Grant peer review is a foundational component of scientific research. In the context of grant review meetings, the review process is a collaborative, socially mediated, locally constructed decision-making task. The current study examines how collaborative discussion affects reviewers' scores of grant proposals, how different review panels score the same proposals, and how the discourse practices of video conference panels differ from in-person panels. Methodologically, we created and videotaped four "constructed study sections," recruiting biomedical scientists with U.S. National Institutes of Health (NIH) review experience and an NIH scientific review officer. These meetings provide a rich medium for investigating the process and outcomes of such authentic collaborative tasks. We discuss implications for research into the peer review process as well as for the broad enterprise of federally funded scientific research.