Flexible behavior under control? Neural and behavioral evidence in favor of a two-component model of task-switching

Abstract

The ability to rapidly change from one course of action to another, i.e. "flexible behavior", is a hallmark of human cognition. Laboratory observations of switch costs, an increase in reaction time and errors when alternating between tasks compared to repeating a task, have been argued to be a measure of endogenous control during flexible behavior. However, alternative models suggest no such reconfiguration processes are necessary to account for performance in these task-switching situations. The first part of this dissertation uses neuroimaging to address whether reconfiguration processes do in fact occur in the explicit cuing variant of the task-switching paradigm. Using a 4:2 mapping between cues and tasks, we found neuroanatomical evidence for a dissociation between cue-switch (left prefrontal and lateral parietal) and task-switch (medial precuneus and cerebellar) related areas, consistent with the claim of endogenous control during task selection. The second portion explores whether automatic, long-term memory (LTM) processes can explain the "switch cost asymmetry", the fact that switch costs are larger when switching into a dominant task rather than into a competing non-dominant task. We modified an alternating runs task-switching paradigm to include either long or short response-to-stimulus intervals (RSIs) after each pair of trials (i.e., AA-AA-BB-BB), thereby inducing selection costs not only at the point of a task-switch (i.e., AA-BB), but also between same-task pairs (i.e., AA-AA). Using spatially compatible versus incompatible response rules and Stroop word versus color naming, we found asymmetric effects not only at task-change transitions, but also at task-repeat transitions when the RSI was long (presumably inducing frequent losses of task set). In two additional experiments, an asymmetry for long RSIs was obtained even when competing tasks were separated into alternating single task blocks, but not when the tasks were compared in a between-subject design. This pattern supports the idea that the asymmetry arises from interference effects occurring in LTM traces. The combined results of this dissertation characterize task-switching processes not as an "either-or" phenomenon in regards to the question of control, but rather as the interplay between top-down, executive functions and bottom-up, long-term memory priming mechanisms.