Content uploaded by Constantine Sedikides
Author content
All content in this area was uploaded by Constantine Sedikides on Apr 06, 2023
Content may be subject to copyright.
Communal Narcissism
Jochen E. Gebauer
Humboldt-Universität zu Berlin Constantine Sedikides
University of Southampton
Bas Verplanken
University of Bath Gregory R. Maio
Cardiff University
An agency-communion model of narcissism distinguishes between agentic narcissists (individuals
satisfying self-motives of grandiosity, esteem, entitlement, and power in agentic domains) and communal
narcissists (individuals satisfying the same self-motives in communal domains). Five studies supported
the model. In Study 1, participants listed their grandiose self-thoughts. Two distinct types emerged:
agentic (“I am the most intelligent person”) and communal (“I am the most helpful person”). In Study
2, we relied on the listed communal grandiose self-thoughts to construct the Communal Narcissism
Inventory. It was psychometrically sound, stable over time, and largely independent of the Narcissistic
Personality Inventory—the standard measure of agentic narcissism. In Studies 3 and 4, agentic and
communal narcissists shared the same self-motives, while crucially differing in their means for need
satisfaction: Agentic narcissists capitalized on agentic means, communal narcissists on communal means.
Study 5 revisited the puzzle of low self–other agreement regarding communal traits and behaviors.
Attesting to the broader significance of our model, this low self–other agreement was partly due to
communal narcissists: They saw themselves as high, but were seen by others as low, in communion.
Keywords: narcissism, communion, agency, Communal Narcissism Inventory, self
Selflessness is the vainest form of selfishness. (Russian proverb)
Altruism is perfected egoism. (Herbert Spencer)
A grandiose narcissist is a person who continuously seeks to
validate his or her self-perceived grandiosity, esteem, entitlement,
and power. Two recent reviews of the narcissism literature concur
with this definition. W. K. Campbell and Foster (2007, p. 115)
stated that “individuals with narcissistic personality possess highly
inflated, unrealistically positive views of the self,” and Morf,
Horvath, and Torchetti (2011, p. 399) maintained that narcissism is
characterized by “grandiose self-views and a relentless addiction-
like striving to continually assert their self-worth and superiority.”
Several other authors have made similar claims (Raskin, Novacek,
& Hogan, 1991b; Sedikides, Cisek, & Hart, 2011; Sedikides &
Gregg, 2001, 2008; Wink, 1991; Zeigler-Hill, Clark, & Pickard,
2008). For example, Raskin et al. (1991b) described narcissism as
a strategy for managing self-esteem via grandiosity.
At the same time, research has documented that narcissists feed
their self-views of grandiosity, esteem, entitlement, and power
with unrealistically positive self-evaluations regarding agency
(e.g., competence, extraversion, uniqueness) rather than commu-
nion (e.g., warmth, agreeableness, relatedness). For example, Ga-
briel, Critelli, and Ee (1994) reported that narcissists perceive
themselves as more intelligent than nonnarcissists. Yet, narcissists
did not differ from their counterparts in their actual IQ. Similarly,
Farwell and Wohlwend-Lloyd (1998) reported that narcissists do
not differ from nonnarcissists on their academic grades. Further,
Rhodewalt and Morf (1998) administered IQ tests to participants.
When narcissists received positive IQ feedback, they exceeded
nonnarcissists in attributing this to ability, but when they received
negative feedback, they exceeded nonnarcissists in attributing this
to bad luck. W. K. Campbell, Reeder, Sedikides, and Elliot (2000)
reported a similar pattern with creativity feedback rather than IQ
feedback. Finally, Paulhus and his colleagues (Paulhus & Harms,
2004; Paulhus, Harms, Bruce, & Lysy, 2003) assessed participants’
knowledge across a broad range of academic topics. Participants were
unaware that some items were real and some were foils. As detailed
later, responses to real and foil items can be used to calculate exag-
geration and accuracy indices. Narcissists inflated their knowledge
across a wide array of academic topics, although they did not differ
from nonnarcissists in knowledge accuracy. Together, these find-
ings indicate that narcissists exaggerate their intelligence, course
grades, creativity, and academic knowledge—exaggerations that
This article was published Online First August 13, 2012.
Jochen E. Gebauer, Institut fu¨r Psychologie, Humboldt-Universität zu
Berlin, Berlin, Germany; Constantine Sedikides, School of Psychology,
University of Southampton, Southampton, United Kingdom; Bas Verplan-
ken, Department of Psychology, University of Bath, Bath, United King-
dom; Gregory R. Maio, School of Psychology, Cardiff University, Cardiff,
Wales.
We thank Jens Asendorpf for helpful comments on an earlier draft of the
manuscript, Barbara Friedrich for serving as the second rater in Study 1,
and John Rauthmann for help with TripleR. We also thank Axel Burger,
Jaap Denissen, and John Krantz for advertising our online studies on their
web portals.
Correspondence concerning this article should be addressed to Jochen E.
Gebauer, Humboldt-Universität zu Berlin, Institut fu¨r Psychologie, Unter
den Linden 6, 10099 Berlin, Germany. E-mail: mail@JochenGebauer
.info
Journal of Personality and Social Psychology © 2012 American Psychological Association
2012, Vol. 103, No. 5, 854– 878 0022-3514/12/$12.00 DOI: 10.1037/a0029629
854
fall into the agentic domain. (For conceptually similar findings, see
Bleske-Rechek, Remiker, & Baker, 2008; Buss & Chiodo, 1991;
Gosling, John, Craik, & Robins, 1998; John & Robins, 1994;
Kernis & Sun, 1994; Paulhus, 1998; Raskin & Novacek, 1991;
Rhodewalt, Tragakis, & Finnerty, 2006; Sedikides, Hart, Cisek, &
Routledge, in press; Zeigler-Hill, Myers, & Clark, 2010.)
In contrast, narcissists do not self-enhance in the communal do-
main. In line with the research described above, W. K. Campbell,
Rudich, and Sedikides (2002) found that narcissists exhibit pro-
nounced better-than-average perceptions of their agentic attributes. Of
importance, these authors found no evidence that narcissists manifest
pronounced better-than-average perceptions of their communal attri-
butes. W. K. Campbell, Bosson, Goheen, Lakey, and Kernis (2007)
replicated conceptually this pattern. In addition, these researchers
reported that narcissists (compared to nonnarcissists) had strong im-
plicit associations (Greenwald, McGhee, & Schwartz, 1998) between
self and agentic traits; however, narcissists and nonnarcissists did not
differ in their strength of implicit associations between self and
communal traits. In conclusion, there is broad consensus that narcis-
sists feed their self-views of grandiosity, esteem, entitlement, and
power through agentic rather than communal means.
These empirical findings are reflected in contemporary models
of grandiose narcissism. Paulhus and John (1998) described nar-
cissism as an egoistic (i.e., agentic) rather than moralistic (i.e.,
communal) bias. Paulhus (2001) suggested that narcissism is best
understood as an extreme form of agency. In line with W. K.
Campbell’s (1999) self-orientation model of narcissism, Sedikides,
Campbell, Reeder, Elliot, and Gregg (2003) argued that narcissists
are so preoccupied with agency that they devalue communion by
endorsing an “others exist for me” illusion. Finally, Campbell and
colleagues (W. K. Campbell, Brunell, & Finkel, 2006; W. K.
Campbell & Foster, 2007) introduced an agency model of narcis-
sism. (For a review of these models, see Bosson et al., 2008.)
In this article, we propose a new and broader formulation: the
agency-communion model of narcissism. According to this model,
past research has described one type of grandiose narcissism, agentic
narcissism. Agentic narcissists satisfy their core self-motives (i.e.,
grandiosity, esteem, entitlement, power) through agentic means.
However, we posit that there is another, equally relevant type of
grandiose narcissism, communal narcissism. Communal narcissists
possess the same core self-motives as agentic narcissists (i.e., gran-
diosity, esteem, entitlement, power). In contrast to agentic narcissists,
however, communal narcissists satisfy these self-motives through
communal means. We report five studies that test this model across 16
samples of participants. The studies address five questions. Is com-
munal narcissism theoretically plausible (Study 1)? Are agentic nar-
cissism and communal narcissism distinct personality traits (Study 2)?
Do communal narcissists have the same self-motives as agentic nar-
cissists (Study 3)? Do communal and agentic narcissists differ in how
they satisfy these self-motives (Study 4)? And, finally, can communal
narcissism explain previous evidence of a puzzlingly weak relation
between self-reported and observer-reported communion (Study 5)?
We elaborate on these questions below.
Is Communal Narcissism Theoretically Plausible?
On first sight, the notion of communal narcissism may seem an
oxymoron. This is because the self-motives for grandiosity, es-
teem, entitlement, and power are agentic themselves. We submit,
however, that the agency-communion distinction can be applied
not only for arranging personality traits in a circular structure
(Abele, Cuddy, Judd, Yzerbyt, 2008; Abele & Wojciszke, 2007;
Bakan, 1966; Helgeson, 1994; Wiggins, 1991) but also for arrang-
ing means to express certain personality traits (here: narcissism)
into such a structure (Gebauer, Leary, & Neberich, in press;
Gebauer, Paulhus, & Neberich, in press; see also Kruglanski et al.,
2002). In particular, we propose that agency and communion are
hierarchically structured in reference to traits and means. Agency
and communion apply to traits at a higher level, and they addi-
tionally apply to means for expressing these traits at a lower level.
Our view that agency and communion span a circular structure that
can simultaneously operate at multiple hierarchical levels is consistent
with evidence that these dimensions are concurrently relevant to
personality traits and social groups, with the latter representing a
higher level in this hierarchy (Fiske, Cuddy, Glick, & Xu, 2002). Our
view is also consistent with a hierarchy of values (Maio, 2010;
Schwartz, 1992). Values can be arranged along the agency and
communion dimensions (Paulhus & Trapnell, 2008; Roberts & Rob-
ins, 2000; Trapnell & Paulhus, 2012), and values can operate at a
personal level (Maio, 2010; Schwartz, 1992) and at a societal level
(Bernard, Gebauer, & Maio, 2006; Schwartz, 1992). Most relevant,
within the personal level, values can operate as conscious theories as
well as even more fundamental entities organizing implicit memory
(Pakizeh, Gebauer, & Maio, 2007). In all, we start from the idea that
the agency-communion distinction can be applied to means for ex-
pressing the trait of grandiose narcissism, and this idea is congruent
with past theory and research.
Agentic narcissism may then be understood as an agency-agency
trait: an agentic trait that is expressed through agentic means. An
inspection of items from the Narcissistic Personality Inventory (NPI;
Raskin & Terry, 1988) indicates that this inventory indeed captures
our reasoning. At the trait level, the NPI assesses the disposition to
hold grandiose self-views—an agentic trait (Morf & Rhodewalt,
2001). At the means level, the NPI is also relevant to agency. In
particular, NPI items reflect grandiose self-views on key components
of agency, such as assertiveness (“I am assertive”), competence (“I
am more capable than other people”), and authority (“People always
seem to recognize my authority”). Paralleling this reasoning, commu-
nal narcissism may be understood as an agency-communion trait: an
agentic trait that is expressed through communal means. Key com-
ponents of communion are helpfulness, interpersonal warmth, and
trustworthiness. Arguably, these traits can serve as means to express
grandiose narcissism (W. K. Campbell & Foster, 2007; Morf et al.,
2011). For example, a communal narcissist would be a person who
grandiosely views himself as “the most helpful person he knows,”
“the most caring person in his social surroundings,” and “extraordi-
narily trustworthy.”
The notion of communal narcissism is primarily dependent on
whether people truly hold communal grandiose self-views and
whether these are largely independent of agentic grandiose self-
views. To test this notion, we asked participants in Study 1 to list
their grandiose self-thoughts in an open-ended fashion. To fore-
shadow the results, participants reported agentic as well as com-
munal grandiose self-thoughts. Next, we examined whether the
quantity and extremity of a person’s agentic grandiose self-
thoughts were related to the quantity and extremity of this person’s
communal grandiose self-thoughts. Lastly, in Study 1, we exam-
ined whether the NPI is a measure of agentic, but not communal,
855
COMMUNAL NARCISSISM
narcissism. If this were indeed the case, we would expect the NPI
to correlate with the quantity and extremity of agentic, but not
communal, grandiose self-thoughts.
Are Agentic Narcissism and Communal Narcissism
Distinct Personality Traits?
Is communal narcissism a personality trait in and of itself (i.e.,
independently of the NPI)? In Study 2, we used the communal
grandiose self-thoughts reported by Study 1 participants (Van de
Vijver & Leung, 2001) to construct candidate items for the Com-
munal Narcissism Inventory (CNI). We implemented exploratory
factor analyses to derive the most appropriate CNI items, and we
tested the CNI’s psychometric properties. To find out if communal
narcissism can be conceptualized as a personality trait, we tested
for its temporal stability. Further, we explored the relation between
the NPI and the CNI. A central tenet of our model is that agentic
and communal narcissism are two distinct facets of grandiose
narcissism. Finally, we examined the correlation between commu-
nal narcissism and clinical/vulnerable forms of narcissism. In
contrast to agentic and (supposedly) communal narcissism, these
latter forms of narcissism are not rooted in grandiosity (W. K.
