Network Switching

What it is

The Distracted Mind calls "multitasking" a borrowed computer-science word. Plenty of processing in the body is genuinely parallel, the authors note, and automated actions can run alongside a task without cost, as in walking while chewing gum. The claim concerns the case where two goals both need cognitive control. Those goals compete for the same limited prefrontal resource, and in the book's account the brain handles the competition by moving between networks.

The book's evidence for the claim at p. 62 is an imaging experiment from the authors' laboratory in which participants held a nature scene in mind across a delay while a face appeared during it and had to be judged. As our reading of the book records it, the working-memory maintenance network was active into the delay, diminished as the face network engaged, and returned once the face was gone. The authors conclude that the behaviour may be multitasking while the brain is network switching, and they say the result is consistent with many other studies.

The book then makes a concession worth keeping. The prefrontal cortex is not the only structure involved. Premotor, parietal and visual cortex, caudate, thalamus and hippocampus all contribute, and parietal volume has been shown to predict individual variability on a self-report measure of distractibility.

In effect

A neural account sitting beside a behavioural one The behavioural side is the switch cost, measured in reaction-time experiments for nearly a century. The Distracted Mind states at p. 79 that task-switching costs persist even when the switch is predictable and grow with the complexity of the task. Rogers and Monsell (1995) and Rubinstein, Meyer and Evans (2001) are the experiments that match that sentence, and they supply the time costs: from 29 to 975 milliseconds across the experiments of the 2001 paper, and a reaction-time cost that remained with 1.2 seconds of preparation in the 1995 work. Network switching offers a reason such costs exist. It does not add a new figure for them.

A 2011 study from the same laboratory Clapp, Rubens, Sabharwal and Gazzaley published a study in the Proceedings of the National Academy of Sciences in 2011. As the University of California, San Francisco described it, participants viewed a natural scene and held it in mind for 14.4 seconds while a face appeared partway through, and they judged its sex and age before recalling the scene. The groups were healthy younger adults, with a mean age of 24.5, and older adults, with a mean age of 69.1. In the younger group, brain imaging showed disengagement from the memory-maintenance network when the face interrupted, a shift of resources to the face, and reconnection with the memory network afterwards. Older adults failed both to disengage from the interruption and to reestablish the memory network, which the authors linked to poorer recall. We have read this study through that description and not in full. We do not know that it is the experiment the book's p. 62 passage describes, since the delay differs from the one in our reading of the book.

The serial bottleneck beneath it The idea that two demanding tasks cannot both be served at once is older than the imaging. Our notes record Pashler (1994) in Psychological Bulletin 116(2) and Dux, Ivanoff, Asplund and Marois (2006) in Neuron 52(6) as the established work on the central bottleneck for response selection.

What it does not say

It does not say that all dual tasks carry a cost. The book's own carve-out for automated actions and for processes that run in parallel is part of the claim.

It does not rest on independent replication to the extent the behavioural cost does. Our notes grade the network-level evidence as largely from one laboratory, with a small sample in the older-adult study, and describe the general principle as broader than its evidence base.

It does not tell you how many minutes of a working day are lost. The imaging experiments last seconds and test small samples of people in a scanner.

Who developed it

The book does not name an originator for the network-switching framing. Our identification is Wesley C. Clapp, Adam Gazzaley and colleagues for the network-level account, with Pashler (1994) as the older source for the serial bottleneck. We hold this at medium confidence, because the attribution is ours and not the book's.

---