Abstract
Our cognitive system must manage two opposing demands: staying focused on one task (stability) and switching between tasks when needed (flexibility). Traditionally, a trade-off has been assumed between these two control modes. Recent studies questioned this idea. However, they do not consider that experimental manipulations can influence stability and flexibility at different levels of information processing and that a trade-off may only exist within the same level. We introduce a novel paradigm designed to investigate flexibility and stability at different levels of information processing (task vs. stimulus level). Participants switched between two categorization tasks while ignoring distractors. Distractors either came from the same (within-task) or the alternative task (between-task) and were either response-congruent or incongruent. In Experiments 1 and 2, participants categorized words superimposed on distractor images representing the generic response categories (task level interference). In Experiment 3, they categorized letters or digits flanked by distractor letters or digits (stimulus level interference). Reduced interference from between-task distractors compared to within-task distractors in Experiments 1 and 2 show shielding at the task level. At the stimulus level (Experiment 3) both distractor-types interfered to the same extent, indicating that the shielding against between-task distractors got bypassed by the automatic retrieval of learnt stimulus-response associations. Across all three experiments, shielding against between-task distractors in the previous trial improved subsequent performance, when the task repeated (i.e. it improved stability), but impaired performance, when the task switched (i.e. it impaired flexibility). This finding supports a trade-off between flexibility and stability at the task level.