Campbell & Miller, 2011; Miller & Campbell, 2008). Hence, we
expected particularly low relations between communal narcissism
and clinical/vulnerable forms of narcissism.
Do Communal Narcissists Have the Same Core
Self-Motives as Agentic Narcissists?
Following theoretical perspectives on grandiose narcissism
(W. K. Campbell & Foster, 2007; Morf et al., 2011; Raskin et al.,
1991b; Sedikides et al., 2002), we have defined narcissists as
people who continuously seek to validate their self-perceived
grandiosity, esteem, entitlement, and power. We settled on these
four constructs because they appear to constitute narcissists’ core
self-motives. In this section, we introduce theory and research that
buttress the relevance of these self-motives.
A heightened desire to validate one’s grandiose self-views has
been called the “most central characteristic” of narcissism (Brown,
Budzek, & Tamborski, 2009, p. 951; see also Buss, 1991). In fact,
Morf et al. (2011) recently labeled NPI narcissism as grandiose
narcissism. Empirical research has indeed documented that agentic
(i.e., NPI) narcissists feel particularly grandiose (Bleske-Rechek et
al., 2008; Brown et al., 2009; Raskin, Novacek, & Hogan, 1991a,
1991b; Rosenthal, Hooley, & Steshenko, 2007).
Excessively high levels of explicit self-esteem have also been
established as an integral part of the narcissistic personality
(Bosson et al., 2008; W. K. Campbell & Foster, 2007; Morf et al.,
2011; Sedikides & Gregg, 2001; Sedikides, Gregg, & Hart, 2007).
Some researchers conceptualize self-esteem as narcissists’ internal
gauge for monitoring their progress in meeting their self-motives
(Morf & Horvath, 2010). Another theoretical view of narcissism
labeled it an addiction to self-esteem (Baumeister & Vohs, 2001).
Finally, Raskin et al. (1991b) understood narcissists’ incessant
validation of their own grandiosity as a means to meet their core
self-need: self-esteem. Not surprisingly then, much empirical re-
search has documented that agentic (i.e., NPI) narcissists report
unusually high self-esteem (Bushman & Baumeister, 1998; W. K.
Campbell, 2001; Emmons, 1984, 1987; Gregg & Sedikides, 2010;
Raskin et al., 1991a; Raskin & Terry, 1988; Rhodewalt & Morf,
1995; Sedikides, Rudich, Gregg, Kumashiro, & Rusbult, 2004).
Narcissists are also said to have a pervasive sense of entitlement,
given their grandiosity and excessively high self-esteem. They feel
unique, special, and entitled to privileged treatment (Emmons, 1984;
Morf & Rhodewalt, 2001; Sedikides et al., 2002). Empirical research
has indeed documented that agentic (i.e., NPI) narcissists report a
heightened sense of entitlement (Brown et al., 2009; W. K. Campbell,
Bonacci, Shelton, Exline, & Bushman, 2004; Emmons, 1987).
Finally, narcissists are believed to have a strong desire to vali-
date their power. For one, their sense of grandiosity, superiority,
and entitlement would seem to legitimize them having power over
others rather than others having power over them. Also, power
could be a means for gaining admiration and verifying their sense
of grandiosity, superiority, and entitlement (W. K. Campbell &
Baumeister, 2006; Rose & Campbell, 2004). Empirical research
has indeed documented that agentic (i.e., NPI) narcissists have a
strong need for power (W. K. Campbell, Foster, & Finkel, 2002;
Carroll, 1987; Horton & Sedikides, 2009; Raskin & Novacek,
1991; Rose & Campbell, 2004). Agentic narcissists also exhibit a
game-playing love style in their intimate relationships (W. K.
Campbell, Foster, & Finkel, 2002; Le, 2005), and this love style
has been described as a means to exert power over intimate
partners (Morf et al., 2011).
To summarize, research and theory converge on the conclusion
that narcissists relentlessly feel the need to validate their self-views
of grandiosity, esteem, entitlement, and power. It is fitting, then, to
refer to these as narcissists’ core self-motives. Importantly, we
hold that these self-motives would not be limited to agentic (i.e.,
NPI) narcissists but rather would apply equally to communal
narcissists. We addressed these issues in Study 3.
Do Communal Narcissists Differ From Agentic
Narcissists in How They Satisfy Their Core
Self-Motives?
A central proposition of the agency-communion model of nar-
cissism is that agentic and communal narcissists differ in the
means they deploy for meeting the same self-motives: The former
rely on agentic means, the latter on communal means. Stated
otherwise, the model proposes a double dissociation between agen-
tic and communal narcissism and between agentic and communal
means. Past research has been consistent with part of this disso-
ciation in showing that agentic (i.e., NPI) narcissists self-enhance
by displaying exaggerated better-than-average judgments on agen-
tic, but not on communal, attributes (W. K. Campbell, Rudich, &
Sedikides, 2002). The agency-communion model additionally hy-
pothesizes that communal (i.e., CNI) narcissists will self-enhance
by judging themselves as better than average on communal, but not
on agentic, attributes. We tested this hypothesis in Study 4.
The better-than-average task capitalizes on social comparison to
capture agentic and communal means for need satisfaction (Hep-
per, Gramzow, & Sedikides, 2010). We complemented this ap-
proach by using a criterion-discrepancy measure to capture com-
munal means for need satisfaction. Specifically, we adapted the
overclaiming task (Paulhus et al., 2003) to assess communal over-
claiming. Agentic (i.e., NPI) narcissists overclaim their knowledge
on academic topics (Paulhus & Harms, 2004; Paulhus et al., 2003),
which are largely agentic in nature (Abele & Wojciszke, 2007;
856 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
Bakan, 1966); agentic narcissists do not overclaim their knowledge
about music artists (Nathanson, Williams, & Paulhus, 2003),
which is a knowledge domain that is not agentic in nature and
hence not ego-relevant for agentic narcissists (Paulhus, 2011).
Based on these results, we proceeded to adapt the overclaiming
task to assess overclaiming on communal topics (e.g., “humani-
tarian aid organizations,” “nature and animal protection organiza-
tions,” and “international health charities”). We hypothesized that
communal, but not agentic, narcissists would overclaim their
knowledge on communal topics.
Can Communal Narcissism Explain the Weak Self–
Other Agreement Regarding Communal Behaviors
and Traits?
We conducted Studies 1–4 with the goal of validating the
agency-communion model of grandiose narcissism. We conducted
Study 5 to test the potential of this model to inform the broader
personality literature. Self–other agreement of traits (Funder &
Colvin, 1988; Paulhus & Reynolds, 1995; Vazire, 2010; D. Wat-
son, 1989) is “a centrally important issue in personality research”
(D. Watson, Hubbard, & Wiese, 2000, p. 546).
One longstanding, well-replicated, and puzzling finding is this:
Self–other agreement is higher for agentic traits than for commu-
nal traits (Allik, Realo, Mo˜ttus, & Kuppens, 2010; Gosling, Rent-
frow, & Swann, 2003; John & Robins, 1993). Communal narcis-
sism may be a piece of this puzzle. Specifically, communal
narcissists should overclaim their communal behaviors and traits
(Study 4). Hence, communal narcissism should be positively re-
lated to self-reports of communal behaviors and traits. In contrast,
communal narcissism should not be positively related to observer
reports of communal behaviors and traits. In fact, communal
narcissists’ unremitting, if not tiring, striving for grandiosity, self-
esteem, entitlement, and power may well lead to a negative rela-
tion between communal narcissism and observer-reported commu-
nal behaviors and traits (Study 3). In other words, for communal
narcissists, we expected a dissociation between self-reports and
observer reports of communal behaviors and traits. Most impor-
tantly, due to this dissociation, communal narcissism should sup-
press self–other agreement regarding communion. Study 5 exam-
ined these hypotheses in a two-wave round-robin design.
Study 1: Is Communal Narcissism Theoretically
Plausible?
In Study 1, we asked participants to report their grandiose
self-thoughts in an open-ended question format. We tested the
hypothesis that the NPI would relate to agentic, but not communal,
grandiose self-thoughts. We also tested the hypothesis that agentic
and communal grandiose self-thoughts would be largely unrelated
to each other, as anticipated by prior literature on the independence
of agency and communion (Abele et al., 2008; Abele & Wojciszke,
2007; Bakan, 1966; Helgeson, 1994; Wiggins, 1991).
Method
Participants. Table 1 contains information about the 16
samples that we used in the five reported studies. As can be seen,
four samples were collected at U.K. universities, one sample was
collected at a German university, and 11 samples were collected
online: 10 of them via an American advertising portal of psycho-
logical research on the net (Krantz, 2012) and one via two German
advertising portals (www.psytests.de, www.forschung-erleben.de).
Sampling decisions were guided by feasibility and convenience.
The advantage of such broad sampling is that results are not only
applicable to a very specific group (such as psychology undergrad-
uate students at a single university). With one exception, there was
no need to exclude any participants from our 16 samples. The
exception was Sample 16 (i.e., Study 5 data). In Sample 16, the
performed analyses necessitated exclusion of participants with
missing data (see Study 5’s method section for a detailed descrip-
tion). We tested Study 1’s hypotheses in Samples 1–3.
Table 1
Sample Characteristics
Sample N(total)
Sex Age
Country of residence:
majority Participants RewardN(female) N(male) MSD
1 95 69 26 23.22 6.65 U.S.A. (87%) Internet none
2 120 94 25 25.51 10.09 U.S.A. (86%) Internet none
3 127 89 38 26.17 11.24 U.S.A. (86%) Internet none
4 195 107 14 — — U.K. (100%) laboratory course credit
5 120 91 28 20.45 6.58 U.S.A. (90%) Internet none
6 322 275 45 19.56 2.33 U.K. (100%) laboratory course credit
7 329 251 74 24.19 7.95 U.S.A. (68%) Internet none
8 224 167 56 23.88 10.52 U.S.A. (88%) Internet none
9 314 220 90 33.06 12.38 Germany (100%) Internet gift voucher
10 102 86 15 19.66 3.24 U.K. (100%) laboratory course credit
11 120 91 28 22.36 7.43 U.S.A. (90%) Internet none
12 103 77 25 22.91 9.02 U.S.A. (87%) Internet none
13 151 123 28 26.73 10.52 U.S.A. (92%) Internet none
14 145 105 38 28.88 11.58 U.S.A. (76%) Internet none
15 201 170 31 20.07 2.45 U.K. (100%) laboratory course credit
16 106 85 21 25.49 7.77 Germany (100%) lecture hall course credit
Note. In cases where numbers do not add up, this is due to missing data. Dashes indicate that no information was available.
857
COMMUNAL NARCISSISM
Materials and procedure. In each of the three samples,
participants completed one of three agentic narcissism measures
and responded to open-ended questions assessing the grandiosity
of their self-thoughts (see below). Different agentic narcissism
measures were chosen to ascertain that results did not hinge on
specifics of one measure (and for the same reason, we have
attempted to use differing measures of the same construct through-
out the whole article). Table 2 provides details about administra-
tion orders, information about which measure was used in which
sample, and internal consistencies of the measures.
Narcissistic Personality Inventory. The 40-item NPI (NPI40;
Raskin & Terry, 1988) asks participants to indicate for each item
whether a narcissistic or nonnarcissistic statement describes them
better. Example items are “I see myself as a good leader” (narcis-
sistic statement) or “I am not sure if I would make a good leader”
(nonnarcissistic statement), and “I am more capable than other
people” (narcissistic statement) or “There is a lot I can learn from
other people” (nonnarcissistic statement).
Narcissistic Personality Inventory—Short Version. The 16-
item short version of the NPI (NPI16; Ames, Rose, & Anderson,
2006) uses the same forced-choice response format as the original
NPI40, and the NPI16 is strongly related to the NPI40, r(766) ⫽
.90, p⫽.001 (Ames et al., 2006).
Narcissistic Personality Inventory—Rating Version. The
40-item rating version of the NPI (Jordan, Spencer, Zanna,
Hoshino-Browne, & Correll, 2001) asks participants to indicate the
degree to which each of the 40 narcissistic statements of the
original NPI40 applies to them (1 ⫽does not apply at all, 7⫽
applies completely). In comparison to the original forced-choice
version, this version has improved psychometric properties (Golec
de Zavala, Cichocka, Eidelson, & Jayawickreme, 2009).
