Meeting minutes (and similar records) provide a cherished window into the internal workings of important bodies, but scholars usually have little option but to trust their veridicality. However, the production of a record of talk as it happens is a difficult task, especially when talk is animated and turn-taking unregimented. I compare recordings of four National Security Council meetings secretly made by Presidents Kennedy and Nixon with minutes and notes taken by NSC principals and staff members. While minute-taking practices differed in level of detail, all minute-takers engaged in processes of preservation, deletion, and transformation as they sought to distill and disambiguate. Moreover, the need to omit some talk made it possible to suppress certain kinds of content, such as evidence of internal disagreement. The loose relationship between talk and its written incarnation is consequential for lay actors, such as subordinates who rely on minutes for insight into their superiors’ wishes and mindsets; for scholars tempted to read minutes as an accurate account of what transpired; and, potentially, for other sorts of investigators looking to apportion responsibility for misbehavior and bad outcomes.