Dark Triad in the HEXACO model
The first part of the present research involved the explication of the Dark Triad constructs. Specifically, we investigated the common and specific characteristics of the Dark Triad using the HEXACO model of personality. As hypothesized, the common variance of the Dark Triad was found to be very highly saturated with the Honesty–Humility factor across two samples involving two different measures of the Dark Triad. The hypothesized relations involving the unique characteristics of each of the Dark Triad variables were generally supported when the Dark Triad variables were measured by the SD3 (Paulhus & Jones, 2011). Specifically, Machiavellianism was associated with (low) HEXACO Agreeableness, Narcissism was associated with Extraversion, and Psychopathy was associated with (low) Emotionality and to a lesser degree with (low) Conscientiousness. Such a pattern of results was not fully observed in the present research when the Dark Triad were measured by the Dirty Dozen (Jonason & Webster, 2010).
The results presented earlier have some implications for the use of Dark Triad measures in empirical research. If one is interested in obtaining a composite measure of the Dark Triad to investigate its relationship with various outcome variables, then both the SD3 and Dirty Dozen might be reasonable choices as measures of the Dark Triad. But if one is interested in measuring each of the Dark Triad variables in its own right, so that unique as well as common variance is emphasized, then the SD3 is the recommended measure.
Big Five, Dark Triad, and HEXACO model
We examined whether the Dark Triad composite can compensate for the limited validity of the Big Five or FFM in predicting behaviours related to the domains of sex, power, and money. The Big Five/FFM as measured by the BFI and NEO-FFI showed rather limited ability to predict these variables, and the Dark Triad composite satisfactorily compensated for the predictive deficit of the Big Five. This finding is as expected given the recent interest in the Dark Triad as predictors of various phenomena in these domains (Hodson et al., 2009; Jonason et al., 2009).
We also compared the predictive validity of the HEXACO factors with that of an ad hoc Big Five-plus-Dark Triad combination. In Samples 1 and 2, we found that the six scales of the HEXACO-60 showed levels of predictive validity that sometimes exceeded those yielded by the combination of the BFI (or NEO-FFI) and the SD3 (or Dirty Dozen). These results can, in part, be attributed to the fact that HEXACO Honesty–Humility sometimes showed higher correlations with the outcome variables than did the Dark Triad measures. The predictive validity of the Dark Triad for some criteria was enhanced when the three Dark Triad variables were used instead of their composite. However, the predictive validity obtained when the Dark Triad variables were entered separately was not generally higher than that achieved by the six HEXACO factors, and substantially lower than that achieved by the four Honesty–Humility facets and the other five HEXACO factors. As far as predictive validity with regard to Sex, Money, and Power is concerned, the results suggest that an ad hoc combination of the Big Five/FFM and Dark Triad would be somewhat less useful than the HEXACO model.
Beyond considerations of predictive validity alone, the HEXACO model may also be preferred over the Big Five/FFM-plus-Dark Triad combination for theoretical reasons. One such reason involves the independence of the factors: Whereas Big Five/FFM Agreeableness overlaps substantially with the Dark Triad (r = −.52 and −.55 in self-report data of Samples 1 and 2), none of the HEXACO scales are so strongly intercorrelated (the strongest correlations being r = .32 and .43 between Honesty–Humility and Agreeableness in the self-report data of Samples 1 and 2). Another reason involves the interpretability of the factors: HEXACO Honesty–Humility, Agreeableness, and Emotionality are suggested to represent dimensions underlying two forms of reciprocal altruism and kin altruism, respectively (see detailed discussion in Ashton & Lee, 2007); in contrast, there are no such conceptual connections among the Dark Triad, Big Five/FFM Agreeableness, and Big Five/FFM Emotional Stability.
The present research also has some implications regarding the use of different measures of the Big Five or FFM. It is well known that FFM measures (such as the NEO-FFI or NEO-PI-R; Costa & McCrae, 1992) include a broad Agreeableness factor that contains a larger portion of the variance related to Honesty–Humility than occurs in measures of ‘classic’ Big Five Agreeableness (Ashton & Lee, 2005; see also Miller, Gaughan, Maples, & Price, 2011). In fact, the inclusion of some Honesty–Humility content within the Agreeableness domain (via the Straightforwardness and Modesty facets in the NEO-PI-R) is a feature of FFM Agreeableness not shared by other widely used measures of Big Five Agreeableness. The NEO-FFI—the shorter version of the NEO-PI-R—used in this study also has an Agreeableness factor that contains some Honesty–Humility content but less than in the longer NEO-PI-R (e.g. Miller et al., 2011). In the present research, the NEO-FFI did not show levels of validity approaching those of the HEXACO model in predicting the Sex, Power, and Money factors. Therefore, the incorporation of some Honesty–Humility content within the Agreeableness factor is unlikely to compensate fully for the predictive deficit of the Big Five or FFM in predicting variables similar to those included in the present research.
