What does and does not alleviate base-rate neglect under direct experience

A descriptive account was sought of when base-rate neglect does and does not occur under direct experience, and a theoretical model proposed of why it occurs, when it occurs. In two experiments, subjects experienced hundreds of trials in which they predicted which of two events would occur. One event occurred more often than the other, and subjects were aided by an imperfect cue. In Experiment 1, base-rate neglect occurred when cues were identical to outcomes but not whey they were arbitrarily related. Additionally, over 1600 trials, choice did not become stable at probability-matching, but tended toward optimality. In Experiment 2, the salience of irrelevant cues was found to have an effect, but to be incidental to base-rate neglect, and response bias effects were minimal. The results are consistent with the hypothesis that subjects have learned that base rates change more frequently than cue accuracies, and are therefore rationally underweighted.