Journal of Research in Personality 73 (2018) 56–62

Contents lists available at ScienceDirect

Journal of Research in Personality

journal homepage: www.elsevier.com/locate/jrp

Full Length Article

A network of dark personality traits: What lies at the heart of

darkness? q

David K. Marcus a,⇑, Jonathan Preszler a, Virgil Zeigler-Hill b

a
Washington State University, United States
b
Oakland University, United States

a r t i c l e i n f o a b s t r a c t

Article history: The question of whether there is a common element at the core of the various dark personality traits (e.g.,
Received 14 August 2017 psychopathy, narcissism, Machiavellianism, spitefulness, aggressiveness) has been the subject of debate.
Revised 31 October 2017 Callousness, manipulativeness, and disagreeableness have all been nominated as possibly serving as the
Accepted 7 November 2017
core of these dark traits. Network analysis, which graphically and quantitatively describes the centrality
Available online 10 November 2017
of various related traits, provides a novel technique for examining this issue. We estimated an association
network and an Adaptive Least Absolute Shrinkage and Selection Operator network for two large sam-
Keywords:
ples, one college student sample (N = 2831) and one mixed college student and Mechanical Turk sample
Network analysis
Psychopathy
(N = 844). Interpersonal manipulation and callousness were the traits that were central to the networks.
Narcissism Ó 2017 Elsevier Inc. All rights reserved.
Machiavellianism
Spitefulness
Dark Triad

1. Introduction Network analysis may provide a novel methodology for examining

Paulhus and Williams (2002) coined the term ‘‘Dark Triad” to
encompass the personality traits of Machiavellianism, narcissism,
and psychopathy, because these three traits are socially aversive
and had received prominent empirical attention. Since the term
was coined, empirical interest has increased considerably with a
PsycINFO search of the term ‘‘Dark Triad” yielding over 330 cita-
tions with most of these articles having been published since
2014 (see Muris, Merckelbach, Otgaar, & Meijer, 2017 for a meta-
analysis, and Furnham, Richards, & Paulhus, 2013, for a narrative
review). The aim of the original Paulhus and Williams (2002) study
was to demonstrate that although these three traits are inter-
correlated, each has distinct characteristics. Although there is con-
sensus that each of these dark traits has unique qualities, the ques-
tion of what is at the core of the Dark Triad remains unresolved.
q
The study data are available as an online supplement. This study was not
preregistered at an independent institutional registry. All three authors contributed
to the study conceptualization and report writing. Marcus and Zeigler-Hill were
responsible for the data collection and preparation. Preszler conducted the data
analysis.
⇑ Corresponding author at: Department of Psychology, Washington State Univer-
sity, Pullman, WA 99164-4820, United States.
E-mail address: david.marcus@wsu.edu (D.K. Marcus).
this issue.
A meta-analysis of studies that assessed all three Dark Triad
traits found a large correlation between Machiavellianism and psy-
chopathy and medium-sized correlations between psychopathy
and narcissism and between Machiavellianism and narcissism
(Muris et al., 2017). These correlations suggest that the shared vari-
ance among these traits may represent a ‘‘core of darkness” (Jones
& Figueredo, 2013, p. 521). Various traits and characteristics have
been proposed as the core of the Dark Triad including low
honesty-humility from the HEXACO model of personality (e.g.,
Book, Visser, & Volk, 2015), low agreeableness (Jakobwitz & Egan,
2006), an exploitive mating strategy (at least in men; Jonason, Li,
Webster, & Schmitt, 2009), callousness (Jones & Paulhus, 2011),
and callousness combined with manipulativeness (Jones &
Figueredo, 2013). A variety of statistical methods have been used
in an attempt to identify the core of the Dark Triad including
canonical correlation analysis (Book et al., 2015), various types of
factor or principal components analysis (Jakobwitz & Egan, 2006),
structural equation modeling (Jones & Figueredo, 2013), and con-
firmatory factor analysis followed by a mediational analysis
(Jonason et al., 2009). Because network analysis visually depicts
the associations among traits and provides measures of centrality,
it may provide unique insight into the core of the Dark Triad (see
Costantini et al., 2015, for a review).

