*Based on work coauthored with Haixin Dang.
Abstract: Assertions are, speaking roughly, descriptive statements which purport to describe some fact about the world. Philosophers have given a lot of attention to the idea that assertions come with special norms governing their behaviour. Frequently, in fact, philosophers claim that for something to count as an assertion it has to be governed by these norms. So what exactly are the norms of assertion? Here there is disagreement. Some philosophers believe assertions are governed by special factive norms, to the effect that an assertion must be true, or known to be true, or known with certainty to be true - or in any case that an assertion is normatively good just in case it meets some condition that entails its truth. Other philosophers place weaker epistemic constraints on good assertion. For instance the claim that an assertion is justified given the assertor's evidence. We use this literature to think through the norms concerning a special class of scientific utterances - namely, the conclusions of scientific papers, or more generally the sort of utterances scientists use to communicate the results of their inquiry. Such utterances might look like paradigm instances of descriptive statements purporting to describe some fact, yet the norms of assertion philosophers have surveyed are systematically inapt for science. Scientific conclusions may justly be put forward even though they are neither known, true, justifiably believed, nor even believed at all. We argue that understanding this has implications for how one understands the significance of the replication crisis in scientific inquiry and how it ought to be responded to. After surveying our argument for this negative claim, we end by suggesting a norm of utterance that would be more appropriate to scientific practice.