Abstract
Performance is generally worse when performing multiple tasks than when performing a single task, but there is debate about whether this multitasking interference arises due to a structural bottleneck that requires serial central processing or due to resource limitations that slow processing of 2 tasks when they are carried out in parallel. The present study used a novel approach of comparing first- and second-task reaction times (RTs) within the psychological refractory period (PRP) paradigm to contrast these 2 possibilities. Counterbalancing task order across participants to control for differences in task difficulty, we found that second-task responses were faster than first-task responses at long stimulus onset asynchronies (SOAs). This second-task advantage is difficult to explain within bottleneck models, which allow the first task to be processed at full speed while the second task waits for access to the bottleneck process. Instead, the effect suggests that processing of the first task is slowed because some cognitive resources are held back in case they are needed for second-task processing. At long SOAs, all resources can be allocated to second-task processing because the first task is already completed. Thus, we propose that cognitive control processes flexibly coordinating the sharing of limited central resources may better explain dual-task performance in the PRP paradigm than bottleneck-based waiting due to structural limitations.
Public Significance Statement It is well known that the concurrent performance of 2 or more tasks is impaired when compared to performing the tasks separately, but there remains debate about the causes of these multitasking costs. The present findings demonstrate that these costs probably do not emerge from structural bottleneck-type limitations in the human cognitive system that require serial processing of 2 tasks. Instead, the results suggest that these costs arise because the system shares limited cognitive processing resources to process 2 tasks in parallel.