https://doi.org/10.1016/j.jrp.2017.11.003
0092-6566/Ó 2017 Elsevier Inc. All rights reserved.
 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62
In network analysis, traits can be represented as nodes and the
associations among pairs of traits are depicted as edges connecting
the nodes, with thicker edges depicting stronger associations. Thus,
a network figure provides a visual representation of a web of
related traits and shows which traits are the most central. Three
metrics are commonly used in personality and psychopathology
research to provide quantitative descriptions of the centrality of
the nodes (McNally, 2016). The strength of a node is the sum of
the correlations between the node and the other nodes in the net-
work. For most purposes, strength is the most relevant of the three
centrality metrics. Closeness is the inverse of the distance between
the node and the other nodes in the network. Betweenness is the
number of times that a node is the shortest path between two
other nodes. Because there will be increased distance between
many nodes of the network if a node with high betweenness is
removed, betweenness is a measure of a node’s connective value
to a network. Thus, if there is a trait in a network of dark person-
ality traits that is high in strength, closeness, and betweenness, it
is a likely candidate to be at the core. When there were negative
edges (i.e., negative associations) between nodes, we also calcu-
lated a fourth centrality metric, expected influence (Robinaugh,
Millner, & McNally, 2016). Expected influence is similar to
strength, but takes into account negative relationships (whereas
strength is calculated using the absolute value of the edges). There
has been a growing interest in using network analysis in psy-
chopathology research to map the relations among sets of psychi-
atric symptoms (e.g., Borsboom & Cramer, 2013).
There have only been a few applications of network analysis to
personality traits. However, these applications have shown pro-
mise in facilitating a novel understanding of personality. For exam-
ple, a network of conscientiousness across two samples revealed
that the industriousness and promotion focus nodes had the high-
est betweenness in the networks of both samples, indicating that
they may be particularly important factors in connecting the differ-
ent aspects of conscientiousness (Costantini & Perugini, 2016).
Although much of the research on dark personality traits has
focused on the Dark Triad, we (Marcus & Zeigler-Hill, 2015;
Zeigler-Hill & Marcus, 2016) have advocated for a broader frame-
work that includes additional antagonistic or dysfunctional traits.
For the current study, we conducted two network analyses exam-
ining the associations among a set of antagonistic dark personality
traits (psychopathy, narcissism, Machiavellianism, spitefulness,
and, in Sample 2, aggressiveness). For psychopathy and narcissism,
which are each multifaceted, we included each facet as a separate
node.
2. Method
2.1. Participants
Network analysis of psychological data is a relatively new ana-
lytic method that has not yet established standards for sample size
or power analysis (Epskamp, Kruis, & Marsman, 2017). However,
simulations conducted by Epskamp et al. (2017) indicate that with
10 nodes, LASSO techniques accurately detected sparse networks
with about 500 participants. Based on these initial findings, both
of our samples were sufficiently large.
Sample 1. This sample included 2831 undergraduate students
(663 men, 2165 women, and 3 who did not disclose their gender)
from a public university in the Midwestern region of the United
States. The original sample included 2971 participants, but 140
participants were excluded because they were missing data for
one or more of the instruments. The mean age of the participants
was 20.11 years (SD = 3.6). Regarding racial/ethnic identity, 76.6%
identified as Caucasian, 8.7% as Black, 5.3% as Asian, 2.9%, as His-
panic, and 6.5% other.
57
Sample 2. Data for this sample were originally collected for a
study of spitefulness (Marcus, Zeigler-Hill, Mercer, & Norris,
2014). This sample included 297 participants who were recruited
through Amazon Mechanical Turk (MTurk) and 547 undergraduate
students from a public university in the Pacific Northwestern
region of the United States. There were originally 853 participants
in this combined sample, but 9 of these participants were missing
one or measures and were exclude from the analyses, leaving 844
total participants. The MTurk participants received $1.00 in
exchange for their participation and the college students received
research credits. The mean age of the MTurk participants was
36.4 (SD = 13.1) and 62.3% were women. The racial/ethnic identifi-
cation of the MTurk sample was 75.4% Caucasian, 9.8% Black, 6.1%
Hispanic, and 4.7% Asian. The mean age of the student participants
was 19.7 years (SD = 2.8) and 79.1% were women. Regarding racial/
ethnic identity, 71.9% identified as Caucasian, 9.0% as Hispanic,
9.0% as Asian, and 4.7%, as Black.
2.2. Materials and procedures
The participants provided informed consent and completed all
of the study measures online. The measures were administered
in random order.
Self-Report Psychopathy Scale (SRP-III; Paulhus, Neumann, &
Hare, in press). The SRP-III is a self-report measure of psychopathy
that was modeled on the Psychopathy Checklist-Revised (PCL-R;
Hare, 2003). Participants indicated their agreement with each of
the 34 statements on scales ranging from 1 (strongly disagree) to
5 (strongly agree). This version of the SRP-III consists of four sub-
scales: Callousness (8 items; e.g., ‘‘I’m not afraid to step on others
to get what I want” [aSample 1 = .77; aSample 2 = .80]), Erratic Lifestyle
(8 items; e.g., ‘‘I enjoy taking risks” [aSample 1 = .77; aSample 2 = .78]),
Interpersonal Manipulation (8 items; e.g., ‘‘I find it easy to manip-
ulate people” [aSample 1 = .71; aSample 2 = .74]), and Criminal Tenden-
cies (10 items; e.g., ‘‘Stole money from my parents” [aSample 1 = .82;
aSample 2 = .80]).
Narcissistic Personality Inventory (NPI; Raskin & Hall, 1981). The
NPI is well-validated and is considered to be the standard measure
of subclinical narcissistic personality features (Paulhus & Williams,
2002). Items on the NPI are in a forced-choice format such that par-
ticipants must choose between a narcissistic and a non-narcissistic
statement (e.g., ‘‘I like having authority over other people” or ‘‘I
don’t mind following orders”). Although there has been contro-
versy regarding the underlying factor structure of the 40-item
NPI, Ackerman et al. (2010) suggested that the NPI consists of three
factors: Leadership/Authority (11 items; e.g., ‘‘If I ruled the world it
would be a much better place” [aSample 1 = .77; aSample 2 = .65]),
Grandiose Exhibitionism (10 items; e.g., ‘‘I really like to be the cen-
ter of attention” [aSample 1 = .75; aSample 2 = .77]), and Exploitative-
ness/Entitlement (4 items; e.g., ‘‘I find it easy to manipulate
people” [aSample 1 = .45; aSample 2 = .50]).
Mach-IV (Christie & Geis, 1970). The Mach-IV is a 20-item
instrument that was developed to assess Machiavellianism (e.g.,
‘‘The best way to handle people is to tell them what they want to
hear”). Participants rated their level of agreement with the items
using scales that ranged from 1 (strongly disagree) to 5 (strongly
agree). The Mach-IV has been found to possess adequate psycho-
metric properties and is the most widely used measure of Machi-
avellianism (McHoskey, Worzel, & Szyarto, 1998). The internal
consistency for the Mach-IV was aSample 1 = .73 and aSample 2 = .67.
Spitefulness Scale (Marcus et al., 2014). The Spitefulness Scale is
a 17-item instrument designed to measure the willingness of a
participant to engage in behaviors that would harm another but
