Proceedings of the Annual Meeting of the Cognitive Science Society

Coherence-based theories propose that decision-makers reinterpret evidence to align with emerging conclusions. Prior research on legal decision-making demonstrated coherence shifts in ratings of agreement, but it remains unclear whether similar effects can be found using normative measures. Across three studies, participants evaluated prosecution and defense arguments in a mock burglary case before and after rendering verdicts. In Study 1, participants exhibited clear coherence shifts: guilty verdicts associated with increased agreement with prosecution arguments and decreased agreement with defense arguments, whereas not-guilty verdicts produced the opposite pattern. In Studies 2a–b, the shift of likelihood-based measures of arguments strength across testing periods did not align with verdicts. A constraint satisfaction model predicted mean shifts from pre- to post-verdict evaluations, and a Bayesian network using participants' post-verdict probability estimates closely approximated empirical guilt judgments. It appears that verdicts and post-verdict evaluations of evidence are consistent with Bayesian inference while driven by coherence.