Professional Documents
Culture Documents
The "Core" of The Dark Triad - A Test of Competing Hypotheses (Vize Et Al, 2019)
The "Core" of The Dark Triad - A Test of Competing Hypotheses (Vize Et Al, 2019)
Colin E. Vize
Katherine L. Collison
Purdue University
Joshua D. Miller
University of Georgia
Donald R. Lynam
Purdue University
Citation: Vize, C.E., Collison, K. L., Miller, J. D., & Lynam, D. R. (in press). The “core” of the
Dark Triad: A test of competing hypotheses. Personality Disorders: Theory, Research, and
Treatment.
Abstract
As research on the Dark Triad (DT; the interrelated constructs of Machiavellianism, narcissism,
and psychopathy) has accumulated, a subset of this research has focused on explicating what
traits may account for the overlap among the DT members. Various candidate traits have been
two as a set). The present study sought to test the leading candidates against one another in their
ability to account for the shared variance among the DT members. Using a pre-registered
analytical plan, we found that Agreeableness (as measured by the IPIP-NEO-120), Honesty-
Humility from the HEXACO, and the SRP-III subscales of Callous Affect and Interpersonal
Manipulation accounted for all or nearly all of the shared variance among the DT members. BFI-
based measures of Agreeableness (BFI and BFI-2) accounted for notably less variance in most
cases. The results were consistent across two large samples (Ns of 627 and 628), and across
various DT measurement approaches. We argue that the most parsimonious explanation for
findings on the core of the DT is that such traits all fall under the umbrella of Antagonism.
Although antagonistic personality traits are common to each DT construct, each DT construct is
purported to possess traits that serve to distinguish each component from the others. Narcissism
psychopathy and Machiavellianism (e.g., Vize, Collison, Miller, & Lynam, 2018). In theory,
psychopathy is distinguished from the other two members of the DT by its inclusion of
disinhibitory traits, and research has shown that it is also most strongly related to Antagonism-
aligned traits (Muris, Merckelbach, Otgaar, & Meijer, 2017). It has been argued that its inclusion
of disinhibitory traits in addition to Antagonistic traits are the reasons that psychopathy tends to
show the most robust relations with various antisocial outcomes (Vize, Miller, & Lynam, 2018).
Machiavellianism is thought to differ from narcissism and psychopathy to the extent that
Machiavellian individuals are more strategic and planful in their deployment of antisocial tactics
(Jones & Paulhus, 2017). Though the theoretical profile of Machiavellianism serves to
differentiate it from the other two DT components, the empirical profile of Machiavellianism
Machiavellianism that would distinguish the construct from psychopathy (Miller et al., 2017;
McHoskey, Worzel, & Szyarto, 1998; Vize, Lynam, Collison, & Miller, 2018).
rapidly since the introduction of the practice in 2002. The present study is focused on a
subdomain of DT research which has sought to elucidate the traits that serve as the “core” of the
1
Throughout the manuscript, when we refer to narcissism we are referring to the grandiose variant since itis
typically the focus of DT research.
THE CORE OF THE DARK TRIAD 4
DT. More specifically, a handful of researchers have aimed to explicitly identify which traits
account for the positive correlations observed among the DT constructs. Various candidate
constructs have been examined in this literature area. Because these candidates are closely
related to one another, it is not clear whether any of the proposed constructs are better in
explaining the shared variance among the DT or if all perform equally well. Additionally, there
has not been a consistent methodology used in this research area, making comparisons across
studies somewhat difficult. Thus, we aim to consolidate this area of DT research, and directly
compare previously proposed candidates against one another in hopes of providing a more
parsimonious understanding of what traits can explain the communality among the DT
constructs.
A number of related constructs have been hypothesized to account for the overlap among
the DT. These candidates include Agreeableness as assessed by the Big Five Inventory (BFI;
Goldberg, 1999) (Paulhus & Williams, 2002), Agreeableness (vs. Antagonism) as assessed by
the NEO-Personality Inventory-Revised (NEO-PI-R; Costa & McCrae, 1992; Stead, Fekken,
Kay, & McDermott, 2012), Honesty-Humility from the HEXACO model of personality (Lee &
Ashton, 2004; Book et al., 2016; Hodson et al., 2018), traits related to interpersonal manipulation
and callousness (Jones & Figueredo, 2013; Marcus, Preszler, & Zeigler-Hill, 2018), and an
exploitative, fast life history strategy which has direct ties to evolutionary theory (Jonason,
Kaufman, Webster, & Geher, 2013; Jonason, Koenig, & Tost, 2010). There has been evidence
provided for each of these candidates, though the evidence for the fast life history strategy as the
individual difference that unites the DT constructs is less compelling as the DT constructs are
less strongly related to measures of fast life strategy than to other candidates such as Honesty-
THE CORE OF THE DARK TRIAD 5
Humility and Agreeableness (Muris et al., 2017). In addition, fast life strategies do not clearly
map onto the theoretical content of Machiavellianism, where the construct is purportedly made
up of some traits that are not in line with a fast life strategy (e.g., being calculating and planful).
assessed by their respective subscales within the Self-Report Psychopathy Scale (SRP; Williams,
Paulhus, & Hare, 2007), have been shown to be strongly related to all DT constructs. For
example, Jones and Figueredo (2013) first estimated a latent DT factor and then examined
whether specific subscales from the Narcissistic Personality Inventory (NPI; Raskin & Hall,
1979), SRP, and MACH-IV (Christie & Geis, 1970) could account for the largest amount of
variance in the latent DT factor. The Callous Affect and Interpersonal Manipulation subscales
from the SRP outperformed all other subscale combinations, accounting for 79% of the variance
in the latent DT factor compared to other subscales (average R2 = .55). Using network analyses in
two large samples, Marcus, Preszler, and Zeigler-Hill (2018) found that the Callousness and
Interpersonal Manipulation subscales were the most central nodes among DT subscales,
complementing the results of Jones and Figueredo (2013). Relatedly, when attempting to
estimate a latent DT factor, Glenn & Sellbom (2015) found that the latent factor could not be
estimated because the entirety of the shared variance among the DT constructs was already
attributed to psychopathy.