that would also entail potential harm to oneself. This harm could
be social, financial, physical, or an inconvenience (e.g., ‘‘I would
be willing to take a punch if it meant that someone I did not like
58 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62
would receive two punches”). Participants rated their level of
agreement with the items of the Spitefulness Scale using scales
that ranged from 1 (strongly disagree) to 5 (strongly agree). The
internal consistency for the Spitefulness Scale was aSample 1 = .91
and aSample 2 = .90.
Forms and Functions of Aggression Scale (AS; Little, Henrich,
Jones, & Hawley, 2003). This measure is based on the idea that
aggressive behavior can be understood as the combination of the
function of aggressive behavior (i.e., pure, reactive, or instrumen-
tal) and the behavioral form of aggression (i.e., overt or relational).
Participants rated their level of agreement with each of the 36
items using scales ranging from 1 (not at all) to 4 (completely true).
The AS was only administered to Sample 2. For the current analy-
sis, we used the total score on the scale (a = .96).
2.3. Analytic strategy
We estimated an association network and an Adaptive Least
Absolute Shrinkage and Selection Operator (LASSO) network for
each sample. Association networks are graphical representations
of zero-order correlations of data and usually result in fully con-
nected networks. Because data sets typically include correlations
that are spurious or not theoretically meaningful, estimating net-
works using partial correlations (i.e., the association that is left
between any two variables after conditioning on all other vari-
ables) is useful for understanding the genuine and meaningful rela-
tionships among the variables. Estimation using adaptive LASSO
induces a sparse network from the data, shrinking small correla-
tions to zero by using a LASSO penalty during estimation
(Costantini et al., 2015). Consequently, the edges are more likely
to reflect direct relationships. Quantitative indices, such as central-
ity measures, can then be used to interpret sparse networks fur-
ther. Therefore, we provide adaptive LASSO networks in the
primary results (the results from the association networks of the
raw correlations are provided in an online supplement). We calcu-
lated four centrality measures (expected influence, strength, close-
ness, and betweenness) and one clustering coefficient for each
adaptive LASSO network to identify nodes that are important to
the network structure. All of these centrality measures were calcu-
lated using the partial correlations. Thus, for example, the strength
index for a node was the sum of the partial correlations between
the node and the other nodes in the network. Similarly, the close-
ness values are a measure of the inverse sum of the distances of a
given node to all other nodes in the partial correlation network
(even nodes with which the given node only has an indirect
connection).
Clustering coefficients measure ‘‘the number of connections
among the neighbors of a focal node over the maximum possible
number of such connections” (Costantini et al., 2015, p. 18). The
clustering coefficient is high when the nodes that a node is con-
nected to are also strongly connected to each other. A high cluster-
ing coefficient often indicates redundancy (i.e., if the node were
removed, the general network structure would remain). If a node
has elevated centrality measures, then an elevated clustering coef-
ficient might indicate that the centrality is artificially inflated. To
assess node clustering, we used the Zhang clustering coefficient
(Zhang & Horvath, 2005) because of its applicability to correla-
tional data. We used the ‘‘signed” version of the Zhang coefficient,
which takes into account negative edge weights. We also calcu-
lated the small-worldness value for each adaptive LASSO network.
Small-worldness is a measure of the tendency for any two nodes
within a network to be easily connected through other nodes. A
small-world network has high transitivity, indicating that each
node quickly affects many of the nodes in the network (i.e., change
to a single node would quickly result in change to the whole net-
work). A small-worldness value above three is considered to have
the small-world property (Humphries & Gurney, 2008). We used
the R package qgraph (Epskamp, Cramer, Waldorp, Schmittmann,
& Borsboom, 2012) to perform the network analytics.
3. Results
Descriptive statistics for the variables measured for both sam-
ples are provided in Table 1. The edge weights (partial correlations
among the variables) are provided in Table 2. The raw bivariate
correlations among the variables are provided in supplementary
Table S1. Across the two samples, the correlation of the edge
weights for the adaptive LASSO networks was .86 and the correla-
tion of the raw bivariate correlations for the association networks
was .93, indicating that the results from the two samples were gen-
erally consistent.
3.1. Network structures
3.1.1. Sample 1 adaptive LASSO network
The adaptive LASSO network for Sample 1 is displayed in Fig. 1.
Sample 1 included 9 traits in the network, making the total possi-
ble number of edges 36 (9  8  0.5). Of these 36 possible edges,
25 (69%) were present in the adaptive LASSO estimation. The four
psychopathic traits connected to make a strong loop, but with no
association between ‘‘Callousness” and ‘‘Erratic Lifestyle,” and only
a small positive association between ‘‘Interpersonal Manipulation”
and ‘‘Criminal Tendencies.” The ‘‘Spitefulness” and ‘‘Machiavellia
nism” nodes clustered strongly with each other and the psycho-
pathic nodes. Regarding narcissism, whereas ‘‘Grandiose Exhibi-
tionism” and ‘‘Leadership/Authority” were strongly connected
with each other, ‘‘Exploitativeness/Entitlement” was only moder-
ately connected with any other trait, including the other two nar-
cissism traits. In addition, the narcissism nodes were only
moderately connected with other nodes in the network, and occu-
pied peripheral positions on the graph. The narcissism traits also
included the strongest negative connection on the graph, that
between ‘‘Leadership/Authority” and ‘‘Machiavellianism.” The
small-worldness value of the Sample 1 network was 1.03, indicat-
ing it does not have the small-world property.
3.1.2. Sample 1 centrality
Centrality measures index the role and relationships a given
node has in a network. Table 3 presents three standardized central-
ity measures (i.e., strength, expected influence, and closeness) as
well as the raw scores for betweenness. ‘‘Interpersonal Manipula-
tion” and ‘‘Callousness” had the highest centrality measures across
all four indices. The other two psychopathic traits (‘‘Erratic Life-
style” and ‘‘Criminal Tendencies”) were average, except for ‘‘Crim-
inal Tendencies” having moderate expected influence and ‘‘Erratic
Lifestyle” having low betweenness. In contrast, the narcissism
nodes were relatively non-central. Specifically, ‘‘Grandiose Exhibi-
tionism” and ‘‘Exploitativeness/Entitlement” had the lowest cen-
trality scores across all four indices, and ‘‘Leadership/Authority”
had above average betweenness (4 connections) but below average
strength and closeness. ‘‘Machiavellianism” had just below average
centrality values except for expected influence, which was lower
than the others, likely because of its negative connection with ‘‘Lea
dership/Authority.” Lastly, ‘‘Spitefulness” had approximately aver-
age values across all centrality indices. Spitefulness did, however,
yield the highest Zhang clustering coefficient values in both this
sample and Sample 2, suggesting that this scale may be redundant
and capturing information that is already measured by the other
scales.
 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62 59

