Five-Factor Model Best Describes Narcissistic Personality Inventory Across Different Item Response Formats

Abstract

The objective of this research was to validate the Narcissistic Personality Inventory across different response formats, given that several factor structures were proposed, ranging from two to seven factors. The original forced-choice format of the Narcissistic Personality Inventory was given to 410 participants and a modified, i.e., Likert format was given to 423 participants from the general population, along with personality and other narcissism measures. The results showed that the five-factor model proposed by Ackerman et al. had the best model fit in both response formats and that a distinction between adaptive (Leadership, Vanity, and Superiority) and some aspects of maladaptive (Manipulativeness and Exhibitionism) narcissism factors could be established. However, the redundancy of items in certain factors could be problematic and further improvements of the Narcissistic Personality Inventory should include more indicators of some proposed factors, especially of Vanity.

Keywords

Narcissism narcissistic personality inventory response format factor structure construct validity adaptive narcissism maladaptive narcissism

Introduction

The most commonly used measure of grandiose narcissism is the Narcissistic Personality Inventory (NPI; Raskin & Terry, 1988), which has undergone many reductions and proposed factor structures since its introduction. Originally, Raskin and Hall (1979) constructed a 54-item forced-choice inventory based on the Diagnostic and Statistical Manual (DSM-III) criteria for the narcissistic personality disorder. Although the NPI was based on the DSM criteria, it was assumed that only extreme manifestations of those criteria constitute pathological narcissism and that milder forms reflect narcissism as a personality trait or subclinical narcissism.

The first author who tested the NPI structure was Emmons (1984, 1987). Based on principal component analysis (PCA) and a principal axis factor analysis on two student samples, he extracted four dimensions: Exploitiveness/Entitlement, Leadership/Authority, Superiority/Arrogance, and Self-Absorption/Self-Admiration. This version of the NPI had 31 items in the final solution. However, Raskin and Terry (1988) reanalyzed the original 54 items in a student sample. They conducted a PCA on tetrachoric correlations, which are more appropriate for binary data, and after the exclusion of 14 items, they extracted seven components: Authority, Self-Sufficiency, Superiority, Exhibitionism, Exploitativeness, Vanity, and Entitlement. To this day, this is the most commonly used version (40 items) and scoring key for the NPI when dimensions of narcissism are of interest. In the meantime, the factor structure of the NPI is continued to be investigated.

Kansi (2003) explored the NPI on a Swedish population sample and performed exploratory approach (PCA and principal-axes method). After exclusion of 14 items, she obtained four dimensions, similar to Emmons’s solution: Leadership/Power, Exhibitionism/Self-admiration, Superiority/Arrogance, and Uniqueness/Entitlement. A later validation of this solution resulted in the exclusion of four more items (Svindseth et al., 2009), but this solution showed characteristics very similar to the ones found in the solution proposed by Kansi (2003). Later, Kubarych, Deary, and Austin (2004) used a combination of the PCA and confirmatory factor analysis (CFA) on the 40-item version in a student sample and suggested two- (Power and Exhibitionism) and three-dimension solutions (Power, Exhibitionism, and Special Person). In both solutions, three items were omitted. They were the first authors to suggest that the response format should be reconsidered for the statistical purpose and the first to use the Likert scale instead of a binary format.

A detailed examination of the proposed NPI models was conducted by Corry, Merritt, Mrug, and Pamp (2008). They used exploratory factor analysis (the minimizing residuals method) on tetrachoric correlations and after omitting 17 items, suggested two factors: Leadership/Authority and Exhibitionism/Entitlement. On a different sample, they conducted the CFA with the diagonally weighted least squares estimation on tetrachoric correlations and tested the all proposed models. Although all models had a satisfactory fit, only the scales from two-factor solutions (one by Kubarych et al., 2004 and another proposed by Corry et al., 2008) had satisfactory internal consistency values.

A further exploration was conducted by Ackerman et al. (2011) on a student sample. They used a combination of exploratory factor analysis and CFA (with weighted least square with mean and variance adjustment—WLSMV estimator). After omitting 15 items, they proposed a three-factor solution: Leadership/Authority, Grandiose Exhibitionism, and Entitlement/Exploitativeness.

Overall, previous findings indicate that different response formats provide different factor structures. For example, Barelds and Dijkstra (2010) validated the Dutch version of the NPI with a five-point Likert scale and concluded that neither the four-dimension nor the seven-dimension solution could be replicated. On a sample that combined community and student subsamples, they conducted the PCA and extracted two components as the most interpretable—Authority/Power and Self-Admiration. Although the first component strongly resembles a dimension that refers to leadership and/or power from previous studies, the second component does not match the dimensions extracted in previous studies. Furthermore, items from the second component had many significant cross-loadings, which is why the authors kept the one-component solution by omitting the four items.

A comprehensive exploration of the effect of the response format on the NPI structure was performed by Ackerman, Donnellan, Roberts, and Fraley (2016). They revealed three factors in the forced-choice format of the NPI, but five in the dichotomous (True/False) and Likert formats. The three-factor solution comprises Leadership, Exhibitionism, and Vanity factors, while the five-factor solution includes the additional Manipulativeness and Superiority factors. However, only the Leadership factor was well replicated across different response formats. Nonetheless, the authors suggested the five-factor model for further validation, based on high average loadings across the dichotomous and Likert formats (23 items in total). All proposed models are presented in Table 1.

Table 1.

Proposed models of Narcissism Personality Inventory.

