How participants to a joint activity come to develop a shared or mutual understanding of what they are perceiving has long been a problematic issue for philosophers, sociologists, and linguists. We examine the abstract model proposed by Clark and Marshall (1981) whereby speakers and hearers construct mutual knowledge and by which discrepancies in definite reference are repaired. We focus in particular on forms of demonstrative reference that depend upon physical co-presence. We examine an attested example of reference repair in the operating room of a teaching hospital. It involves learning to recognize pertinent structures within endoscopic surgeries, that is surgeries in which internal spaces are rendered visible by inserting a fiber-optic lens into the body of the patient. Clark and Marshall provide a useful vocabulary for discussing referential practices in this applied setting. We are left with some questions about how to interpret certain features of their model, however. We conclude that further theoretical framing is required before we develop a full appreciation of how reference and reference repair is accomplished in day-to-day interaction.