Grandiose self-thought listing. We instructed participants in
all three samples as follows: “People have all kinds of private
thoughts about themselves. One type of thoughts all of us sometimes
have are grandiose self-thoughts. That is, we have thoughts of our
own grandiosity. From person to person, these grandiose self-thoughts
can vary quite a lot in content.” Further, we asked participants to “take
some time and think about the grandiose self-thoughts you occasion-
ally have. Please list these thoughts in the textboxes below. Use one
textbox for each grandiose self-thought.”
Results
Qualitative analyses. We proceeded with coding of the 1,167
grandiose self-thoughts that participants listed across all three
samples. Coding took place in two steps. In the content-
classification step, the first author classified each self-thought as
either agentic or communal based on accepted definitions of
agency and communion (Abele et al., 2008; Bakan, 1966; Helge-
son, 1994; Wiggins, 1991). Given that the circumplex structure of
Table 2
Tasks/Measures per Sample
Sample Tasks/measures in order of administration (Cronbach’s alpha in parentheses)
1 Demographics 3NPI40 (.82) 3grandiose self-view listing 3debriefing.
2 Demographics 3NPI16 (.53) 3grandiose self-view listing 3debriefing.
3 Demographics 3ratingNPI40 (.93) 3grandiose self-view listing 3debriefing.
4 Demographics 3candidate CNI items 3debriefing.
5 Demographics 3candidate CNI items 3debriefing.
6 (Wave 1) Demographics 3SLSC (.89) 3NPI16 (.73) 3CNI.
6 (Wave 2) Eight-week lag: NPI16 (.74) 3CNI 3debriefing.
7 Demographics 3at random: {NPI16 (.68) & CNI} 3debriefing.
8 Demographics 3at random: {NPI40 (.84), CNI, NPD-PDQ-4 (.59), & HNS (.73)} 3debriefing.
9 Demographics 3at random: {NPI16 (.68) 3HIAGT (.95), CNI 3HICGT (.95), UC (.80), BSRI (Masc [.90], Fem
[.89]), COS (.79), EOS (.74), BFI (E [.88], O [.85], A [.76], C [.84], N [.89]), IAS (PA [.86], BC[.87], DE [.91], FG
[.89], HI [.88], JK [.66], LM [.88], NO [.87])} 3debriefing.
10 Demographics 3NPI16 (.68) 3CNI 3PES (.86) 3NGS (.95) 3SLSC (.93) 3AG- (.72) & CO- (.81) BTAT 3
debriefing.
11 Demographics 3NPI16 (.77) 3CNI 3PowVal (.65) 3PowNeed (.88) 3debriefing.
12 Demographics 3CNI 3NPI16 (.72) 3AG- (.71) & CO- (.70) BTAT 3debriefing.
13 Demographics 3CNI 3CO-OCQ (.89) 3debriefing.
14 Demographics 3CNI & NPI16 (.74) (at random) 3CO-OCQ (.88) 3debriefing.
15 Demographics 3at random: {CNI, NPI40 (.84), AG-OCQ (.71), & CO-OCQ (.79)} 3debriefing.
16 (self-report) Demographics 3at random: {CO-traits (self-report [.72]) & CNI}.
16 (Round-Robin 1) CO-traits (peer report [.54]) 3CO-behaviors (peer report [.80]).
16 (Round-Robin 2) CO-traits (peer report [.57]) 3CO-behaviors (peer report [.83]) 3debriefing.
Note. NPI40 ⫽40-item Narcissistic Personality Inventory; NPI16 ⫽16-item Narcissistic Personality Inventory; ratingNPI40 ⫽rating scale version of
the NPI40; CNI ⫽Communal Narcissism Inventory; SLSC ⫽Self-Liking/Self-Competence Scale; NPD-PDQ-4 ⫽Narcissistic Personality Disorder
subscale of the Personality Diagnostic Questionaire–4; HNS ⫽Hypersensitive Narcissism Scale; HIAGT ⫽Habitual Index of Agentic Grandiose Thinking;
HICGT ⫽Habitual Index of Communal Grandiose Thinking; UC ⫽Unmitigated Communion Scale; BSRI ⫽Bem Sex Role Inventory (Masc ⫽
Masculinity subscale; Fem ⫽Femininity subscale); COS ⫽Communal Orientation Scale; EOS ⫽Exchange Orientation Scale; BFI ⫽Big Five Inventory
(E ⫽Extraversion subscale; O ⫽Openness subscale; A ⫽Agreeableness subscale; C ⫽Conscientiousness subscale; N ⫽Neuroticism subscale); IAS ⫽
Interpersonal Adjectives Scales (PA ⫽Assured-Dominant subscale; BC ⫽Arrogant-Calculating subscale; DE ⫽Cold-Hearted subscale; FG ⫽
Aloof-Introverted subscale; HI ⫽Unassured-Submissive subscale; JK ⫽Unassuming-Ingenuous subscale; LM ⫽Warm-Agreeable subscale; NO ⫽
Gregarious-Extraverted subscale); PES ⫽Psychological Entitlement Scale; NGS ⫽Narcissistic Grandiosity Scale; AG-BTAT ⫽agentic better-than-
average task; CO-BTAT ⫽communal better-than-average task; PowVal ⫽Power Values Scale; PowNeed ⫽Need for Power Scale; AG-OCQ ⫽Agentic
Overclaiming Questionnaire; CO-OCQ ⫽Communal Overclaiming Questionnaire; CO-traits ⫽communal traits (i.e., agreeableness and conscientious-
ness); CO-behaviors ⫽communal behaviors in work groups.
858 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
personality traits allows some traits to be simultaneously high on
agency and on communion, a portion of self-thoughts (18%) were
classified as not clearly attributable to either the agentic or com-
munal domain and were thus excluded from further analyses. From
the 950 clearly agentic or communal grandiose self-thoughts, 640
were classified as agentic (66%) and 310 were classified as com-
munal (33%). It is not surprising that agentic grandiose self-
thoughts outnumbered communal grandiose self-thoughts. This is
because the NPI preceded the listing task, with NPI items consist-
ing of agentic, but not communal, content. Hence, completion of
the NPI should have primed agentic grandiose self-thoughts
(Strack & Schwarz, 2007). Nonetheless, participants listed a suf-
ficiently large amount of communal grandiose self-thoughts to
proceed with the analyses. On average, each participant listed M⫽
1.87 (66%; SD ⫽2.03) agentic grandiose self-thoughts and M⫽
.91 (33%; SD ⫽1.47) communal grandiose self-thoughts. Con-
sidering the means, the standard deviations of agentic and com-
munal grandiose self-thoughts were large and comparable. This
suggests considerable individual differences in the variety of gran-
diose self-thoughts that people hold, and thus provides the first
indication for meaningful individual differences not only in agen-
tic narcissism but also in communal narcissism.
In the grandiosity-rating step, the first author rated each self-
thought according to the extent of grandiosity reflected by this
self-thought (1 ⫽not at all grandiose, 7⫽extremely grandiose).
We computed indices of agentic and communal grandiose self-
thoughts for each participant. With these indices, we wanted to
capture both the variety of grandiose self-thoughts and the extent
of grandiosity reflected by each self-thought. Thus, for each par-
ticipant, we computed an agentic grandiose self-thoughts index by
multiplying the number of this participant’s agentic grandiose
self-thoughts (reflecting variety) by the mean grandiosity of this
participant’s agentic grandiose self-thoughts (reflecting extent).
We used the same formula to compute a communal grandiose
self-thoughts index. To ensure validity of these indices, a second
judge also provided content classifications followed by grandiosity
ratings for one sample. Interrater reliability was high, r(89) ⫽.81,
p⫽.001, and both raters were unaware of participants’ NPI
scores.
Quantitative analyses. In support of the hypothesis that the
NPI assesses agentic, but not communal, narcissism, (a) the rela-
tion between the NPI and agentic grandiose self-thoughts was
positive in all three samples (.36 ⬍r⬍.55, all ps ⬍.01), and (b)
the relation between the NPI and communal grandiose self-
thoughts was not significant in all three samples (⫺.13 ⬍r⬍.04,
all ps⬎.17). Also, in support of the hypothesis that agentic and
communal grandiose self-thoughts constitute two independent di-
mensions, agentic and communal grandiose self-thoughts were
uncorrelated with each other in all three samples (⫺.13 ⬍r⬍
⫺.01, all ps⬎.19).
Discussion
The strength of the relation between the NPI and agentic gran-
diose self-thoughts is remarkably high, given that the NPI is a
self-report scale and the self-thoughts were assessed with an open-
ended measure and then rated by judges (who were unaware of
NPI scores). Study 1’s results are important for two reasons. First,
they provide direct evidence that the NPI assesses agentic narcis-
sism only. Second, the presence of communal grandiose self-
thoughts suggests the presence of communal narcissism.
Study 2: Are Agentic Narcissism and Communal
Narcissism Distinct Personality Traits?
In Study 2, we used the communal grandiose self-thoughts listed
by Study 1 participants as candidate items for a Communal Nar-
cissism Inventory (CNI). We conducted exploratory factor analy-
ses to select appropriate CNI items and then conducted confirma-
tory factor analyses to verify the CNI’s factor structure. Moreover,
we examined whether the CNI assesses a relatively stable person-
ality trait and tested our agency-communion model’s assumption
that agentic and communal narcissism are largely independent.
Furthermore, we examined the relation between communal narcis-
sism and NPI subfacets (Ackerman, Witt, Donnellan, Trzesni-
ewski, Robins, & Kashy, 2011). In continuation of Study 1’s
research question, we wanted to further explore the prevalence of
communal grandiose self-thoughts (compared to agentic grandiose
self-thoughts). Hence, we tested for the frequency/habituality
(Verplanken, Friborg, Wang, Trafimow, & Woolf, 2007) with
which agentic and communal narcissists hold agentic and commu-
nal grandiose self-thoughts, respectively. We also examined the
relation between communal narcissism and clinical/vulnerable
forms of narcissism (W. K. Campbell & Miller, 2011) as well as
unmitigated communion (Helgeson & Fritz, 1998). We further
investigated sex differences in agentic and communal narcissism
and tested for the relations between agentic and communal narcis-
sism and sex roles (Bem, 1974). In addition, we tested the relation
between agentic and communal narcissism on the one hand and
communal and exchange orientations in interpersonal relationships
on the other hand (Chen, Lee-Chai, & Bargh, 2001). Finally, we
explored the relation of communal narcissism to the five-factor
model of personality (Digman, 1990; McCrae & John, 1992;
Saucier & Goldberg, 1996) and to the eight factors of the Inter-
personal Adjective Lists (Wiggins, 1991).
Method
Participants. We tested Study 2’s hypotheses in Samples
4–9 (see Table 1).
Materials and procedure. Below, we describe those Study 2
measures that have not been used—and thus described—in Study
1. As a reminder, Table 2 provides details about administration
orders, information about which measure was used in which sam-
ple, and internal consistencies of the measures.
Candidate CNI items. We considered all communal grandi-
ose self-thoughts listed by Study 1 participants as candidate CNI
items. We deleted redundant communal grandiose self-thoughts,
and we expressed in abstract form concrete communal grandiose
self-thoughts in order to increase their relevance to participants at
large (e.g., “I am the best mother on this planet” 3“I am [going
to be] the best parent on this planet”). This resulted in a total of 34
communal grandiose self-thoughts as CNI candidate items. Partic-
ipants proceeded to rate each item (1 ⫽disagree strongly, 7⫽
agree strongly).
Subfacets of the NPI40. The NPI40 contains three subfacets
(Ackerman et al., 2011). The 11-item Leadership/Authority sub-
scale assesses extremely positive self-views largely in the agentic
859
COMMUNAL NARCISSISM
domain (e.g., “I am a born leader,” “I am an extraordinary per-
son”). The 10-item Grandiose Exhibitionism subscale assesses a
preoccupation with one’s looks and associated praise from others
(e.g., “I like to look at my body,” “I get upset when people don’t
notice how I look when I go out in public”). The four-item
Exploitativeness/Entitlement subscale assesses an exploitative in-
terpersonal orientation explicitly directed towards one’s personal
advantage (e.g., “I find it easy to manipulate people,” “I insist
upon getting the respect that is due to me”).
Frequency/habituality of grandiose self-thoughts. On the
page following the CNI, participants saw a list of the 16 CNI
items and were instructed to indicate how frequently/habitually
they had thoughts that paralleled the CNI items. Frequency/
habituality was assessed using the Mental Habits Index (Ver-
planken et al., 2007). Example items are “To think about myself
in the above described way, is something . . .” “. . . I do fre-
quently,” “. . . I do automatically,” and “. . . that’s typically
‘me’.” Participants responded using a 7-point scale (1 ⫽
strongly disagree, 7⫽strongly agree). On the page following
the NPI16, participants saw a list of the 16 agentically narcis-
sistic responses of the NPI16 and completed the Mental Habits
Index for these agentic grandiose self-thoughts.
