Abstract: For collaborative environments to be successful, it is critical that participants have the ability to generate effective references. Given the heterogeneity of the objects and the myriad of possible scenarios for collaborative augmented reality environments, generating meaningful references within them can be difficult. Participants in co-located physical spaces benefit from non-verbal communication, such as eye gaze, pointing and body movement; however, when geographically separated, this form of communication must be synthesized using computer-mediated techniques. We have conducted an exploratory study using a collaborative building task of constructing both physical and virtual models to better understand inter-referential awareness -- or the ability for one participant to refer to a set of objects, and for that reference to be understood. Our contributions are not necessarily in presenting novel techniques, but in narrowing the design space for referencing in collaborative augmented reality. This study suggests collaborative reference preferences are heavily dependent on the context of the workspace.