Supervision, category structure, and selective attention in category learning: A comparative approach.

Abstract

We investigated the interaction between supervision and category structure in visual categorization across two species that we assumed to have different abilities to selectively attend. Specifically, we compared humans and pigeons in a category learning task, in which the category structure (dense vs. sparse) and the level of supervision by corrective feedback (low vs. high) were manipulated. Results showed a benefit of supervision across species, which was particularly strong in learning the sparse categories. Moreover, both species benefited from category structures that had multiple category-relevant dimensions (i.e., dense categories). In addition, humans, who have a more advanced ability to attend selectively, showed faster learning and better generalization overall. Finally, attention was optimized to the category-relevant dimension in the sparse category only in humans. Subsequent computational simulation of the data indicated that these patterns were well explained by a parameter that controls the organism's ability to flexibly switch attention to category-relevant dimensions. (PsycInfo Database Record (c) 2026 APA, all rights reserved).