Readers might wonder whether the advantage of the HEXACO model over the Big Five/FFM in the prediction of Sex, Power, and Money is attributable to any direct overlap of item content between predictors and criteria. In the case of the Sex criteria, the predictive advantage is clearly not attributable to any such overlap, as none of the HEXACO-PI-R items refer to sexual behaviour. The same conclusion applies in the case of the Power criteria; note that although the Modesty facet of HEXACO-PI-R Honesty–Humility contains some items referring to social status, it does not include any items describing a desire for power over others. With regard to the money criteria, some content overlap does exist, insofar as the fairness and greed-avoidance facets of HEXACO-PI-R Honesty–Humility do contain items referring to money and/or material goods. If a reduced Honesty–Humility scale is computed from the items of the remaining two facets (sincerity and modesty), then the predictive advantage of the HEXACO variables over the Big Five/FFM variables is substantially reduced, but not eliminated.4 This result suggests that item content related to money and material goods is partly but not entirely responsible for the advantage of the HEXACO model over the Big Five/FFM in predicting the money criterion.
We should note that some previous studies have examined the relative validity of the Big Five/FFM and HEXACO factors in predicting some variables within the Sex and Money domains (e.g. sexual quid pro quos and materialism in Ashton & Lee, 2008). However, the present research was much broader in scope: Not only did it include a wider array of variables within those two domains (e.g. game-playing ‘love styles’ and conspicuous consumption) but it also extended these comparisons to the domain of power. Moreover, the present studies have provided the first joint comparison of the predictive validity of the Dark Triad, the Big Five/FFM, and the HEXACO factors.
Sex, Power, Money, and Personality
We found that the criterion variables involving money (e.g. conspicuous consumption, materialism), power (e.g. desire for power), and sex (e.g. short-term mating strategy and instrumental sexual behaviours) correlated more strongly with Honesty–Humility and the Dark Triad composite than with the other personality variables included in this study. All three outcome variables were particularly strongly correlated with Honesty–Humility. Thus, although these three criterion domains were only modestly intercorrelated, their common element appears to be low Honesty–Humility.
Why should low Honesty–Humility be the common element underlying these three domains? On the one hand, low Honesty–Humility represents the willingness to gain at the expense of others. On the other hand, the Money, Power, and Sex factors of this study all involve the motivation to have more resources than others (rather than being content merely to have equal resources): The Money factor represents a drive to consume more material resources than do others, the Power factor represents a drive for superior social status, and the Sex factor emphasized the motivation to have many uncommitted, short-term sexual partners.5 Thus, through their emphasis on having more resources than do others, the Money, Power, and Sex factors have a close conceptual link with the exploitation that characterizes low Honesty–Humility.
There are several limitations in the present research. First, the criterion variables were assessed with self-reports rather than objective measures of actual behaviours. Although we avoided the problem of common rating source variance by including peer reports of personality, it would have been useful to have included some reliable and construct valid objective criteria (cf., Hershfield, Cohen, & Thompson, 2012).
As with some studies in this field, we relied on convenience samples of psychology undergraduate students. One of the characteristics of such samples includes the imbalance in gender ratio. In the present research, about 70% of the participants were women. It may well be that the relationships of Honesty–Humility and the Dark Triad with some criterion variables (e.g. short-term mating tendencies) would differ depending on participants' gender. Owing to the relatively small absolute number of male participants in our studies, we did not pursue within-gender analyses.
In assessing the Dark Triad, we used two recently developed short instruments. These measures were developed with reference to the NPI (Raskin & Hall, 1979), the SRP-III (Paulhus et al., ), and the Mach-IV (Christie & Geis, 1970). One might question whether the results of the current study would generalize to the original measures of the three constructs. We believe that the results involving the SD3 (Paulhus & Jones, 2011) are likely to be highly generalizable for two reasons. First, as shown by Paulhus and Jones, the SD3 scales displayed strong convergence with the original measures of the constructs. Second, the SD3 generally showed the hypothesized relationships with the HEXACO variables, even though these hypotheses were largely derived from the content of the original measures of the constructs. With regard to the Dirty Dozen, this instrument appears to tap the common core of the Dark Triad reasonably well, but it appears not to capture fully the unique features of the original Dark Triad measures. Some caution is therefore warranted regarding the generalizability of results obtained from the Dirty Dozen.