Our research explores predictions that people make in a simple environment consisting of sequences of a binary signal followed by two possible outcomes. In order to optimize their prediction success, respondents should use a very simple decision rule, called maximization, whereby they consistently predict according to the signal. In line with past research, our findings show that even respondents who realized after the experiment that maximization is optimal failed to use it during the experiment itself. We discuss conditions that weaken or reinforce behaving according to the optimal rule in a repeated choice situation. Experiment 1 shows that individuals who are forced to plan their strategy and justify their actions are more likely to discover and use the optimal rule than those not forced to do so. Thinking about the appropriateness of one's performance can be done in two different orientations: focusing on the past (justifying past actions) or on the future (planning future action). Experiment 2 shows that planning induces rule-base thinking, while justifying fails to do so. These findings are discussed within a theoretical framework which suggest an interplay between the experiential and the rational modes of processing. Copyright © 2003 John Wiley & Sons, Ltd.