Multitasking · threaded cognition
When does doing two things at once work?
People can steer a car while listening to the radio, yet struggle to dial a number at the same time. Others can talk to a passenger and still drive, until the conversation needs real thought. Threaded cognition explains the pattern: tasks run in parallel as threads, but they share a small set of mental resources that can serve only one thread at a time.
The idea
Threads run in parallel but share the mind's resources
Every task is a thread of steps. A step uses a resource: vision to look, hearing to listen, the hands to act, memory retrieval to recall a fact, and the problem state, a single working-memory slot for the intermediate result you are holding on to. Different resources work in parallel, so a thread that needs only hearing does not bother a thread that needs only the hands. Interference appears when two threads need the same resource at the same time: the one that asked first goes first, and the other waits.
My inaugural lecture (in Dutch) develops this idea under the image of a "brain kitchen" (hersenkeuken).
Try it
Pick two tasks and watch them compete
Each row below is a mental resource; each coloured block is a thread using it. Dashed outlines show a thread waiting for a resource that is busy. The figures compare how often each task gets a step done alone and together.
Durations are round, illustrative numbers chosen to show the principle, not fitted data. Real threaded-cognition models, built in the ACT-R architecture, predict timing in milliseconds from experiment-based parameters.
The surprising bottleneck
The problem state: a bottleneck you cannot see
Try Mental arithmetic together with Remembering a route. Neither task looks at anything or uses the hands, yet together they slow down far more than steering with dialling. Both need the problem state, and each time the other thread has used it, the thread must first restore its own contents, which costs a few hundred milliseconds (the hatched blocks). Together with Jelmer Borst and Hedderik van Rijn we showed this bottleneck experimentally, and in fMRI studies it is associated with activity in the parietal cortex.
Whether multitasking works depends less on how difficult the tasks are than on which resources they claim at the same moment. That is why a conversation can interfere with driving at the moment the road suddenly demands thought: both then need the same working memory.
The other question
Why do people multitask at all?
To support people in their multitasking it helps to know why they do it. Threaded cognition predicts that a second task is productive when it uses different resources from the first, but do people actually use that as a criterion? In experiments where participants choose when to switch, we ask what prompts those choices and whether they are rational, both locally and globally. This work was supported by an ERC Starting Grant.
Funded by an ERC Starting Grant, Towards safe and productive human multitasking.
Further reading
References
Salvucci, D. D., & Taatgen, N. A. (2008). Threaded cognition: An integrated theory of concurrent multitasking. Psychological Review, 115(1), 101–130.
Borst, J. P., Taatgen, N. A., & van Rijn, H. (2010). The problem state: A cognitive bottleneck in multitasking. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(2), 363–382.
Salvucci, D. D., & Taatgen, N. A. (2011). The Multitasking Mind. Oxford University Press.
Implementation and example models are available from Dario Salvucci's threaded cognition page. See also the research overview and the time perception page: time itself is one more resource in this theory.