Humility traits) as comprising the core of the DT. Using canonical correlations across two
samples, Book, Visser, and Volk (2015) found that the HEXACO model accounted for a
significantly larger amount of variance in the DT measures compared to a BFI-based Big Five
model and also a fast life history model. The results also showed that adding the Callous Affect
THE CORE OF THE DARK TRIAD 6
subscale of the SRP into the model with the HEXACO did not improve performance. Using the
same methodological approach in a separate study, a similar pattern of findings emerged (i.e., the
HEXACO model was superior to other models) even when including sadism with the other DT
latent DT factor from subscales of the Short Dark Triad (SD3; Jones & Paulhus, 2014) and a
latent Honesty-Humility factor from the facet scales of Sincerity, Fairness, Greed Avoidance,
and Modesty. The latent correlation between the factors was nearly perfectly inverse (r = -.95)
and this strong inverse relation was also observed when relying on peer-reports in smaller
samples.
In sum, various researchers have attempted to investigate which traits can explain the
overlap among DT constructs. There is evidence that can be drawn on to support many of the
candidates as accounting for the core of the DT. In cases where candidates are tested against one
Agreeableness), the tests have often focused on Agreeableness as assessed by the BFI, which
likely biases the test against Agreeableness given that the BFI contains significantly less content
related to honesty and modesty (Miller, Gaughan, Maples, & Price, 2011). Indeed, Vize, Miller,
and Lynam (2019) found that the amount of overlap accounted for by Agreeableness among the
used.
Given that there is evidence that various traits can account for the communality among
the DT, it is necessary to more directly test competing hypotheses against one another. Thus,
THE CORE OF THE DARK TRIAD 7
there are three primary questions the current study seeks to answer. First, when using the same
methodology, is there a specific set of traits that consistently outperforms other traits in
accounting for the shared variance of the DT? Second, in regard to assessments of
Agreeableness, does the assessment of Agreeableness matter (i.e., BFI instruments vs. NEO
instruments) when trying to account for the maximum amount of variance in the DT? Last, does
the assessment approach of the DT influence results on what traits can account for the
interrelations among the DT? This last question has received little attention despite the fact that
research has shown important differences among popular assessments of the DT (Miller et al.,
2012).
Method
All study procedures (i.e., participant inclusion/exclusion criteria and planned analyses)
were preregistered using the AsPredicted.org template. The preregistered protocol can be found
here: https://osf.io/xhbqg. The protocol was preregistered before data collection occurred.
platform. After the initial posting of the HIT, a total of 230 participants’ work was rejected based
on our exclusion criteria (i.e., completion time, attention checks, and re-captcha items). The HIT
was made available again, and 22 out of the 230 participants’ work was rejected based on our
exclusion criteria resulting in N=1,378. Participants were paid $2.00 for completing the protocol.
Participant data were then inspected for invalid responding using preregistered cutoff scores on
the Infrequency and Too Good to Be True validity scales of the Elemental Psychopathy
Assessment (EPA; Lynam et al., 2013). An additional 123 participants were excluded resulting
in a final sample of N=1,255. The average age in the final sample was 38.95 yrs. old (SD=11.88)
THE CORE OF THE DARK TRIAD 8
and the sample was predominantly Euro American (79.3%). The sample was 41% male. As
planned in the study preregistration, the final sample was then split randomly into two samples
(Sample 1 N = 627; Sample 2 N = 628) so that all analyses could be replicated across samples.
Measures
Short Dark Triad (SD3). The SD3 (Jones & Paulhus, 2014) is a 27-item self-report
measure of the Dark Triad. Nine-item subscales assess each construct within the Dark Triad
(Machiavellianism, narcissism, and psychopathy). Each item was rated along a 5-point scale
ranging from 1-“Strongly Disagree” to 5-“Strongly Agree”. The internal consistency of the
subscales was similar in both samples (Sample 1 α range = .82 to .85; Sample 2 α range = .80
to .83).
Dirty Dozen (DD). The DD (Jonason & Webster, 2010) is a 12-item measure of the Dark
Triad designed to efficiently assess each component of the DT. Each DT construct is measured
by 4-item subscales using a 5-point scale (1-“Strongly Disagree” to 5-“Strongly Agree”). The
internal consistency of the DD subscales was adequate in both samples (Sample 1 α range = .68
MACH-IV. The MACH-IV (Christie & Geis, 1970) is a 20-item self-report measure
designed to assess the core traits related to Machiavellianism. Although researchers have made
use of factor analyses to empirically identify subscales within the MACH-IV (e.g., Miller, Smart,
& Rechner, 2015), we only make use of the total score in the present study. The internal
consistency of the total score was adequate in both Sample 1 (α = .84) and Sample 2 (α = .82).