Item number in Raskin and Terry (1988)	Emmons (1987)	Raskin and Terry (1988)	Kansi (2003)	Kubarych et al. (2004)	Kubarych et al. (2004)	Corry et al. (2008)	Barelds and Dijkstra (2010)	Ackerman et al. (2011)	Ackerman et al. (2016)
1	L/A	A	L/P	P	P	L/A	A/P	L/A	M
2	–	Ex	U/E	Ex	Ex	–	A/P	–	–
3	S/A	Ex	–	Ex	SP	Ex/En	A/P	–	–
4	S/S	S	E/S	P	SP	Ex/En	SA	GE	–
5	–	Ent	L/P	P	P	–	A/P	L/A	–
6	S/A	Exp	–	P	P	–	A/P	–	M
7	L/A	Ex	–	Ex	Ex	Ex/En	A/P	GE	Ex
8	–	A	L/P	P	P	L/A	A/P	–	S
9	S/S	S	U/E	P	SP	–	SA	–	S
10	L/A	A	L/P	P	P	L/A	A/P	L/A	L
11	L/A	A	L/P	P	P	L/A	A/P	L/A	L
12	L/A	A	–	P	P	L/A	A/P	L/A	L
13	E/E	Exp	–	P	P	–	A/P	En/Ex	M
14	E/E	Ent	U/E	P	SP	Ex/En	–	En/Ex	–
15	S/S	V	E/S	–	–	Ex/En	SA	GE	V
16	S/A	Exp	S/A	P	P	–	–	–	M
17	–	S-S	L/P	P	SP	–	A/P	–	L
18	–	Ent	U/E	Ex	Ex	–	SA	–	–
19	S/S	V	E/S	Ex	SP	Ex/En	SA	GE	V
20	S/A	Ex	E/S	Ex	Ex	Ex/En	SA	GE	Ex
21	S/A	S-S	S/A	P	P	–	A/P	–	–
22	–	S-S	S/A	–	–	–	–	–	–
23	S/A	Exp	–	P	SP	–	A/P	–	–
24	E/E	Ent	–	P	P	Ex/En	SA	En/Ex	–
25	E/E	Ent	U/E	P	P	Ex/En	SA	En/Ex	–
26	S/S	S	E/S	Ex	Ex	–	SA	GE	–
27	E/E	Ent	L/P	P	P	L/A	A/P	L/A	–
28	–	Ex	U/E	Ex	Ex	Ex/En	SA	GE	–
29	S/S	V	E/S	–	–	Ex/En	SA	GE	V
30	L/A	Ex	–	Ex	Ex	Ex/En	–	GE	Ex
31	–	S-S	S/A	P	SP	–	A/P	–	–
32	L/A	A	–	P	P	L/A	A/P	L/A	L
33	L/A	A	L/P	P	P	L/A	A/P	L/A	L
34	S/S	S-S	U/E	P	SP	–	SA	L/A	S
35	S/A	Exp	S/A	P	SP	–	A/P	–	M
36	S/A	A	U/E	P	P	L/A	A/P	L/A	L
37	–	S	–	Ex	Ex	–	SA	–	–
38	E/E	Ex	E/S	Ex	Ex	Ex/En	SA	GE	Ex
39	E/E	S-S	S/A	P	SP	Ex/En	A/P	–	–
40	S/S	S	–	P	SP	–	SA	L/A	S

Note: L/A: leadership/authority; S/A: superiority/arrogance; S/S: self-absorption/self-admiration: E-E: exploitiveness/entitlement; A: authority; Ex: exhibitionism; S: superiority; Ent: entitlement; Exp: exploitativeness; V: vanity; S-S: self-sufficiency; P: power; SP: special person; Ex/En: exhibitionism/entitlement; SA: self-admiration; A/P: authority/power; GE: grandiose exhibitionism; M: manipulativenss; L/P: leadership/power; E/S: exhibitionism/self-admiration; U/E: uniqueness/entitlement.

The aim of this study was to test all the proposed NPI models on community samples in order to determine which model had the best fit indices. In previous research, different factor structures were suggested, ranging from two to seven factors. Although in two studies (Ackerman et al., 2011; Corry et al., 2008), different models were tested via the CFA, the authors also proposed their own models, which need to be further validated. Furthermore, the results from some previous studies suggested that the item response format could influence the different factor structure of the NPI (Ackerman, Donnellan, et al., 2016; Barelds & Dijkstra, 2010). However, the forced-choice and Likert versions were substantially interrelated, except in the case of the entitlement aspect (Miller et al., 2018; Wetzel, Roberts, Fraley, & Brown, 2016). Improved performance was found in the Likert format of the Entitlement/Exploitativeness scale from the Ackerman et al. (2011) three-factor model, compared to the forced-choice format (Miller et al., 2018). Namely, the Likert version of Entitlement/Exploitativeness scale manifested a higher alpha coefficient and stronger convergent and criterion validity correlations. However, in another study, even with a Likert scale, Entitlement/Exploitativeness scale did not obtain satisfactory psychometric properties (Grosz et al., 2017). Therefore, we wanted to test if the same model solution had the best model fit regardless of the response format.

A further validation of the NPI scales included an exploration of convergent and discriminant validity. Previous research showed that the leadership/authority/power/exhibitionism aspect of narcissism had a strong negative correlation with Agreeableness and had a positive correlation with Extraversion from the Big Five or Five-Factor Model and captured adaptive narcissism. On the other hand, the Entitlement/Exploitativeness aspect correlated negatively with Agreeableness and positively with Neuroticism, Machiavellianism, and psychopathy, thus capturing maladaptive narcissism (Ackerman et al., 2011; Foster, Shiverdecker, & Tirner, 2016; Kubarych et al., 2004; Miller & Campbell, 2008; Stanton et al., 2017). The adaptive-maladaptive distinction is important because these aspects of narcissism are related to different outcomes. Adaptive or normal narcissism refers to positive feelings toward the self and it is associated with subjective well-being, lack of interpersonal problems, and lack of psychological distress, while maladaptive or pathological narcissism is associated with indicators of dysfunctional adjustment (e.g., Ackerman et al., 2011). However, some authors questioned the validity of the NPI in capturing maladaptive narcissism (e.g., Pincus et al., 2009; Rosenthal & Hooley, 2010). Moreover, a recent meta-analysis showed that a number of NPI items poorly differentiate narcissism from self-esteem, self-confidence, and similar constructs (Rosenthal, Montoya, Ridings, Rieck, & Hooley, 2011). Thus, the positive relation between NPI and psychological health may result from the presence of the abovementioned confounding items (Rosenthal et al., 2011). Therefore, the conclusion about the healthy side of narcissism captured by the NPI is questionable.

In this study, we tested the correlations with the six-factor HEXACO model and expected that the sixth factor, Honesty-Humility, like Extraversion, should be most strongly related to narcissism, followed by Agreeableness (Lee & Ashton, 2005; Miller et al., 2018). Honesty-Humility is considered to be the “core” of maladaptive or dark traits (e.g., Book, Visser, & Volk, 2015) which include Machiavellianism, psychopathy, and narcissism (called Dark Triad, see Paulhus & Williams, 2002). Therefore, we expected that the NPI scales would obtain significant correlations with the rest of the dark traits, as well as pathological narcissism, which would provide a better insight into adaptive-maladaptive aspects of the NPI scales.

Method

Participants and procedure

Sample 1 included 410 participants (50% were males) from the general population, aged between 18 and 79 years (M = 39.92, SD = 14.06), about half of whom were highly educated (37.3% were undergraduate students and 22.7% had a university degree) and the rest were high school graduates. About half of the participants were from an urban area (54.9%).

Sample 2 included 423 participants (49.2% were males) from the general population, aged between 18 and 85 years (M = 31.59, SD = 14.24), about half of whom were highly educated (40% were undergraduate students and 25.5% had a college or university degree) and the rest were high school graduates.

A majority of both samples were Caucasian (95%). Both samples were collected by trained undergraduate students as a part of their preexam activity. In order to collect data from a heterogeneous sample, each student collected data from six participants, in accordance with the given gender and age quotas (three age groups: 18–25, 26–35, 36 years and older, with both male and female participants in each age group). Both studies were approved by the Ethical Committee of the Faculty of Philosophy, University of Novi Sad, Serbia, which is the Second Instance Commission of the Ethical Committee of the Serbian Psychological Society.