Table 1
Descriptive Statistics.

Symptom Sample 1 (N = 2831) Sample 2 (N = 845)

Skewness Kurtosis Mean (SD) Skewness Kurtosis Mean (SD)
Psychopathy
Interpersonal manipulation 0.47 0.01 2.15 (0.65) 0.32 0.06 2.29 (0.69)
Callousness 0.95 0.38 1.75 (0.59) 0.57 0.53 1.91 (0.62)
Criminal tendencies 1.30 1.33 1.64 (0.64) 0.71 0.12 1.90 (0.70)
Erratic lifestyle 0.25 0.54 2.44 (0.74) 0.01 0.48 2.61 (0.75)
Narcissism
Exploitativeness/entitlement 1.06 0.55 0.84 (0.97) 1.11 0.50 0.88 (1.04)
Grandiose exhibitionism 0.57 0.46 3.25 (2.47) 0.52 0.63 3.26 (2.59)
Leadership/authority 0.12 0.92 5.26 (2.92) 0.00 .096 5.39 (2.53)
Machiavellianism, spitefulness, & aggression
Machiavellianism 0.27 0.22 2.62 (0.42) 0.49 2.47 2.81 (0.38)
Spitefulness 0.84 0.12 1.82 (0.67) 0.55 0.00 1.94 (0.61)
Aggression – – – 1.78 3.09 1.39 (0.42)

Note. Aggression items were not used in Sample 1.

