Keywords

The mask model of narcissism explains the narcissist’s overinflated, positive self-views as a protective mask against deep-seated insecurities. This conceptualization of narcissism emerges from psychoanalytic origins (Freud, 1914; Kohut, 1966) and is most evident in the work of Kernberg (1986), who called narcissistic grandiosity a defense against an underlying “empty self.” Tests of the mask model in social-personality psychology often conceptualize it as fragile self-esteem that assumes one of several different forms. First, the most literal interpretation of the mask model holds that grandiose narcissism is characterized by high explicit (conscious, deliberative) self-esteem that masks underlying low implicit (automatic, overlearned) self-esteem. We refer to this as the discrepant self-esteem hypothesis because it posits high surface self-esteem that is discrepant from low underlying self-esteem. Second, the unstable self-esteem hypothesis posits that grandiose narcissism is characterized by fragile high self-esteem that is unstable and prone to fluctuation; that is, the individual’s self-esteem is high on average, but it plummets at times in response to contextual factors. Third, the contingent self-esteem hypothesis operationalizes the mask as overblown self-esteem that is fragile due to its contingence on achievement in agentic domains. Although distinct, these three forms of fragile self-esteem are associated with each other and with verbal defensiveness, suggesting that they may all indicate a similar factor (Kernis, Lakey, & Heppner, 2008). Here, we review the evidence for each of these interpretations of the mask model.

Note that the evidence reviewed here derives from research on grandiose narcissism in nonclinical samples. While narcissism is viewed as a personality disorder within clinical psychology, social-personality psychologists are often more interested in nonclinical, trait-like narcissism (Miller & Campbell, 2008). This personality approach defines grandiose narcissism as an enduring pattern of arrogance, entitlement, self-absorption, and superiority that is measured with the Narcissistic Personality Inventory (NPI ; Raskin & Hall, 1979, 1981) or other self-report scales (see Chap. 12 by Foster et al., this volume), but that does not reach clinical levels.

The Discrepant Self-Esteem Hypothesis

Whereas explicit self-esteem reflects conscious feelings of self-worth that are measured via self-reports, implicit self-esteem reflects relatively automatic self-evaluations that are overlearned, difficult to verbalize, and arise in response to self-relevant stimuli (Greenwald & Banaji, 1995; Zeigler-Hill & Jordan, 2010). As noted, the discrepant self-esteem hypothesis predicts that larger discrepancies between favorable explicit self-esteem and unfavorable implicit self-esteem – usually indexed by a statistical interaction between explicit and implicit self-esteem – should be associated with higher grandiose narcissism scores .

Early tests of the discrepant self-esteem hypothesis offered promise. Two studies found that individuals high in grandiose narcissism scored high in explicit self-esteem but low in implicit self-esteem (Jordan, Spencer, Zanna, Hoshino-Browne, & Correll, 2003; Zeigler-Hill, 2006). Though not a direct test of the mask model, another study found that people with high explicit and low implicit self-esteem displayed the highest levels of unrealistic self-enhancement, which is a feature of grandiose narcissism (Bosson, Brown, Zeigler-Hill, & Swann, 2003). Similarly, Kernis and colleagues found that people responded in a more narcissistic manner after being primed with either positive or negative implicit self-esteem if the valence of the implicit self-esteem prime was discrepant from their trait self-esteem (Kernis et al., 2005).

These effects proved inconsistent and difficult to replicate, however. A series of studies failed to find evidence that discrepancies between high explicit and low implicit self-esteem predict grandiose narcissism (Bosson & Prewitt-Freilino, 2007; Campbell, Bosson, Goheen, Lakey, & Kernis, 2007; Gregg & Sedikides, 2010). A meta-analysis that examined data from both published and unpublished studies found no overall association between grandiose narcissism and the combination of high explicit and low implicit self-esteem, regardless of how implicit self-esteem was assessed (Bosson et al., 2008).

