An original experiment explored the differing predictions of the Proteus effect (Yee & Bailenson) and behavioral compensation processes (Bond) in dyadic computer-mediated interaction. The experiment randomly assigned male dyad members to see an attractive, unattractive, or no avatar representation of his female partner, while female dyad members were assigned to see either attractive, unattractive, or no avatar representation of themselves. Results supported the hypothesized behavioral compensation effect such that both partner and naïve observer reports of relational communication suggested that females who saw unattractive avatars of themselves behaved more positively toward their partners than those who saw no avatar or saw an attractive avatar. These results, their theoretical implications, and future directions are discussed.