Table 2
Edge weights (partial correlations) of each network.

M L/A GE E/E C EL IM CT S
Machiavellianism (M) .09 – .08 .12 .01 .15 – .28
Leadership/authority (L/A) .09 .36 .12 – .05 .12 – .01
Grandiose exhibitionism (GE) .05 .45 .13 – .08 – – –
Exploitativeness/entitlement (E/E) – .04 .12 .12 .02 .09 – .04
Callousness (C) .01 – – .08 – .28 .32 .22
Erratic lifestyle (EL) .13 .07 .12 – – .25 .35 –
Interpersonal manipulation (IM) .15 .20 – .11 .30 .24 .05 .08
Criminal tendencies (CT) – – – – .23 .36 – .19
Spitefulness (S) .08 .01 – – .35 – .15 .11
Aggression .06 .05 – .20 .23 .07 – .17 .16

Note. Values above the diagonal are for Sample 1 (N = 2831) and those below the diagonal are for Sample 2 (N = 845). ‘‘–” indicates the LASSO estimation shrunk the
correlation to ‘‘0”. There is no column for ‘‘Aggression” because Aggression was not measured in Sample 1.

3.1.3. Sample 2 adaptive LASSO network 3.1.4. Sample 2 centrality

The adaptive LASSO network for Sample 2 is displayed in Fig. 1.
Sample 2 included 10 traits in the network, making the total pos-
sible number of edges 45 (10  9  0.5). Of these 45 possible edges,
29 (64.4%) were estimated by the adaptive LASSO procedure. The
psychopathic traits made a loop with each other (i.e., ‘‘Criminal
Tendencies” – ‘‘Callousness” – ‘‘Interpersonal Manipulation” –
‘‘Erratic Lifestyle” – ‘‘Criminal Tendencies”), with no connections
between ‘‘Callousness” and ‘‘Erratic Lifestyle” or ‘‘Criminal Tenden-
cies” and ‘‘Interpersonal Manipulation” (the apparent negative
connection is actually between ‘‘Erratic Lifestyle” and ‘‘Aggres-
sion”). Of these psychopathic traits, ‘‘Interpersonal Manipulation”
and ‘‘Callousness” seem to occupy a central role within the graph,
where they both have numerous moderate-to-strong connections
with other nodes. Specifically, ‘‘Callousness” is connected strongly
with ‘‘Spitefulness” and ‘‘Aggression,” whereas ‘‘Interpersonal
Manipulation” is connected with ‘‘Machiavellianism” and ‘‘Leader
ship/Authority.” This formation almost mimicked the loop found
in Sample 1, but without the weak connection between ‘‘Interper-
sonal Manipulation” and ‘‘Criminal Tendencies.” Of note is that the
narcissistic traits are on the periphery of the graph, with a strong
connection between ‘‘Grandiose Exhibitionism” and ‘‘Leadership/
Authority,” but with few strong connections besides these two
nodes and the rest of the network. Three relatively weak negative
connections were present: between ‘‘Machiavellianism” and ‘‘Lea
dership/Authority,” between ‘‘Leadership/Authority” and ‘‘Aggres-
sion,” and between ‘‘Erratic Lifestyle” and Aggression.” The small-
worldness value of the Sample 2 network was 0.99, indicating it
does not have the small-world property.
The centrality measures for Sample 2 are displayed in Table 4.
Consistent with their central locations on the graphical display,
‘‘Interpersonal Manipulation” and ‘‘Callousness” had the highest
centrality index scores across all four indices. No other nodes were
noticeably above average on strength or expected influence. ‘‘Ma
chiavellianism,” ‘‘Grandiose Exhibitionism,” and ‘‘Exploitative
ness/Entitlement” had well below average centrality scores across
the indices, congruent with their peripheral position on the graph.
Both the ‘‘Erratic Lifestyle” and ‘‘Criminal Tendencies” nodes hov-
ered around average values across all four indices.
3.2. Supplementary analyses
3.2.1. Association networks
Consistent with the assertion that ‘‘the correlations between
pairs of symptoms that remain when all other symptoms are con-
trolled for. . .may be considered to provide clues about the causal
skeleton of a network” (Borsboom & Cramer, 2013, p. 105), our pri-
mary results focused on the LASSO networks of partial correlations.
However, it may be worthwhile to compare these findings to the
results from association networks based on raw bivariate correla-
tions. These networks are displayed in Supplementary Fig. S1 and
the centrality measures for these association networks are pro-
vided in Supplementary Tables S2 and S3. Comparisons of Table 3
with Table S2 and Table 4 with Table S3 indicate that the few dif-
ferences in strength and closeness between the LASSO and associ-
ation networks in either sample were in magnitude and not in the
ordering of the traits. Interpersonal Manipulation and Callousness
60 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62