Narcissistic Personality Inventory (NPI-40). The NPI-40 (Raskin & Terry, 1979) is a
40-item forced choice self-report assessment instrument that primarily assesses the grandiose
variant of narcissism (Miller, Gaughan, Pryor, Kamen, & Campbell, 2009). Similar to the
THE CORE OF THE DARK TRIAD 9
MACH-IV, subscales of the NPI have been empirically identified (Ackerman et al., 2011) but we
only make use of the total score in the present study. The internal consistency of the NPI total
Self-Report Psychopathy Scale (SRP-III). The SRP-III (Williams & Paulhus, 2003) is a
64-item self-report assessment of psychopathy that was developed out of the conceptualization of
Revised (PCL-R; Hare, 2003). The SRP-III is composed of four subscales: Interpersonal
Manipulation, Callousness, Erratic Lifestyle, and Antisocial Behavior. The internal consistency
of the total score was excellent in both samples (Sample 1 and 2 α = .94) and the SRP subscales
also showed high internal consistency (Sample 1 α range = .81 to .87; Sample 2 α range = .82
to .87).
instrument that assesses the six domains of the HEXACO model of personality. In addition to the
domain scales, each domain includes four facet scales. In the present study, we only
administered the 16 items assessing the Honesty-Humility domain. The internal consistency of
the domain scores in both samples was adequate (Sample 1 α = .87; Sample 2 α = .88) as were
the four facet scales of Sincerity, Fairness, Greed-avoidance, and Modesty (Sample 1 α range
(Maples, Guan, Carter, & Miller, 2014) is a freely available, 120-item self-report measure
assessing the five domains of the Five-factor Model of personality (McCrae & Costa, 1987). The
IPIP-NEO-120 also allows for six facet scales to be computed for each domain. Although the
entire IPIP-NEO-120 was administered to the sample, we only make use of the Agreeableness
THE CORE OF THE DARK TRIAD 10
domain/facet scales in the present study. The internal consistency of the domain scores in both
samples was adequate (Sample 1 and 2 α = .87) as were the six Agreeableness facet scales of
Big Five Inventory (BFI). The BFI (Goldberg, 1999) is a 44-item self-report measure
that assesses each of the Big Five domains. Only the Agreeableness items (nine total) were
administered in the present study. The internal consistency of the Agreeableness total score was
Big Five Inventory-2 (BFI-2). The BFI-2 (Soto & John, 2017) is an expanded version of
the BFI, including a total of 60 items to assess each domain of the Big Five. In addition, the BFI-
2 includes three facet scales for each domain. We only administered the 12 Agreeableness items
from the BFI-2, which allows for a total domain score and three facet scales (Compassion,
both samples (Sample 1 and 2 α = .86) as did the three facet scales (Sample 1 α range = .69
Preregistered Analyses
relation between latent DT factors (estimated from either the SD3 subscales, the DD subscales,
or the MACH-IV/NPI/SRP-III scales) and latent factors estimated from subscales of the various
candidate traits purported to account for the core of the DT (i.e., Honesty-Humility, IPIP-
Callousness and Interpersonal Manipulation). A representative graphical model is shown for the
model involving Honesty-Humility facets and the SD3 subscales in Figure 1. Our primary
THE CORE OF THE DARK TRIAD 11
outcome of interest was the latent correlation and variance accounted for in the latent DT factors
by the respective candidate traits. The primary reason for making use of structural equation
methods to test for which traits best account for the overlap among the DT is because it is the
most straightforward and applicable test for our primary research question and is commonly used
to test such questions in this research area (e.g., Hodson et al., 2018; Jones & Figueredo, 2013).
preregistered analyses focusing on IPIP Agreeableness, we reported that we would first estimate
the correlation between the latent DT factor and a latent Agreeableness factor estimated from the
six IPIP-NEO Agreeableness facets. Next, we reported that we would estimate the same model,
but also include paths between the residuals of the Straightforwardness and Modesty facets of the
IPIP-NEO Agreeableness domain. Our goal in including this additional specification was to
determine whether additional variance in the latent DT factor could be determined by the content
unique to those specific facets given their theoretical overlap with the dishonest, manipulative
nature of DT constructs. However, a more appropriate analysis2 was substituted for these
preregistered analyses in which a latent Agreeableness factor was estimated from four
Agreeableness facet scales, excluding the Straightforwardness and Modesty facets. Next, a latent
Agreeableness factor was estimated using all facet scales to examine the contribution
Straightforwardness and Modesty add to accounting for the variance in the latent DT factor.
Even though the latter analyses were more appropriate for our research questions, code and
results for the originally planned analyses are available on the OSF page for the project.
Exploratory Analyses
2
Specifically, our preregistered analyses did not actually provide the estimates that tested our primary research
question. Thus, we undertook a revised analysis that allowed for us to examine the change in variance accounted for
in the latent DT factor when Straightforwardness and Modesty were added to the model as a different means of
testing the same research question.
THE CORE OF THE DARK TRIAD 12
Though a variety of studies have been conducted on the core of the DT, little work has
explored the degree to which item overlap between DT measures and candidate measures (e.g.,
HEXACO-Honesty Humility and NPI narcissism) may inflate the relations between constructs.
