Multi-User Virtual Environments (MUVE) are increasingly being used in education and provide environments where users can manipulate minute details of their avatar's appearance including those traditionally associated with gender and race identification. The ability to choose racial and gender characteristics differs from real-world educational environments and raises interesting questions regarding the applicability of previous racial and gender research findings. Specifically, do racial and gender categorizations found in traditional classroom research convey to virtual worlds where gender and race are controllable? To explore this issue research related to racial and gender characteristics in traditional and MUVEs environments is considered. Additionally, the theories of classification and mental categorization, media equation theory, equalization hypothesis, and Social Identification Model of Deindividuation Effects are examined as potential foundations of understanding. Results of two pilot studies conducted to determine associations of avatar appearance with gender and racial classifications are discussed in relation to the development of a theoretical framework. Implications for future investigations are discussed.