In this paper we discuss the use of models of space in the building of mixed-reality systems. By model of space we mean a geometric or symbolic description associated with a physical space. We outline several types of model that exist, how they are surveyed and authored, how they are represented to the users and how they are supported by middleware and sensors. We show that systems often contain numerous models of space and we discuss the issues in maintaining or reifying assumptions about transformations between models. We illustrate these ideas by describing the implementation of a collaborative mixed-reality system that allows users to experience a museum in three modalities: physically co-located visitor with personal digital assistant guide, virtual reality visitor and web visitor.