Instruments

On sample 1, four instruments were administered: (1) NPI (Raskin & Terry, 1988), which consists of 40 forced-choice items (e.g., A. I prefer to blend in with the crowd, B. I like to be the center of attention); (2) Brief HEXACO (BHI: deVries, 2013), which is a short measure of six traits from the lexical HEXACO model of personality; (3) Mach IV (Christie & Geis, 1971), which measures Machiavellianism; and (4) Psychopathic Personality Trait Scale (Boduszek, Debowska, Dhingra, & DeLisi, 2016), which measures psychopathy as characterized by interpersonal manipulation, egocentricity, and a lack of affective and cognitive responsiveness. All measures, except for the NPI, contain a five-point Likert scale for answering.

On sample 2, three instruments were administered: (1) NPI (Raskin & Terry, 1988) with 40 six-point Likert-type items (from 0=not like me to 5 = very much like me), which contains only narcissistic options from the forced-choice format. Several studies have already applied the strategy of using the NPI with a Likert scale (e.g., Clarke, Karlov, & Neale, 2014; Egan & Lewis, 2011); (2) HEXACO-60 (Ashton & Lee, 2009), which is a short form of HEXACO-PI-R, the original measure of HEXACO model of personality, with five-point Likert-type items; and (3) Pathological Narcissism Inventory (PNI: Pincus et al., 2009), which measures seven dimensions, three of which constitute grandiose narcissism (Self-Sacrificing Self-Enhancement, Exploitative, and Grandiose Fantasy) and four of which constitute vulnerable narcissism (Contingent Self-Esteem, Hiding the Self, Devaluing, and Entitlement Rage), with scoring suggested by Wright, Lukowitsky, Pincus, and Conroy (2010). The PNI features the same response format as the modified NPI used on this sample. Descriptives and reliabilities for all used instruments are given in Table 4.

Data analysis

First, data cleaning was conducted. There were 0.17% of missing data on all items on sample 1 and 0.22% on sample 2. Since less than 5% of the data were missing, imputation by expectation–maximization algorithm was applied (Tabachnick & Fidell, 2013). Expectation–maximization algorithm was run in IBM SPSS 23. On sample 1, there were five univariate outliers (z > ± 3.33) and one multivariate outlier (Machalanobis χ²(13) = 34.53, p < .001), which resulted in a total of 404 participants after the elimination of outliers. On sample 2, there were three univariate and nine multivariate outliers (Machalanobis χ²(18) = 42.31, p < .001), which resulted in a total of 411 participants after the elimination of outliers.

Second, the CFA was used to test model fit of the proposed models. In the case of forced-choice items, diagonally weighted least squares estimation was used as an adequate solution for binary or ordinary data up to four rating points (Kline, 2016). In the case of Likert-type items (six points), maximum likelihood estimation was used. Due to significant multivariate kurtosis (z = 49.07, p < .001), a robust maximum likelihood estimator was used with robust standard errors and a Satorra-Bentler scaled test statistic (MLM). Analyses were run in “lavaan” package (Rosseel, 2012). In line with recommendations made by Hu and Bentler (1999), several fit indices were used for determining model fit: χ², comparative fit index—CFI, Tucker-Lewis index—TLI, root mean square error of approximation—RMSEA, and standardized root mean residual—SRMR. Although there are no absolute standards, the determination of model fit requires a consideration of the range of fit indices that may evidence a good fit (e.g., RMSEA and SRMR < .06, TLI and CFI > .95) or an acceptable fit (e.g., RMSEA and SRMR < .08, TLI and CFI > .90). For the comparison of nested models, Δχ² was used. For the comparison of non-nested models, Akaike information criterion—AIC ( $χ_{M}^{2} + 2 q,$ where q is the number of parameters) and Bayesian information criterion—BIC ( $χ_{M}^{2} + qln (N)$ ) were used (Kline, 2016). After obtaining which model had the best fit, further analyses were run on mean scores and include testing the gender differences (t test) and validity correlations. For determining the effect size of gender differences, Cohen’s d was calculated, and the value of 0.41 was interpreted as the minimum effect size representing a “practically” significant effect for social science data (Ferguson, 2009). Since some of the used scales had lower alphas (e.g., BHI), all correlations with validity variables were corrected for attenuation. To compare NPI scales’ associations, i.e., their profiles based on validity variables, profile similarity was calculated. Cronbach and Gleaser’s (1953) D statistics based on Euclidian distances was used as a measure of profile similarity. Thus, lower values indicated greater profile similarity, and D could be interpreted as Cohen’s d. In addition, the significance of difference between dependent correlations was tested by Steiger’s Z test.

Results

Confirmatory factor analysis

NPI with forced-choice items

All models had acceptable fit, with models 2, 4, 5, 8, and 9 having a good fit (CFI and TLI >.95, see Table 2). According to the AIC and the BIC, the five-factor model proposed by Ackerman et al. (2016) had the best fit (note that the model proposed by Corry et al. (2008) had the same number of parameters, but somewhat higher AIC and BIC). However, the correlation between Leadership and Manipulativeness factor scores in this model is high (.86), indicating that the distinctions between these factors would be hard to establish. Therefore, the four-factor solution with Leadership and Manipulativeness factors combined resulted in a worse model fit (Δχ²(4) = 14.81, p = .005, model fit: χ²(224) = 370.59, p < .001, CFI = .96, TLI = .96, RMSEA = .04 (90% CI: .03, .05), SRMR = .06, AIC = 388.28, BIC = 596.35). The final five-factor model is presented in Table 3.

Table 2.

Model fit indices for proposed models of NPI with different formats.

Models	Nf	Ni	χ²(df)	CFI	TLI	RMSEA (90% CI)	SRMR	AIC	BIC
NPI with forced-choice items (n = 404)
1. Emmons (1984)	4	31	822.91(428)	.93	.93	.05 (.04, .05)	.07	958.91	1231.01
2. Raskin and Terry (1988)	7	40	1023.02(719)	.96	.96	.03 (.03, .04)	.06	1225.02	1629.16
3. Kansi (2003)	4	29	605.72(371)	.94	.93	.04 (.03, .05)	.06	733.72	989.81
4. Kubarych et al. (2004)	2	37	951.92(628)	.95	.95	.04 (.03, .04)	.06	1101.92	1402.03
5. Kubarych et al. (2004)	3	37	924.45(626)	.96	.96	.03 (.03, .04)	.06	1078.45	1386.56
6. Corry et al. (2008)	2	23	434.98(229)	.94	.94	.05 (.04, .05)	.07	528.98	717.05
7. Barelds and Dijkstra (2010)	2	36	951.42(593)	.94	.94	.04 (.03, .04)	.07	1097.42	1389.52
8. Ackerman et al. (2011)	3	25	473.60(272)	.95	.95	.04 (.04, .05)	.07	579.60	791.68
9. Ackerman et al. (2016)	5	23	269.46(220)	.99	.99	.02 (.02, .03)	.05	381.46	605.54
NPI with Likert-type items (n = 411)
1. Emmons (1984)	4	31	1638.83(428)	.82	.81	.08 (.08, .09)	.06	41273.81	41635.65
2. Raskin and Terry (1988)	7	40	1948.47(719)	.86	.84	.07 (.06, .07)	.06	53092.08	53658.70
3. Kansi (2003)	4	29	1301.14(371)	.83	.81	.08 (.07, .08)	.07	39236.31	39610.04
4. Kubarych et al. (2004)	2	37	2000.71(628)	.82	.81	.07 (.07, .08)	.06	49630.30	50080.38
5. Kubarych et al. (2004)	3	37	1988.48(626)	.82	.81	.07 (.07, .08)	.07	49611.01	50069.13
6. Corry et al. (2008)	2	23	875.66(229)	.88	.86	.08 (.08, .09)	.06	30310.65	30591.95
7. Barelds and Dijkstra (2010)	2	36	1956.96(593)	.82	.81	.08 (.07, .08)	.06	48242.36	48680.38
8. Ackerman et al. (2011)	3	25	1085.34(272)	.86	.84	.09 (.08, .09)	.07	33151.86	33465.31
9. Ackerman et al. (2016)	5	23	674.98(220)	.92	.91	.07 (.07, .08)	.05	29675.70	29993.17

