Avatars are extensions of users' identities in virtual environments (VEs), yielding substantial influence over self-perception, affect, and behavior. However, extant research has largely focused on the implications associated with creating, and interacting with, avatars representing desirable aspects of the self (ideal self), overlooking other aspects of the self. Across several studies I examine the psychological and emotional effects of (a) creating avatars representing anxiety within the user (anxious self), and (b) interacting with such “anxiety avatars”. Results from three experiments consistently show that creating, and subsequently destroying, an anxiety avatar in various contexts reduces anxiety, effects explained by self-discrepancy theory's self-regulatory mechanisms. Follow-up studies examine interactions with anxiety avatars in custom-created PC and VR games to identify other potential mechanisms, unique to immersive VEs, contributing to these outcomes. Preliminary results support conclusions from the initial studies. Herein, I review pertinent literature, discuss findings, and outline ongoing experiments.