Thus, we sought to more thoroughly explore how much influence item overlap may have on the
relations between the latent DT factors and latent factors of Honesty-Humility, IPIP-
Preregistered Hypotheses
We had four primary hypotheses for the present study based on previous research in this
area and past work that has examined the personality content of the various candidate traits. We
note that Hypothesis 1b is no longer applicable given the changes to our preregistered analyses,
Hypothesis 1a. A latent factor estimated from the HEXACO-HH facets will show a near
perfect correlation (i.e., | r >.90 |) with a latent DT factor. A similar model will be tested using a
latent factor estimated from IPIP-NEO-A facet scales. We expect the IPIP-NEO-A latent model
to be strongly related to the latent DT factor (i.e., | r > .60 |), but it will account for a smaller
amount of variance in the latent DT factor compared to the variance accounted for in the latent
Hypotheses 1b. An additional model will be estimated using the IPIP-NEO-A facets, but
will also include two estimated paths between the residuals of the Straightforwardness and
Modesty facets to the latent DT factor. Using this model, we expect that the total variance
accounted for in the latent DT factor will be comparable to the variance accounted for in the
Hypothesis 2. The latent correlation between the latent DT factor and latent BFI-
smaller than the latent correlations observed for the latent H/H and IPIP-NEO-Agreeableness
factors. This will be due to the BFI instruments containing very little content that explicitly
Hypothesis 3: The latent correlation between the latent DT factor and latent SRP-Factor,
estimated using the two subscales of Interpersonal Manipulation and Callousness, will be large
(i.e., r >.60) but smaller than the latent correlations observed for IPIP-NEO-A and HEXACO-
HH due to the latter latent factors including more breadth of agreeableness-related traits.
Results
All data and R code needed to reproduce the results can be found at the OSF page for the
(means, standard deviations, and zero-order correlations of primary study variables) can also be
found on the OSF page. The ‘lavaan’ package (Rosseel, 2012) was used to conduct all primary
analyses.3 All models were estimated using maximum likelihood estimation with robust standard
errors and Satorra-Bentler scaled test statistics. Although not of primary interest given our
research questions, fit indices for the models can be attained using the R code provided on the
OSF page.
All indicator loadings for each of the latent factors are presented in Supplemental
Materials (Tables S1 and S2). Across both samples, standardized loadings for the DT latent
factors ranged from .47 - .94 (SD-3), .87 - .90 (Dirty Dozen), and .52 - .93 (single construct
measures). For the HEXACO-H/H, factor loadings ranged from .59 - .77. For the IPIP-NEO-
3
We note that our original preregistered analyses involving IPIP-Agreeableness could not be estimated in lavaan,
but could be estimated in Mplus (Muthén & Muthén, 2007). Mplus syntax to estimate the preregistered models is
available on the OSF page, along with R code to estimate the models using the ‘MplusAutomation’ package
(Hallquist & Wiley, 2018).
THE CORE OF THE DARK TRIAD 14
Agreeableness factor without Straightforwardness and Modesty, loadings ranged from .36 - .87.
When all facets were included as indicators, factor loadings for the latent IPIP-NEO-
Agreeableness factor ranged from .27 - .88. Factor loadings ranged from .48 - .70 and .60 - .84
for the BFI-Agreeableness and BFI-2-Agreeableness factors, respectively. For the SRP latent
factor, factor loadings ranged from .76 - .92. Tucker’s congruence coefficient (Lorenzo-Seva &
ten Berge, 2006), which indexes the similarity of factor loadings between two samples, was
calculated for all of the factor loadings from all 18 SEM models across Sample 1 and Sample 2.
The coefficient (TCC = .999) indicated near perfect congruence in factor loadings across the
samples.
Correlations between latent factors in Samples 1 and 2 are presented in Table 1. Profile
similarity coefficients, which index the similarity of correlations between the latent factors,
indicate that the correlational profiles for each of the latent DT core factors was identical across
samples (ranging from .996 - .999). The HEXACO-H/H latent factor was highly negatively
correlated with all three latent DT factors, with rs ranging from -.83 to -.99. In both samples, the
HEXACO-H/H was most strongly related to the latent DT factor comprising the SD-3 scales,
followed by single construct latent factor and then by the Dirty Dozen latent factor. The IPIP-
NEO-Agreeableness latent factor was similarly highly negatively correlated with all three DT
factors; these correlations were even higher than those observed for HEXACO-HH. Latent rs
ranged from -.89 to -1.00, with the SD-3 and single construct DT factors bearing the strongest
Straightforwardness and Modesty facets tended to yield smaller relations with DT factors, but
4
In some cases, the estimated latent correlation was slightly above 1 (e.g., 1.04). Despite this, no model
convergence issues arose. We denote in Table 1 the three cases in which the latent correlations were outside the
boundary of 1.00 to -1.00.
THE CORE OF THE DARK TRIAD 15
not always (e.g., when the latent DT factor was estimated from Single Construct scales). The
latent rs ranged from -.56 to -.97. Similar to the results of the model that included all IPIP-NEO-
Agreeableness facets, this modified IPIP-NEO-Agreeableness factor bore the smallest relations
Correlations between the latent DT factors and latent BFI-Agreeableness and BFI-2-
Agreeableness factors were more modest than the latent correlations observed for the latent
-.64 to -.76 across samples. For BFI-2-Agreeableness, rs ranged from -.72 to -.87. In both cases,
the largest relations were found with the DT latent factor comprised of the single construct
measures.