Note: Nf: number of factors; Ni: number of items; CFI: comparative fit index; TLI: Tucker-Lewis index; RMSEA: root mean square error of approximation; SRMR: standardized root mean residual; AIC: Akaike information criterion; BIC: Bayesian information criterion.

Estimator on forced-choice items was DWLS, and on Likert scale was MLM. All χ² are significant at p < .001.

Table 3.

Parameters of five-factor model of NPI with different formats.

	Forced choice				Likert type
Item number in Raskin and Terry (1988)	Unstandard parameter	SE	Standard parameter	Unstandard parameter	SE	Standard parameter
Leadership
10	1.00		.68	1.00		.82
11	0.74	.05	.53	0.62	.05	.56
12	0.71	.05	.49	0.98	.04	.83
17	0.32	.04	.27	0.68	.05	.60
32	0.65	.05	.48	0.79	.05	.71
33	0.90	.06	.68	1.10	.05	.88
36	0.42	.03	.45	1.04	.06	.86
Exhibitionism
7	1.00		.83	1.00		.85
20	0.31	.03	.39	0.84	.05	.73
30	0.98	.07	.82	1.04	.04	.89
38	0.19	.02	.31	0.53	.05	.54
Vanity
15	1.00		.44	1.00		.76
19	2.46	.30	.73	1.14	.07	.84
29	3.61	.44	.88	1.16	.07	.80
Manipulativeness
1	1.00		.55	1.00	.06	.77
6	0.75	.06	.43	0.99	.06	.72
13	0.65	.05	.49	1.11	.07	.82
16	0.90	.07	.49	0.69	.06	.50
35	0.87	.06	.48	1.01		.79
Superiority
8	1.00		.53	1.00		.77
9	1.02	.09	.58	1.05	.07	.72
34	0.49	.06	.27	0.91	.07	.69
40	1.07	.09	.61	0.84	.08	.92

Note: SE: standard error.

Table 4.

Descriptives, reliability, and homogeneity of all measures.

		M	SD	Range	Skewness	Kurtosis	α	No. of items	MIC
Sample 1 (n = 404)
NPI (forced-choice items)^a	Leadership	0.30	0.27	1	0.69	−0.43	.85	7	.44
	Exhibitionism	0.23	0.28	1	0.82	−0.57	.82	4	.54
	Vanity	0.20	0.31	1	1.33	0.49	.93	3	.81
	Manipulativeness	0.38	0.29	1	0.38	−0.84	.77	5	.40
	Superiority	0.39	0.31	1	0.47	−0.81	.69	4	.36
BHI	H	3.61	0.75	3.75	−0.45	0.06	.51	4	.21
	E	2.89	0.72	4.00	0.14	−0.05	.42	4	.16
	X	3.88	0.69	3.00	−0.36	−0.40	.63	4	.32
	A	3.09	0.62	3.75	−0.05	0.04	.33	4	.12
	C	3.57	0.76	4.00	−0.40	−0.10	.64	4	.32
	O	3.66	0.73	3.75	−0.38	0.04	.55	4	.24
Mach IV	Machiavellianism	2.89	0.43	2.90	0.09	0.29	.72	20	.11
PPTS	Psychopathy	2.49	0.44	2.40	0.16	−0.12	.72	20	.11
Sample 2 (n = 411)
NPI (Likert-type items)	Leadership	2.27	1.22	5	0.02	−0.89	.90	7	.56
	Exhibitionism	1.70	1.23	5	0.25	−0.80	.83	4	.55
	Vanity	1.79	1.39	5	0.37	−0.86	.84	3	.64
	Manipulativeness	2.02	1.19	5	0.18	−0.79	.84	5	.52
	Superiority	2.80	1.18	5	−0.45	−0.33	.79	4	.49
HEXACO-60	H	3.46	0.65	3.50	−0.23	−0.22	.72	10	.20
	E	3.17	0.64	3.40	−0.10	−0.01	.73	10	.21
	X	3.27	0.59	3.30	0.00	−0.12	.73	10	.21
	A	3.02	0.49	3.00	−0.30	0.53	.55	10	.11
	C	3.48	0.62	3.40	0.03	−0.21	.75	10	.23
	O	3.43	0.74	3.80	−0.07	−0.39	.78	10	.27
PNI	SSSE	2.59	1.00	4.83	−0.17	−0.51	.72	6	.30
	EXP	2.15	1.07	5.00	0.10	−0.68	.77	5	.39
	GF	2.26	1.23	5.00	−0.11	−0.79	.87	7	.49
	NG	2.34	0.92	4.50	−0.18	−0.70	.89	18	.30
	CSE	1.75	1.06	4.83	0.35	−0.62	.91	12	.45
	HS	2.42	0.96	4.86	−0.34	−0.47	.71	7	.26
	DEV	1.78	1.02	4.86	0.10	−0.75	.80	7	.37
	ER	2.12	1.05	4.75	0.07	−0.63	.83	8	.39
	NV	1.98	0.90	4.24	0.06	−0.61	.94	34	.32
	PNI total	2.11	0.85	4.23	−0.06	−0.57	.95	52	.29

Note: MIC: mean inter-item correlation as measure of homogeneity; H: honesty–humility; E: emotionality; X: extraversion; A: agreeableness; C: conscientiousness; O: openness to experience; SSSE: self-sacrificing self-enhancement; EXP: exploitativeness; GF: grandiose fantasy; NG: narcissistic grandiosity (SSSE+EXP+GF); CSE: contingent self-esteem; HS: hiding the self; DEV: devaluing; ER: entitlement rage; NV: narcissistic vulnerability (CSE+HS+DEV+ER); NPI: Narcissistic Personality Inventory; PPTS: psychopathic personality traits scale; PNI: Pathological Narcissism Inventory.

^aalphas and MIC were calculated on tetrachoric correlations.

