Improving Characterization of Psychopathy within the DSM-5 AMPD: Creation and Validation of PID-5 Triarchic Scales

Abstract

The latest edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) bridges a significant gap in psychiatric nosology by providing trait-based characterizations of psychopathy along with antisocial personality disorder (ASPD) within the Section III alternative model for personality disorders (AMPD). However, the representation of psychopathy in the AMPD has met with some criticisms (e.g., Few, Lynam, Maples, MacKillop, & Miller, 2015; Crego & Widiger, 2014). The current study was undertaken to establish an improved means for characterizing psychopathy in DSM-5 Section III terms, by creating scale measures of triarchic psychopathy dimensions using items from the best-established assessment instrument for the AMPD, the Personality Inventory for DSM-5 (PID-5). Using data from a sample of community adults (N = 210) we employed a construct rating and psychometric refinement approach to develop item-based PID-5 Triarchic (PID-5-Tri) scales for measuring psychopathy dimensions of boldness, meanness, and disinhibition. The validity of the PID-5-Tri scales was then evaluated in relation to criteria including other scale measures of the triarchic constructs and psychopathy, self-reported antisocial behavior and substance use, empathy, internalizing and other clinical problems, and personality within the development sample and a separate independent sample of adults (N = 240) recruited to have elevated psychopathic traits. Results of this work provide a foundation for improved characterization of psychopathy in terms of the AMPD trait system and provide a mechanism for future research oriented toward clarifying the developmental interface between childhood conduct disorder and psychopathy, as well as identifying neurobiological correlates of dimensions of psychopathy.

Traditional category-based models of personality disorders (PDs) suffer from well-documented limitations including low reliability, arbitrary symptom thresholds for diagnosis, and high comorbidity among PDs (Livesley & Jang, 2000; Trull & Durrett, 2005; Widiger & Clark, 2000). A further issue with categorical PDs as represented in the official diagnostic system for North America, the Diagnostic and Statistical Manual of Mental Disorders (DSM; American Psychiatric Association [APA], 2013) is that some are specified in ways that do not effectively represent historic descriptions. Perhaps the most notable example is the categorical diagnosis of antisocial personality disorder (ASPD), introduced in the third edition of the DSM (DSM-III; APA, 1980) to represent psychopathy – a PD classically seen to involve interpersonal-emotional features such as superficial charm, egocentricity, persuasiveness, low anxiousness, absence of remorse, and incapacity for love along with pervasive behavioral deviancy (Cleckley, 1976; Hare, 2003; McCord & McCord, 1964). Following publication of DSM-III, some experts voiced concerns about weaknesses in the ASPD diagnosis as a representation of psychopathy (e.g., Millon, 1981) – in particular its overemphasis on behavioral deviance criteria and poor representation of interpersonal affective features – and these concerns persisted in relation to somewhat amended versions of the diagnosis in revised-third and fourth editions of the DSM (e.g., Hart & Hare, 1996).

Though the categorical diagnosis of ASPD was carried over unchanged from DSM-IV (APA, 2000) to the main diagnostic section of DSM-5 (APA, 2013), the current fifth edition also contains an alternative trait-dimensional system for diagnosing PDs – within Section III, Emerging Models and Measures. This alternative model for personality disorders (AMPD) provides improved coverage of classic interpersonal-affective features in a trait-based definition of ASPD that includes a “psychopathy specifier.” However, the AMPD psychopathy specifier has been criticized on certain grounds (Few et al., 2015), including its reliance on reversed trait facets (i.e., low anxiousness, low withdrawal; Crego & Widiger, 2014). The current study sought to address such criticisms and refine the AMPD characterization of psychopathy by applying a well-established scale construction approach to the Personality Inventory for DSM-5 (PID-5; Krueger, Derringer, Markon, Watson, & Skodol, 2012) to create item-based measures of psychopathy facets as described in the triarchic model (Patrick, Fowles, & Krueger, 2009), a construct-based model not bound to any particular measure. The development of triarchic scales for the PID-5 provides a means for indexing psychopathy subdimensions in AMPD trait terms in a way that connects closely (i.e., harmonizes) with other established triarchic scale measures (e.g., Brislin et al., 2015; Drislane et al., 2015; Drislane, Brislin, Jones, & 2018; Hall et al., 2014; Sellbom et al., 2016; see also Drislane & Patrick, 2017). Further, because the triarchic model constructs are conceptualized in neurobehavioral trait terms (Patrick & Drislane, 2015; Patrick, Durbin, & Moser, 2012), the current work also provides a means for interfacing the DSM-5’s trait-based conception of psychopathy with neural concepts and measures (Wygant, Pardini, Marsh, & Patrick, 2018).

DSM-5 Dimensional-Trait System for Personality Pathology: Description and Quantification

Traditional categorical diagnoses of personality disorders have been retained in the primary Diagnostic Criteria and Codes section (Section II) of DSM-5 (APA, 2013), and within this section, are defined in precisely the same manner as in DSM-IV (APA, 2000). However, in response to critiques of categorical PD diagnosis by a number of investigative groups (Livesley & Jang, 2000; Trull & Durrett, 2005; Widiger & Mullins-Sweatt, 2009), Section III of DSM-5 now provides an alternative dimensional system – the AMPD – for characterizing personality pathology in terms of extremes along maladaptive trait continua (25 in all, organized into 5 broad domains), in connection with dysfunction in self- or interpersonal functioning (i.e., impairments in identity, self-direction, empathy, and/or intimacy). Included within the AMPD are trait-based definitions for 6 of the 10 personality disorders listed in Section II: antisocial, borderline, narcissistic, avoidant, obsessive-compulsive, and schizotypal.

To anchor this new system for PDs to an empirical measurement model, Krueger, Derringer, Markon, Watson, and Skodol (2012) developed the self-report based Personality Inventory for DSM-5 (PID-5; Krueger et al., 2012). Results from this empirical work fed back into the Section III system, influencing final characterizations of the traits and their assignment to domains (Krueger, 2019). The PID-5 assesses the 25 lower-order traits of the Section III AMPD within five broad domains: Negative Affect, Detachment, Antagonism, Disinhibition, and Psychoticism. Research has supported the construct validity of the PID-5 in relation to other assessment inventories for personality disorders (Anderson, et al., 2013; Few et al., 2013) and in relation to comprehensive measures of normal-range personality (deFruyt et al., 2013; Gore & Widiger, 2013). The five broad domains of the PID-5 also map on to the Personality Psychology–Five (PSY-5) dimensions measured by the Minnesota Multiphasic Personality Inventory-2 Restructured Form (Anderson et al., 2013).

