Categorization is a decision making process where we selectively attend to the most distinguishing features of the categories. During learning we transfer the category-specific response to all members of the category, by eliminating individual differences and focusing on similarities.

We investigated the nature of the abstraction process during which participants learned the categorization rule in a supervised category-learning paradigm. An information integration task was used with naturalistic Gestalt-like stimuli, where all the exemplars (72 different items during learning) also possessed additional idiosyncratic features.

The learning strategies, the exemplar effect, and the retention of the categorization rule were tested in behavioral experiments. Our developmental study compared 7-8 year-olds and adults, and we conducted an electrophysiological (ERP) experiment to understand the learning mechanism better.

Hit rates and reaction time results show that participants were able to learn the complex categorization rule without realizing that they have learnt it. Moreover, this general knowledge was stable a week later. Their memory for individual exemplars in the immediate test was as weak as a week later. Behavioral data was inaccurate in case of learning strategies, but the ERP components were sensitive to the changes in them. Later components didn’t, but the response and feed-back related components (ERN and FRN) indicated the changes during the three learning blocks. The differences between school children and adults showed different learning strategies as less of children than adults were able to learn the categorization rule. This implies that children are not able to inhibit the explicit strategies as successfully as adults do.

Generally, our results show that the standard category learning paradigm is extendable to more complex and naturalistic stimuli. With our method we could test the long-term retention of the learned information, and the memory for individual exemplars both in children and adults.