Machiavellianism

Grandiose
Exhibitionism

Leadership/
Authority

Grandiose
Exhibitionism

Erratic

Exploitativeness/
Entitlement
Leadership/
Authority

Machiavellianism
Interpersonal

Criminal

Interpersonal

Callous
Spite Callous
Erratic

Exploitativeness/
Entitlement

Spite

Criminal

Machiavellianism

Leadership/
Authority

Grandiose
Exhibitionism

Erratic

Interpersonal

Criminal

Exploitativeness/
Entitlement

Callous

Aggression

Spite

Fig. 1. Adaptive LASSO network graphs. Top left graph = Sample 1 (N = 2831). Top right Graph = Sample 2 (N = 845) without aggression included. Bottom graph = Sample 2
with aggression included. Blue = Psychopathic traits; Gray = Narcissistic traits; Green = Spite; Red = Machiavellianism; Yellow = Aggression. The nodes represent personality
traits and the edges represent the adaptive LASSO partial correlations between the items. Thicker edges represent stronger associations, with green edges representing
positive associations and red edges representing negative associations. The Fruchterman-Reingold (specified as ‘‘spring” in the qgraph package) layout is used to display the
results. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

remained most central on both of these metrics and they both in these networks, suggesting that when raw correlations are
appear to be central to the networks in Fig. S1. Because association examined, Spitefulness may provide some unique information.
networks involve many more direct connections between nodes,
there was less betweenness (i.e., fewer nodes could serve as the 3.2.2. Dark Triad
shortest path between pairs of nodes) and Interpersonal Manipula- In an attempt to broaden the network of dark personality traits,
tion accounted for all of the betweenness (7) in Sample 1 and the primary analyses included spitefulness (Samples 1 and 2) and
almost all of the betweenness (6) in Sample 2. Thus, whether aggression (Sample 2). However, given past interest in identifying
examining association (raw correlations) or LASSO (partial correla- the core of the Dark Triad, we also ran the analyses with just the
tions) networks, Interpersonal Manipulation and Callousness Dark Triad traits. These LASSO networks are depicted in Supple-
appear to be at the core of these dark traits. Interestingly, Spiteful- mentary Fig. S2 and the centrality measures for these networks
ness no longer yielded especially high Zhang clustering coefficients are provided in Supplementary Tables S4 and S5. Interpersonal
 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62 61

Table 3
Sample 1 (N = 2831) centrality and clustering measures for adaptive LASSO network.

Trait Strength EI Closeness Betweenness Zhang

Psychopathy
Interpersonal manipulation 1.29 1.30 1.49 5 0.38
Callousness 1.55 1.52 1.39 6 0.61
Criminal tendencies 0.60 0.72 0.43 3 0.16
Erratic lifestyle 0.30 0.21 0.20 1 1.56
Narcissism
Exploitativeness/entitlement 1.14 0.90 0.85 0 0.39
Grandiose exhibitionism 1.38 0.92 1.47 0 0.13
Leadership/authority 0.35 0.90 0.83 4 1.47
Machiavellianism & spitefulness
Machiavellianism 0.38 0.93 0.19 2 0.21
Spitefulness 0.12 0.32 0.16 2 1.91

Note. Betweenness values are raw scores. All other values are z-scores. EI = Step1 Expected Influence. The Zhang clustering coefficient is used because of its usability with
correlational data.

Table 4
Sample 2 (N = 845) centrality and clustering measures for adaptive LASSO network.

Trait Strength EI Closeness Betweenness Zhang

Psychopathy
Interpersonal manipulation 1.25 1.43 1.82 11 0.27
Callousness 1.47 1.62 1.38 8 0.82
Criminal tendencies 0.07 0.29 0.32 3 0.02
Erratic lifestyle 0.52 0.21 0.33 4 0.95
Narcissism
Exploitativeness/entitlement 1.50 0.95 0.97 1 0.45
Grandiose exhibitionism 0.61 0.17 0.69 1 0.67
Leadership/authority 0.17 0.67 0.39 4 0.98
Machiavellianism, spitefulness, & aggression
Machiavellianism 1.49 1.64 1.38 0 0.92
Spitefulness 0.10 0.18 0.32 0 2.20
Aggression 0.37 0.30 0.10 3 0.16

Note. Betweenness values are raw scores. All other values are z-scores. EI = Step1 Expected Influence. The Zhang clustering coefficient is used because of its usability with
correlational data.