Psychopathy in the DSM-5 AMPD: Relationship to the Triarchic Model

As noted at the outset, criticisms have been voiced about the diagnosis of ASPD as a representation of psychopathy in the third through fourth editions of the DSM, and by extension, in Section II of the current fifth edition (for a recent review and discussion of issues, see Crego & Widiger, 2015; Widiger & Crego, 2018). A particular source of concern has been with the lack of specificity of the diagnosis. The criteria for ASPD are polythetic and can be fulfilled in many different ways, resulting in wide heterogeneity among individuals assigned the diagnosis. This heterogeneity is a problem for clinical decision-making (e.g., choice of treatment) and prediction (e.g., risk for violence or suicide). It is also a problem for research directed at understanding causal factors contributing to psychopathy and antisocial behavior. Indeed, the importance of interpersonal-affective features for delineating an etiologically distinct subset of individuals exhibiting persistent impulsive deviancy has been a central recurring theme in the psychopathy literature over the years (e.g., Cleckley, 1976; Fowles, 1980; Frick, Ray, Thornton, & Kahn, 2014; Lynam, 1998).

The trait-based approach to diagnosing ASPD in the DSM-5 AMPD includes markedly enhanced representation of interpersonal-affective features relative to the Section II categorical diagnosis (Anderson et al., 2014; Crego & Widiger, 2014; Few et al., 2015; Strickland et al., 2013; Wygant et al., 2016). The AMPD diagnostic criteria for ASPD include characteristic deficits in self- and interpersonal functioning (i.e., egocentrism; failure to conform to the law; lack of remorse; incapacity for mutually intimate relationships) along with elevations on distinct personality traits from three domains: callousness, manipulativeness, and deceitfulness (domain of Antagonism); impulsivity, risk-taking, and irresponsibility (domain of Disinhibition); and hostility (domain of Negative Affectivity; APA, 2013). Three of the four characteristic deficits in functioning (egocentrism, lack of remorse, incapacity for intimacy) and 3 of the 7 designated traits (those from the domain of Antagonism) reflect interpersonal-affective features. Additionally, the AMPD characterization of ASPD includes a “psychopathy specifier” encompassing attention-seeking (Antagonism domain) along with low withdrawal (Detachment domain) and low anxiousness (Negative Affectivity domain). This specifier is intended to provide coverage of interpersonal-affective features (i.e., social ease/efficacy, imperturbability) beyond those represented in the main dysfunction and trait criteria for ASPD that have been emphasized in influential historic accounts of psychopathy (Cleckley, 1976 [per Crego & Widiger, 2016]; Fowles, 1980; Lykken, 1957).

The traits included in the AMPD definition of ASPD and those comprising the psychopathy specifier can be operationalized using the PID-5. Several empirical studies have been conducted to evaluate the coverage of psychopathic features provided by these AMPD traits as assessed by the PID-5 (Anderson, Sellbom, Wygant, Salekin, & Krueger, 2014; Crego & Widiger, 2014; Few et al., 2015; Fossati, Krueger, Markon, Borroni, & Maffei, 2013; Miller, Lamkin, Maples-Keller, Sleep, & Lynam, 2017; Strickland, Drislane, Lucy, Krueger, & Patrick, 2013; Wygant et al., 2016). Most of these studies have evaluated coverage in terms of representation of facets of psychopathy represented in the triarchic model (Patrick et al., 2009). The triarchic model provides a valuable referent for evaluating the coverage of alternative assessment instruments because the constructs of this model – boldness, meanness, and disinhibition – were deduced from the psychopathy literature as a whole, and are represented to varying degrees in all conceptualizations and measures of psychopathy (for a review of relevant evidence, see Patrick & Drislane, 2015). In addition, the triarchic model constructs are conceptualized in biobehavioral terms and thus are helpful for linking psychopathy symptoms to neurobiological concepts and measures (Patrick, 2018).

The triarchic construct of boldness encompasses dominant and insouciant/fearless elements of psychopathy and is theorized to have a basis in the brain’s defensive (fear) system; meanness reflects callous, exploitative, and disaffiliative elements and is hypothesized to reflect dysfunction in neural circuits for empathy and attachment; and disinhibition, which encompasses impulsive, irresponsible, and reckless elements, is theorized to have a basis in frontal control (executive) systems of the brain. Studies conducted to date that have examined the representation of triarchic model constructs in PID-5 traits comprising Section III APSD and the AMPD psychopathy specifier (Anderson et al., 2014; Crego & Widiger, 2014; Miller et al., 2017; Strickland et al., 2013; Wygant et al., 2016) have used the 58-item Triarchic Psychopathy Measure (TriPM; Drislane, Patrick, & Arsal, 2014). These have consistently found that the AMPD traits designated for ASPD provide highly effective coverage of the meanness and disinhibition constructs. In particular, callousness from the PID-5 domain of Antagonism shows strong convergence with triarchic meanness as indexed by the TriPM, and impulsivity and irresponsibility from the PID-5 domain of Disinhibition converge very strongly with TriPM-assessed disinhibition. Two of the other ASPD-designated traits, deceitfulness and hostility, relate at similar levels to TriPM-assessed meanness and disinhibition (but not to boldness), and the remaining two ASPD traits demonstrate convergence with all three triarchic constructs as indexed by the TriPM – manipulativeness at similar levels with all three, and risk-taking more strongly with boldness than with either disinhibition or meanness. As for the AMPD psychopathy specifier, moderate-level associations have been found for attention seeking and withdrawal (positive and negative, respectively) with TriPM-assessed boldness, along with a moderately high correlation (negative) for anxiousness. Taken together, these specifier traits appear to provide fairly effective coverage of the boldness facet of psychopathy as described in the triarchic model. Strickland et al. (2013), for example, reported an omnibus prediction coefficient (R) of .7 when these three traits were used together to predict TriPM-assessed boldness.

