Patients' understanding of the origins and consequences of their medical problems (self-diagnosis) has historically been viewed by social scientists as either an asymmetry in role relations (physicians are high status and have access to technical understanding; patients are of variable status and do not really understand what is wrong with them), or as an effect of cognition (health beliefs) on health care processes and outcomes. Recent efforts to understand the medical encounter as a speech event have yielded important insights about how physicians and patients communicate with one another. The role of patient self-diagnosis in the encounter remains under-researched, however. Using a case example, in which an unstated difference in perspectives between a patient and provider regarding her diagnosis was followed by the patient's suicide, three social psychological theories of physician–patient communication are reviewed to see how they deepen understanding of the case. Based on interactional evidence from the third approach, micro-interactional analysis, two key observations are offered. The first comes from evidence based on the initial data gathering segment of the encounter. Here, in an experimental manipulation involving standardized patients being interviewed by second-year medical students, it is shown that eliciting patients' self-diagnosis (attribution) systematically leads to more complete and accurate diagnoses. The second observation is that physicians' delivery of diagnostic ‘information’ at the conclusion of the visit is contingent upon the patient's initial statement of concerns, including attribution whether stated or unstated, and the range of questions and topics pursued by the clinician between the statement of the problem and the delivery of diagnosis. A lack of agreement or alignment between the problem statement and the proposed solution can result in outright or unstated rejection of the diagnostic news, as detailed analysis of two cases reveals. From the evidence provided, the article concludes that interaction analysis is a useful tool for understanding the importance of patient self-diagnosis in the medical encounter. It also provides the best insight to date on the costs and consequences of not addressing self-diagnosis issues at the beginning and the end of the encounter.