Base-rate neglect based on base-rates in experience-based contingency learning

Predicting criterion events based on probabilistic predictor events, humans often lend excessive weight to predictor event information and insufficient weight to criterion event base-rates. Using the matching-to-sample paradigm established in studies on experience-based contingency learning in animals, Goodie and Fantino (1996) showed that human judges exhibit base-rate neglect when sample cues are associated with response options through similarity relations. In conceptual replications of these studies, we demonstrated similar effects when sample cues resemble the response options in terms of base-rates skewed in the same direction rather than physical similarity. In line with the pseudocontingency illusion (Fiedler & Freytag, 2004), predictions were biased toward the more (less) frequently rewarded response option following the more (less) frequently presented sample cue. Thus, what is a demonstration of base-rate neglect from one perspective turns out to reflect the judges' sensitivity to the alignment of skewed base-rate distributions.