Rigid perfectionism correlated more strongly with Neuroticism in both samples (rs = .34 and .41, respectively). These findings suggest that the trait perhaps is better viewed as an indicator of Negative Affectivity, as is Perseveration, the other facet in the DSM-5 model specifically developed to capture variance associated with obsessive-compulsive PD.
Limitations and Future Directions
Our study has some notable strengths. First, we used a variety of general trait scales that were designed to assess either the Big-Three or Big-Five models of personality. Second, we replicated our key findings across both a patient and a community sample. Third, we measured personality pathology using the PID-5, which was developed by some members and consultants of the DSM-5 Personality and Personality Disorders Work Group to operationalize the Work Group’s trait model. As noted earlier, the PID-5 greatly facilitates analyses that compare and contrast the DSM-5 trait scheme with other models of personality.
At the same time, however, our study also suffers from two key (and related) limitations. First, our assessment of personality pathology as specifically delineated in DSM-5 was limited to a single instrument, the PID-5. As we have discussed, we obtained some unexpected findings that ultimately may reflect idiosyncratic aspects of this particular inventory (i.e., rather than the DSM-5 model per se). Second, our results all are based on a single, common method, namely, self-report. It will be important to examine these relations in other types of data, including informant reports and interview/clinician ratings.
Despite these limitations, our study has contributed to the literature by clarifying how the DSM-5 trait model relates to the higher order dimensions contained in the prominent Big-Three and Big-Five models of personality. Although we obtained some unexpected results, we also were able to establish that three domains in the DSM-5 trait model (viz., Negative Affectivity, Disinhibition, and Antagonism) clearly represent pathological variants of general personality dimensions (neuroticism, conscientiousness, and agreeableness, respectively). Thus, these domains represent a “Big Three” that is common to both normal and pathological personality. We hope that future studies will build on these findings to articulate a fully integrated trait structure that subsumes both normal and pathological manifestations of personality.
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 National Institute of Mental Health Grant R01-MH083830 to Lee Anna Clark.