The lay perception of those who work with interpreters in their professional conversations (and sometimes of interpreters themselves) is that translation of what they and their patients/clients say should be as close as possible. The very idea of close translation may seem simple, but the practice of rendering spontaneous talk-in-interaction as close as possible is quite complex, and requires much more from the mediator than textually close renditions. In this paper, we discuss what is involved in the accomplishment of close rendering in talk. We focus on question-answer sequences with clinicians and patients, when details about the patients’ history and their symptoms are collected. We show that meaning is achieved and rendered through processes that may go rather beyond the single words and utterances and that involve the contribution of all the interlocutors. We show three types of sequences taken from a collection of authentic data audio-recorded in health care services in Italy and transcribed to allow for analysis. The sequences show three different forms of talk, found in the data, where participants pursue close, precise rendition of medical details. The analysis highlights that the Italian intercultural mediators (just like those trained and assigned as public service interpreters in other countries), even in these cases which are peculiarly focused on precise details, do need to take responsibility not only for translating the textual items but also for coordinating turns at talk. It is responsible turn-coordination which leads to rendering meaning closely.