The SRP latent factor accounted for all the variance in the SD-3 latent DT factors in
Samples 1 and 2. Latent correlations between the SRP latent factor and the Dirty Dozen factor
In order to examine the influence of item overlap on the relations between latent factors,
we correlated all specific items of the DT measures (SD3, DD, and SC) with the specific items
2-Agreeableness, and the SRP subscales of Callousness and Interpersonal Manipulation). All
item correlations are available on the OSF page for the project. An inspection of the item
correlations showed that the largest item-level correlation was -.71. We adopted an empirical
cut-off of r = |.60| as our indicator of item overlap. All overlapping items at or above the cutoff
Across all scales, a total of 37 item correlations greater than or equal to |.60| were
identified.5 We dropped items in such a way that no single scale was inordinately affected in any
analyses. For example, if there were two items from the IPIP-A-Modesty facet scale that were
correlated above |.60| with two items from the DD-Psychopathy subscale, one item would be
dropped from the Modesty facet while the other would be dropped from the DD-Psychopathy
subscale. After dropping overlapping items, the SEM analyses were re-run in order to examine
the influence of item overlap on the latent correlations. The results showed very little change in
the latent correlations (See Table S3 in Supplementary Material)—across all latent correlations,
the largest change was a decrease of .05 with most latent correlations being only slightly reduced
(in the .01 to .03 range). Thus, item overlap appeared to have very little influence on the original
latent correlations.
Discussion
The present study tested whether any of the previously proposed candidates for the core
of the DT outperformed the other candidates in accounting for the shared variance among the DT
the Callousness and Interpersonal Manipulation subscales of the SRP-III outperformed BFI-
Agreeableness and BFI-2 Agreeableness. The differences were more pronounced for the BFI
compared to the BFI-2. Exploratory analyses showed that the large observed latent correlations
were not attributable to item overlap among the DT and candidate measures.
Most preregistered hypotheses were supported by the data, with some important caveats.
Hypothesis 1a was partially supported, as the results for Honesty-Humility were extremely
similar to previous work that has examined Honesty-Humility as the core of the DT (Hodson et
5
An Excel file that includes all item correlations is available on the OSF page for the project, and all correlations
greater than or equal to |.60| have been highlighted.
THE CORE OF THE DARK TRIAD 17
al., 2018). However, the correlation between latent IPIP-NEO-Agreeableness and the latent DT
factors was larger than hypothesized, and in most cases slightly larger than the latent correlation
between the Honesty-Humility factor and the latent DT factors. Interestingly, there were some
cases in which the latent IPIP-NEO-Agreeableness latent factor performed similarly to the
Honesty-Humility latent factor in accounting for variance in the DT factor even when the
Straightforwardness and Modesty facets were not included as indicators (e.g., when the latent DT
factor was estimated from the Single Construct measures and the SD3 subscales). Hypothesis 2
was also supported—the BFI-Agreeableness factors performed less well compared to the other
candidate factors. Notably, the BFI-2-Agreeableness latent factor accounted for more variance in
the DT factors compared to BFI-Agreeableness. However, in some cases (e.g., when the DT
factor was estimated from the Single Construct scales), the BFI-2 accounted for similar amounts
factors. Last, Hypothesis 3 was not supported. Though we expected a large latent factor
Manipulation factor would not account for as much variance in the DT latent factors compared to
variance in the latent DT factors estimated from the DD and SD3 scales. The present results have
important implications for DT research but also personality research focused on antagonism-
One research question we sought to address was whether different measures of the DT
yielded different results in regard to how much shared variance among DT components could be
accounted for by the various candidates. Our results suggest that the choice of DT measures does
THE CORE OF THE DARK TRIAD 18
have an effect on the observed relations between latent factors, with the DT factor estimated
from the Dirty Dozen subscales showing smaller relations with the candidate personality factors
compared to models in which the latent DT factor was estimated from the SD3 subscales or the
Single Construct scales. Thus, the use of the Dirty Dozen scales results in an underestimation of
the overlap between the core of the DT and the candidates. This is likely due to the limited
breadth of the Dirty Dozen scales, which assess each DT construct with four items each.
Research has shown that the Dirty Dozen subscales produces attenuated relations with relevant
outcomes (Maples, Lamkin, & Miller, 2014; Miller et al., 2012) compared to other DT measures.
