Recent studies of the pragmatics of politeness have drawn on conversational data. I argue that their model can be extended to some genres of written texts. There have been two obstacles to such an extension: the lack of a definite addressee for published texts, and the dificulty of defining relevant cultural variables. Taking a corpus of articles by molecular geneticists, I assume a simple model of a two-part audience, and focus on two kinds of impositions: claims and denials of claims. With this framework, one can see politeness claims and denials of claims. With this framework, one can see politeness strategies in regularities of scientific style—such as the use of pronouns and of passives—that are usually explained in terms of conventions. The analysis also accounts for some otherwise unexplained stylistic features, such as the use of adverbs in establishing solidarity, and the use of personal attribution in hedging. With these positive and negative politeness strategies in mind, we can understand better the social significance of the occasional instances in which the writer makes an imposition without redress, or makes the imposition indirectly or chooses not to make it at all. Comparisons with popularizations, a genre in which the writer has a different kind of relation to the reader, and thus uses different kinds of politeness devices, show that these devices arise in response to the interaction embodied in the text.