In an adaptation of the discrepant self-esteem hypothesis, Campbell and colleagues asked whether people high in grandiose narcissism have high explicit self-views in agentic domains and low implicit self-views in communal domains (Campbell et al., 2007). This logic is based on the tendency for grandiose narcissists to inflate self-reports of their agentic, but not their communal, traits and tendencies. However, Campbell et al. found instead that grandiose narcissism correlated positively with both explicit and implicit agency and not at all with implicit communion, a pattern that did not support the discrepant self-esteem hypothesis.

Another approach to testing the discrepant self-esteem hypothesis utilizes a bogus pipeline (an ostensible lie detector) to assess people’s underlying self-esteem. Using this method, Myers and Zeigler-Hill (2012) found that women higher in grandiose narcissism reported lower self-esteem in the bogus pipeline condition than they did in a control condition. This is consistent with the hypothesis that people high in narcissism mask their fragility behind exaggeratedly positive self-reports. However, Brunell and Fisher (2014) used a similar bogus pipeline procedure (and a much larger sample size) and found that neither men nor women high in grandiose narcissism modified their reports of their high self-esteem across conditions, thus failing to replicate earlier findings.

In a more recent study, grandiose narcissism was only associated with higher explicit self-esteem among people whose implicit self-esteem was either moderate or high; among those with low implicit self-esteem, grandiose narcissism was unrelated to explicit self-esteem (Di Pierro, Mattevelli, & Gallucci, 2016). Again, these findings are inconsistent with the mask model , because they do not indicate that people high in grandiose narcissism have high explicit self-esteem that masks hidden feelings of low self-regard.

The Unstable Self-Esteem Hypothesis

Unstable high self-esteem is characterized by exaggeratedly positive views of the self that require validation, are vulnerable to threat, and fluctuate regularly in response to self-relevant feedback and events (Kernis, 2003). The unstable self-esteem hypothesis states that individuals with frequently fluctuating high trait self-esteem will score higher in grandiose narcissism.

Evidence in support of the unstable self-esteem hypothesis is inconsistent. While some studies find that narcissism is associated with unstable self-esteem (Rhodewalt, Madrian, & Cheney, 1998; Zeigler-Hill, Chadha, & Osterman, 2008a), others fail to do so (Webster, Kirkpatrick, Nezlek, Smith, & Paddock, 2007; Zeigler-Hill, 2006). A meta-analysis of 11 datasets representing 1349 respondents indicated no overall relationship between unstable self-esteem and grandiose narcissism (Bosson et al., 2008).

An updated version of the unstable self-esteem hypothesis posits that people high in grandiose narcissism do not display uniform self-esteem reactivity, but instead, demonstrate reactivity to specific events. For instance, one study showed that people high in narcissism reported lower state self-esteem on days that they experienced more negative achievement-related events (Zeigler-Hill, Myers, & Clarke, 2010), and this pattern did not hold for positive achievement events or for negative or positive intimacy-related events. Similarly, individuals who scored high in grandiose narcissism demonstrated stronger anger responses to achievement failures compared to interpersonal threats (Besser & Priel, 2010). Note that these patterns reflect the agency-communion distinction discussed earlier: Accruing evidence suggests that agentic and achievement-related events have special significance for the self-esteem of people high in grandiose narcissism, while some types of interpersonal and social events appear irrelevant to their self-esteem.

People who score high in grandiose narcissism may not report as much self-esteem reactivity to interpersonal events because, although they like attention, they are relatively unconcerned with gaining others’ approval (Morf & Rhodewalt, 2001). Similarly, people high in grandiose narcissism tend to consider others in terms of their utility rather than as sources of affiliation (Campbell, 1999). Thus, interpersonal events that involve getting attention and controlling others may affect the self-worth of people high in grandiose narcissism, while those that involve connectedness, intimacy, and warmth may not.

However, other findings suggest that the effects of interpersonal events on the self-esteem of individuals high in narcissism may differ as a function of the specific narcissism facet under examination. For example, individuals who score high on the Entitlement/Exploitativeness subscale of the NPI demonstrate lower state self-esteem on days that they experience social rejection. Alternatively, the self-esteem of those who score high on the NPI subscales of Leadership/Authority and Grandiose/Exhibitionism is largely unrelated to negative interpersonal events (Zeigler-Hill & Besser, 2013). That is, their self-esteem remains high even on days that they experience rejection or exclusion. Thus, when grandiose narcissism is measured as a unidimensional construct, people who score high on it appear more reactive to negative achievement-related than interpersonal events. However, when each facet is examined in isolation, scores on Entitlement/Exploitativeness capture a facet of grandiose narcissism that is fragile in response to negative interpersonal events. Interestingly, this facet of grandiose narcissism – perhaps because it involves mistreatment of others – may also be most strongly predictive of experiencing interpersonal difficulties in the first place.

