Change one category at a time: Sequence effects beyond interleaving and blocking.

Research on study sequences has not considered the cross-classification of to-be-learned categories. In two experiments, we utilized cross-classified exemplars, which simultaneously belonged to categories of two orthogonal dimensions. Experiment 1 addressed the question of how interleaving one category dimension while simultaneous blocking another category dimension affects the induction of the simultaneously blocked category dimension. Experiment 2 examined our proposed mechanism by manipulating the degree of change (one-category vs. cross-category change) and the frequency of change (high vs. low) in the presentation sequence of exemplars with cross-classified characteristics. In Experiment 1, sequences that combined interleaving one dimension while blocking the other dimension were superior to sequences that provided no comparison opportunities when classifying both interleaved and blocked categories. This revealed a carry-over effect of interleaving: blocked and interleaved categories were equally well classified. Our findings are incompatible with the discriminative contrast hypothesis and the attentional bias framework, where interleaving is not assumed to support within-category comparisons. We explain the results according to the principle of change one category at a time (COCAT). Interleaving exemplars on one category dimension, but blocking them on another category dimension, enables learners to reliably map the distinctive features onto the covarying categories and the shared features onto the constant category. In contrast, there is a risk of confounding common characteristics when no comparison opportunities are given. Likewise, pure interleaving impedes category induction by confounding changing characteristics. Accordingly, Experiment 2 demonstrated that as long as a sequence implemented the COCAT principle, learners accurately identify diagnostic feature sets. (PsycInfo Database Record (c) 2021 APA, all rights reserved).