However, the rank-ordering of the candidate traits in their ability to account for the shared
variance among the DT constructs was relatively consistent across the DT measurement
approaches. These findings suggest that the candidate traits that outperform the others in
accounting for the shared variance among the DT components is not highly dependent on a
Past research has highlighted that the BFI does not explicitly include content related to
the tendency to value honesty or to act humbly—this exclusion likely attenuates the relations
Agreeableness (Miller et al., 2011). One goal regarding the recent development of the BFI-2
(Soto & John, 2017) was to increase the content breadth of each Big Five domain. In regard to
Agreeableness, the facet scales of Compassion, Respectfulness, and Trust were designed to
capture the affective, behavioral, and cognitive elements of Agreeableness, respectively. Our
results show that the BFI-2 does outperform the original BFI in terms of capturing the shared
THE CORE OF THE DARK TRIAD 19
variance among personality constructs with robust ties to antagonism, suggesting that the aim of
Despite this notable improvement, the BFI-2 still did not capture as much of the shared
subscales of Callousness and Interpersonal Manipulation. This is likely due to the lack of explicit
inclusion of content related to honesty and humility within the BFI-2. However, recent work
focused on empirically identifying the hierarchical structure of Agreeableness suggests that even
though traits related to honesty and humility are not included in BFI scales of Agreeableness,
they cohere well with the other traits within the domain. Specifically, Crowe, Lynam, & Miller
(2019) identified 5 factors that underlie the broader domain using 22 self-report Agreeableness
scales (initial item pool of N=131). The results showed that five factors (named Compassion,
Morality, Modesty, Affability, and Trust) emerged at the final level of the hierarchy where the
factors remained interpretable and contained relatively homogenous content. These factors
aligned strongly with the conceptual facets of the NEO PI-R, with the exception that the
Altruism and Tendermindedness facets of the NEO PI-R were subsumed within the Compassion
factor. Despite the inclusion of 11 Honesty-Humility items from the HEXACO-100, as well as
many other relevant items from various measures, no distinct Honesty-Humility factor emerged
It is also important to consider the presence of reverse-coded items when comparing the
contains items like “Cheat to get ahead”, both the BFI and BFI-2 Agreeableness scales contain
the item, “Is sometimes rude to others”, and the HEXACO-H/H scale contains items like “I think
that I am entitled to more respect than the average person is.” However, in the case of the BFI
THE CORE OF THE DARK TRIAD 20
and HEXACO-H/H scales. Regarding BFI-Agreeableness items, 4/9 items are reverse coded
while 6/12 BFI-2 items are reverse coded. However, most of the items for IPIP-Agreeableness
and HEXACO-H/H are reverse coded, with 15/24 items being reverse coded for the IPIP-
Agreeableness scale and 10/16 items being reverse coded for the HEXACO-H/H scale. Given
that the various DT scales contain many items keyed in the antagonism direction, differences in
the number of reverse-scored items across the Agreeableness and H/H scales may have also
influenced latent correlations, in addition to the BFI instruments lacking specific content related
some of the core traits of antagonism-related constructs (e.g., the DT), NEO-based measures of
Agreeableness and Honesty-Humility from the HEXACO are the best candidates as they assess
important aspects of Agreeableness that are not covered by existing BFI instruments. Attenuated
relations are likely to be observed for the BFI-2 and BFI, though our results suggest the
Theoretical Considerations
Empirically, the present results show there is little difference among IPIP-NEO-
Agreeableness, H/H, and the SRP-III subscales of Callousness and Interpersonal Manipulation in
their ability to account for the shared variance among the DT components. In addition, more
accounting for the shared variance among the DT components. Indeed, the present results offer a
few pieces of evidence in support of this view. First are the strong negative relations between the
scope compared to IPIP-NEO-Agreeableness, it suggests that the more specific traits within
Honesty-Humility can account for nearly all the shared variance among the DT. Second, when
including the Modesty and Straightforwardness facets as indicators of latent Agreeableness, there
were consistent increases in the variance accounted for in the latent DT factors. When
negatively related to all facets of Agreeableness, but most strongly negatively related to the facet
of Straightforwardness of the NEO-PI-R (e.g., Lynam & Miller, 2015). Indeed, in both samples
of the current study, the relations between the Interpersonal Manipulation subscale and
Straightforwardness were also very large (rs of -.78 and -.80). Related factors in other
psychopathy measures (e.g., the Grandiose/Manipulative factor of the Youth Psychopathic Traits
Inventory; Andershed, Ker, Stattin, & Levander, 2002) also show that although there is a
negative manifold between these psychopathic traits and Agreeableness facets, the strongest
negative relations are with the Straightforwardness and Modesty facets (Sherman, Lynam, &
Heyde, 2014). Last, BFI measures of Agreeableness, which differ from the other candidates in
their lack of content focused on honesty and humility, accounted for smaller amounts of variance
These points notwithstanding, the results highlight that there is still substantial shared
variance among the DT components that is accounted for by the shared variance among
Agreeableness-related scales more broadly, and not solely scales assessing honesty and humility.
For example, latent DT factors estimated from the SD3 subscales and the Single Construct
measures were strongly related to the latent IPIP-Agreeableness factor that did not include the
Given that multiple candidate traits are empirically equivalent in accounting for the
shared variance among the DT constructs, other factors may help decide which candidate should
be taken to be the best explanation as to what traits underlie the DT. Our view is that the most
parsimonious account of the core of the DT is that it is Antagonism (vs. Agreeableness) from the
Five-factor Model. This perspective offers a straightforward account of current findings on the
DT core and can accommodate all previous findings. Specifically, all traits that have been
investigated as different candidates can be thought of as various tests of related traits that fall
under the broader umbrella of Agreeableness (e.g., Honesty-Humility) vs. Antagonism (e.g.,
SRP-III subscales). There are also practical reasons for adopting a view that what underlies the
shared variance of the DT is Antagonism. Primarily, it connects DT research to the much broader
personality research on the Five-factor Model and Antagonism, where antagonism is seen as the
Agreeableness, can be brought to bear on research on the DT. This research spans multiple areas
that are likely of interest to DT researchers, including developmental research (e.g., Cumberland-
li, Eisenberg, & Reiser, 2004; Jensen-Campbell et al., 2002), research on antisocial behavior
(e.g., Vize, Miller, & Lynam, 2018), and the neurobiological underpinnings of personality traits
including Agreeableness (e.g., DeYoung et al., 2010). Additionally, with regard to Antagonism,
research focused on the Section III model of personality disorders within the Diagnostic and
Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2011) can be
brought to bear on DT-related research, with a focus on the Antagonistic (vs. Agreeableness)
traits within Section III which include manipulativeness, deceitfulness, grandiosity, attention
seeking, callousness, and hostility. Hierarchical frameworks developed for clinical nosology,
THE CORE OF THE DARK TRIAD 23
Last, given the clinical implications of the work on the DT, reframing “dark traits” or
“dark factors” of personality as simply manifestations of Antagonism offers a less pejorative way
to think about such traits. The use of the term “dark” ultimately serves to do relatively little to
explicate the nature of these constructs while inadvertently stigmatizing antagonistic individuals
Limitations
A few limitations are worthy of mention. First, as previously noted, our analytical
strategy is not the only analytical strategy that has been used to try and account for the core of
the DT. Though the strategy used in the current paper is a straightforward test of the main
research question in this literature area (i.e., which traits account for the core of the DT) other
analytical strategies are also defensible and may offer additional insights not offered by the
Second, our analyses focused solely on the three DT components that are most commonly
studied. There has been less work that has expanded the scope of the DT to include constructs
like vulnerable narcissism and sadism; the inclusion of the latter alongside the DT has been
termed the dark tetrad (Book et al., 2016; Paulhus, 2014). Work that has taken a broader
constructs with robust ties to antagonism (e.g., Moshagen, Hilbig, & Zettler, 2018).