Clinical/vulnerable forms of narcissism. The nine-item Nar-
cissistic Personality Disorder subscale of the Personality Diagnos-
tic Questionnaire–4 (Hyler, 1994) is the most frequently used
self-report measure of narcissistic personality disorder (Miller &
Campbell, 2008; C. Watson & Bagby, 2011). Items (e.g., “I have
accomplished far more than others give me credit for,” “Some
people think that I take advantage of others”) are completed using
a yes/no response format.
The 10-item Hypersensitive Narcissism Scale (Hendin & Cheek,
1997) is the most frequently used self-report measure of vulnerable
narcissism (Tamborski & Brown, 2011). Example items are “My
feelings are easily hurt by ridicule or the slighting remarks of
others,” and “I dislike sharing the credit of an achievement with
others.” Participants responded using a 5-point scale (1 ⫽disagree
strongly, 5⫽agree strongly).
Unmitigated communion. Jacobs and Scholl’s (2007) nine-
item German version of the Unmitigated Communion Scale (Hel-
geson & Fritz, 1998) is the best validated measure of unmitigated
communion. Example items are “I always place the needs of others
above my own,” and “I never find myself getting overly involved
in others’ problems” (reverse-scored). Participants responded us-
ing a 7-point scale (1 ⫽strongly disagree, 7⫽strongly agree).
Sex roles. T. Campbell, Gillaspy, and Thompson’s (1997)
20-item short form of the Bem Sex Role Inventory (Bem, 1981)
contains a 10-item Masculinity subscale (e.g., “I am a person, who
is . . .” “. . . assertive,” “. . . dominant”) and a 10-item Femininity
subscale (e.g., “I am a person, who is . . .” “. . . tender,” “. . .
compassionate”). Participants responded using a 7-point scale (1 ⫽
strongly disagree, 7⫽strongly agree).
Communal orientation. The 14-item Communal Orienta-
tion Scale (M. S. Clark, Ouellette, Powell, & Milberg, 1987)
assesses the degree to which a person values communally
oriented interpersonal relationships by behaving communally
him- or herself (e.g., “I often go out of my way to help another
person”) as well as by seeking others who behave communally
(e.g., “ I expect people I know to be responsive to my needs and
feelings”). Participants responded using a 7-point scale (1 ⫽
strongly disagree, 7⫽strongly agree).
Exchange orientation. The nine-item Exchange Orientation
Scale (Mills & Clark, 1994) assesses the degree to which a person
values equaling out gives and takes in interpersonal relationships
by expecting adequate returns from others (e.g., “When I give
something to another person, I generally expect something in
return”) as well as by providing adequate returns him- or herself
(e.g., “When someone buys me a gift, I try to buy that person as
comparable a gift as possible”). Participants responded using a
7-point scale (1 ⫽strongly disagree, 7⫽strongly agree).
Five factors of personality. Lang, Lu¨dtke, and Asendorpf’s
(2001) German version of the Big Five Inventory (BFI; Benet-
Martı´nez & John, 1998) contains an eight-item Extraversion sub-
scale (e.g., “I see myself as someone who . . .” “. . . has an
assertive personality,” “. . . is outgoing, sociable”), a 10-item
Openness subscale (e.g., “I see myself as someone who . . .” “. . .
is original, comes up with new ideas,” “. . . is curious about many
different things”), an eight-item Agreeableness subscale (e.g., “I
see myself as someone who . . .” “. . . is considerate and kind with
almost everyone,” “. . . likes to cooperate with others”), a nine-
item Conscientiousness subscale (e.g., “I see myself as someone
who . . .” “. . . is a reliable worker,” “. . . perseveres until the task
is finished”), and a seven-item Neuroticism subscale (e.g., “I see
myself as someone who . . .” “. . . is depressed, blue,” “. . . worries
a lot”). Participants responded using a 7-point scale (1 ⫽strongly
disagree, 7⫽strongly agree).
Interpersonal adjectives. The Interpersonal Adjective
Scales–Revised (Wiggins, 1991) contain an eight-item Assured-
Dominant subscale (e.g., “assertive,” “forceful”), an eight-item
Arrogant-Calculating subscale (e.g., “crafty,” “boastful”), an
eight-item Cold-Hearted subscale (e.g., “ruthless,” “cruel”), an
eight-item Aloof-Introverted subscale (e.g., “antisocial,” “intro-
verted”), an eight-item Unassured-Submissive subscale (e.g.,
“bashful,” “unauthoritative”), an eight-item Unassuming-
Ingenuous subscale (e.g., “unargumentative,” “uncunning”), an
eight-item Warm-Agreeable subscale (e.g., “charitable,” “kind”),
and an eight-item Gregarious-Extraverted subscale (e.g., “perky,”
“extraverted”). Participants responded using a 7-point scale (1 ⫽
strongly disagree, 7⫽strongly agree).
Results and Discussion
Exploratory factor analyses. We used Sample 4 data (pre-
dominantly British undergraduate participants) in order to conduct
the first exploratory factor analysis on the 34 CNI candidate items.
As expected, a one-factor solution emerged (Cattell, 1966). This
single factor had an eigenvalue of 13.85 and explained 40.72% of the
total variance. In order to match the length of the CNI with its agentic
counterpart (i.e., the NPI16), we chose 16 CNI items that (a) loaded
highly on the single factor (ⱖ.62), and (b) covered as many differing
communal traits as possible (e.g., helpfulness, interpersonal warmth,
trustworthiness). Table 3 shows the factor loadings of the 16 CNI
items. The Appendix displays the CNI, including instructions. The
correlation between the CNI and the mean of all 34 candidate items
was very high, r(195) ⫽.98, p⫽.001.
It is considered good practice in scale construction to repeat the
initial exploratory factor analysis within a different sample in order
to test whether the same items emerge for the final scale (Gorsuch,
860 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
1974). To this end, we repeated the exploratory factor analysis
described above using Sample 5 data (predominantly American
Internet participants). The single factor had an eigenvalue of 15.47
and explained 45.50% of the total variance. As shown in Table 3,
all 16 items previously selected for the CNI again loaded highly on
the single factor. The correlation between the CNI and the mean of
all 34 candidate items was again very high, r(120) ⫽.98, p⫽.001.
Table 3 also shows the factor loadings of the 16 CNI items from
an exploratory factor analysis that combined Samples 4 and 5.
Such an analysis is useful because the larger overall sample size
(N⫽315) reveals more stable estimates of individual items’ true
factor loadings.
In all, exploratory factor analyses in samples derived from two
countries revealed that the 34 communal grandiose self-thoughts
listed by Study 1 participants loaded on a single factor. This
supports the presence of a global trait that we labeled communal
narcissism. Based on these analyses, we chose 16 items for the
final CNI. The 16-item version correlated highly with the mean of
all 34 candidate items. This high correlation suggests that the CNI
captures communal narcissism broadly and exhaustively, given
that the candidate items reflect the entire range of communal
grandiose self-thoughts listed by Study 1 participants.
Confirmatory factor analyses. Confirmatory factor analy-
ses, using maximum-likelihood estimation, formed the next step of
scale construction. This step enabled us to address three questions.
First, do the 16 CNI items fit the expected one-factor solution
sufficiently well? Second, do the 16 CNI items form a factor
independent of the NPI16 items? Finally, are the results of the
confirmatory factor analyses the same across British undergradu-
ate participants and (predominantly) American Internet partici-
pants?
We conducted the first confirmatory factor analysis using Sam-
ple 6 data (predominantly British undergraduate students). We first
specified a structural model in which each of the 16 CNI items
served as an indicator (manifest variable) of communal narcissism
(latent variable). Following recommendations by Schumacker and
Lomax (2004), we allowed the error variances of the eight present-
focused communal grandiose self-thoughts as well as the eight
future-focused communal grandiose self-thoughts to correlate with
each other (Trope & Liberman, 2003). In line with expectations,
this one-factor model fit the data well (
2
/df ⫽3.21, comparative
fit index [CFI]⫽.96, root-mean-square error of approximation
[RMSEA]⫽.08; cf. Hu & Bentler, 1999).
1
Next, we added a second latent variable to the model. We
labeled this latent variable agentic narcissism and defined it with
16 observed variables—the items of the NPI16. We allowed the
two latent variables to intercorrelate. The latent-variable correla-
tion between agentic and communal narcissism was not significant
(r⫽.12, p⫽.11), supporting our prediction about the indepen-
dence of these two constructs. Due to well-known problems with
the internal consistency of the NPI (Brown & Tamborski, 2011),
the resultant two-factor model fit the data somewhat weakly
1
In this analysis, we assumed that differences between present- and future-
focused CNI items are due to systematic differences in the error variances of
these items. Alternatively, differences between these items may be due to a
meaningful two-dimensional structure underlying communalnarcissism. Spe-
cifically, the CNI may entail two components: present- and future-focused
communal narcissism. To test for this possibility, we compared the fit of the
original one-factor model of communal narcissism with the fit of a two-factor
model consisting of two correlated factors (i.e., present- and future-focused
communal narcissism). This two-factor model fit the data rather poorly,
2
/df ⫽4.76, CFI ⫽.85, RMSEA ⫽.11, whereas the one-factor model fit the
data significantly better, ⌬
2
⫽320.32, ⌬df ⫽50, p⫽.001. To examine this
issue further, we repeated twice all analyses reported in this article. In the first
set of analyses, we replaced the full 16-item CNI with the mean of the eight
present-focused CNI items (Items 1, 3, 6, 8, 10, 12, 13, and 15). In a second
set of analyses, we replaced the full 16-item CNI with the mean of the eight
future-focused CNI items (Items 2, 4, 5, 7, 9, 11, 14, and 16). The results
involving the present-focused CNI items were virtually identical to the results
involving the future-focused CNI items. Together, we found no support for the
utility of distinguishing between present- and future-focused CNI items. In-
deed, the confirmatory factor analyses show that the one-factor model of
communal narcissism is psychometrically superior to a distinction between
present- and future-focused communal narcissism.
Table 3
Items of the Communal Narcissism Inventory and Its Factor Loadings
Item
number Item
Factor loadings
Sample 4 Sample 5 Samples 4 and 5 combined
1 I am the most helpful person I know. .65 .68 .68
2 I am going to bring peace and justice to the world. .68 .65 .70
3 I am the best friend someone can have. .64 .71 .71
4 I will be well known for the good deeds I will have done. .79 .76 .79
5 I am (going to be) the best parent on this planet. .64 .68 .68
6 I am the most caring person in my social surrounding. .65 .67 .68
7 In the future I will be well known for solving the world’s problems. .67 .76 .73
8 I greatly enrich others’ lives. .72 .73 .76
9 I will bring freedom to the people. .68 .72 .73
10 I am an amazing listener. .68 .69 .68
11 I will be able to solve world poverty. .65 .65 .66
12 I have a very positive influence on others. .64 .75 .71
13 I am generally the most understanding person. .64 .74 .69
14 I’ll make the world a much more beautiful place. .74 .80 .79
15 I am extraordinarily trustworthy. .62 .52 .60
16 I will be famous for increasing people’s well-being. .70 .81 .77
861
COMMUNAL NARCISSISM
(
2
/df ⫽2.32, CFI ⫽.84, RMSEA ⫽.06).
2
(It has been argued that
the NPI’s low internal consistency reflects its role as a broad and
divergent construct; Miller & Campbell, 2011.) For the purpose of
validating the CNI, this is not a problem as long as the fit of our
agency-communion model of narcissism provides a better fit than
the fit of a rival one-factor model subsuming all 32 items of the
NPI16 and the CNI. In line with the agency-communion model,
the two-factor solution fit the data significantly better than the
one-factor solution (⌬
2
⫽259.16, ⌬df ⫽1, p⫽.001; fit statistics
for the one-factor model:
2
/df ⫽2.94, CFI ⫽.76, RMSEA ⫽.08).
Next, we tested the replicability of these findings using Sample
7 data (predominantly American Internet participants). The anal-
yses were identical to those mentioned above. Again, the 16 CNI
items described the communal narcissism factor well (
2
/df ⫽
2.62, CFI ⫽.96, RMSEA ⫽.07). And, again, our agency-
communion model of narcissism fit the data sufficiently well
(
2
/df ⫽2.28, CFI ⫽.84, RMSEA ⫽.06). More important, the
latent variable correlation between agentic and communal narcis-
sism in our model was marginal (r⫽.15, p⫽.07), while the
agency-communion model fit significantly better than the rival
one-factor model (⌬
2
⫽257.61, ⌬df ⫽1, p⫽.001; fit statistics
for the one-factor model:
2
/df ⫽2.90, CFI ⫽.75, RMSEA ⫽.08).
Overall, confirmatory factor analyses in samples derived from
two countries supported the CNI’s one-dimensional factor struc-
ture, and showed that the CNI and NPI items describe two distinct
facets of narcissism. Thus, the CNI complements the NPI. To-
gether, they are likely to cover the full range of grandiose narcis-
sism, because the agency and communion dimensions span a
circular structure including an exhaustive list of personality traits
(Abele et al., 2008; Helgeson, 1994; Wiggins, 1991).
