A comparative investigation of task-switching performance in category learning paradigms

Recent research has demonstrated that task switching-which has traditionally been attributed to declarative rule use and executive control-can also be supported by associative learning mechanisms. However, whereas declarative task switching is characterized by large switch costs and small congruency effects, associative mechanisms may produce small (or no) switch costs and large congruency effects. Here, we asked whether humans (who possess both declarative and associative learning mechanisms) and pigeons (which have thus far shown no evidence of possessing declarative learning mechanisms) would display different patterns of performance when switching either between two rule-based (RB) subtasks, which should encourage declarative rule use, or between two information integration (II) subtasks, which should encourage associative learning. The pigeons showed no switch costs in either task condition, consistent with the view that they depend entirely on associative mechanisms to solve both RB and II tasks. Conversely, the humans showed strong switch costs in both task conditions. These data raise two possibilities: (a) that human learners may have used declarative mechanisms to solve both RB and II tasks and (b) that among humans, associative learning mechanisms might not reliably preclude switch costs. The theoretical implications of each possibility are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).