Manipulation and Callousness remain the most central nodes on
each of the centrality metrics in both samples, and visually Inter-
personal Manipulation was at the center of both networks. Overall,
the results from these supplementary analyses confirm the central-
ity of Interpersonal Manipulation and Callousness to both the Dark
Triad and to a slightly broader network of dark traits.
4. Discussion
Two components of psychopathy, interpersonal manipulation
and callousness, occupied central positions in both sets of network
analyses. In fact, interpersonal manipulation and callousness were
the only two traits to yield above average values on all four central-
ity metrics across both samples. Furthermore, as depicted in Fig. 1,
interpersonal manipulation and callousness are the only traits that
are located centrally in the network and not along the periphery. It
is noteworthy that these two traits were often the ones that linked
the other dark personality traits, accounting for roughly half of the
betweenness values in the adaptive LASSO networks (11 of 23 in
Sample 1 and 19 of 35 in Sample 2). When only the Dark Triad
was analyzed, they accounted for more than half of the between-
ness values in each sample (11 of 17 in Sample 1, 11 of 18 in Sam-
ple 2). Removing interpersonal manipulation and callousness from
these dark networks would leave the other nodes much less den-
sely connected. Overall, these results suggest that interpersonal
manipulation and callousness are the dark traits that hold the Dark
Triad (or an expanded set of dark traits) together.
These results are strikingly consistent with the conclusion of
Jones and Figueredo (2013) that the combination of manipulative-
ness and callousness form the ‘‘dark core” (p. 528) of a malevolent
personality. The current study was similar to Jones and Figueredo
(2013) in that both studies used the same instruments to measure
the Dark Triad personality traits, and both used a combination of
college student and MTurk participants. Still, the consistency in
results across these two studies is remarkable because Jones and
Figueredo used a very different analytic strategy (i.e., multi-
sample structural equation modeling). Furthermore, the findings
were consistent even though the current study added spitefulness
and (in Sample 2) aggressiveness to the network of dark personal-
ity traits, whereas Jones and Figueredo added social dominance
orientation to their analyses for their second study. Thus, these
consistent findings across samples and diverse analytic methods
may be seen as an example of ‘‘critical multiplism” (Cook, 1985)
that enhances our confidence that interpersonal manipulation
and callousness are at the dark core of antagonistic traits. Given
recent concerns about the reproducibility of psychological findings
and the need for replication in research (Open Science
Collaboration, 2015), finding such consistency is both reassuring
and valuable.
Jones and Figueredo (2013) made a compelling argument why
the combination of both interpersonal manipulation and callous-
ness is necessary for a malevolent or antagonistic personality.
Essentially, an individual who is callous or unemotional but not
manipulative may simply be socially disconnected (e.g., schizoid
or autistic). Many manipulative individuals, who are not callous,
may direct their manipulation toward benign pursuits. In contrast,
individuals who are both callous and manipulative are positioned
to force others into harmful outcomes. In the current networks,
62 D.K. Marcus et al. / Journal of Research in Personality 73 (2018) 56–62
interpersonal manipulation and callousness had strong connec-
tions with one another, and each was not only connected to the
other psychopathy nodes, but to nodes from each of the other dark
traits.
One limitation of the current study is that the networks did not
include other dark personality traits, most notably sadism, which
has been identified as being a component of the ‘‘Dark Tetrad”
(Chabrol, Van Leeuwen, Rodgers, & Séjourné, 2009). Future studies
may also consider including additional malevolent traits such as
greed (Seuntjens, Zeelenberg, van de Ven, & Breugelmans, 2015)
and authoritarianism (Ludeke & Krueger, 2013). Furthermore, the
current study focused on the relations among specific dark traits
and did not include broadband personality traits such as agreeable-
ness from the Big Five or honesty-humility from the HEXACO per-
sonality model. Network analysis studies that include these
broader traits may further clarify the nature of the dark core. Dark
Triad research has focused on ‘‘subclinical” manifestations of the-
ses antagonistic personality traits and the current study had the
advantage of including both college and national (MTurk) samples.
Still, it would be worthwhile to examine whether these network
structures replicate in samples with higher rates of antisocial or
malevolent behavior, such as criminal offenders or other forensic
samples. It would also be beneficial to examine the consistency
of these network structures in specific contexts such as highly
competitive or stressful environments.
Acknowledgement
Jonathan Preszler’s work was supported by the Anthony Mar-
chionne Foundation for the Scientific Study of Human Relations
and Psychological Processes Endowed Graduate Fellowship for
