This paper provides a unique quantitative analysis of the social dynamics of three chat rooms in the Microsoft V-Chat graphical chat system. Survey and behavioral data were used to study user experience and activity. 150 V-Chat participants completed a web-based survey, and data logs were collected from three V-Chat rooms over the course of 119 days. This data illustrates the usage patterns of graphical chat systems, and highlights the ways physical proxemics are translated into social interactions in online environments. V-Chat participants actively used gestures, avatars, and movement as part of their social interactions. Analyses of clustering patterns and movement data show that avatars were used to provide nonverbal cues similar to those found in face-to-face interactions. However, use of some graphical features, in particular gestures, declined as users became more experienced with the system. These findings have implications for the design and study of online interactive environments.