In this paper we investigate strategies to support knowledge sharing in distributed, synchronous collaboration. Our goal is to propose, justify, and assess a multiple view approach to support common ground in geo-collaboration within multi-role teams. We argue that a collaborative workspace, which includes multiple role-specific views coordinated with a team view, affords a clear separation between role-specific and shared data, enables the team to filter out role-specific details and share strategic knowledge, and allows serendipitous learning about knowledge and expertise within the team. We discuss some key issues that need to be addressed when designing multiple views as a collaborative visualization. We illustrate the design features of a geo-collaborative prototype that address these issues in the context of two collaborative scenarios. We finally describe a laboratory method for investigating how multi-role teams establish common ground while the amount of prior shared knowledge and the type of visualization are experimentally manipulated.