NPI with Likert-type items

None of the proposed models achieved good fit indices, but the only model with acceptable fit was the five-factor model by Ackerman et al. (2016) (Table 2). This model also had the smallest AIC and BIC. Correlations between factor scores in all the proposed models were high, most of them higher than .80. Correlations between factor scores in the Ackerman et al. (2016) model were also high, especially between Leadership and Manipulativeness factor scores (.81). However, the four-factor model combining Leadership and Manipulativeness factors resulted in a worse model fit (Δχ²(4) = 143.1, p < .001, model fit: χ²(224) = 856.26, p < .001, CFI = .89, TLI = .87, RMSEA = .08 (90% CI: .08, .09), SRMR = .06, AIC = 29900.28, BIC = 30201.67). The final five-factor model is presented in Table 3.

Descriptives, psychometric properties, and gender differences

NPI with forced-choice items

Skewness and kurtosis were in the recommended range for normal distribution (±2.00, see Gravetter & Wallnau, 2014). The reliabilities of scales were acceptable or good and the mean inter-items correlations suggested that there should not be any redundant items, except in the case of Vanity (Table 4). Gender differences were significant in the case of Leadership (Cohen’s d = 0.33) and Exhibitionism (Cohen’s d = 0.25), with male participants having higher scores. However, considering the recommended cut-off criteria provided by Ferguson (2009), all gender differences were negligible.

NPI with Likert-type items

Skewness and kurtosis met the cut-off criteria for normal distribution. The reliabilities of the NPI scales were good, but the homogeneity index mean inter-items correlations indicated that the redundancy of items in the same scale could be a problem (an acceptable range for Pearson correlations is .20–.50, see Clark & Watson, 1995), especially in the case of Vanity (Table 4). There were significant gender differences in Leadership (Cohen’s d = 0.40) and Manipulativeness (Cohen’s d = 0.20), with male participants having higher scores. However, all gender differences were negligible considering the effect size (Ferguson, 2009).

Validity correlations

NPI with forced-choice items

All NPI scales had the highest correlations with Honesty-Humility, followed by Extraversion (Table 5). Vanity obtained the smallest correlations, which may be due to its narrow variability. D statistic indicated that Leadership and Manipulativeness manifested a very similar pattern of relations with HEXACO, Machiavellianism, and psychopathy traits. Therefore, the significant difference between attenuated correlations was tested. Manipulativeness obtained significantly higher negative correlation with Agreeableness (Z = 2.01, p < .05) and positive with Machiavellianism (Z = −1.98, p < .05) and psychopathy (Z = − 2.46, p < .01). On the other hand, Leadership obtained significantly higher negative correlation with Emotionality (Z = − 2.49, p < .01) and positive with Conscientiousness (Z = 5.25, p < .001). These results suggest that although there are similarities between Leadership and Manipulativeness, the content of these two scales could be distinguished. Furthermore, similar profiles manifested Manipulativeness and Exhibitionism, with the only difference in stronger relations between Manipulativeness and Emotionality (Z = 4.95, p < .001) and Openness (Z = − 2.49, p < .01).

Table 5.

Inter-correlations between NPI scales (forced-choice items) and correlations with HEXACO, Machiavellianism, and psychopathy (n = 404).

	NPI scales
	Leadership	Exhibitionism	Vanity	Manipulativeness	Superiority
Leadership	1	.43	.20	.57	.43
Exhibitionism	0.20 (0.29)	1	.38	.39	.30
Vanity	0.24 (0.69)	0.23 (0.35)	1	.27	.30
Manipulativeness	0.05 (0.13)	0.15 (0.15)	0.28 (0.56)	1	.37
Superiority	0.13 (0.20)	0.15 (0.34)	0.17 (0.51)	0.16 (0.32)	1
Honesty–Humility	−.40 (−.61)	−.39 (−.60)	−.27 (−.39)	−.40 (−.64)	−.35 (−.59)
Emotionality	−.30 (−.50)	−.08 (−.14)	.01 (.02)	−.23 (−.40)	−.13 (−.24)
Extraversion	.29 (.40)	.31 (.43)	.15 (.20)	.28 (.40)	.20 (.30)
Agreeableness	−.10 (−.19)	−.17 (−.33)	−.06 (−.11)	−.14 (−.28)	.03 (.06)
Conscientiousness	.17 (.23)	−.08 (−.11)	−.08 (−.10)	−.00 (−.01)	.10 (.15)
Openness to experience	.22 (.32)	.14 (.21)	.10 (.14)	.22 (.34)	.25 (.41)
Machiavellianism	.09 (.12)	.10 (.13)	.09 (.11)	.16 (.21)	.03 (.04)
Psychopathy	.28 (.36)	.32 (.42)	.16 (.20)	.36 (.48)	.16 (.23)

Note: NPI: Narcissistic Personality Inventory.

All correlations above ± 16 are significant at p < .001. Correlations between NPI scales are above diagonal. Profile similarity D statistics between NPI scales are below diagonal, and D statistics based on attenuated correlations are in parentheses. Attenuated correlations with validity variables are in parentheses.

NPI with Likert-type items

Among HEXACO traits, all NPI scales had the highest correlations with Honesty-Humility, followed by Extraversion (Table 6). Among pathological narcissism dimensions, the NPI scales had the highest associations with Narcissistic Grandiosity, especially with Exploitativeness and Grandiose Fantasy. Within the narcissistic vulnerability domain, somewhat higher correlations were obtained with Entitlement Rage. Due to larger D statistics, the differences between all scales were more conspicuous. It is noticeable that Manipulativeness obtained an almost perfect correlation with Exploitativeness, which is expected given that this PNI scale contains four items from NPI (out of five in total). Exhibitionism obtained the largest correlation with Honesty-Humility and the PNI.

Table 6.

Inter-correlations between NPI scales (Likert-type items) and correlations with HEXACO and PNI scales (n = 411).

	NPI
	Leadership	Exhibitionism	Vanity	Manipulativeness	Superiority
Leadership	1	.66	.48	.69	.67
Exhibitionism	0.37 (0.66)	.1	.69	.64	.60
Vanity	0.30 (0.28)	0.32 (0.46)	1	.51	.55
Manipulativeness	0.24 (0.25)	0.24 (0.33)	0.29 (0.38)	1	.52
Superiority	0.31 (0.31)	0.35 (0.46)	0.15 (0.21)	0.32 (0.39)	1
Honesty–Humility	−.41 (−.51)	−.54 (−.70)	−.39 (−.50)	−.47 (−.60)	−.31 (−.41)
Emotionality	−.21 (−.26)	−.05 (.−06)	−.04 (−.05)	−.15 (−.19)	−.03 (−.04)
Extraversion	.51 (.63)	.34 (.44)	.27 (.34)	.38 (.49)	.32 (.42)
Agreeableness	−.21 (−.30)	−.21 (−.31)	−.12 (−.18)	−.23 (−.34)	−.12 (−.18)
Conscientiousness	.09 (.11)	−.17 (−.22)	−.11 (−.14)	−.01 (−.01)	.01 (.01)
Openness to experience	.17 (.20)	.07 (.09)	.07 (.09)	.11 (.14)	.19 (.24)
Self-sacrificing self-enhancement	.39 (.48)	.47 (.61)	.34 (.44)	.40 (.51)	.45 (.60)
Exploitativeness	.63 (.76)	.62 (.78)	.48 (.60)	.85 (.99)	.52 (.67)
Grandiose fantasy	.53 (.60)	.64 (.75)	.47 (.55)	.50 (.58)	.59 (.71)
Narcissistic grandiosity (SSSE+EXP+GF)	.62 (.69)	.70 (.81)	.52 (.60)	.68 (.79)	.63 (.75)
Contingent self-esteem	.25 (.28)	.51 (.59)	.31 (.35)	.35 (.40)	.33 (.39)
Hiding the self	.30 (.38)	.36 (.47)	.28 (.36)	.40 (.52)	.34 (.45)
Devaluing	.19 (.22)	.39 (.48)	.27 (.33)	.29 (.35)	.31 (.39)
Entitlement rage	.40 (.46)	.58 (.70)	.44 (.53)	.47 (.56)	.47 (.58)
Narcissistic vulnerability (CSE+HS+DEV+ER)	.33 (.36)	.54 (.61)	.38 (.43)	.43 (.48)	.41 (.48)
PNI total	.46 (.50)	.64 (.72)	.46 (.51)	.55 (.62)	.53 (.61)

