Narcissism is a personality trait associated with an inflated, grandiose self-concept and a lack of intimacy in interpersonal relationships. A popular assumption is that narcissists’ positive explicit (conscious) self-views mask implicit (nonconscious) self-loathing. This belief is typically traced to psychodynamic theory, especially that of Kohut (1966; Morrison, 1983). Empirically, this view predicts that narcissists will reveal negative self-views when these are measured with unobtrusive instruments—such as the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998)—that record people’s automatic, uncontrolled responses. Using the IAT, however, researchers found no simple relation between narcissism and implicit selfesteem (rs 5 .13 and .03; Jordan, Spencer, Zanna, HoshinoBrowne, & Correll, 2003; Zeigler-Hill, 2006). According to another line of thought, narcissists’ explicit selfviews are not uniformly positive; rather, narcissism is associated with positive self-views in agentic domains (e.g., status, intelligence), but not in communal domains (e.g., kindness, morality). Evidence for this idea comes from both explicit trait ratings, which show an association between narcissism and positive selfviews only on agentic traits (Campbell, Rudich, & Sedikides, 2002), and from analyses showing that narcissism is particularly strongly associated with self-esteem measures that capture dominance (Brown & Zeigler-Hill, 2004). Bradlee and Emmons (1992) and Paulhus and Williams (2002) have also reported personality data supporting this distinction. As narcissists do not evaluate themselves uniformly positively across all dimensions—and the self-esteem IAT measures the strength of cognitive associations between the self and an evaluative dimension—the lack of correlation between narcissism and implicit self-esteem might reflect the words used in the IAT. Specifically, IATs employing more agentic words may correlate positively with narcissism, whereas those using more communal words may correlate negatively or not at all with narcissism. Indeed, this pattern is seen in narcissists’ implicit responses on the Thematic Apperception Test. On this test, narcissism correlates positively with nPower and negatively with nIntimacy and nAffiliation (Carroll, 1987). Researchers often use IAT words that activate respondents’ communal self-views more than their agentic ones. For example, the IAT words used by Jordan et al. (2003; friend, gift, happy, holiday, joy, love, party, smile, sunshine, warmth, agony, cockroach, death, disaster, disease, evil, garbage, pain, stink, and vomit) and by Zeigler-Hill (2006; happy, joy, paradise, pleasure, smile, sunshine, agony, death, grief, pain, sickness, and tragedy) include several communal terms and few agentic terms. Twentyfour pilot respondents rated both word lists along agentic and communal dimensions. Both lists conveyed significantly more communion than agency, t(23)s > 4.50, preps > .997, which might explain the weak associations these researchers observed between narcissism and IAT scores. We tested our logic that narcissism correlates positively with implicit agency and negatively or not at all with implicit communion in two studies. Using the IAT words from Jordan et al. (2003), we tested the link between narcissism and implicit selfesteem in a sample of undergraduates. Next, we created separate IATs to measure agentic and communal implicit self-views and tested their associations with narcissism.