One of the most persistent questions in criminal evidence relates to the use of (unchallenged) expert evidence. What does it mean to accept or reject (unchallenged) expert evidence? To what extent can, and should, an expert enter jurisprudential territory? Is the traditional model of trial by jury viable in our complex world? In order to clarify these pressing questions, we will examine the evidential structure underpinning expert witness testimony. We will show that what we usually and, at the cost of oversimplification, call ‘evidence’, comprises three distinct questions: (i) What does the data show? (ii) What should we believe? (iii) What should we do? From this insight, a number of corollaries fall into place. First, although decisions have to be informed through reasoned inferential procedures, they cannot be reduced to scientific propositions. As a result, fact-finders do not need to cede their decision-making prerogative as some proponents of expert-driven decision-making suggest. Secondly, criminal liability is not a scientific conclusion. Rather, so our argument, it is an individualistic normative construction that involves an inferential leap which is not warranted by any scientific (i.e. general) proposition. For the rectitude of the criminal verdict (or indeed any legal decision) does not map logically onto the possible treatment of scientific findings, that is, acceptance/rejection. Thirdly, our clarification of this evidential structure, which we call coherent decisionalism, provides a conceptual framework to understand and stabilise case law on expert witness testimony.