This paper argues for the importance of the relevance of testimony for its evidential value. When testimony is used as a source of knowledge, the primary expectation is that it is credible. Indeed, what matters first and foremost is whether the content of a testimony is true. However, this is not sufficient for testimony to serve as reliable evidence - that is, evidence on which one can build safely and efficiently in order to make progress in one's epistemic inquiry. By studying testimony within the framework of evidential reasoning under uncertainty - with a focus on the figure of the fact-finder in criminal investigations and trials - this paper explores the complexity and subtlety of the inferential processes that agents draw from information conveyed by others. Through an exploration of the notion of epistemic relevance, as inspired by common law scholarship on criminal evidence, it shows that the complexity inherent in uncertain evidential reasoning is multiplied by the pragmatic aspects of human communication. Detailed analysis of examples drawn from a fictional case highlights the many possible routes to genuine error that can arise from misunderstandings about the relevance of true and honest testimony - and thus opens up as many routes to intentional deception.