Notwithstanding these results, some authors have questioned the utility and effectiveness of the AMPD psychopathy specifier. Few and colleagues (2015), for example, noted that scores on the psychopathy specifier are minimally associated with measures of externalizing behaviors. This finding relates to broader debates in the field about the relevance of boldness/fearless dominance to the psychopathy construct (see, e.g., Crego & Widiger, 2016; Marcus, Fulton, & Edens, 2012; Miller & Lynam, 2012; Lilienfeld et al., 2012; Patrick, Venables, & Drislane, 2013; Venables, Hall, & Patrick, 2014). Crego and Widiger (2014) questioned the AMPD specifier more on methodological and conceptual grounds. Calling attention to the finding of weaker correlations for the withdrawal and attention seeking scales of the PID-5 with measures of boldness than for PID-5 anxiousness, these authors suggested that these two traits may be less relevant to defining this dimension. Crego and Widiger further noted that the AMPD traits of withdrawal and anxiousness are treated in reverse for the specifier, and questioned whether the absence of a proclivity (e.g., anxiousness) that is opposite to a boldness-related proclivity (e.g., imperturbability) necessarily implies the presence of the boldness-related proclivity. They suggested, for example, that individuals who do not endorse maladaptive anxiousness may simply be calm or relaxed without necessarily being fearless or bold. Crego and Widiger also expressed concern about the lack of discriminant validity for the domains of Disinhibition and Antagonism as indexed by the PID-5. For example, they presented empirical evidence that scores on PID-5 Antagonism correlate as highly with measures of impulsivity and disinhibition as they do with measures of callousness and coldheartedness.

Current Study

The current study was undertaken to address the concerns raised by Crego and Widiger (2014) and improve the AMPD characterization of psychopathy, by developing PID-5 based scales anchored to the constructs of the triarchic model. To accomplish this, we employed a now well-established scale construction method to configure relevant sets of items from the PID-5 to align clearly with the three triarchic model constructs. In particular, we sought to: (1) map individual items of the PID-5 to triarchic constructs based on conceptual relevance and empirical coherence, (2) include items indicative of both high and low poles of each construct, and (3) demonstrate convergence of the resultant PID-5 based triarchic scales with other triarchic scale measures developed and validated in prior work.

Based on findings from prior published work as reviewed above, our specific hypotheses were as follows: Scale development and refinement efforts would result in an item-based measure of boldness containing items from PID-5 scales that assess the traits of the AMPD psychopathy specifier (i.e., Anxiousness, Attention Seeking, and Withdrawal) along with items providing coverage of fearless/venturesome tendencies (i.e., from PID-5 Risk-Taking) and presence versus absence of a self-assured, dominant interpersonal style (e.g., Grandiosity, Submissiveness ). The PID-5 based Meanness scale was expected to be composed primarily of items from the PID-5 Callousness scale, with inclusion of some items from other Antagonism traits specified for ASPD (Deceitfulness, Manipulativeness) and some items from scales indexing unemotionality (Restricted Affectivity) and lack of/ disdain for close relationships (Intimacy Avoidance). Finally, the PID-5 based measure of the third triarchic model construct, disinhibition, was expected to be composed mainly of items from PID-5 scales with greatest specific relevance to the concept of general externalizing proneness, including Impulsivity, Irresponsibility, Hostility, and Suspiciousness (Krueger et al., 2007).

Following development of the PID-5-Tri scales, we utilized two separate samples to evaluate their convergent and discriminant validity in relation to criteria of various kinds, including measures of externalizing problems (i.e., substance use and antisocial behavior), other psychopathy measures, internalizing problems, and personality traits. A listing of the criterion measures available for each sample (with accompanying citations) and specific hypotheses for each measure are presented in Table A of the Online Supplement. Additionally, extending previous work (Drislane et al., 2018; Drislane & Patrick, 2017), we predicted that the PID-5-Tri scales would operate as effective indicators of latent triarchic constructs when included along with counterpart scales from the TriPM and MPQ in a confirmatory factor analysis (see Figure A of Online Supplement).

Method

Participants

Development Sample (N = 210).

The development sample¹ (M age = 20.8, SD = 4.22; 102 women) included undergraduates from a southeastern university (recruited through campus notices) along with young adults from the surrounding urban community (recruited via Craigslist ads). Participants were 79% White, 11.9% African American, 5.7% Asian American, 1.4% of mixed race, and 1.9% undisclosed; 31% reported being Hispanic or Latino. Prior to participating in the study individuals completed the Disinhibition and Boldness scales of the TriPM (Drislane et al., 2014) as pre-screening measures; those scoring in the highest and lowest quartiles on these scales were oversampled for testing, with some representation of individuals in the middle 50% as well (see Online Supplement for further details). Test subjects received either payment (at $10/hr) or, in the case of undergraduates, course credit for participating. Procedures for the study were approved by the university’s Institutional Review Board and all participants provided written informed consent prior to initiation of testing.

Validation Sample (N = 240).

The validation sample was collected independently by another research team and consisted of adults (M age = 26.9, SD = 10.1; 100 women) from a different urban community in the southeastern United States, recruited using Craigslist ads calling for “adventurous, fearless, charming, and carefree people who’ve led exciting lives.” This recruitment strategy has been shown in prior research to draw in individuals with elevated levels of psychopathic traits (e.g., DeMatteo, Heilbrun & Marczyk, 2006; Widom, 1977). Participants in this validation sample were 55% White, 36% African American, and 9% other or mixed ethnicity. All participants provided informed written consent and were paid $75 for their participation.

Measures and Procedure

The PID-5 (Krueger et al., 2012) was administered to participants in both samples. This self-report measure, which consists of 220 items designed to assess maladaptive personality traits specified within the DSM-5 AMPD (APA, 2013), served as the source of candidate items for the triarchic scales developed in the current study. Item responses are coded 0–3 and collated to yield scores for 25 specific traits and five broad domains (Negative Affectivity, Detachment, Disinhibition, Antagonism, and Psychoticism). As noted earlier, a sizable body of evidence has accrued for the validity of the PID-5 in terms of relations with criterion measures of personality and personality pathology (Few et al., 2015; Strickland et al., 2013).

Participants in both of the current study samples were also administered the TriPM (Drislane et al., 2014), which contain 58 items organized into three scales (Boldness, Meanness, Disinhibition) specifically designed to index the triarchic model dimensions. Scores on the TriPM correlate highly with other well-established measures of psychopathy in undergraduate, community, and correctional samples (Drislane et al., 2014; Sellbom & Phillips, 2013; Stanley et al., 2013; Wygant et al., 2016; see Sellbom, Lilienfeld, Fowler, & McCrary, 2018, for a review). Differing sets of criterion measures were also available for the two samples of the current study, as listed in Supplemental Table A. Those for the development sample consisted of questionnaire inventories administered along with the PID-5 in the context of an in-person laboratory testing session. Criterion measures for the validation sample included questionnaire inventories along with responses to questions about use and problems experienced with alcohol, marijuana, and other drugs from a background interview (administered by a trained graduate student) that also included questions about education, medical history, and legal problems.