Psychometric properties, descriptive statistics, and temporal
stability. Table 4 shows that the internal consistencies of the
CNI were excellent in all 13 applicable samples (.86 ⱕ␣sⱕ.95).
Table 4 also displays the means and standard deviations of the
CNI. As can be seen, the mean scores ranged around the midpoint
of the rating scale (3.30 ⱕMⱕ4.21; overall M⫽3.76), and the
standard deviation ranged around one (.79 ⱕSD ⱕ1.22; overall
SD ⫽1.05). In line with this, the CNI had a near-perfect normal
distribution across all participants. Next, we examined the tempo-
ral stability of the CNI relative to the NPI in Sample 6. The CNI’s
8-week test–retest reliability was high, r(322) ⫽.71, p⫽.001, and
similar to the NPI’s 8-week test–retest reliability, r(322) ⫽.79,
p⫽.001.
In all, the psychometric properties of the CNI were commend-
able. The internal consistency of the scale was excellent and its
temporal stability was high. Together, these findings suggest that
the CNI truly captures a personality trait.
Relations to the NPI and NPI40 subfacets. The relation
between the CNI and the NPI was moderately positive across all
10 applicable samples, omnibus correlation: r(1,971) ⫽.27, p⫽
.001 (see Table 4). This is consistent with our proposition that both
measures cover facets of grandiose narcissism.
Next, we examined the relation between the CNI and the three
subfacets of the NPI40 (Ackerman et al., 2011). For these analy-
ses, we aggregated all samples that included the CNI as well as the
NPI40 (i.e., Samples 8 and 15). The leadership/authority facet
concerns highly positive self-views and thus most strongly taps
into narcissistic self-motives, as indicated by a substantial relation
with self-esteem in Ackerman et al.’s (2011) research, (3,349) ⫽
.42, p⬍.001. Hence, we expected (see also Study 3) and found a
positive relation between communal narcissism and the leadership/
authority facet, (423) ⫽.41, p⬍.001 (following Ackerman et
al., 2011, the other two NPI subfacets were controlled). The
grandiose exhibitionism facet concerns physical appearance,
which is largely agentic (Gebauer, Leary, & Neberich, in press).
Also, this facet is unrelated to narcissistic self-motives, as indi-
cated by its independence to self-esteem in Ackerman et al.’s
research, (3,349) ⫽.03, ns. Hence, we expected (see also Study
3) and found no relation between communal narcissism and the
grandiose exhibitionism facet, (423) ⫽.07, p⫽.13 (controlling
for the other two NPI subfacets). Finally, the exploitativeness/
entitlement facet concerns explicitly devaluing others. Thus, com-
munal narcissists should explicitly reject such exploitativeness. In
line with this, we found a negative relation between communal
narcissism and the exploitativeness/entitlement facet, (423) ⫽
⫺.10, p⫽.03.
Finally, we also tested whether the strength of correlation be-
tween the CNI and the NPI16 depends on administration order.
Sample 7 participants who were randomly assigned to complete
the CNI immediately before the NPI showed a correlation of
r(153) ⫽.22, p⫽.006, between the two measures. Sample 7
participants who were randomly assigned to complete the CNI
immediately after the NPI showed a correlation of r(174) ⫽.27,
p⫽.001 between the two measures. A regression analysis, with
dummy coding for measure order, revealed that order did not
moderate the strength of correlation between the CNI and NPI,
(321) ⫽⫺.08, p⫽.63.
In all, correlations between the CNI and NPI further corroborate
that agentic narcissism and communal narcissism are distinct con-
structs. Yet the moderately positive correlation between the two
measures is in line with our theoretical argument that both mea-
sures assess facets of narcissism. Further, the relations between the
CNI and NPI40 subfacets were similar to, but somewhat weaker
than, the relations between self-esteem and NPI40 subfacets. This
provides support for our hypotheses that agentic and communal
narcissism partially overlap because both forms of narcissism are
based on the same narcissistic self-motives (e.g., for self-esteem;
Baumeister & Vohs, 2001). Finally, CNI and NPI administration
order did not moderate their intercorrelation; these measures were
robust to order effects.
Frequency/habituality of communal grandiose self-thoughts.
In Study 1, we used the occurrence of communal grandiose self-
thoughts as an indicator of communal narcissism’s existence. But
how frequently/habitually do communal narcissists hold commu-
nal grandiose self-thoughts? In particular, are communal grandiose
self-thoughts among communal narcissists more or less frequent/
habitual than agentic grandiose self-thoughts among agentic nar-
cissists? To answer this question, we compared (in Sample 9) the
frequency/habituality of communal grandiose self-thoughts among
communal narcissists (i.e., participants scoring above the theoret-
ical midpoint on the CNI) with the frequency/habituality of agentic
2
We conducted all analyses using the CNI twice, once as reported in the
ensuing main text and once including sex as a moderator of CNI relations.
We consistently found no moderation by sex in Study 2 (all s⬍.09, all
ps⬎.16), Study 3 (all s⬍|.11|, all ps⬎.23), Study 4 (all s⬍|.09|, all
ps⬎.34), and Study 5 (all s⬍.17, all ps⬎.12).
862 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
grandiose self-thoughts among agentic narcissists (i.e., participants
scoring above the theoretical midpoint on the NPI). The results
revealed that communal narcissists possess communal grandiose
self-thoughts as frequently/habitually (M⫽3.87, SD ⫽1.33) as
agentic narcissists possess agentic grandiose self-thoughts (M⫽
4.04, SD ⫽1.49, t⫽1.30, p⫽.21). More precisely, frequency/
habituality indices among communal narcissists and among agen-
tic narcissists significantly differed from the low point (i.e., 1) of
the Mental Habits Index (ts⬎14.08, ps⬍.001). This result
indicates that communal narcissists frequently/habitually think
about themselves in communally grandiose ways, whereas agentic
narcissists frequently/habitually think about themselves in agenti-
cally grandiose ways. Moreover, frequency/habituality indices
among communal narcissists and among agentic narcissists did not
differ significantly from the theoretical midpoint (i.e., 4) of the
Mental Habits Index (ts⬍|.91|, ps⬎.36). This result indicates
that communal and agentic narcissists both think about themselves
in grandiose ways with considerable frequency/habituality (Ver-
planken et al., 2007).
Relations to clinical/vulnerable narcissism and unmitigated
communion. The agency-communion model describes agentic
and communal narcissism as two complementary forms of gran-
diose narcissism. At the same time, past literature has described
grandiose narcissism as the intrapersonally adaptive form of nar-
cissism (W. K. Campbell, 2001; Sedikides et al., 2004). In con-
trast, the clinical and vulnerable forms of narcissism have been
described as intrapersonally maladaptive (Hendin & Cheek, 1997;
Miller & Campbell, 2008). Hence, we expected that communal
narcissism would relate more strongly to agentic narcissism than to
clinical and vulnerable narcissism, with relatively low relations
between communal narcissism and clinical and vulnerable narcis-
sism. In support of these hypotheses, communal narcissism related
moderately to agentic narcissism, r(224) ⫽.39, p⫽.001, but
weaker to clinical narcissism, r(224) ⫽.23, p⫽.001, and even
slightly negatively to vulnerable narcissism, r(224) ⫽⫺.14, p⫽
.04 (Sample 8). The correlation between communal and agentic
narcissism was significantly higher than the correlations between
communal and clinical/vulnerable narcissism (2.48 ⬍zs⬍5.56,
ps⬍.02).
Unmitigated communion is defined as “a focus on and involve-
ment with others to the exclusion of the self” (Helgeson & Fritz,
1998, p. 173). Given this definition, is unmitigated communion
identical to communal narcissism? Theoretically, communal nar-
cissists and individuals scoring high on unmitigated communion
should share the conviction that they focus on and are particularly
involved with others. In contrast to individuals scoring high on
unmitigated communion, however, communal narcissists should
not show this other-focus to the exclusion of the self. To the
contrary, claims of extreme communion among communal narcis-
sists should serve as means to bolster their unrealistic and inflated
self-views. Together, we expected a modest positive relation be-
tween unmitigated communion and communal narcissism. In con-
trast, we expected a modest negative relation between unmitigated
communion and agentic narcissism, because agentic narcissists not
only are self-focused (as communal narcissists are) but also tend to
devalue communal other-focus (Sedikides et al., 2002). We tested
these hypotheses in Sample 9, and found a positive relation be-
tween unmitigated communion and communal narcissism,
r(313) ⫽.20, p⫽.001, and a negative relation between unmiti-
gated communion and agentic narcissism, r(314) ⫽⫺.14, p⫽.02.
Thus, communal narcissism and unmitigated communion are
clearly distinct from each other.
Sex differences and relations to sex roles. A specific pattern
of sex differences in agentic and communal narcissism follows the
agency-communion model of narcissism. In particular, the agency-
communion model conceptualizes agentic narcissism as an
agency-agency trait. Given that men score higher on agency than
women (Bem, 1974), the agentic mean component as well as the
agentic trait component of agentic narcissism should lead to higher
agentic narcissism in men than in women. Indeed, much prior
research supports this prediction (Farwell & Wohlwend-Lloyd,
1998; Foster, Campbell, & Twenge, 2003; Gabriel et al., 1994;
Tschanz, Morf, & Turner, 1998). What predictions does the
agency-communion model make regarding communal narcissism?
The agency-communion model conceptualizes communal narcis-
sism as an agency-communion trait. Because men score higher on
agency than women, but women score higher on communion than
men (Bem, 1974), the agentic traits component should lead to
Table 4
Descriptive Statistics of the Communal Narcissism Inventory
Sample Reliability (␣) Relation with NPI (r)MSD
4 .92 — 3.11 0.98
5 .94 — 3.93 1.11
6 (Wave 1) .89 .20
ⴱⴱⴱ
3.59 0.79
6 (Wave 2) .90 .30
ⴱⴱⴱ
3.68 0.89
7 .91 .24
ⴱⴱⴱ
4.21 0.98
8 .92 .39
ⴱⴱⴱ
4.08 1.12
9 .95 .39
ⴱⴱⴱ
3.30 1.11
10 .89 .29
ⴱⴱⴱ
3.71 0.74
11 .86 .00 4.18 0.77
12 .94 .32
ⴱⴱⴱ
3.52 1.22
13 .91 — 4.18 0.97
14 .91 .13 4.06 1.04
15 .91 .20
ⴱⴱ
3.52 0.86
16 .91 — 3.86 0.99
Note. Dashes indicate that no information was available. NPI ⫽Narcissistic Personality Inventory.
ⴱⴱ
pⱕ.01.
ⴱⴱⴱ
pⱕ.001.
863
COMMUNAL NARCISSISM
higher communal narcissism in men than in women, but the
communal means component should lead to higher communal
narcissism in women than in men. Overall, then, men and women
should not differ in communal narcissism. To examine these
predictions, we first tested for independent main effects of sex on
agentic narcissism scores across all applicable samples. We aggre-
gated available data because of small numbers of male participants
in most samples (see Table 1). A one-factor analysis of variance
yielded a main effect of sex on agentic narcissism. As in prior
research, men (M⫽.36, SD ⫽.21; scale: 0–1) scored higher on
the NPI than women (M⫽.28, SD ⫽.18), F(1, 1961) ⫽61.58,
p⫽.001. Next, we repeated this analysis with the CNI as the sole
dependent variable. No main effect of sex on the CNI emerged,
F(1, 2345) ⫽.13, p⫽.72. Men (M⫽3.80, SD ⫽1.10; scale: 1–7)
and women (M⫽3.78, SD ⫽1.02) did not differ on communal
narcissism.
Do relations to sex roles parallel these sex differences? Our
evidence (from Sample 9) suggests that they do. Specifically,
agentic narcissism was positively related to masculinity, r(313) ⫽
.43, p⫽.001, and unrelated to femininity, r(313) ⫽⫺.01, p⫽.88,
rendering agentic narcissism more masculine than feminine (z⫽
6.90, p⫽.001). Communal narcissism, however, was positively
related to masculinity, r(313) ⫽.28, p⫽.001, as well as to
femininity, r(313) ⫽.32, p⫽.001, rendering communal narcis-
sism not any more masculine than feminine (or vice versa; z⫽.67,
p⫽.50).
Relations to communal and exchange orientations in inter-
personal relationships. M. S. Clark and colleagues distin-
guished between two types of interpersonal relationship orien-
tations—namely, a communal orientation (M. S. Clark et al.,
1987) and an exchange orientation (Mills & Clark, 1994).