Note: NPI: Narcissistic Personality Inventory; PNI: Pathological Narcissism Inventory; SSSE: self-sacrificing self-enhancement; EXP: exploitativeness; GF: grandiose fantasy; CSE: contingent self-esteem; HS: hiding the self; DEV: devaluing; ER: entitlement rage.

Discussion

The results showed that the five-factor model proposed by Ackerman et al. (2016) best describes the NPI structure across different response formats. These factors are Leadership, Exhibitionism, Vanity, Manipulativeness, and Superiority. As could be noticed in some previous studies (Ackerman, Donnellan, & Robins, 2012; Ackerman et al., 2016), the entitlement aspect of grandiose narcissism is not clearly represented in the NPI structure. The entitlement factor was singled out only in the original Raskin and Terry (1988) study, but in all other models, it formed a factor with other aspects, such as Exploitativeness or Exhibitionism (e.g., Ackerman et al., 2011; Corry et al., 2008; Emmons, 1987). According to numerous authors, the central features of narcissism are grandiosity and entitlement (e.g., Ackerman, Hands, Donnellan, Hopwood, & Witt, 2016; Brown, Budzek, & Tamborski, 2009; Rosenthal & Hooley, 2010). The lack of the entitlement aspect provokes concern because it is characterized as a “socially toxic” aspect of maladaptive narcissism (Ackerman et al., 2011), and it showed positive correlations with some negative outcomes, such as counterproductive school behaviors, and a direct expression of hostile feelings and aggression (Ackerman et al., 2011; Maxwell, Donnellan, Hopwood, & Ackerman, 2011; Raskin & Terry, 1988; Reidy, Zeichner, Foster, & Martinez, 2008). Therefore, without the entitlement scale, it seems that the NPI lacks of some important maladaptive forms of narcissism, which limits the interpretation of the results.

The five factors had satisfactory reliabilities, given the small number of items in most of the factors. However, parameters of homogeneity showed that the redundancy of items in some of the scales could be a problem, which was also pointed out in previous studies (Ackerman et al., 2012; Corry et al., 2008). The most obvious redundancy was obtained in the Vanity scale across different response formats. It is important to note that Vanity scale comprises only three items. This result could mean that this aspect of narcissism is narrower and could be a subordinate dimension of other factors (e.g., Exhibitionism, as Ackerman et al., 2016 suggested), or that this scale does not have items in the NPI that could capture more various indicators of this dimension. In one previous study (Popov & Sekulić Bartoš, 2016), one of the extracted factors of the NPI contained items from both Exhibitionism and Vanity. However, in the present study, Exhibitionism and Vanity showed profiles that were distinct enough for them to be considered as separate factors.

All five factors are highly inter-correlated. However, some distinctions between factors could be made in order to establish their discriminant validity. For example, although Leadership and Manipulativeness correlated highly, it seems that Manipulativeness captures more maladaptive aspects of narcissism, showing stronger negative relations with Honesty–Humility and positive with psychopathy, Machiavellianism, and pathological narcissism. On the other hand, Leadership seems to capture more adaptive aspects of narcissism, showing stronger positive relations with Extraversion. This is in line with the theoretical assumption that Exploitativeness (or Manipulativeness in NPI five-factor model) and Entitlement represent the maladaptive content of narcissism, and leadership represents the adaptive content of narcissism (Emmons, 1987; Pincus et al., 2009; Von Känel, Herr, Van Vianen, & Schmidt, 2017; Watson, Grisham, Trotter, & Biderman, 1984). Therefore, Leadership was more closely related to positive affect, social confidence, and dominance, which are indicators of positive outcomes, e.g., well-being. Manipulativeness was more closely related to features, such as callousness, instrumental motivation, willingness to cheat, and break the rules, showing positive relations with negative outcomes and opposite or non-significant relations with positive outcomes (e.g., Ackerman et al., 2011; Foster et al., 2016; Wetzel et al., 2016).

The status of the rest of the factors in regard to the adaptive-maladaptive distinction is not as clear. One group of authors suggests that Superiority and Vanity are less maladjusted, compared to Exhibitionism which is more pathological (Ansell, 2005; Raskin & Terry, 1988), whereas the second group suggests that both Exhibitionism and Vanity are maladaptive (Popov & Sekulić Bartoš, 2016). Finally, the third group suggests that Exhibitionism is more adaptive (Emmons, 1984; Horton, Bleau, & Drwecki, 2006; Watson et al., 1984). In the present study, Exhibitionism showed the strongest positive correlations with most of the pathological narcissism scales and the strongest negative correlations with Honesty–Humility, which indicates that Exhibitionism could be interpreted as a maladaptive aspect of narcissism. Among HEXACO traits, what makes this scale different is a somewhat higher negative relation with Conscientiousness, indicating that the lack of self-control is a characteristic of this dimension, which is also noted in previous study (e.g., Ansell, 2005). This association also suggests that Exhibitionism is more likely to capture maladaptive narcissism. It could be concluded that, among all the NPI factors, Manipulativeness seems the most closely related to maladaptive narcissism, followed by Exhibitionism. This means that the five-factor model of the NPI dominantly captures adaptive aspects, which limits the information obtained from it.

There are several limitations to this study. First, for direct comparison of different response formats, the same instrument with different response formats should be given to one sample in reverse order with a time interval of a few months in order to eliminate the “memory” effect. Second, validity measures in samples 1 and 2 are not the same, although some of them refer to the same constructs (e.g., HEXACO traits). Further studies should include the same measures for validation, and full-length instruments of validity measures are recommended. Third, the alphas of some of the used scales (e.g., the BHI) are low, but in the same range as the one originally reported (see de Vries, 2013). In the case of the BHI, low alphas are expected, given that each item in the scale measures different facets, so the scales cover a broader domain of a trait. The advantage of the BHI over some other short six-factor measures is that it is a valid measure of the HEXACO model and the matching scales highly correlate with the original HEXACO-PI-R inventory (Dinić, 2018). Fourth, although both samples are from the general population, the student population is dominant. Thus, a further validation should include a sample that is more representative of the general population. Fifth, all used instruments are self-report measures. Therefore, future studies should include behavioral indicators as specific outcome criteria regarding adaptability (e.g., self-esteem, well-being, or presence of personality pathology, for example, depression, anxiety, and antisocial and borderline personality disorders) for a further validation of the NPI scales.