Scale Development and Data Analytic Approach

Triarchic scales were constructed from items of the PID-5 using a construct rating and scale refinement approach documented in other published work (e.g., Drislane et al., 2018; Hall et al., 2014; Sellbom et al., 2016). Five evaluators (2 PhD researchers and 3 clinical psychology graduate students) rated each item of the PID-5 for its relevance to the three triarchic model constructs, using narrative descriptions of each construct developed from information provided by Patrick et al. (2009) as referents² (see Online Supplement for the full Construct Definition Form provided to raters). Ratings for each item triarchic construct were assigned on a 5-point scale. Considering the construct of disinhibition, for example, PID-5 items were each rated as follows: unrelated to disinhibition, strongly represents HIGH disinhibition, somewhat represents HIGH disinhibition, somewhat represents LOW disinhibition, or strongly represents LOW disinhibition. The overall inter-rater reliability of these assignments was high (ICC_{Boldness All Items} = .89, ICC_{Final Boldness Scale} = .95; ICC_{Meanness All Items} = .94, ICC_{Final Meanness Scale} = .97; ICC_{Disinhibition All Items} = .90, ICC_{Final Disinhibition Scale} = .95). As an initial step, a set of candidate items was identified for each scale, consisting of those rated as highly relevant to one particular triarchic construct by at least 4 of the 5 raters. This yielded 34 candidate items for Boldness, 51 for Meanness, and 40 for Disinhibition. Next, items were dropped from a particular candidate set if their deletion improved internal consistency (Cronbach’s alpha), if their content was redundant with other items in that set, or if they correlated appreciably with the items of another set. After applying these criteria, 16 candidate items remained for the Boldness scale, 23 for Meanness, and 20 for Disinhibition.

As a next step following the item selection and refinement phase, other available items that exhibited moderate construct relevance (according to evaluators’ ratings) were considered for inclusion based on criteria of good convergence with items in a particular set and lower convergence with items of other sets. This procedure led to the addition of 1 item to each of the Boldness and Disinhibition scale sets; no further items were added to the Meanness scale. As a final step, candidate items for each scale were examined in terms of item response theory (IRT) properties using Mplus (version 5; Muthen & Muthen, 2007). Items that showed weak IRT discrimination were dropped, resulting in deletion of 1 item from the Disinhibition scale and 2 each from the Boldness and Meanness scales. The final PID-5 triarchic (PID-5-Tri) scales therefore comprise a total of 55 items – 15 in the Boldness scale (7 of them reverse-keyed), 21 in the Meanness scale (3 reverse-keyed), and 19 in the final Tri-Disinhibition scale (2 reverse-keyed).³ As noted previously, the foregoing scale construction procedures were applied to data for the development sample only.

Following completion of this scale construction and refinement process, the resultant PID-5-Tri scales were evaluated for validity by examining their associations with criterion measures of psychopathy and related constructs in two samples: the community/undergraduate sample in which the scales were developed (N = 210), and the separate community validation sample recruited to have elevated psychopathic traits (N = 240). The types of measures used to evaluate convergent and discriminant validity of the final PID-5-Tri scales are summarized above and listed individually in Supplemental Table A. Regression analyses were also performed in order to examine the unique contribution of each PID-5-Tri scale to prediction of the various criterion measures. A somewhat conservative threshold for significance (p < .01) was used to balance issues of experimentwise error and statistical power.

In addition to bivariate correlations and regression analyses, we drew on recent published work (Drislane et al., 2018; Drislane & Patrick, 2017) to further evaluate these new scales by testing the fit of a confirmatory factor analytic (CFA) model of the triarchic dimensions utilizing the PID-5-Tri scales along with other established triarchic scale measures.

Results

Content of Final Scales

The content coverage of the items comprising the final PID-5-Tri scales appears largely consistent with study hypotheses (see Table 1). The Boldness scale includes five items reflecting traits of attention seeking, manipulativeness, and grandiosity (from the PID-5 domain of Antagonism), three items indicative of risk-taking (PID-5 domain of Disinhibition), and five reverse-keyed items pertaining to dispositional anxiousness and submissiveness (domain of Negative Affect). PID-5 Boldness also includes two reverse-keyed items from the domain of Detachment, reflecting social engagement (low withdrawal) and positive agency (lack of anhedonia). PID-5 Meanness items primarily reflect traits of callousness and grandiosity from the domain of Antagonism (12 items), and traits of restricted affectivity, hostility, and (low) emotional lability from the domain of Negative Affect (7 items). The PID-5 Meanness scale also includes two items from the Detachment domain, reflecting social withdrawal and intimacy avoidance. Lastly, PID-5 Tri-Disinhibition is composed primarily of items reflecting impulsivity, irresponsibility, and risk-taking from the PID-5 domain of Disinhibition (13 items), along with two items each reflecting traits of hostility, suspiciousness, and deceitfulness (from domains of Negative Affect, Detachment, and Antagonism, respectively).

Table 1.

Items Comprising PID-5-Triarchic Scales

PID-5-Tri Scale	PID-5 Domain	PID-5 Trait	PID-5 Item Number
Boldness	Antagonism	Attention Seeking	111,211
	Antagonism	Grandiosity	65
	Antagonism	Manipulativeness	107,180
	Negative Affect	Anxiousness	95 (−), 96, 130 (−)
	Negative Affect	Submissiveness	15(−),202(−)
	Disinhibition	Risk-Taking	7 (−), 87 (−), 195
	Detachment	Anhedonia	155
	Detachment	Withdrawal	186 (−)
Meanness	Antagonism	Callousness	13,54, 72, 73,90 (−),
			153, 166, 183,200,
			207,208
	Antagonism	Grandiosity	40
	Detachment	Intimacy Avoidance	97(−)
	Detachment	Withdrawal	10
	Negative Affect	Hostility	116
	Negative Affect	Restricted Affectivity	8,84,91, 167, 184
	Negative Affect	Emotional Lability	102 (−)
Disinhibition	Disinhibition	Impulsivity	4, 16, 17, 58(−), 204
	Disinhibition	Irresponsibility	31, 129, 156, 160,
			171, 201, 210 (−)
	Disinhibition	Risk-Taking	3
	Negative Affect	Hostility	28,158
	Detachment	Suspiciousness	103,190
	Antagonism	Deceitfulness	126, 134

Note: Items are keyed in the direction of higher scores on triarchic dimensions. (−) indicates reverse-scored item.