The Contingent Self-Esteem Hypothesis

Contingent self-esteem is self-esteem that is contingent on specific achievements or achievements in specific domains (Crocker & Wolfe, 2001). According to the contingent self-esteem hypothesis, grandiose narcissism should be related to self-esteem that is contingent on achievement in agentic domains.

At first glance, there appears to be a great deal of overlap between “unstable self-esteem that is reactive to negative achievement-related events” as described in the prior section and self-esteem that is contingent on achievement in agentic domains. In fact, these two hypotheses are conceptually similar, but they emerged out of different research traditions and utilize different measurement approaches. Self-esteem instability refers to individual differences in short-term fluctuations of self-worth and is usually assessed by measuring broadly defined state self-esteem – without reference to specific precipitating events – multiple times per day (Kernis, 2005). Contingent self-esteem can reflect either differences in the degree to which people derive self-worth from performance in specific domains (Deci & Ryan, 1995) or differences in the specific domains on which people stake their self-esteem (Crocker & Wolfe, 2001). Moreover, contingent self-esteem is usually assessed by asking people to report on the degree to which their self-esteem depends on achievements or events within different domains (e.g., competition, family support, school competence; Crocker & Wolfe, 2001). Importantly, unstable self-esteem and contingent self-esteem are weakly correlated, r = 0.32, p < 0.01, indicating that they are distinct (see Kernis et al., 2008). Note that a correlation of 0.32 means that unstable self-esteem accounts for only about 10% of the variance in contingent self-esteem.

In support of the contingent self-esteem hypothesis, people high in grandiose narcissism appear to link their self-worth to events and achievements within primarily agentic domains. For example, scores on the NPI were positively associated with contingent self-esteem in domains of competition and appearance, they were negatively associated with contingent self-esteem in the domains of others’ approval and virtue (being a good and moral person), and they were unrelated to contingent self-esteem in domains of academic competence, family support, and God’s love (Crocker, Luhtanen, Cooper, & Bouvrette, 2003). In a later study, grandiose narcissism scores correlated positively with contingent self-esteem in the competition domain and negatively with contingent self-esteem in domains of approval and family support (Zeigler-Hill, Clark, & Pickard, 2008b).

Straying a bit from the mask model, a large literature indicates that people high in grandiose narcissism report exaggeratedly positive self-views in agentic domains. For instance, high grandiose narcissism is associated with inflated self-reports of intelligence and power and a tendency to display the “better-than-average effect” for agentic but not communal traits (Campbell, Rudich, & Sedikides, 2002). Similarly, as described earlier, Campbell et al. (2007) found that grandiose narcissism is positively related to implicit agency, but unrelated to implicit communion. Moreover, grandiose narcissism is closely tied to social dominance, which is a tendency toward social status and leadership over others (Brown & Zeigler-Hill, 2004; Hodson, Hoggs, & MacInnis, 2009). Similarly, adolescents who score higher in grandiose narcissism also report more agentic goals that reflect power and status (Findley & Ojanen, 2013), and narcissism predicts increases in agentic, but not communal, goals over time. These findings do not directly indicate that the self-esteem of people high in grandiose narcissism is contingent on performance in agentic domains. However, they do suggest that grandiose narcissists emphasize goals related to social status and power more strongly than goals related to warmth, intimacy, and being a “good” person.

