Abstract. In the classic preference reversal (PR) phenomenon, individuals often choose the safer of two gambles but assign a higher monetary evaluation to the riskier one. A key distinction between choice and evaluation tasks is that the former typically require comparative processing of the two bets, while the latter can be conducted in isolation for each bet. If this task-mode difference plays a causal role in generating PRs, then the frequency of PRs across individuals should correlate with differences in comparative cognitive processes across tasks. Assuming that sequences of visual fixations reflect cognitive comparisons, we test this hypothesis using eye-tracking during a classic PR experiment. Specifically, we compare the prevalence of across-bet transitions, eye movements between the probabilities or prizes of different bets, with within-bet transitions, movements between attributes of the same bet. As expected, participants exhibit more across-bet transitions in Choice than in Evaluation. However, variation in transition patterns does not predict behavior or PR incidence, challenging context-dependent explanations of classical PRs.