Psychometric Properties

Intercorrelations among the final PID-5-Tri scales are presented, along with basic descriptives (Ms/SDs) and internal consistencies for each individual scale, in Table B of the Online Supplement. The PID-5-Tri scales demonstrated acceptable to good internal consistencies, with Cronbach’s alphas (α) ranging from .83 – .89 in the development sample and from .66 – .89 in the validation sample. Within the development sample, PID-5 Meanness showed a moderate positive correlation (.48) with PID-5 Tri-Disinhibition and a modest positive r (.20) with PID-5 Boldness, with PID-5 Tri-Disinhibition showing a weaker, nonsignificant positive r with PID-5 Boldness. In the validation sample, PID-5 Meanness also showed significant positive correlations with PID-5 Tri-Disinhibition and Boldness (.51 and .27, respectively), with PID-5 Tri-Disinhibition also showing a significant positive r (.26) with PID-5 Boldness. The observed intercorrelations among the PID-5-Tri scales appear similar to those reported for alternative scale measures of the triarchic constructs (Drislane & Patrick, 2017), albeit with a somewhat stronger positive r between PID-5 Boldness and Tri-Disinhibition than is generally observed (for an exception, see Drislane et al., 2015).

External Validation: Development Sample

Relations with Alternative Triarchic Measures.

Within the development sample, the PID-5-Tri scales evidenced expected patterns of convergent and discriminant validity with other measures of the triarchic constructs (see Table 2). At the simple bivariate level, the PID-5-based triarchic scales showed strong convergence with counterpart TriPM (rs = .67 - .79) and MPQ-Tri (rs = .68 - .76) scales. Scores on the PID-5-Tri scales also displayed small positive rs with non-counterpart scales of other triarchic measures; in particular, scores on PID-5 Tri-Disinhibition were moderately correlated with Meanness scores for the TriPM and the MPQ. However, within regression models using the PID-5-Tri scales as predictors of TriPM and MPQ-Tri scales, robust predictive associations were observed only for counterpart scales (βs = .62 - .82), with the exception of MPQ-Meanness, which showed a small but significant unique predictive relationship with PID-5 Tri-Disinhibition (β = .18, p<.01).

Table 2.

Development Sample (N=210): Pearson rs (& regression βs) for PID-5 Triarchic Scales with Criterion Measures

	PID-5 Triarchic Scale
	Boldness r(β)	Meanness r(β)	Disinhibition r(β)	Full Model R(R2)
TriPM
Boldness	.79* (.79*)	.17 (.05)	.07 (−.06)	.79* (.62)
Meanness	.24* (.12)	.70* (.63*)	.42* (.09)	.71* (.50)
Disinhibition	.16 (.09)	.26* (−.10)	.67* (.71*)	.68* (.46)
MPQ-Tri
Boldness	.76* (.75*)	.18 (.05)	.06 (−.05)	.76* (.58)
Meanness	.05 (−.09)	.68* (.62*)	.47* (.18*)	.71* (.50)
Disinhibition	.08 (.02)	.24* (−.16*)	.74* (.82*)	.76* (.58)
Trait Fear Inventory	−.79* (−.78*)	−.20* (−.06)	−.11 (.02)	.79 (.62)
Substance Use Problems
Short Drug Abuse Screening Test	.16 (.12)	.16 (−.08)	.42* (.44*)	.44* (.19)
Alcohol Dependence Scale	.25* (.22*)	.14 (−.07)	.35* (.36*)	.42* (.18)
Positive and Negative Affect Schedule
Positive Affect	.49* (.53*)	−.13 (−.11)	−.21* (−.22*)	.57* (.32)
Negative Affect	−.36* (−.40*)	.09 (−.07)	.38* (.46*)	.56* (.31)
State-Trait Anxiety Inventory	−.47* (−.52*)	.12 (−.02)	.38* (.45*)	.64* (.41)
STAXI Total Score	−.06 (−.11)	.24* (.11)	.34* (.31*)	.37* (.14)
Beck Depression Inventory	−.08 (−.14)	.25* (.02)	.50* (.51*)	.52* (.27)

Note. r = Pearson correlation coefficient; β = standardized beta coefficient from regression model incorporating scores on the three PID-5 triarchic scales as predictors; Full Model R = multiple R for regression model. TriPM= Triarchic Psychopathy Measure; MPQ = Multidimensional Personality Questionnaire; STAXI= State-Trait Anger Expression Inventory.

^*p < .01.

To further examine the convergence between scores on the PID-5 triarchic scales with other operationalizations of the triarchic constructs, we evaluated the fit of a correlated three-factor model (Drislane & Patrick, 2017) in which higher-order factors of Boldness, Meanness, and Tri-Disinhibition were modeled using the PID-5-Tri scales along with the TriPM and MPQ-Tri scales as indicators. A modified 3-factor model provided adequate fit to the data and results for this model (shown in Table C of the Online Supplement, which also includes results for alternative 1- and 2-factor models, and supplemental Figure A, which depicts the model schematically) indicated that the PID-5-Tri scales loaded at comparable levels with their counterpart (TriPM, MPQ-Tri) scales on latent triarchic dimensions, further supporting their effectiveness as indicators of the triarchic model constructs.

Relations with Internalizing and Externalizing Problems.

Table 2 also presents results from analyses examining associations between PID-5-Tri scales and measures of dispositional fear (cf. Kramer et al., 2012), substance use, and emotionality. PID-5 Boldness demonstrated a strong negative association with Trait Fear inventory scores both at the zero-order level and in the context of regression. PID-5 Meanness also showed a modest negative r with Trait Fear, but did not predict unique variance in this measure. With respect to substance abuse, PID-5 Tri-Disinhibition was the only unique predictor of self-reported drug abuse, whereas both PID-5 Boldness and Tri-Disinhibition emerged as unique, positive predictors of self-reported alcohol abuse (Nelson et al., 2016).

For self-report emotionality measures (i.e., positive and negative affect, anxiety, anger, and depression; see Table 2, lower part), PID-5 Boldness and Tri-Disinhibition showed opposing directional associations, with both rs and regression Bs significant in 3 of 5 cases for Boldness (i.e., PANAS Positive Affect [+] and Negative Affect [–], and STAI trait anxiety [–]) and in all cases for Tri-Disinhibition. By contrast, PID-5 Meanness showed significant rs only with STAXI trait anger and BDI depression, which were rendered nonsignificant in regression models.

Validation Sample

Relations with Psychopathy Measures.

As in the development sample, PID-5 Boldness emergedas the only unique predictor of TriPM Boldness, and PID-5 Tri-Disinhibition was the only predictor of unique variance in TriPM Tri-Disinhibition, when accounting for overlap among the PID-5-Tri scalesin the context of regression. However, contrary to hypotheses and findings within the development sample, all three PID-5-Tri scales displayed moderate to high positive correlations with scores on TriPM Meanness, and each predicted unique variance in this scale; however, the magnitude of prediction was strongest for PID-5 Meanness. In examining relations between scores on the PID-5-Tri scales and PPI-Tri scales, each PID-5-Tri scale emerged as the strongest unique predictor of its counterpart PPI-Tri scale (βs = .60 - .76 for counterpart scales and −.16 to +.16 for non-counterpart scales).