Research at Washington State University.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.jrp.2017.11.003.
References
Ackerman, R. A., Witt, E. A., Donnellan, B., Trzesniewski, K. H., Robins, R. W., & Kashy,
D. A. (2010). What does the narcissistic personality inventory really measure?
Assessment, 18, 67–87.
Book, A., Visser, B. A., & Volk, A. A. (2015). Unpacking ‘‘evil”: Claiming the core of the
dark triad. Personality and Individual Differences, 73, 29–38.
Borsboom, D., & Cramer, A. O. (2013). Network analysis: An integrative approach to
the structure of psychopathology. Annual Review of Clinical Psychology, 9,
91–121.
Chabrol, H., Van Leeuwen, N., Rodgers, R., & Séjourné, N. (2009). Contributions of
psychopathic, narcissistic, Machiavellian, and sadistic personality traits to
juvenile delinquency. Personality and Individual Differences, 47, 734–739.
Christie, R., & Geis, R. L. (1970). Studies in machiavellianism. New York, NY: Academic
Press.
Cook, T. D. (1985). Postpositivist critical multiplism. In L. Shotland & M. M. Mark
(Eds.), Social science and social policy (pp. 21–62). Beverly Hills, CA: SAGE.
Costantini, G., Epskamp, S., Borsboom, D., Perugini, M., Mottus, R., Waldorp, L. J., &
Cramer, A. O. J. (2015). State of the aRt personality research: A tutorial on
network analysis of personality data in R. Journal of Research in Personality, 54,
13–29.
Costantini, G., & Perugini, M. (2016). The network of conscientiousness. Journal of
Research in Personality, 65, 68–88.
Epskamp, S., Cramer, A. O., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D.
(2012). Qgraph: Network visualizations of relationships in psychometric data.
Journal of Statistical Software, 48, 1–18.
Epskamp, S., Kruis, J., & Marsman, M. (2017). Estimating psychopathological
networks: Be careful what you wish for. PLoS ONE, 12(6), e0179891.
Furnham, A., Richards, S. C., & Paulhus, D. L. (2013). The dark triad of personality: A
10 year review. Social and Personality Psychology Compass, 7, 199–216.
Hare, R. D. (2003). Manual for the revised psychopathy checklist (2nd ed.). Toronto,
ON: Multi-Health Systems.
Humphries, M. D., & Gurney, K. (2008). Network ‘small-world-ness’: A quantitative
method for determining canonical network equivalence. PLoS ONE, 3(4),
e0002051.
Jakobwitz, S., & Egan, V. (2006). The dark triad and normal personality traits.
Personality and Individual Differences, 40, 331–339.
Jonason, P. K., Li, N. P., Webster, G. D., & Schmitt, D. P. (2009). The dark triad:
Facilitating a short-term mating strategy in men. European Journal of Personality,
23, 5–18.
Jones, D. N., & Figueredo, A. J. (2013). The core of darkness: Uncovering the heart of
the dark triad. European Journal of Personality, 27, 521–531.
Jones, D. N., & Paulhus, D. L. (2011). Differentiating the dark triad within the
interpersonal circumplex. In L. M. Horowitz & S. Strack (Eds.), Handbook of
interpersonal psychology: Theory, research, assessment, and therapeutic
interventions (pp. 249–268). New York, NY: Wiley.
Little, T. D., Henrich, C. C., Jones, S. M., & Hawley, P. H. (2003). Disentangling the
‘whys’ from the ‘whats’ of aggressive behaviour. International Journal of
Behavioral Research, 27, 122–133.
Ludeke, S. G., & Krueger, R. F. (2013). Authoritarianism as a personality trait:
Evidence from a longitudinal behavior genetic study. Personality and Individual
Differences, 55, 480–484.
Marcus, D. K., & Zeigler-Hill, V. (2015). A big tent of dark personality traits. Social
and Personality Psychology Compass, 9, 434–446.
Marcus, D. K., Zeigler-Hill, V., Mercer, S., & Norris, A. L. (2014). The psychology of
spite and the measurement of spitefulness. Psychological Assessment, 26,
563–574.
McHoskey, J. W., Worzel, W., & Szyarto, C. (1998). Machiavellianism and
psychopathy. Journal of Personality and Social Psychology, 74, 192–210.
McNally, R. J. (2016). Can network analysis transform psychopathology? Behaviour
Research and Therapy, 86, 95–104.
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, Machiavellianism, and psychopathy). Perspectives on
Psychological Science, 12, 183–204.
Open Science Collaboration (2015). Estimating the reproducibility of psychological
science. Science, 349, aac4716.
Paulhus, D. L., Neumann, C. S., & Hare, R. D. (in press). Manual for the Self-Report
Psychopathy Scale (4th ed.). Toronto, ON: Multi-Health Systems.
Paulhus, D. L., & Williams, K. M. (2002). The dark triad of personality: Narcissism,
Machiavellianism, and psychopathy. Journal of Research in Personality, 36,
556–563.
Raskin, R. N., & Hall, C. S. (1981). The narcissistic personality inventory: Alternate
form reliability and further evidence of construct validity. Journal of Personality
Assessment, 45, 159–162.
Robinaugh, D. J., Millner, A. J., & McNally, R. J. (2016). Identifying highly influential
nodes in the complicated grief network. Journal of Abnormal Psychology, 125,
747–757.
Seuntjens, T. G., Zeelenberg, M., van de Ven, N., & Breugelmans, S. M. (2015).
Dispositional greed. Journal of Personality and Social Psychology, 108, 917–933.
Zeigler-Hill, V., & Marcus, D. K. (2016). The dark side of personality: Science and
practice in social, personality, and clinical psychology. Washington, DC: American
Psychological Association.
Zhang, B., & Horvath, S. (2005). A general framework for weighted gene co-
expression network analysis. Statistical Applications in Genetics and Molecular
Biology, 4(1), 17.