Individuals high on communal orientation value interpersonal
relationships in which self and others behave in an other-
profitable manner. In contrast, individuals high on exchange
orientation value interpersonal relationships in which self and
others behave in a tit-for-tat manner. To what degree do agentic
and communal narcissists value these interpersonal orienta-
tions? We explored this question in Sample 9. We expected that
communal narcissists, but not agentic narcissists, would report
higher communal and exchange orientations because both ori-
entations resonate to some degree with communal means for
goal attainment. Indeed, we found that communal narcissism
was positively related to a communal orientation, r(313) ⫽.24,
p⫽.001, and to an exchange orientation, r(313) ⫽.16, p⫽
.005. At the same time, agentic narcissism was related neither
to a communal orientation, r(314) ⫽⫺.02, p⫽.76, nor to an
exchange orientation, r(314) ⫽.04, p⫽.50. Again, the strength
of these relations illustrates that communal narcissism is dis-
tinct from communal and exchange orientations.
Relations to the five-factor model of personality and inter-
personal adjectives. How is communal narcissism related the
five-factor model of personality (Digman, 1990; McCrae & John,
1992; Saucier & Goldberg, 1996)? The agency-communion model
of narcissism makes clear predictions regarding these relations.
Communal narcissism contains agentic elements (at the trait
level—i.e., narcissism) as well as communal elements (at the
means level—i.e., communion). As such, and in line with the
above results for sex roles, we expected communal narcissism to
be positively related to agentic factors in the five-factor model
(i.e., Extraversion and Openness; W. K. Campbell et al., 2004;
Paulhus & John, 1998), and we expected communal narcissism to
be positively related to communal factors in the five-factor model
(i.e., Agreeableness and Conscientiousness; W. K. Campbell et al.,
2004; Paulhus & John, 1998). Further, we expected a negative
relation to Neuroticism, due to Neuroticism’s strong link to self-
esteem (Judge, Erez, Thoresen, & Bono, 2002), with self-esteem
being one of narcissists’ presumed self-motives (Baumeister &
Vohs, 2001). The upper part of Table 5 shows that these expec-
tations were met (using Sample 9 data). Table 5 also shows the
relations between agentic narcissism and the five factors of the
five-factor model. These relations square with much prior research
(Ames et al., 2006; W. K. Campbell, Rudich, & Sedikides, 2002;
Paulhus & John, 1998; Sedikides et al., 2004) as well as with the
theoretical predictions derived from the agency-communion
model, which conceptualizes agentic narcissism as an agency-
agency trait.
Table 5
Relations to the Five-Factor Model of Personality and Interpersonal Adjectives
Criterion
Zero-order correlation Multiple regression
CNI NPI CNI NPI
Extraversion .25
ⴱⴱⴱ
.39
ⴱⴱⴱ
.12
ⴱ
.33
ⴱⴱⴱ
Openness .23
ⴱⴱⴱ
.27
ⴱⴱⴱ
.14
ⴱ
.21
ⴱⴱⴱ
Agreeableness .17
ⴱⴱ
⫺.18
ⴱⴱ
.26
ⴱⴱⴱ
⫺.27
ⴱⴱⴱ
Consientiousness .16
ⴱⴱ
.12
ⴱ
.12
ⴱ
.07
Neuroticism ⫺.21
ⴱⴱⴱ
⫺.21
ⴱⴱⴱ
⫺.14
ⴱⴱ
⫺.16
ⴱⴱ
Assured-dominant .27
ⴱⴱⴱ
.45
ⴱⴱⴱ
.10
†
.40
ⴱⴱⴱ
Arrogant-calculating .23
ⴱⴱⴱ
.29
ⴱⴱⴱ
.13
ⴱ
.24
ⴱⴱⴱ
Cold-hearted ⫺.12
ⴱ
.11
ⴱ
⫺.18
ⴱⴱ
.18
ⴱⴱ
Aloof-introverted ⫺.27
ⴱⴱⴱ
⫺.17
ⴱⴱ
⫺.23
ⴱⴱⴱ
⫺.07
Unassured-submissive ⫺.11
ⴱ
⫺.34
ⴱⴱⴱ
.02 ⫺.32
ⴱⴱⴱ
Unassuming-ingenuous .07 ⫺.27
ⴱⴱⴱ
.20
ⴱⴱⴱ
⫺.33
ⴱⴱⴱ
Warm-agreeable .25
ⴱⴱⴱ
⫺.11
†
.31
ⴱⴱⴱ
⫺.23
ⴱⴱⴱ
Gregarious-extraverted .34
ⴱⴱⴱ
.23
ⴱⴱⴱ
.30
ⴱⴱⴱ
.11
†
Note. CNI ⫽Communal Narcissism Inventory; NPI ⫽Narcissistic Personality Inventory.
†
pⱕ.10.
ⴱ
pⱕ.05.
ⴱⴱ
pⱕ.01.
ⴱⴱⴱ
pⱕ.001.
864 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
How is communal narcissism related with the eight factors in
the Interpersonal Adjective Lists (Wiggins, 1991)? Our predic-
tions again follow the agency-communion model of narcissism
and square with our findings regarding sex roles and the five
factors from the five-factor model. Specifically, communal
narcissism should be positively related to factors reflecting high
agency (i.e., assured-dominant) and high communion (i.e.,
warm-agreeable). Conversely, communal narcissism should be
negatively related to factors reflecting low agency (i.e.,
unassured-submissive) and low communion (i.e., cold-hearted).
The Interpersonal Adjective Lists also contain a factor simul-
taneously reflecting high agency and high communion (i.e.,
gregarious-extraverted), and it follows from the conceptualiza-
tion of communal narcissism (as an agency-communion trait)
that it should be particularly positively related to this factor.
Conversely, the Interpersonal Adjective Lists also contain a
factor simultaneously reflecting low agency and low commu-
nion (i.e., aloof-introverted), and it follows from the conceptu-
alization of communal narcissism (as an agency-communion
trait) that it should be particularly negatively related to this
factor. The lower part of Table 5 shows that these expectations
were met (using Sample 9 data). Table 5 also illustrates the
relations between agentic narcissism and the Interpersonal Ad-
jective Lists. These relations square with much prior research
(Bradlee & Emmons, 1992; Gurtman, 1992; Miller, Price, Gen-
tile, Lynam, & Campbell, 2012) as well as with the theoretical
predictions derived from the agency-communion model, which
conceptualizes agentic narcissism as an agency-agency trait.
Study 3: Do Communal and Agentic Narcissists Have
the Same Self-Motives?
On theoretical grounds, narcissists should differ from nonnar-
cissists in regard to their core self-motives. NPI narcissists indeed
have strong desires for grandiosity (Brown et al., 2009; Raskin et
al., 1991b; Rosenthal et al., 2007), self-esteem (W. K. Campbell,
2001; Gregg & Sedikides, 2010; Sedikides et al., 2004), entitle-
ment (Brown et al., 2009; W. K. Campbell et al., 2004; Morf &
Rhodewalt, 2001), and power (W. K. Campbell, Foster, & Finkel,
2002; Horton & Sedikides, 2009; Raskin & Novacek, 1991; Rose
& Campbell, 2004). Our model anticipates that communal narcis-
sists would share these core self-motives with agentic (i.e., NPI)
narcissists. Study 3 tested this hypothesis.
We began with an attempt to replicate past research showing
that agentic narcissists have heightened self-motives for grandios-
ity, self-esteem, entitlement, and power. We also aimed to show
that these self-motives are the same for communal narcissists. In
addition, we tested whether agentic and communal narcissism each
explains unique variance in the four core self-motives. Finally, we
examined the relation between NPI and CNI on the one hand and
the four self-motives on the other. Individuals who score highly on
both the NPI and the CNI should have particularly strong self-
motives for grandiosity, self-esteem, entitlement, and power. This
effect may be due to an additive effect of agentic and communal
narcissism (i.e., two independent main effects) or alternatively due
to a multiplicative effect (i.e., an interaction between the NPI and
the CNI). We put these two possibilities to an empirical test.
Method
Participants. We tested Study 3’s hypotheses in Samples 7,
8, and 10 (see Table 1).
Materials and procedure. Below, we describe those Study 3
measures that have not been used—and thus described—in Studies
1 and 2. As a reminder, Table 2 provides details about adminis-
tration orders, information about which measure was used in which
sample, and internal consistencies of the measures.
Grandiosity. The 16-item Narcissistic Grandiosity Scale
(Rosenthal et al., 2007; see also Brown et al., 2009) assesses
grandiose self-views. Example items, preceded by the stem “I
usually feel that I am . . .”, are “. . . perfect,” “. . . superior,” “. . .
glorious.” Participants responded using a 7-point scale that ranged
from 1 (definitely wrong)to7(definitely right).
Self-esteem. The 16-item Self-Liking/Self-Competence
Scale—Revised (Tafarodi & Swann, 2001) assesses self-esteem
broadly (Tafarodi & Milne, 2002; Verplanken et al., 2007). Ex-
ample items are “I feel great about who I am” and “I sometimes
fail to fulfill my goals” (reverse-scored). Participants responded
using a 7-point scale from 1 (definitely wrong)to7(definitely
right).
Entitlement. The nine-item Psychological Entitlement Scale
(W. K. Campbell et al., 2004) assesses beliefs about own entitle-
ment. Example items are “I honestly feel I’m just more deserving
than others” and “I do not necessarily deserve special treatment”
(reverse-scored). Participants responded using a 7-point scale from
1(definitely wrong)to7(definitely right).
Power. Riketta’s (2008) 20-item adaptation of the Need for
Power Scale (Schmidt & Frieze, 1997) assesses a desire for power.
Example items are “I find satisfaction in having influence over
others” and “I would like doing something important where people
looked up to me.” Participants responded using a 7-point scale
from 1 (definitely wrong)to7(definitely right).
Additionally, we administered Bernard et al.’s (2006) five-item
version of the Power Values Scale (Schwartz, 1992). This scale
assesses the importance that people place on power values. Exam-
ple items are “authority (the right to lead or command)” and
“influential (having an impact on people and events).” Participants
rated the importance of each value as a guiding principle in their
life (⫺1⫽opposed to my values, 7⫽extremely important).
Results
Table 6 displays the zero-order correlations between the four
core self-motives and the NPI (Data Column 1) as well as the CNI
(Data Column 2) in Samples 7, 8, and 10. Replicating past re-
search, agentic narcissists reported heightened levels of self-
motives for grandiosity, self-esteem, entitlement, and power. More
important, communal narcissists also reported heightened levels of
these self-motives. Next, we examined the relations between the
four self-motives and the NPI independently of the CNI (see Table
6’s Data Column 3), as well as the relations between these self-
motives and the CNI independently of the NPI (see Table 6’s Data
Column 4). As hypothesized, agentic and communal narcissists,
independently of each other, reported heightened levels of the core
self-motives.
Finally, we examined the interaction between NPI and CNI on
each of the four self-motives. To this end, we conducted a series of
865
COMMUNAL NARCISSISM
multiple regression analyses, simultaneously regressing each self-
motive on the NPI (centered), the CNI (centered), and their inter-
action term. Table 6’s final data column displays the results of
these interactions. As can be seen, the standardized regression
coefficients of the interactions were consistently close to zero. The
plevels were far from significance.
Discussion
This study showed that agentic (i.e., NPI) and communal (i.e.,
CNI) narcissists share the same self-motives for grandiosity, self-
esteem, entitlement, and power. Agentic and communal narcis-
sism, independently of each other, explained variance in these
self-motives. In other words, agentic and communal narcissism
exhibited additive effects on grandiosity, self-esteem, entitlement,
and power. We additionally tested whether the two facets of
narcissism interact in predicting the four core self-motives. We
consistently found no evidence for such an interaction. Nonethe-
less, narcissists scoring high on the NPI and the CNI also scored
high on each self-need, but this was due to additive rather than
multiplicative effects of the NPI and the CNI.
Study 4: Do Agentic and Communal Narcissists Differ
in How They Satisfy Their Self-Motives?
A central tenet of the agency-communion model is that agentic
and communal narcissists share core self-motives, but differ in the
means by which they strive to meet them. Agentic narcissists rely
on agency but not on communion, whereas communal narcissists
rely on communion but not on agency. We tested this tenet in
Study 4.
As a reminder, W. K. Campbell, Rudich, and Sedikides (2002)
showed that agentic (i.e., NPI) narcissists manifested higher better-
than-average judgments on agentic, but not on communal, attri-
butes. We attempted to replicate this finding, while also expecting
that communal (i.e., CNI) narcissists would exhibit pronounced
better-than-average judgments on communal, but not on agentic,
attributes. In addition, we hypothesized that overclaiming knowl-
edge on communal topics would be related to communal, but not
agentic, narcissism, whereas overclaiming knowledge on agentic
topics would be related to agentic, but not communal, narcissism.
Such overclaiming results would conceptually replicate the better-
than-average results, using a criterion-based rather than a social
comparative measure of self-enhancement. Overclaiming results in
line with our hypotheses would also complement and extend the
finding that agentic narcissists overclaim knowledge on agentic
topics (Paulhus & Harms, 2004; Paulhus et al., 2003).