The availability of the PPI–R in the validation sample also allowed for examination of relations between PID-5-Tri scales and the PPI’s standard factors (Lilienfeld & Widows, 2005). In the context of regression, PPI-R Fearless Dominance scores were associated uniquely and to a strong positive degree with PID-5 Boldness. All three PID-5-Tri scales showed significant positive correlations with PPI-R Self-Centered Impulsivity scores at the zero-order level; however, PID-5 Tri-Disinhibition emerged as the sole unique predictor of variance in PPI-R Self-Centered Impulsivity when accounting for overlap among the PID-5-Tri scales through regression. Consistent with hypotheses, PID-5 Meanness displayed a moderate positive association with scores on PPI-R Coldheartedness, with PID-5 Boldness showing a more modest positive r. The correlation for PID-5 Tri Disinhibition was negligible. In the context of regression, the positive predictive association for Meanness with PPI-R Coldheartedness increased somewhat, whereas the association for PID-5 Boldness decreased to some degree, and the association for Tri-Disinhibition became significantly negative, indicating the presence of a suppressor effect at the zero-order level.

All three PID-5-Tri scales showed positive bivariate rs with scores on the LSRP as a whole and its Egocentric and Factor 1 (“Primary”) subscales (see Table 3). However, in regression models, only PID-5 Meanness and Tri-Disinhibition evidenced unique prediction of these LSRP scales. For the LSRP’s Antisocial and Factor 2 (“Secondary”) subscales, PID-5 Meanness and Tri-Disinhibition (but not Boldness) showed significant associations at the zero-order level, with Tri-Disinhibition emerging as the only unique predictor in regression analyses. The Callous scale of the LSRP showed significant rs and βs with both PID-5 Tri-Disinhibition and Meanness.

Table 3.

Validation Sample (N=192): Pearson rs (& regression βs) for PID-5 Triarchic Scales with Psychopathy Related Criterion Measures

	PID-5 Triarchic Scale
	Boldness r(β)	Meanness r(β)	Disinhibition r(β)	Full Model R(R2)
TriPM
Boldness	.74* (.78*)	.19* (.04)	.11 (−.12)	.74* (.55)
Meanness	.43* (.20*)	.75* (.54*)	.61* (.27*)	.82* (.66)
Disinhibition	.23* (.04)	.39* (.02)	.73* (.70*)	.73* (.54)
PPI-Tri
Boldness	.74* (.72*)	.14 (−.07)	.17 (.00)	.74* (.55)
Meanness	.33* (.16*)	.68* (.60*)	.41* (.07)	.70* (.49)
Disinhibition	.19* (.03)	.23* (−.16*)	.69* (.76*)	.70* (.50)
PPI-R
Total	.63* (.48*)	.50* (.13*)	.68* (.51*)	.83* (.69)
Fearless Dominance	.78* (.77*)	.17 (−.10)	.27* (.11)	.79* (.62)
Self-Centered Impulsivity	.30* (.09)	.47* (.07)	.79* (.72*)	.80* (.64)
Coldheartedness	.23* (.16)	.45* (.57*)	.02 (−.33*)	.54* (.29)
LSRP
Total (19 item)	.24* (−.01)	.62* (.32*)	.77* (.61*)	.81* (.66)
Egocentric	.28* (.06)	.64* (.41*)	.65* (.42*)	.74* (.55)
Callous	.15 (−.01)	.44* (.30*)	.42* (.25*)	.50* (.25)
Antisocial	.09 (−.10)	.32* (−.05)	.73* (.74*)	.74* (.54)
Factor 1 (26 item)	.26* (.04)	.65* (.44*)	.64* (.42*)	.74* (.55)
Factor 2 (26 item)	.18 (−.03)	.40* (.00)	.79* (.70*)	.79* (.62)
IRI
Perspective Taking	−.12 (.03)	−.44* (−.31*)	−.40* (−.23*)	.49* (.24)
Fantasy	−.17 (−.15)	−.21* (−.26*)	−.01 (.17)	.28* (.08)
Empathic Concern	−.24* (−.06)	−.68* (−.65*)	−.33* (.04)	.69* (.47)
Personal Distress	−.49* (−.53*)	−.18 (−.24*)	.15 (.41*)	.61* (.37)
Fear Questionnaire	−.39* (−.40*)	−.14 (−.07)	−.09 (.05)	.40* (.16)
ABQ	.29* (.16)	.33* (.13)	.41* (.28*)	.44* (.16)
History of Alcohol Use	.28* (.15)	.22* (.01)	.40* (.34*)	.41* (.19)
History of Marijuana Use	.23* (.15)	.22* (.01)	.34* (.29*)	.36* (.13)
History of Other Drug Use	.18 (.14)	.14 (−.07)	.28* (.19*)	.27* (.07)

Note. r = Pearson correlation coefficient; β = standardized beta coefficient from regression model incorporating scores on the three PID-5 triarchic scales as predictors. TriPM= Triarchic Psychopathy Measure, PPI-Tri= PPI based triarchic scales, PPI-R = Psychopathic Personality Inventory-Revised, LSRP = Levenson Self Report Psychopathy Scales, IRI = Interpersonally Reactivity Index, ABQ = Antisocial Behavior Questionnaire, PID-5 = Personality Inventory for the DSM-5.

^*p < .01.

Relations with Empathy and Fear

. Consistent with hypotheses, scores on PID-5 Meanness showed significant negative associations with scores on all four subscales of the IRI, a self-report index of empathy (see Table 3), with the most robust association observed for the Empathic Concern subscale (r/β = −.68 / −.65). After accounting for their overlap with PID-5 Meanness in the context of regression, PID-5 Boldness and Tri-Disinhibition displayed more selective associations with particular facets of empathy as measured by the IRI. PID-5 Tri-Disinhibition contributed uniquely to prediction of scores on IRI Perspective Taking (in a negative direction), whereas PID-5 Boldness and Tri-Disinhibition evidenced opposing relations with IRI Personal Distress (negative and positive, respectively). Findings for fearfulness as indexed by the Fear Questionnaire were similar to those for the Trait Fear inventory in the development sample: A moderate negative bivariate association was evident for PID-5 Boldness, which remained robust in the regression model, together with a weaker negative r for PID-5 Meanness (−.14, p = .048), which became even weaker in the regression model.

