Where do we store tasks while switching tasks? Is this Working Memory?

Flexible, goal-directed behavior requires maintaining and switching between task sets in response to changing demands. A central debate concerns how multiple task sets remain available for action. One possibility is that only the currently relevant task is held in the capacity-limited procedural working memory (WM), with other tasks stored in the activated part of long-term memory and retrieved as needed. Alternatively, several task sets can be concurrently active in WM. Evidence for the latter comes from mixing costs and congruency effects in task-switching paradigms, but these findings admit alternative explanations based on response-level conflict or stimulus-driven activation, rather than genuine concurrent maintenance of task sets in a capacity-limited store. A cleaner test of the multiple maintenance account requires showing that switching among a greater number of tasks imposes a higher WM load. The present experiment provides such a test by comparing performance in blocks involving three tasks (3-mixed blocks) versus two tasks (2-mixed blocks). The paradigm is designed so that non-relevant tasks cannot be offloaded to long-term memory, ruling out dynamic retrieval as an explanation, and uses univalent, trial-unique stimuli to prevent stimulus-based task activation or the formation of stimulus–response associations that could bypass task selection. Tasks change frequently within blocks, and one "common" task is performed far more often than the remaining "rare" task(s), minimizing practice and learning effects. If performance declines as the number of tasks to switch between increases, this would indicate a higher WM load hence that multiple tasks are maintained in WM. Conversely, no such decline would indicate that procedural WM maintains only a single task set at a time.