Summary and Conclusions

Is narcissistic personality characterized by a grandiose mask that hides underlying fragility? The answer appears to depend on how one operationalizes key variables. Support for the discrepant self-esteem hypothesis is lacking, with the exception of a promising (but unreplicated) recent study that utilized a bogus pipeline paradigm (Myers & Zeigler-Hill, 2012). The lack of consistent support for this hypothesis may arise because researchers conceptualize self-esteem overly broadly (i.e., as general feelings of self-worth) rather than examining specific self-esteem domains (e.g., competence versus self-liking; Tafarodi & Swann, 1995), or fail to link discrepant self-esteem to achievements in agentic versus communal domains. Another possibility is that many researchers seek evidence of the discrepant self-esteem hypothesis in a statistical interaction between explicit and implicit self-esteem (but see Gregg & Sedikides, 2010, for an exception). However, interaction effects are often small and difficult to detect, and implicit self-esteem scores tend to contain a lot of measurement error. These factors combined may make it exceedingly difficult to find clear and consistent support for the mask model. Next, the unstable self-esteem hypothesis is not supported in its general form, but there is support for a more specific version of it that proposes that individuals high in grandiose narcissism have unstable self-esteem regarding negative achievement-related events (Zeigler-Hill et al., 2010). Thus, while grandiose narcissism is not generally unstable, it does fluctuate in response to events in agentic domains. That said, scores on one facet of grandiose narcissism – Entitlement/Exploitativeness – are related to self-esteem that is unstable in the face of interpersonal threats.

Finally, the contingent self-esteem hypothesis receives consistent and strong support, possibly because it expressly proposes that grandiose narcissists stake their self-worth on agentic – and not communal – goals. This hypothesis is consistent, moreover, with a large literature in which people higher in grandiose narcissism routinely self-report inflated agentic self-views and claim that agentic goals are especially important to them. From the perspective of the psychodynamic mask model, it is interesting that scores on the NPI are positively associated with endorsement of statements such as “My self-worth is influenced by how well I do on competitive tasks” and negatively correlated with endorsement of statements like “My self-esteem would suffer if I did something unethical” (Crocker et al., 2003). In contrast to the more literal interpretation of the mask model – that posits that the fragility of the underlying self must be kept from consciousness because it is too threatening – the contingent self-esteem hypothesis proposes and finds that people high in grandiose narcissism are aware that their self-esteem is linked to their achievements. That is, grandiose narcissists admit that their self-esteem depends on competition and achievement, while denying that it depends on virtuosity and family support. Whether this pattern indicates a truly “fragile” self, however, is another question altogether. After all, if people high in grandiose narcissism routinely convince themselves that their own achievements surpass everyone else’s, then admitting that their self-esteem is dependent on achievements may not reveal much genuine vulnerability.

Going forward, researchers interested in the mask model of narcissism may profit from focusing on the links between narcissistic self-esteem and outcomes in agentic and achievement-related domains as opposed to communal domains. It may be important, moreover, to distinguish between different facets of grandiose narcissism (e.g., Entitlement/Exploitativeness vs. other facets). In the evolution of research testing each mask model hypothesis, gains are made when researchers operationalize variables more precisely and specifically as opposed to globally.

In distinguishing between agentic and communal domains of functioning, we caution researchers not to conflate communal domains with social ones and agentic domains with nonsocial ones. Recall that some research finds that grandiose narcissism is unrelated to reactivity to social- and intimacy-related events (Zeigler-Hill et al., 2010), while other research finds that grandiose narcissism is associated with leadership and social status goals (Brown & Zeigler-Hill, 2004; Findley & Ojanen, 2013). Thus, it is not the sociality of a given event, but the opportunity that an event affords for demonstrating agency and leadership, that is relevant to the self-esteem of people high in grandiose narcissism. Social contexts that offer opportunities for dominance, assertiveness, and competition should have self-esteem relevance for people high in grandiose narcissism, while those that do not should be less relevant to the self-esteem of grandiose narcissists.

In conclusion, the mask model has come a long way, and each of its various iterations represents an important improvement on older versions. The most promising version of the mask model is one that views the self-esteem of those high in grandiose narcissism as “fragile” insofar as it is contingent on performance and outcomes in domains that are valued by grandiose narcissists: those that offer opportunities for demonstrating agentic superiority and status over others. Researchers who continue this work are encouraged to use specific and precise measures of key constructs, as global measures that gloss over important distinctions may obscure the complex relationships between grandiose narcissism and self-esteem.