Proceedings of the Annual Meeting of the Cognitive Science Society

Traditional models of cognitive control account for a host ofclassic findings, but these classic tasks have limited our abil-ity to test a broader range of model predictions. In particu-lar, such models predict that control should vary parametricallyin response to cognitive demands and that control adjustmentsshould be targeted towards task-relevant stimulus features. Wedeveloped a task to probe these predictions across two exper-iments. Participants responded to one dimension of a stim-ulus while ignoring the other, and we parametrically variedthe conflict between those dimensions and the predictability ofthis conflict across trials. We found that control adjustments(1) varied parametrically in response to cognitive demands,(2) were sensitive to the predictability of those demands, and(3) were primarily targeted towards task-irrelevant dimensions.These results raise interesting questions about the structure ofcognitive control and demonstrate the utility of rich tasks forconstraining model predictions.