Jury members are confronted with highly complex, ill-defined problems. Coherence-based reasoning (Pennington & Hastie 1992; Simon 2004), which partially relies on intuitive-automatic processing, empowers them to nonetheless make meaningful decisions. These processes, however, have a downside. We tested possible negative effects in a set of studies. Specifically, we investigated whether stricter standards of proof are suppressed by stronger coherence shifts and whether the probative value of the evidence is properly taken into account. We found that U.S. model jury instructions for preponderance of the evidence and beyond a reasonable doubt influence conviction rates in the intended direction and are not undermined by coherence shifts, although probabilistic estimations of these standards are inappropriate. However, even massive changes in explicitly stated probabilities, while holding the overall constellation of facts constant, did not influence conviction rates or the estimated probability of conviction. We argue that reforms in legal procedure should focus on measures to reduce the negative side effects of coherence-based reasoning in general but, more specifically, to make probabilistic information better evaluable for decisionmakers in law.