Conclusion

The results of this study indicated that the advantages of the NPI should be obtained if it is scored on a scale level, whereby the scoring model proposed by Ackerman et al. (2016) had the best fit indices. To the best of our knowledge, this is the first study in which all proposed models of the NPI were tested and the first study in which validation of Ackerman et al. (2016) model was examined. This model includes five dimensions of narcissism—Leadership, Exhibitionism, Vanity, Manipulativeness, and Superiority. Although most of the authors agree that the leadership aspect of narcissism is adaptive and the Entitlement/Exploitativeness aspect is maladaptive (e.g., Brown et al., 2009; Emmons, 1987), the status of the rest of the narcissism dimensions was not clear. Based on correlations with HEXACO traits, Machiavellianism, psychopathy, and pathological narcissism, it was confirmed that Leadership is more adaptive, and Manipulativeness (exploitativeness) is more maladaptive. However, the results of this study shed light on the status of other dimensions, indicating that Exhibitionism is more likely to capture maladaptive narcissism and exhibits a profile pattern that is most similar to the profile pattern of Manipulativeness, while Vanity and Superiority dimensions are more likely to capture adaptive narcissism.

The use of scales instead of a total score is recommended, given that there are some concerns that the NPI total score refers only to the adaptive aspects of narcissism (e.g., leadership and attention seeking, see Ackerman et al., 2012) and that it is not justified to combine adaptive and maladaptive aspects of narcissism (Brown et al., 2009; Stanton et al., 2017). However, the lack of the entitlement scale brings into question the NPI’s suitability for capturing the maladaptive aspects of narcissism. Thus, the content of the five-factor NPI model had narrower scope of maladaptive forms of narcissism, which were limited to forms of manipulativeness and exhibitionism. Further improvement is needed given that the content of items in some scales is highly similar (e.g., in Vanity).

Footnotes

Acknowledgments

The authors would like to thank Goran Knežević for significant contribution in the translation of the NPI and Mach IV, Janko Međedović in the translation of the Psychopathic Personality Trait Scale, and Ashley Lauren Watts for providing procedure for calculating D statistic.

Article Notes

References

Ackerman

R. A.

Donnellan

M. B.

Roberts

B. W.

Fraley

R. C.

(2016) The effect of response format on the psychometric properties of the Narcissistic Personality Inventory: Consequences for item meaning and factor structure. Assessment 23(2): 203–220. doi:10.1177/1073191114568113.

Ackerman

R. A.

Donnellan

M. B.

Robins

R. W.

(2012) An item response theory analysis of the Narcissistic Personality Inventory. Journal of Personality Assessment 94(2): 141–155. doi:10.1080/00223891.2011.645934.

Ackerman

R. A.

Hands

A. J.

Donnellan

M. B.

Hopwood

C. J.

Witt

E. A.

(2016) Experts’ views regarding the conceptualization of narcissism. Journal of Personality Disorder 30: 1–16. doi:10.1521/pedi_2016_30_254.

Ackerman

R. A.

Witt

E. A.

Donnellan

M. B.

Trzesniewski

K. H.

Robins

R. W.

Kashy

D. A.

(2011) What does the Narcissistic Personality Inventory really measure? Assessment 18(1): 67–87. doi:10.1177/1073191110382845.

Ansell, B. E. (2005). The structure of narcissistic personality: Adaptive and maladaptive dimensions as an integrated model of narcissism (Doctoral dissertation). Ann Arbor, MI: The Pennsylvania State University. Retrieved from eDT database at https://etda.libraries.psu.edu/files/final_submissions/1706.

Ashton

M. C.

Lee

(2009) The HEXACO-60: A short measure of the major dimensions of personality. Journal of Personality Assessment 91(4): 340–345. doi:10.1080/00223890902935878.

Barelds

D. P. H.

Dijkstra

(2010) Narcissistic Personality Inventory: Structure of the adapted Dutch version. Scandinavian Journal of Psychology 51(2): 132–138. doi:10.1111/j.1467-9450.2009.00737.x.

Boduszek

Debowska

Dhingra

DeLisi

(2016) Introduction and validation of Psychopathic Personality Traits Scale (PPTS) in a large prison sample. Journal of Criminal Justice 46: 9–17. doi:10.1016/j.jcrimjus.2016.02.004.

Book

Visser

B. A.

Volk

A. A.

(2015) Unpacking “evil”: Claiming the core of the Dark Triad. Personality and Individual Differences 73: 29–38. doi:10.1016/j.paid.2014.09.016.

10.

Brown

R. P.

Budzek

Tamborski

(2009) On the meaning and measure of narcissism. Personality and Social Psychology Bulletin 35(7): 951–964. doi:10.1177/0146167209335461.

11.

Christie

Geis

F. L.

(1971) Studies in Machiavellianism, New York, NY: Academic Press.

12.

Clark

A. D.

Watson

(1995) Construct validity: Basic issues in objective scale development. Psychological Assessment 7(3): 309–319.

13.

Clarke

Karlov

Neale

(2015) The many faces of narcissism: Narcissism factors and their predictive utility. Personality and Individual Differences 81: 90–95. doi:10.1016/j.paid.2014.11.021.

14.

Corry

Merritt

R. D.

Mrug

Pamp

(2008) The factor structure of the Narcissistic Personality Inventory. Journal of Personality Assessment 90(6): 593–600. doi:10.1080/00223890802388590.

15.

Cronbach

L. J.

Gleser

G. C.

(1953) Assessing similarity between profiles. Psychological Bulletin 50: 456–473. doi:10.1037/h0057173.

16.

de Vries

R. E.

(2013) The 24-item Brief HEXACO Inventory (BHI). Journal of Research in Personality 47(6): 871–880. doi:10.1016/j.jrp.2013.09.003.

17.

Dinić

B. M.

(2018) Comparison of three short six-factor personality instruments. Primenjena Psihologija 11(2): 189–206. doi:10.19090/pp.2018.2.189-206.

18.

Egan

Lewis

(2011) Neuroticism and agreeableness differentiate emotional and narcissistic expressions of aggression. Personality and Individual Differences 50(6): 845–850. doi:10.1016/j.paid.2011.01.007.

19.

Emmons

R. A.

(1984) Factor analysis and construct validity of the Narcissistic Personality Inventory. Journal of Personality Assessment 48: 291–300. doi:10.1207/s15327752jpa4803_11.

20.

Emmons

R. A.

(1987) Narcissism: Theory and measurement. Journal of Personality and Social Psychology 52(1): 11–17.

21.

Ferguson

C. J.

(2009) An effect size primer: A guide for clinicians and researchers. Professional Psychology: Research and Practice 40(5): 532–538. doi:10.1037/a0015808.