Relations with Externalizing Problems.

Associations between scores on the PID-5-Tri scales and measures of antisocial behavior and substance use problems are also presented in Table 3. All PID-5-Tri scales showed significant zero-order rs with self-reported levels of antisocial behavior as measured by the ABQ, with regression analysis results indicating the greatest magnitude of unique prediction for PID-5 Tri-Disinhibition. Similarly, at the bivariate level, all three PID-5-Tri scales were positively related to alcohol, marijuana, and drug use as assessed by interview. However, in the context of regression, Tri-Disinhibition emerged as the only unique predictor of marijuana and other drug use, whereas alcohol use history was positively predicted by both PID-5 Tri-Disinhibition and (to a more modest degree) PID-5 Boldness.

Discussion

This study was undertaken to provide more effective representation of psychopathy in terms of the DSM-5 AMPD dimensional-trait system. This was done by identifying item sets from the purpose-built assessment instrument for the AMPD, the PID-5 inventory, that provide effective measurement of the dimensions of the triarchic model (Patrick et al., 2009; Patrick & Drislane, 2015), a conceptualization that draws upon historic and contemporary writings to characterize psychopathy in biobehavioral trait terms. Our PID-5 based scale measures of these dimensions, developed using a construct rating approach, demonstrated patterns of convergent and discriminant validity with conceptually-relevant criteria (including measures of psychopathy, externalizing problems, internalizing problems, and personality) across two independent samples that conformed largely with predictions based on findings for other triarchic scale measures in previous work (see, e.g., Patrick & Drislane, 2015; Drislane et al., 2018; Sellbom et al., 2016). As such, these scales provide a means for quantifying distinct facets of psychopathy within the formal psychiatric nosology and facilitating research directed at interfacing DSM-based psychopathy assessments with neural systems concepts and measures (cf. Patrick & Drislane, 2015).

Addressing Criticisms of the DSM-5 AMPD Characterization of Psychopathy

The current study also sought to improve the DSM-5 AMPD characterization of psychopathy by addressing concerns raised by Crego and Widiger (2014) about the AMPD psychopathy specifier and the discriminant validity of the PID-5 Antagonism and Disinhibition domains. We addressed the issue of reversed traits in the psychopathy specifier (intended to index boldness) by including both positively and negatively worded items in each PID-5-Tri scale (including Boldness), and coded each such that higher scores reflect the presence of the target trait rather than the absence of an opposing trait. Nevertheless, the majority of PID-5-Boldness items from the domains of Negative Affect (6 of 8) and Detachment (1 of 2) continue to require reverse-coding, highlighting an inherent limitation of the PID-5 item set – that it does not appear to index certain traits (e.g., anxiousness vs. stress immunity) in a sufficiently bipolar manner (Crego & Widiger, 2014). However, the current work allows boldness to be assessed within the AMPD framework in a more psychometrically sound way, which can help to facilitate research aimed at addressing ongoing debates in the literature regarding the relevance of boldness to psychopathy (Crego & Widiger, 2014, 2016; Lilienfeld et al., 2012; Lynam & Miller, 2012; Miller, Lamkin, Maples-Keller, & Lynam, 2016). Of relevance to these debates, PID-5 Boldness evidenced modest positive associations with antisocial behavior and alcohol use across separate samples. It may be the case that inventories such as the PID-5 that contain items generally indicative of pathology will yield boldness scales with more deviant content (for a discussion of this point, see Drislane et al., 2015). Investigators primarily interested in predicting externalizing outcomes may wish to consider using the PID-5 measures of the triarchic model constructs, as PID-5 Boldness appears to predict both adaptive and maladaptive outcomes.

Regarding the issue of discriminant validity of the PID-5 Antagonism and Disinhibition domains raised by Crego & Widiger (2014), the construct rating approach to scale development we used (see, e.g., Drislane et al, 2018; Hall et al., 2014) emphasized selection of items that maximally differentiated the three triarchic dimensions while eliminating construct-irrelevant content. This allowed for selection of items of maximal construct relevance without being constrained by traits as specified in the AMPD or as configured in the PID-5 (Anderson et al., 2014; Crego & Widiger, 2014). Whereas scores on PID-5 Antagonism and Disinhibition are highly correlated (r > .7; Crego & Widiger), PID-5-Tri Meanness and Disinhibition are only moderately correlated (rs = .48 and .51 for the development and validation samples, respectively), consistent with other triarchic model operationalizations (Drislane et al., 2017), and they exhibit distinct criterion correlates. PID-5 Meanness was preferentially associated with measures of callousness and (lack of) empathy, whereas PID-5 Tri-Disinhibition was preferentially associated with measures of disconstraint, non-aggressive externalizing, and substance abuse – and exclusively associated with negative emotionality and internalizing measures. Nevertheless, the moderate correlations observed between PID-5 Tri-Disinhibition and Meanness scales from the PID-5, TriPM, and MPQ suggest some overlap in what these dimensions measure. Meanness and disinhibition are critically important themes in the psychopathy literature (Patrick et al., 2009) and likely have unique etiological influences, as they display distinct neurophysiological correlates (Brislin et al., 2018; Venables et al., 2018). As such, it will be important to continue exploring how best to operationalize and further differentiate these constructs in both adults in children. Given that multiple measures of the triarchic dimensions have now been developed (Drislane et al., 2017), one way to address this issue would be to use IRT analyses to evaluate properties of items composing alternative scales and identify items that best differentiate meanness and disinhibition.

Content Coverage of PID-5-Tri Scales

With respect to content coverage, many of the items included in the PID-5-Tri scales were drawn from traits specified as diagnostic of ASPD in the DSM-5 AMPD and those associated with the psychopathy specifier. Nearly half of the PID-5 Meanness items come from the Callousness scale of the PID-5 (Antagonism domain), and PID-5 Tri-Disinhibition contains items from each of the Disinhibition domain traits specified for ASPD (Impulsivity, Irresponsibility, Risk-Taking). Likewise, PID-5 Boldness contains items from scales indexing the three traits of the AMPD psychopathy specifier (Anxiousness, Attention-Seeking, Withdrawal). However, the PID-5-Tri scales also contain representation of traits apart from those specified for ASPD and psychopathy in the AMPD. For example, Boldness includes items from the PID-5 Risk-Taking scale that tap venturesomeness, Meanness includes items from PID-5 Restricted Affectivity reflecting unemotionality, and Tri-Disinhibition includes items from the PID-5 Suspiciousness scale that index alienation. The composition of the PID-5-Tri scales also suggests that the Risk-Taking scale of the PID-5 may be misplaced within the Disinhibition domain, as this scale primarily contains items that index fearless adventure-seeking (i.e., boldness-related content) with lesser representation of items that index reckless risk-taking behaviors (i.e., disinhibition-related content).