THE CORE OF THE DARK TRIAD 24
The present results suggest that the traits that best explain the overlap among the DT
constructs fall under the umbrella of Antagonistic traits, with the more specific traits of
manipulativeness and arrogance being central to the DT constructs. In fact, when considering the
overlap among the DT components, these latter traits appear to be sufficient in accounting for
DT overlap. Broader measures of Agreeableness that contain sufficient content related to these
traits appear to capture all of the overlapping variance among the DT constructs. Importantly, not
all measures of Agreeableness appear to adequately capture some of the core features of the DT
components—BFI instruments, particularly the original BFI, lack important content that is
central to the DT constructs. Additionally, given that the relations varied somewhat by DT
measure, these findings suggest that DT researchers should be thoughtful when choosing a scale
Antagonism can help connect DT research to the broader personality literature that has explicitly
focused on Antagonism.
THE CORE OF THE DARK TRIAD 25
References
Ackerman, R. A., Witt, E. A., Donnellan, M. B., Trzesniewski, K. H., Robins, R. W., & Kashy,
D. A. (2011). What does the narcissistic personality inventory really measure? Assessment,
American Psychiatric Association. (2011). Diagnostic and statistical manual of mental disorders
Andershed, H., Ker, M., Stattin, H., & Levander, S. (2002). Psychopathic traits in non- referred
youths: A new assessment tool. In E. Blaauw & L. Sheridan (Eds.), Psychopaths: Current
Book, A., Visser, B. A., Blais, J., Hosker-Field, A., Methot-Jones, T., Gauthier, N. Y., …
D’Agata, M. T. (2016). Unpacking more “evil”: What is at the core of the dark tetrad?
https://doi.org/10.1016/j.paid.2015.11.009
Book, A., Visser, B. a, & Volk, A. a. (2015). Unpacking ‘“ evil ”’: Claiming the core of the Dark
https://doi.org/10.1016/j.paid.2014.09.016
Christie, R., & Geis, F. (1970). Studies in Machiavellianism. New York, NY: Academic Press,
Inc.
Collison, K. L., Vize, C. E., Miller, J. D., & Lynam, D. R. (2018). Development and preliminary
Costa, P. T., & McCrae, R. R. (1992). Professional manual: revised NEO personality inventory
Assessment Resources.
Crowe, M. L., Lynam, D. R., & Miller, J. D. (2017). Uncovering the Structure of Agreeableness
https://doi.org/10.1111/jopy.12358
Cumberland-li, A., Eisenberg, N., & Reiser, M. (2004). Relations of Young Children’s
Agreeableness and Resiliency to Effortful Control and Impulsivity. Social Development, 13,
193–212.
DeYoung, C. G., Hirsh, J. B., Shane, M. S., Papademetris, X., Rajeevan, N., & Gray, J. R.
(2010). Testing predictions from personality neuroscience. Brain structure and the big five.
Glenn, A. L., & Sellbom, M. (2015). Theoretical and empirical concerns regarding the Dark
https://doi.org/10.1521/pedi_2014_28_162
structures from the top down. Journal of Research in Personality, 40, 347–358.
https://doi.org/10.1016/j.jrp.2006.01.001
Hallquist, M. N., & Wiley, J. F. (2018). MplusAutomation: An R Package for Facilitating Large-
Hodson, G., Book, A., Visser, B. A., Volk, A. A., Ashton, M. C., & Lee, K. (2018). Is the Dark
THE CORE OF THE DARK TRIAD 27
Jensen-campbell, L. A., Rosselli, M., Workman, K. A., Santisi, M., Rios, J. D., & Bojan, D.
Jonason, P. K., Kaufman, S. B., Webster, G. D., & Geher, G. (2013). What lies beneath the Dark
Triad Dirty Dozen: Varied relations with the Big Five. Individual Differences Research, 11,
81–90.
Jonason, P. K., Koenig, B. L., & Tost, J. (2010). Living a fast life: The Dark Triad and Life
Jonason, P. K., & Webster, G. D. (2010). The dirty dozen: a concise measure of the dark triad.
Jones, Daniel N., & Paulhus, D. L. (2017). Duplicity among the dark triad: Three faces of deceit.
https://doi.org/10.1037/pspp0000139
Jones, Daniel N, & Paulhus, D. L. (2014). Introducing the short Dark Triad (SD3): a brief
https://doi.org/10.1177/1073191113514105
Jones, Daniel Nelson, & Figueredo, A. J. (2013). The Core of Darkness: Uncovering the Heart of
https://doi.org/10.1002/per.1893
Kotov, R., Krueger, R. F., Watson, D., Achenbach, T. M., Althoff, R. R., Bagby, R. M., ... &
454-477.
