This dissertation builds upon previous studies showing how a person's behavior in virtual environments is altered by the characteristics of the digital avatar they control, an effect coined as the Proteus Effect by Yee and Bailenson (2007). The goal of the dissertation is to extend this research by evaluating the Proteus Effect in a naturally occurring context, rather than a controlled laboratory setting. The community chosen for this study is one which has formed around the game Dota 2, a popular game in the multiplayer online battle arena genre. Whereas the majority of Proteus Effect research to date has focused on the effects of an avatar's aesthetic appearance on player behavior, the retention of local context in the present study allows for a wider range of avatar characteristics to be studied. This study also uses two distinct corpora of linguistic behavior, one composed of the language players use in the game itself, the other taken from a popular online forum dedicated to the Dota 2 community. The differences between these two interactional sites allows for a meaningful comparison of the conditions necessary to trigger the Proteus Effect. Findings show that the Proteus Effect does take place in naturally occurring environments, with linguistic differences found between players controlling avatars of different genders, and avatars fulfilling different functional roles within the game environment. Comparisons between the two corpora show that conditions with increased levels of embodiment lead to increased Proteus Effect differences, confirming the hypothesis proposed by Fox et al. (2013).