Method
Participants. We tested Study 4’s hypotheses in Samples 10
and 12–15 (see Table 1).
Materials and procedure. Below, we describe those Study 4
measures that have not been used—and thus described—in Studies
1–3 (see also Table 2).
Better-than-average task. W. K. Campbell, Rudich, and
Sedikides’s (2002) agency versus communion adaptation of the
better-than-average task (Alicke, 1985; Alicke & Govorun, 2005)
asks participants to indicate their perceived stand on several attri-
butes relative to the average person (0 ⫽much less than the
average person, 8⫽much more than the average person). In
Sample 10, the four attributes covering agency were “assertive,”
“ambitious,” “dominant,” and “blunt,” whereas the four attributes
covering communion were “helpful,” “caring,” “understanding,”
and “trustworthy.” In Sample 12, the six attributes covering
agency were “successful,” “ambitious,” “extraverted,” “stupid”
(reverse-scored), “shy” (reverse-scored), and “quiet” (reverse-
scored), whereas the six attributes covering communion were
“helpful,” “kind,” “honest,” “rude” (reverse-scored), “immoral”
(reverse-scored), and “cold” (reverse-scored).
Overclaiming task. In the original overclaiming task (Paulhus
et al., 2003), participants indicated their familiarity with specific
instantiations of several academic topics. For example, within the
topic of “physical sciences,” participants rated their familiarity
with “alloy,” “photon,” and “ultra-lipid.” Participants were not told
that some of the items are foils (here: “ultra-lipid”). Paulhus et al.
(2003) computed overclaiming and knowledge scores for each
participant based on the familiarity ratings of these foils relative to
the real items (Macmillan & Creelman, 1991).
To assess overclaiming on communal topics, we presented par-
ticipants with four communal topics in Samples 13 and 14: “hu-
manitarian aid organizations,” “nature and animal protection or-
Table 6
Narcissism and the Four Core Narcissistic Self-Motives
Criterion
Predictors
Zero-order correlation Multiple regression
NPI CNI NPI CNI NPI ⫻CNI
Grandiosity .36
ⴱⴱⴱ
.37
ⴱⴱⴱ
.27
ⴱⴱ
.29
ⴱⴱ
⫺.09
Self-esteem
Sample 6 .30
ⴱⴱⴱ
.30
ⴱⴱⴱ
.25
ⴱⴱⴱ
.25
ⴱⴱⴱ
⫺.04
Sample 10 .42
ⴱⴱⴱ
.47
ⴱⴱⴱ
.31
ⴱⴱⴱ
.38
ⴱⴱⴱ
⫺.01
Entitlement .27
ⴱⴱ
.38
ⴱⴱⴱ
.18
†
.33
ⴱⴱⴱ
.08
Power
Need for power .35
ⴱⴱⴱ
.47
ⴱⴱⴱ
.35
ⴱⴱⴱ
.47
ⴱⴱⴱ
.04
Power values .30
ⴱⴱⴱ
.37
ⴱⴱⴱ
.30
ⴱⴱⴱ
.37
ⴱⴱⴱ
.09
Note. NPI ⫽Narcissistic Personality Inventory; CNI ⫽Communal Narcissism Inventory.
†
pⱕ.10.
ⴱⴱ
pⱕ.01.
ⴱⴱⴱ
pⱕ.001.
866 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
ganizations,” “parenting and childcare,” and “international health
charities.” Participants rated their familiarity with 17 instantiations
of each topic (0 ⫽I never heard of it, 6⫽I am very familiar with
it). Thirteen of these instantiations were real and four were foils.
Examples of “humanitarian aid organization” items are “Red Cross
International” (real), “Doctors Without Borders/Me´decins Sans
Frontie`res” (real), and “International Well-Being Fund (IWBF)”
(foil). Examples of “nature and animal protection organization”
items are “Greenpeace” (real), “WWF International” (real), and
“WildlifeProtected” (foil). Examples of “parenting and childcare”
items are “Overparenting” (real), “Declaration of the Rights of the
Child” (real), and “UN Act Against Childism (UNAC)” (foil).
Lastly, examples of “international health charity” items are “The
Stroke Association” (real), “International Children’s Heart Foun-
dation” (real), and “Asch AIDS Aid (AAA)” (foil). We used
Paulhus et al.’s (2003) commonsense signal detection formulae to
calculate overclaiming and true knowledge scores for each of the
four communal topics.
Based on the results of Samples 13 and 14, we devised a 12-item
short form of our communal overclaiming task and complemented
this short form with a parallel 12-item agentic overclaiming task.
This allowed us to replicate conceptually the results obtained with
the agentic versus communal better-than-average task, and we
attempted this replication in Sample 15. The short form of the
communal overclaiming task contained the same four communal
topics as did the long form described above. However, each topic
only included three items: two real items and one foil. The items
of the short form are those given as example items for the long
form above. The parallel agentic overclaiming task consisted of the
following four agentic topics: “international stock market” (items:
“Nikkei” [real], “blue chips” [real], “Alpha Centauri Index [ACI]”
[foil]), “chemistry & physics” (items: “the theory of general rela-
tivity” [real], “thermodynamics” [real], “the Mander periodical
equation” [foil]), “market principles” (items: “Nash equilibrium”
[real], “game theory” [real], “Satured market hub” [foil]), and
“leading educational institutions” (items: “Massachusetts Institute
of Technology [MIT]” [real], “London School of Economics and
Political Science [LSE]” [real], “the Wall Institute Berlin [WIB]”
[foil]).
Results
Better-than-average judgments. Table 7 displays the rela-
tions between narcissism and better-than-average judgments. First,
Data Column 1 shows the results of Sample 10. In line with the
agency-communion model, communal narcissism was positively
related to communal better-than-average judgments, but not to
agentic better-than-average judgments, and the difference between
the two effects was marginal (z⫽1.70, p⫽.09). At the same time,
agentic narcissism was positively related to agentic better-than-
average judgments, but not to communal better-than-average judg-
ments, and the difference between the two effects was significant
(z⫽4.78, p⫽.001).
Second, Data Column 2 shows the results of Sample 12. Again,
communal narcissism was positively related to communal better-
than-average judgments, but not to agentic better-than-average
judgments, and the difference between the two effects was signif-
icant (z⫽3.70, p⫽.001). At the same time, agentic narcissism
was again positively related to agentic better-than-average judg-
ments, but not to communal better-than-average judgments, and
the difference between the two effects was also significant (z⫽
7.92, p⫽.001).
Overclaiming. Table 8 displays the relations between narcis-
sism and overclaiming. First, Data Columns 1 and 2 show the
results of Sample 13, which only included measures of communal
narcissism and communal overclaiming. In line with the agency-
communion model, communal narcissism was positively related to
communal overclaiming, but not to better knowledge in the com-
munal domain, and the difference between the two effects was
significant (z⫽3.36, p⫽.001).
Second, Data Column 2 shows the results of Sample 14, which
included measures of agentic and communal narcissism and com-
munal overclaiming. Replicating Sample 13 results, communal
narcissism was positively related to communal overclaiming, but
negatively related to better knowledge in the communal domain,
and the difference between the two effects was significant (z⫽
3.64, p⫽.001). Furthermore, agentic narcissism was related
neither to communal overclaiming nor to communal knowledge,
and the difference between communal and agentic narcissisms’
relations to communal overclaiming was significant (z⫽3.63, p⫽
.001).
Third, Data Column 3 shows the results of Sample 15, which
included measures of agentic and communal narcissism and agen-
tic and communal overclaiming. Replicating Sample 13 results,
communal narcissism was positively related to communal over-
claiming, but unrelated to knowledge in the communal domain,
and the difference between the two effects was significant (z⫽
2.14, p⫽.03). Furthermore, replicating Sample 14 results, agentic
narcissism was negatively related to communal overclaiming and
Table 7
Narcissism and Better-Than-Average Judgments
Criterion
Simultaneous predictors
Sample 10 Sample 12
CNI NPI CNI NPI
Communal better-than-average task .33
ⴱⴱⴱ
⫺.14 .41
ⴱⴱⴱ
⫺.38
ⴱⴱⴱ
Agentic better-than-average task .11 .48
ⴱⴱⴱ
⫺.03 .58
ⴱⴱⴱ
Note. CNI ⫽Communal Narcissism Inventory; NPI ⫽Narcissistic Personality Inventory.
ⴱⴱⴱ
pⱕ.001.
867
COMMUNAL NARCISSISM
unrelated to communal knowledge, and the difference between
communal and agentic narcissisms’ relations to communal over-
claiming was significant (z⫽3.85, p⫽.001). Finally, agentic
narcissism was positively related to agentic overclaiming, but
unrelated to knowledge in the agentic domain, and the difference
between the two effects was significant (z⫽3.09, p⫽.002). At
the same time, communal narcissism was related neither to agentic
overclaiming nor to agentic knowledge, and the difference be-
tween agentic and communal narcissisms’ relations to agentic
overclaiming was significant (z⫽2.04, p⫽.04).
Discussion
Across five independent samples, communal narcissists mani-
fested exaggerated better-than-average judgments on communal,
but not on agentic, attributes, and they overclaimed their knowl-
edge on communal, but not on agentic, topics. In addition, there
was no evidence that communal narcissists actually possessed
more communal knowledge than nonnarcissists. In fact, in all three
studies, communal narcissists possessed less communal knowl-
edge than nonnarcissists, albeit this effect was only significant in
Sample 14. This pattern is consistent with the anecdotal know-it-
all, who sees no need for knowledge acquisition because he knows
everything (including foils) anyway. Irrespective, the overall result
pattern shows that communal narcissists’ self-ratings reflect style
rather than substance: Communal narcissists believe that they are
particularly advanced in communal matters, but they are not. This
parallels the pattern of results obtained with agentic (i.e., NPI)
narcissists’ beliefs about agentic matters (Paulhus & Harms, 2004;
Paulhus et al., 2003; Paulhus & Williams, 2002). Finally, in the
current study, agentic narcissism was unrelated to overclaiming
and true knowledge on communal topics. This finding furthers the
agency-communion model’s tenet that agentic and communal nar-
cissists differ in their means of satisfying their core self-motives.
Study 5: Can Communal Narcissism Explain Weak
Self–Other Agreement Regarding Communal
Behaviors and Traits?
Studies 1–4 support the agency-communion model of grandiose
narcissism. These studies show that the fundamental distinction
between agency and communion helps to explain more fully
grandiose narcissism. Yet how relevant to the broader personality
literature is the construct of communal narcissism? We argue that
communal narcissism can help in solving longstanding puzzles in
the personality literature. A relatively weak self–other agreement
regarding communal behaviors and traits is one such example
(Allik et al., 2010; Gosling et al., 2003; John & Robins, 1993). We
have reasoned, in the introduction, that communal narcissism may
weaken self–other agreement regarding communal behaviors and
traits. This should be the case because communal narcissists over-
claim their communality in self-reports (Study 4), but are likely to
be seen as less communal by others (Study 3). In Study 5, then, we
expected a positive relation between communal narcissism and
self-reported communality, but a negative relation between com-
munal narcissism and observer-reported communality. Hence,
communal narcissism should suppress the self–other agreement
regarding communal traits. We tested these hypotheses using a
round-robin design (Warner, Kenny, & Stoto, 1979). To assure
robustness and consistency of our findings, we assessed observer
reports twice, with a time lag of 7 weeks in between.
Method
Participants and procedure. In the first week of the semes-
ter, 131 newly enrolled Humboldt-Universität zu Berlin (Berlin,
Germany) first-year undergraduate psychology students completed
the CNI and measures of self-reported communal traits. Three
weeks later, each participant was randomly allocated into one of 20
work groups with a maximum of seven students per group. From
that week onwards (until the end of the study), each work group
met on average at least once per week for group work as part of a
personality psychology lecture. Seven weeks after the study’s
beginning (i.e., the self-report portion), 123 of the original partic-
ipants (reparticipation rate: 94%) completed the first round-robin
wave (i.e., observer reports of their group members’ communal
behavior during group discussion, observer reports of their group
members’ communal traits). Another 7 weeks later (i.e., 14 weeks
after the self-report part), 115 of the original participants (repar-
ticipation rate: 88%) completed the second round-robin wave,
which was identical to the first one. (None of the round-robin
waves included the assessment of self-report measures.) Overall,
Table 8
Agentic and Communal Narcissism and Agentic and Communal Overclaiming
Criterion
Simultaneous predictors
Sample 13 Sample 14 Sample 15
CNI NPI CNI NPI CNI NPI
Communal overclaiming
Overclaiming .26
ⴱⴱ
— .26
ⴱⴱ
⫺.13 .18
ⴱⴱ
⫺.17
ⴱⴱ
True knowledge ⫺.10 — ⫺.17
ⴱ
⫺.10 ⫺.05 ⫺.11
Agentic overclaiming
Overclaiming — — — — .02 .20
ⴱⴱⴱ
True knowledge — — — — ⫺.02 ⫺.07
Note. Dashes indicate that no information was available. CNI ⫽Communal Narcissism Inventory; NPI ⫽
Narcissistic Personality Inventory.
