Humans often process several tasks concurrently. Such multitasking can differ in whether one switches between different task sets (task switching, TS) or processes them simultaneously (dual tasking, DT).
The present study investigated the context-specificity vs. -generality of control states in multitasking, by testing whether control states implemented in one type of multitasking (i.e., TS) transfer to another (i.e., DT).
For this, we randomly interspersed single DT trials into TS blocks. Control states were established by manipulating the frequency of task switches between participants (Experiments 1 to 4). Frequent switches should increase the concurrent activation of both task sets in working memory, which should result in increased crosstalk in DT and easier shifts from Task 1 to Task 2 component processing. Frequent switches reduced switch costs (performance difference between switch and repetition trials) in TS, but did not affect DT performance.
In Experiment 5, control states were established by manipulating the proportion of response conflict in TS. Low conflict frequency should decrease task shielding (increased response congruency effect). These control states were successfully implemented in TS. Importantly, they transferred to DT: Response congruency effects in Task 1 as a marker of crosstalk were larger if TS featured many congruent than many incongruent stimuli.
To our knowledge, this is the first study to demonstrate that control adjustments can generalize across multitasking paradigms. That adjustments only transferred if induced by varied task conflict indicates that the level of processing at which control states are implemented determines whether control states are context-specific or -generic.