While some previous studies have demonstrated robust, fairly straightforward domain-level associations between scores on the TriPM and PID-5 (i.e., TriPM Meanness with PID-5 Antagonism, TriPM Disinhibition with PID-5 Disinhibition; Crego & Widiger, 2014), each of the PID-5-Tri scales includes a mix of items from different trait domains of the PID-5, rather than mapping neatly onto one PID-5 domain or another. These findings are consistent with recent work demonstrating that triarchic scales developed from the NEO-PI-R reflect a blending of five-factor model (FFM) domains (Drislane et al., 2018). Likewise, the UPPS model of impulsivity conceptualizes this attribute in terms of expressions of conscientiousness, extraversion, and neuroticism rather than conscientiousness alone (Whiteside & Lynam, 2001), and meta-analytic work has shown that proneness toward externalizing problems is associated with multiple FFM domains and trait facets (Ruiz, Pincus, & Schinka, 2008).

Considering that the PID-5 is a maladaptive variant of the FFM (Gore & Widiger, 2013), these studies highlight key differences between FFM conceptualizations and the triarchic model constructs. Specifically, the dimensional constructs of the triarchic model were inferred from the extant literature on models and measures of psychopathy and their known correlates, including neurobiological correlates. The construct of boldness, for example, draws on findings for the Fearless Dominance factor of the PPI, including its distinctive associations with physiological indices of threat reactivity (e.g., Dvorak-Bertsch, 2009), along with a structural model of scale measures of fear and fearlessness known to predict aversive startle potentiation (Kramer et al., 2012). Constructs of meanness and disinhibition derive in part from literatures on callous-unemotional traits (Kimonis et al., 2015) and general externalizing proneness (Krueger et al., 2007) and their distinctive neurophysiological correlates (e.g., Marsh et al., 2008; Venables et al., 2018). The FFM, by contrast, is a lexical model that was not directly informed by neurobiological research findings. This does not represent a shortcoming on the part of the FFM, which has proven highly effective for a wide range of investigative purposes. However, as we have discussed elsewhere (e.g., Patrick & Drislane, 2015; Patrick, Iacono, & Venables, 2019), efforts to characterize relations between the neurobehavioral traits of the triarchic model and FFM lexical traits are likely to be of value for interfacing FFM traits more effectively with neurobiological systems and measures.

Implications for Representing Psychopathy within the DSM-5 AMPD

The representation of psychopathy in DSM-5 marks a significant moment in the history of psychiatric nosology, considering that psychopathy has not been formally included in prior editions of the DSM despite decades of research documenting its clinical importance (Crego & Widiger, 2015; Skeem, Polaschek, Patrick, & Lilienfeld, 2011). The alternative dimensional approach to characterizing ASPD in DSM-5 has the potential to advance developmental theories of antisocial behavior. Compared with traditional behavioral criteria for ASPD that primarily index disinhibitory tendencies (Venables et al., 2014; Wall et al, 2015), the trait-based AMPD definition of this condition provides balanced coverage of interpersonal-affective and behavioral deviance features of psychopathy (Anderson et al., 2013; Few et al., 2015; Strickland et al., 2013; Wygant et al., 2016). Along similar lines, the Section II diagnosis of childhood conduct disorder (CD) now includes a “limited prosocial emotions” (LPE) specifier for indexing callous-unemotional tendencies emphasized within the child psychopathy literature (Kimonis et al., 2015). A valuable direction for future research will be to examine relations between the LPE specifier and PID-5 Meanness, cross-sectionally (in younger-aged literate samples) and longitudinally (in participants tested at both younger and older ages), in order to clarify how childhood callousness relates to adult meanness and to what extent these dimensions can be considered developmental counterparts.

Further, the current study demonstrates how callous-unemotional and impulsive-externalizing proclivities can be measured as distinct dimensions via the PID-5 Meanness and Tri-Disinhibition scales. In addition to a dimensional measure of LPE, it will be valuable to consider how disinhibitory features of CD can be represented dimensionally in future editions of the DSM. Representing disinhibition and meanness as distinct constructs within the DSM will allow for more nuanced investigations on the developmental trajectories and etiological bases of these dimensions, and may point to further areas of convergence and divergence in these components of the triarchic model. Further, the phenotypic expression of boldness or fearlessness in children and youths and how it relates to adult psychopathic traits remains an understudied area (Klingzell et al., 2016).

Limitations and Future Directions

Our results need to be interpreted in light of certain limitations. First, the two samples consisted of adults from the community. Although efforts were made to enhance representation of participants with elevated psychopathic traits, the validity of the PID-5-Tri scales should be further evaluated in correctional or forensic samples where levels of deviancy are higher. Further, the study’s focus on adults limits our ability to directly address developmental questions about psychopathy. Future studies that include adolescents would inform the extent to which the PID-5-Tri scales function similarly in younger samples and allow for a more direct comparison of youth and adult conceptions of antisocial behavior disorders. Ultimately, longitudinal studies spanning a large developmental range will be needed to elucidate how the triarchic model constructs change over time and how they intersect with other psychological constructs, such as limited prosocial emotions, for example.

A further limitation is that criterion measures in the current study were limited to self-report inventories. As a consequence, shared method variance between the PID-5-Tri scales and criterion measures may have inflated some of the observed associations. On a broader level, it will be important for future studies utilizing the PID-5-Tri scales to include criteria from other measurement domains, such as task performance and physiological response. For example, scores on PID-5 Tri-Disinhibition would be expected to correlate with reduced P3 brain response and impaired performance on executive functioning tasks (Venables et al., 2018).

These findings also point to some limitations of the PID-5 (i.e., high intercorrelations among trait scales and domains, insufficient bipolarity of some scales, potential misplacement of Risk-Taking in the domain of Disinhibition). More broadly, we acknowledge that any effort to develop triarchic scales from existing inventories will be limited by the content coverage and quality of their constituent items. Nevertheless, the current work is valuable in terms of providing for improved characterization of psychopathy in AMPD trait terms. This work can facilitate ongoing research on psychopathy within existing studies that have administered the PID-5, including studies of clinical patients and large-scale normative datasets. Related to this, the PID-5 scale measures of the triarchic model constructs developed in the current work can provide a basis for interfacing personality pathology as represented in the AMPD with neurobiological concepts and measures (Patrick & Drislane, 2015).