In her Exemplarist Moral Theory, Linda Zagzebski argues that we can empirically discover the meaning of moral terms like 'virtue' and 'the good life' by direct reference to moral exemplars -- those people we admire as morally exceptional. Her proposal is promising, because (1) moral exemplars play an important motivating role in moral education, and (2) her use of direct reference means we may be able to avoid the contentious descriptivism that accompanies moral terms like 'good' and 'virtue'. In this article, I argue that Zagzebski's theory fails regarding (2), because her direct reference method must use presupposed descriptions and leads to circular identification of moral exemplars.