Is there a justified presumption that a speaker is testifying sincerely? Anti-reductionism about testimony claims that there is, absent reasons to the contrary. Yet why believe this, given the actuality and prevalence of lies and deception? I examine one argument that may be appropriated to meet this challenge, David Lewis's claim that truthfulness is a convention. I argue that it fails, and that the supposition that there is a presumption of sincerity remains unsupported. The failure of Lewis's argument is instructive, however, for it shows us a better way of approaching language use than the standard anti-reductionist treatment. As speech is an intentional action, so a presumption of the sincerity or otherwise of others' testimony must be explicable in the terms we normally use to explain action.