We investigate the age-old questions of what makes us who we are and what features of identity, if changed, would make us a different person. Previous approaches to identity have suggested that there is a type of feature that is most defining of identity (e.g., autobiographical memories or moral qualities). We propose a new approach to identity that suggests that, like concepts in general, more causally central features are perceived as more defining of the self-concept. In three experiments, using both measured and manipulated causal centrality, we find that changes to features of identity that are perceived as more causally central are more disruptive to both the identity of the self and others.