Lee, K., & Ashton, M. C. (2004). Psychometric Properties of the HEXACO Personality
57–64. https://doi.org/10.1027/1614-2241.2.2.57
Lynam, D. R., & Miller, J. D. (2015). Psychopathy from a basic trait perspective: The utility of a
https://doi.org/10.1111/jopy.12132
Lynam, D. R., Sherman, E. D., Samuel, D., Miller, J. D., Few, L. R., & Widiger, T. A. (2013).
659–669. https://doi.org/10.1177/1073191113502072
Maples, J. L., Guan, L., Carter, N. T., & Miller, J. D. (2014). A test of the international
personality item pool representation of the revised NEO personality inventory and
Maples, J. L., Lamkin, J., & Miller, J. D. (2014). A test of two brief measures of the dark triad:
The Dirty Dozen and Short Dark Triad. Psychological Assessment, 26, 326–331.
https://doi.org/10.1037/a0035084
Marcus, D. K., Preszler, J., & Zeigler-Hill, V. (2018). A network of dark personality traits: What
lies at the heart of darkness? Journal of Research in Personality, 73, 56–62. https://doi.org/
10.1016/J.JRP.2017.11.003
THE CORE OF THE DARK TRIAD 29
McCrae, R. R., & Costa, P. T. (1987). Validation of the five-factor model of personality across
instruments and observers. Journal of Personality and Social Psychology, 52, 81–90.
https://doi.org/10.1037/0022-3514.52.1.81
Miller, B. K., Smart, D. L., & Rechner, P. L. (2015). Confirmatory factor analysis of the
https://doi.org/10.1016/j.paid.2015.03.022
Miller, J. D., Gaughan, E. T., Pryor, L. R., Kamen, C., & Campbell, W. K. (2009). Is research
https://doi.org/10.1016/j.jrp.2009.02.001
Miller, Joshua D., Few, L. R., Seibert, L. A., Watts, A., Zeichner, A., & Lynam, D. R. (2012). an
examination of the Dirty Dozen measure of psychopathy: A cautionary tale about the costs
https://doi.org/10.1037/a0028583
Miller, Joshua D., Hyatt, C. S., Maples-Keller, J. L., Carter, N. T., & Lynam, D. R. (2017).
Miller, Joshua D, Gaughan, E. T., Maples, J., & Price, J. (2011). A Comparison of
Agreeableness Scores From the Big Five Inventory and the NEO PI-R: Consequences for
https://doi.org/10.1177/1073191111411671
Moshagen, M., Hilbig, B. E., & Zettler, I. (2018). The dark core of personality. Psychological
Muris, P., Merckelbach, H., Otgaar, H., & Meijer, E. (2017). The Malevolent Side of Human
Nature: A meta-analysis and critical review of the literature on the Dark Triad (narcissism,
https://doi.org/10.1177/1745691616666070
Muthén, L. K.., & Muthén, B. O. (2007). Mplus User's Guide (Sixth Edition). Los Angeles, CA:
Paulhus, Delroy L., & Williams, K. M. (2002). The Dark Triad of personality: Narcissism,
Raskin, R. N., & Hall, C. S. (1979). A narcissistic personality inventory. Psychological Reports,
Sherman, E. D., Lynam, D. R., & Heyde, B. (2014). Agreeableness accounts for the factor
structure of the youth psychopathic traits inventory. Journal of Personality Disorders, 28,
262–280. https://doi.org/10.1521/pedi_2013_27_124
Soto, C. J., & John, O. P. (2017). The next big five inventory (BFI-2). Journal of Personality
Stead, R., Cynthia Fekken, G., Kay, A., & McDermott, K. (2012). Conceptualizing the Dark
Vize, C. E., Miller, J. D., & Lynam, D. R. (2019). Antagonism in the Dark Triad. In Miller, J. D.,
THE CORE OF THE DARK TRIAD 31
Vize, C.E., Collison, K. L., Miller, J. D., & Lynam, D. R. (2018). Examining the effects of
controlling for shared variance among the dark triad using meta-analytic structural equation
Vize, C. E., Miller, J. D., & Lynam, D. R. (2018). FFM facets and their relations with different
67–75. https://doi.org/10.1016/j.jcrimjus.2018.04.004
Vize, C. E., Lynam, D. R., Collison, K. L., & Miller, J. D. (2018). Differences among Dark Triad
Williams, K. M., & Paulhus, D. L. (2003). Factor structure of the Self-Report Psychopathy scale
Williams, K. M., Paulhus, D. L., & Hare, R. D. (2007). Capturing the four-factor structure of
psychopathy in college students via self-report. Journal of Personality Assessment, 88, 205–
219. https://doi.org/10.1080/00223890701268074
THE CORE OF THE DARK TRIAD 32
Figure 1
Graphical Illustration of Model Used to Test Latent Correlation Between SD3-DT Factor and
HEXACO-Honesty Humility Factor
Note: DT=Dark Triad; SD3-M=Short Dark Triad-Machiavellianism; SD3-N= Short Dark Triad-
Narcissism; SD3-P=Short Dark Triad-Psychopathy; H-H=Honesty-Humility; HH-Sinc=Honesty-
Humility-Sincerity Facet; HH-Fair=Honesty-Humility-Fairness Facet; HH-Greed=Honesty-
Humility-Greed Avoidance Facet; HH-Modes=Honesty-Humility-Modesty Facet.
THE CORE OF THE DARK TRIAD 33
Table 1