Discussion
The results show a strong effect of category structure on performance in an indirect category-learning task. First,participants in rule-based conditions were fairly accurate in making same–different judgments, and their responses were mostly consistent with an optimal categorization strategy, suggesting that they learned the correct category representations. In contrast, participants in the informationintegration
condition performed only slightly better than chance throughout the whole experiment, and the modelfitting analyses suggested that they did not base their same–different responses on accurate category representations.Yet, previous research has shown that participants can reliably learn all three category structures in direct classification
(for reviews, see Ashby & Maddox, 2005; Maddox & Ashby, 2004). The failure of the information-integration participants is consistent with at least two different
hypotheses. One possibility is that these participants learned little or nothing about the information-integration categories, but an alternative is that there was some category learning, but the participants were unable to apply that knowledge to the same–different judgment required in the task. Experiment 2 tests between these two possibilities.