22.

Gravetter

Wallnau

(2014) Essentials of statistics for the behavioral sciences, 8th ed. Belmont, CA: Wadsworth.

23.

Grosz, M. P., Emons, W. H. M., Wetzel, E., Leckelt, M., Chopik, W. J., Rose, N., & Back, M. D. (2017). A comparison of unidimensionality and measurement precision of the Narcissistic Personality Inventory and the Narcissistic Admiration and Rivalry Questionnaire. Assessment. Advance online publication. doi:10.1177/1073191116686686.

24.

Foster

J. K.

Shiverdecker

Turner

(2016) What does the Narcissistic Personality Inventory measure across the total score continuum? Current Psychology 35(2): 207–219. doi:10.1007/s12144-016-9407-5.

25.

Horton

R. S.

Bleau

Drwecki

(2006) Parenting narcissus: What are the links between parenting and narcissism? Journal of Personality 74(2): 345–376. doi:10.1111/j.1467-6494.2005.00378.x.

26.

Bentler

P. M.

(1999) Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling 6(1): 1–55. doi:10.1080/10705519909540118.

27.

Kansi

(2003) The Narcissistic Personality Inventory: Applicability in a Swedish population sample. Scandinavian Journal of Psychology 44: 441–448.

28.

Kline

R. B.

(2016) Principles and practice of structural equation modeling, 3rd ed. New York, NY: The Guilford Press.

29.

Kubarych

T. S.

Deary

I. J.

Austin

E. J.

(2004) The Narcissistic Personality Inventory: Factor structure in a non-clinical sample. Personality and Individual Differences 36(4): 857–872. doi:10.1016/S0191-8869(03)00158-2.

30.

Lee

Ashton

M. C.

(2005) Psychopathy, Machiavellianism, and Narcissism in the five-factor model and the HEXACO model of personality structure. Personality and Individual Differences 38(7): 1571–1582. doi:10.1016/j.paid.2004.09.016.

31.

Maxwell

Donnellan

M. B.

Hopwood

C. J.

Ackerman

R. A.

(2011) The two faces of narcissus? An empirical comparison of the Narcissistic Personality Inventory and the Pathological Narcissism Inventory. Personality and Individual Differences 50(5): 577–582. doi:10.1016/j.paid.2010.11.031.

32.

Miller

J. D.

Campbell

W. K.

(2008) Comparing clinical and social-personality conceptualizations of narcissism. Journal of Personality 76(3): 449–476. doi:10.1111/j.1467-6494.2008.00492.x.

33.

Miller

J. D.

Gentile

Crowe

Carter

N. T.

Hoffman

B. J.

Campbell

W. K.

(2018) A comparison of the nomological networks associated with forced-choice and Likert formats of the Narcissistic Personality Inventory. Journal of Personality Assessment 100(3): 259–267. doi:10.1080/00223891.2017.1310731.

34.

Paulhus

D. L.

Williams

(2002) The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy. Journal of Research in Personality 36: 556–568. doi:10.1016/S0092-6566(02)00505-6.

35.

Pincus

A. L.

Ansell

E. B.

Pimentel

C. A.

Cain

N. M.

Wright

A. G. C.

Levy

K. N.

(2009) Initial construction and validation of the Pathological Narcissism Inventory. Psychological Assessment 21(3): 365–379. doi:10.1037/a0016530.

36.

Popov

Sekulić Bartoš

(2016) Narcizam u kontekstu psihološkog zdravlja adolescenata [Narcissism in the context of the psychological health of adolescents]. Psihološka Istraživanja 19(2): 147–163.

37.

Raskin

R. N.

Hall

C. S.

(1979) A Narcissistic Personality Inventory. Psychological Reports 45(2): 590–590. doi:10.2466/pr0.1979.45.2.590.

38.

Raskin

R. N.

Terry

(1988) A principal-components analysis of the Narcissistic Personality Inventory and further evidence of its construct validity. Journal of Personality and Social Psychology 54: 890–902. doi:10.1037/0022-3514.54.5.890.

39.

Reidy

D. E.

Zeichner

Foster

J. D.

Martinez

M. A.

(2008) Effects of narcissistic entitlement and exploitativeness on human physical aggression. Personality and Individual Differences 44: 865–875. doi:10.1016/j.paid.2007.10.015.

40.

Rosenthal

S. A.

Hooley

J. M.

(2010) Narcissism assessment in social–personality research: Does the association between narcissism and psychological health result from a confound with self-esteem? Journal of Research in Personality 44: 453–465. doi:10.1016/j.jrp.2010.05.008.

41.

Rosenthal

S. A.

Montoya

R. M.

Ridings

L. E.

Rieck

S. M.

Hooley

J. M.

(2011) Further evidence of the Narcissistic Personality Inventory’s validity problems: A meta-analytic investigation—Response to Miller, Maples, and Campbell (this issue). Journal of Research in Personality 45: 408–416. doi:10.1016/j.jrp.2011.06.004.

42.

Rosseel

(2012) lavaan: An R package for structural equation modeling. Journal of Statistical Software 48: 1–36.

43.

Stanton

Daly

Stasik-O’Brien

S. M.

Ellickson-Larew

Clark

L. A.

Watson

(2017) An integrative analysis of the Narcissistic Personality Inventory and the Hypomanic Personality Scale: Implications for construct validity. Assessment 24(6): 695–711. doi:10.1177/1073191115625801.

44.

Svindseth

M. F.

Sørebø

Ø.

Nøttestad

J. A.

Roaldset

J. O.

Wallin

Dahl

A. A.

(2009) Psychometric examinatio and normative data for the Narcissistic Personality Inventory 29 item version. Scandinavian Journal of Psychology 50: 151–159. doi:10.1111/j.1467-9450.2008.00686.x.

45.

Tabachnick, B. G., & Fidell, L. S. (2013). Using Multivariate Statistics, 6th ed. Boston: Pearson.

46.

Von Känel

Herr

R. M.

Van Vianen

A. E. M.

Schmidt

(2017) Association of adaptive and maladaptive narcissism with personal burnout: Findings from a cross-sectional study. Industrial Health 55(3): 233–242. doi:10.2486/indhealth.2016-0136.

47.

Watson

P. J.

Grisham

S. O.

Trotter

M. V.

Biderman

M. D.

(1984) Narcissism and empathy: Validity evidence for the Narcissistic Personality Inventory. Journal of Personality Assessment 48(3): 301–305. doi:10.1207/s15327752jpa4803_12.

48.

Wetzel

Roberts

B. W.

Fraley

R. C.

Brown

(2016) Equivalence of Narcissistic Personality Inventory constructs and correlates across scoring approaches and response formats. Journal of Research in Personality 61: 87–98. doi:10.1016/j.jrp.2015.12.002.

49.

Wright

A. G. C.

Lukowitsky

M. R.

Pincus

A. L.

Conroy

D. E.

(2010) The higher order factor structure and gender invariance of the Pathological Narcissism Inventory. Assessment 17(4): 467–483. doi:10.1177/1073191110373227.