This article investigates the academic speech of humanities and natural science instructors and students in 32 lectures and interactional classes at a U.S. university. It examines how structural markers, questions, question tags, and turn-initial response tokens contribute to variations of style in response to academic division, context, gender, and communicative role in academic discourse. Data analysis couples qualitative discourse analytic methods with a quantitative sociolinguistic analysis. The quantitative analysis shows the factors of communicative role, academic discipline, and speech mode – not gender – to be the most influential in the use of the structures investigated. It is argued that the lack of significant results for gender arise from global discourse restrictions in academic speech. However, despite the global restrictions shown by quantitative analysis, qualitative analysis suggests that such restrictions can be overridden, especially in contexts of structural breaks and disruptions of information flow, and that features that contribute to more interactional and cooperative speech styles, frequently linked to females, can emerge.