ⴱ
pⱕ.05.
ⴱⴱ
pⱕ.01.
ⴱⴱⴱ
pⱕ.001.
868 GEBAUER, SEDIKIDES, VERPLANKEN, AND MAIO
106 participants completed all measures in all three waves of
measurement (i.e., [a] self-report wave, 关b兴Round-Robin Wave 1,
[c] Round-Robin Wave 2) and hence qualified for statistical anal-
yses (see Table 1, Sample 16). All assessments took place in a
large lecture theatre, and participants were rewarded with course
credit.
Materials.
Self-report. In the self-report part, we assessed communal
narcissism with the CNI, and we assessed communal traits with the
BFI’s Agreeableness and Conscientiousness subscales (Lang et al.,
2001). This operationalization of communal traits follows W. K.
Campbell, Rudich, and Sedikides (2002), who described the ag-
gregate of agreeableness and conscientiousness as prototypical for
communion and ideally suited to capture moralistic (i.e., commu-
nal) bias (Paulhus & John, 1998). Indeed, factor analyses have
confirmed that agreeableness and conscientiousness are appropri-
ately subsumed by a larger communion factor (Saucier, 2009). The
aggregate of the Agreeableness subscale (eight items; e.g., “I see
myself as someone who . . .” “. . . is helpful and unselfish with
others,” “. . . is considerate and kind to almost everyone”; rating
scale from 0 [not at all]to4[very much]) and the Conscientious-
ness subscale (nine items; e.g., “I see myself as someone who . . .”
“. . . can be somewhat careless” [reverse-scored], “. . . is a reliable
worker”; rating scale from 0 [not at all]to4[very much]) was
internally consistent (␣⫽.72).
Round-robin. In both round-robin waves, participants judged
the communality of all their group members (i.e., each participant
judged up to six other participants). First, participants judged their
group members’ communal traits, using Rammstedt and John’s
(2007) German short forms of the BFI’s Agreeableness and Con-
scientiousness subscales (Benet-Martı´nez & John, 1998). In line
with the conceptualization of communion as a broad construct, the
internal consistency of the aggregated Agreeableness subscale
(two items; “Person X . . .” “. . . is generally trusting,” “. . . tends
to find fault with others” [reverse-scored]; rating scale from 0 [not
at all]to4[very much]) and Conscientiousness subscale (two
items; “Person X . . .” “. . . does a thorough job,” “. . . tends to be
lazy” [reverse-scored]; rating scale from 0 [not at all]to4[very
much]) indicated measurement breadth in Wave 1 (␣⫽.54) and in
Wave 2 (␣⫽.57; L. A. Clark & Watson, 1995). Importantly, an
adequate test of the measure’s reliability—namely, a test–retest
correlation between Wave 1 and Wave 2—strongly supported the
measure’s reliability, r(106) ⫽.72, p⫽.001 (Cronbach & Meehl,
1955). Second, participants judged their group members’ commu-
nal behaviors on five items from Moskowitz’s (1994) behavioral
self-report items. These items were chosen because they are com-
munal in nature, while being applicable to the participants’ group
discussions: “Person X . . .” “. . . listens attentively to the others,”
“. . . complements or praises group members,” “. . . smiles and
laughs with others,” “. . . makes concessions to avoid unpleasant-
ness,” and “. . . expressed reassurance.” The internal consistency
of the measure was high in Wave 1 (␣⫽.80) and in Wave 2 (␣⫽
.83), and the test–retest correlation between Wave 1 and Wave 2
was also high, r(106) ⫽.71, p⫽.001.
Results
From Sample 16’s round-robin data, we extracted scores for
each participant that indicated how this participant had been
judged by his or her group members regarding (a) communal
behaviors and (b) communal traits. In the language of the social
relations model (Kenny, 1994), we estimated partner effects for
each participant regarding communal behaviors and communal
traits. For these estimations, we used the TripleR package (Schön-
brodt, Back, & Schmukle, 2012) for the statistical software R (R
Development Core Team, 2008), while taking the multiple group
design into account.
Once partner effects for communal behaviors and traits were
obtained, we used multilevel modeling (HLM 6.06; Raudenbush,
Bryk, & Congdon, 2004) to test our hypotheses. Because partici-
pants were nested in groups but group variables were of no interest
to us, all reported effects take place on Level 1. (Nonetheless,
HLM appropriately takes the nested data structure into account.)
First, we tested the hypothesis that (self-reported) communal
narcissism is positively related to self-reported communal traits,
but is negatively related to observer-reported communal traits. To
this end, we simultaneously regressed communal narcissism on
self-reported communal traits and on observer-reported communal
traits from the first round-robin wave. In line with the hypothesis,
communal narcissists judged themselves as particularly high on
communal traits, b⫽.69, SE ⫽.21, t(103) ⫽3.36, p⫽.001,
whereas communal narcissists were judged as particularly low on
communal traits by observers, b⫽⫺.69, SE ⫽.20, t(103) ⫽
⫺3.51, p⫽.001.
Second, we tested whether these results held when observer-
reported communal traits from the first round-robin wave were
replaced with observer-reported communal traits from the second
round-robin wave (7 weeks later). And indeed they did so. Again,
communal narcissists were judged as particularly low on commu-
nal traits by observers, b⫽⫺.69, SE ⫽.16, t(103) ⫽⫺4.33, p⫽
.001.
Third, we tested whether the same pattern held when replacing
observer-reported communal traits with observer-reported commu-
nal behaviors. Using observer reports of communal behavior from
the first round-robin wave, communal narcissists showed particu-
larly little communal behavior in the eyes of their group members,
b⫽⫺.47, SE ⫽.16, t(103) ⫽⫺2.97, p⫽.004. Yet this effect did
not reach significance when using observer reports of communal
behavior from the second round-robin wave, b⫽⫺.28, SE ⫽.21,
t(103) ⫽⫺1.38, p⫽.17, albeit the effect was in the same
direction as before.
Fourth, we tested whether the observer reports of communal
behavior mediated the link between communal narcissism and
observer reports of communal traits. Stated otherwise, did the
observed communal behaviors constitute cues that functioned as
the process linking high communal narcissism to low observer-
reported communal traits? In line with this hypothesis, a Sobel
(1982) test revealed a significant path from higher communal
narcissism over lower observer-reported communal behavior
(Round-Robin Wave 1) to lower observer-reported communal
traits (Round-Robin Wave 1; z⫽⫺2.63, SE ⫽.02, p⫽.008).
Similarly, a Sobel test revealed a significant path from higher
communal narcissism over lower observer-reported communal be-
havior (Round-Robin Wave 1) to lower observer-reported com-
munal traits (Round-Robin Wave 2; z⫽⫺2.66, SE ⫽.02, p⫽
.008).
These results constitute necessary preconditions for Study 5’s
main hypothesis: Does communal narcissists’ communal self-
869
COMMUNAL NARCISSISM
aggrandizement suppress self–other agreement in the communal
domain? To test this hypothesis, we first examined self–other
agreement using observer-reported communal traits from Round-
Robin Wave 1. That is, we regressed self-rated communal traits on
observer-rated communal traits. We obtained a marginal positive
relation, b⫽.20, SE ⫽.11, t(104) ⫽1.94, p⫽.06. Given that we
previously found communal narcissism to relate positively to
self-reported communion, but to relate negatively to observer-
reported communion, it is likely that controlling for communal
narcissism boosts this marginal self–other agreement. Indeed, this
was the case. Controlling for communal narcissism rendered self–
other agreement regarding communal traits significant, b⫽.27,
SE ⫽.10, t(103) ⫽2.73, p⫽.008, and the increase in the
self–other agreement itself was also significant (z⫽⫺2.12, SE ⫽
.03, p⫽.03).
Finally, we tested the robustness of our findings by using
observer-reported communal traits from Round-Robin Wave 2. To
begin with, we regressed self-reported communal traits on
observer-reported communal traits. Self–other agreement did not
reach significance, B⫽.19, SE ⫽.13, t(104) ⫽1.45, p⫽.15.
Again, however, when controlling for communal narcissism, self–
other agreement became significant, B⫽.26, SE ⫽.12, t(103) ⫽
2.14, p⫽.03, and, again, this increase was significant (z⫽⫺2.29,
SE ⫽.02, p⫽.02).
Discussion
Study 5 confers two crucial contributions. First, it buttresses the
conceptualization of communal narcissism as a personality trait
that describes individuals who exaggerate their communal quali-
ties (Study 4). In line with this conceptualization, communal
narcissists described themselves as particularly high on commu-
nion, but their acquaintances described them as particularly low on
communion. Further, the link between communal narcissism at
Time 1 and particularly low levels of observer-reported commu-
nion at Time 3 was mediated by observer reports of relatively low
communal behaviors at Time 2. These behaviors were concrete
actions observed during many hours of group work as part of a
personality psychology lecture. The resultant high ecological va-
lidity is a noteworthy strength of the present study.
Over and above this first merit, the present study provides one
example for the usefulness of introducing communal narcissism to
personality theory and research. The study shows that puzzlingly
low relations between self-reported and observer-reported commu-
nion are attributable, in part, to the existence of communal narcis-
sism: Communal narcissists overclaim their communal traits,
while being seen as low in communion (behaviors and traits) by
observers, and this dissociation is the reason why communal
narcissism suppresses the otherwise substantial self–other agree-
ment in the communal domain. This effect was obtained both 7 and
14 weeks after the self-report phase, attesting to the robustness of
this effect.
General Discussion
Much theory and research have been devoted to examining the
nature of grandiose narcissism. These efforts have culminated in
the conclusion that grandiose narcissism is what we have currently
labeled an agency-agency trait: an agentic trait (here: self-
aggrandizement; W. K. Campbell & Foster, 2007; Morf et al.,
2011) that is expressed through agentic means. For example, W. K.
Campbell, Rudich, and Sedikides (2002) reported that narcissists
display exaggerated better-than-average judgments on agentic, but
not on communal, attributes. Based on this and similar findings,
contemporary models of narcissism (W. K. Campbell, 1999; W. K.
Campbell & Foster, 2007; Paulhus, 2001; Paulhus & John, 1998;
Sedikides et al., 2002) converge in positing that narcissists meet
their core self-motives (i.e., grandiosity, esteem, entitlement, and
power) through agentic means.
The current approach aimed to complement and extend prior
theorizing and research. In particular, we have proposed an
agency-communion model of narcissism. The model describes
agentic narcissists as individuals who meet their core self-motives
through agentic means. In keeping with prior narcissism theories
(W. K. Campbell & Foster, 2007; Paulhus, 2001) and research
(W. K. Campbell, Rudich, & Sedikides, 2002; Paulhus et al.,
2003), we accepted the utility of the NPI as a suitable measure of
agentic narcissism (Miller & Campbell, 2011). More important,
the model includes a second, previously unidentified form of
narcissism. Communal narcissists have the same core self-motives
as agentic narcissists. However, communal narcissists meet these
self-motives through communal means. Five studies across 16
samples lent support to the model.
Overview of the Findings
In Study 1, participants listed their grandiose self-thoughts in an
open-ended measure. Content analyses revealed that participants
listed both agentic and communal grandiose self-thoughts. More-
over, NPI narcissists reported a higher amount and extremity of
agentic, but not of communal, grandiose self-thoughts. This is
consistent with prior evidence that NPI narcissists show elevated
better-than-average judgments on agentic, but not on communal,
attributes (W. K. Campbell, Rudich, & Sedikides, 2002; see also
W. K. Campbell et al., 2007). At the same time, this finding
supports our claim that the NPI only assesses agentic narcissism
(see also W. K. Campbell & Foster, 2007). In Study 2, we used the
communal grandiose self-thoughts listed by Study 1 participants as
candidate items for the CNI. We carried out exploratory and
confirmatory factor analyses to construct the final 16-item CNI
(see the Appendix). This self-report measure has good psychomet-
ric properties and high temporal stability, and it is largely inde-
pendent of the NPI. Thus, these results provide additional support
for our model, while providing support for the CNI as an adequate
measure of communal narcissism. Further corroborating the
model, Study 3 demonstrated that agentic and communal narcis-
sists share the same core self-motives for grandiosity, esteem,
entitlement, and power. Study 4 backed the model’s proposition
that agentic and communal narcissists differ in their means for
need satisfaction. Agentic narcissists manifested higher better-
than-average judgments on agentic, but not on communal, attri-
butes. Conversely, communal narcissists manifested higher better-
than-average judgments on communal, but not on agentic,
attributes. Furthermore, agentic narcissists overclaimed their
knowledge regarding agentic, but not communal, topics. Con-
versely, communal narci