Although talking is one of the most common activities in social VR, there is little empirical work identifying what people say and how
they communicate in immersive, virtual settings. The current paper addresses this opportunity by performing automated text analysis on over 4,800 minutes of in-VR, small group conversations. These conversations took place over the span of two months during a university course where 171 students attended discussion sections via head-mounted displays. We provide a methodology for analyzing verbal communication along two dimensions: content and structure. We implement methods to describe linguistic patterns from the class and introduce a preliminary VR Dictionary