The Cost of Switching. Part Three: Can You Practise It Away?

Abstract

The first two parts established that switching costs are real, separable, and produced by extra effortful processing rather than a passive queue. This part asks the question that follows: can enough practice make the cost disappear?

It has been tested at a scale unusual for cognitive psychology. Steyvers, Hawkins, Karayanidis and Brown, publishing in PNAS in 2019, analysed task-switching performance from users of a commercial brain-training platform — including people who had completed thousands of sessions.

The answer has two halves that point in opposite directions, and reporting only one of them is how this literature usually gets misrepresented. Practice helps considerably. It does not appear to abolish the cost, and whether any of it carries beyond the trained task is a separate question with a discouraging answer.

1.Why Commercial Data Was Worth Using

Laboratory training studies face a structural limit: nobody can afford to have participants practise for thousands of sessions. The longest are measured in weeks, which leaves the interesting question — what happens with really extensive practice — permanently out of reach.

The authors worked around this by analysing data from a large commercial platform. They describe the advantage as being able to select samples by demographic characteristics, engagement levels and training histories, which conventional recruitment cannot match.

Three samples were examined: a large adult-lifespan group who trained for up to sixty sessions each; a group of the most active users who had trained for thousands of sessions; and a group of older adults with at least a thousand sessions. That third sample is the one that makes the study distinctive, because it allows a question nobody had been able to ask properly.

The authors state it directly: can extensive practice make an older person functionally similar to a younger person? Whatever one thinks of brain-training products, that is a serious question and the data to address it did not previously exist.

2.Decomposing the Behaviour

Rather than tracking reaction times alone, the authors fitted a computational model that decomposes task switching into underlying cognitive processes and describes how each changes both within a session and across sessions.

This matters more than it might sound. Reaction time is a sum: how fast you can execute the basic task, how quickly you can activate the relevant task set, how well you suppress the irrelevant one, plus the mechanics of responding. A person can get faster because any of those improved, and the everyday interpretation differs enormously depending on which.

Separating the switching dynamics from the response mechanism, as the authors put it, makes it possible to compare performance across age groups on latent characteristics — the ability to perform the basic tasks, the rate of switching between them, and the degree to which the relevant task can be activated while the irrelevant one is suppressed.

3.What Practice Did

The headline result is positive and, unusually for this field, unambiguous. Task switching in this paradigm can be substantially improved by practice across all age ranges.

The pattern by age is more interesting than the headline. Older age groups had generally slower response times and lower accuracy — the expected finding — but also showed more pronounced improvements with practice. The people who started furthest behind gained the most.

That is worth stating clearly because the opposite is widely assumed. Plasticity in this measure did not vanish with age in this dataset; if anything the room for improvement was larger.

4.What Practice Did Not Do

Now the other half. The literature this paper sits within includes work whose title makes the point without needing a summary: extensive practice does not eliminate human switch costs. Performance improves; the cost of switching relative to repeating does not disappear.

This is the distinction that popular coverage of brain training almost always loses. Getting faster at a task is not the same as removing the underlying cost, and the second is what anyone actually cares about.

There is also a sampling problem the design cannot escape. These are paying users of a commercial platform who chose to train, and the most extreme sample consists of people who chose to train thousands of times. They are not a random slice of the population, and whatever makes someone complete a thousand sessions may also be associated with how they improve.

5.The Transfer Question, Which Is the Real One

Improving at a trained task is the least interesting form of improvement. The question that matters is whether it carries over to anything else — what the field calls transfer.

Baniqued and colleagues addressed this in a 2015 trial published in PLOS ONE, with a title that telegraphs the conclusion: transfer effects, limitations, and great expectations. Their assessment of the wider literature is that computer-based training paradigms, from video games to laboratory regimens, yield improvement in the trained tasks but limited transfer to other related abilities — including abilities similar to the ones trained.

They are equally direct about why the field's optimistic findings should be discounted. Many promising experiments face methodological shortcomings involving small sample sizes, single tests of cognitive transfer, and the absence of a comparable active control group.

That last one deserves emphasis. If a training group is compared against people who did nothing, any difference may reflect having been given something to do. The authors also raise the expectancy problem themselves, noting that awareness of being trained has been argued to produce subconscious expectations and therefore placebo effects — the same mechanism that dominated the microdosing literature discussed elsewhere in this library.

6.What We Take From the Three Parts Together

The defensible summary across this series is narrower than the popular version and more useful.

Switching has two separable costs, and the larger one is often paid for merely holding a second task available rather than for switching to it. The cost arises from additional effortful processing — administration — not from a passive bottleneck. Dual-tasking and task-switching are substantially different operations in the brain, sharing only a small parietal, premotor and insular core. Preparation time reduces the switch cost and appears to shift the brain from reactive to prepared control. Practice improves performance at all ages, with the largest gains in those who start slowest, but does not eliminate the cost and shows limited evidence of transferring anywhere else.

The practical implication follows from the shape of that list rather than from any single finding. The reliable levers are structural — fewer live task sets, longer uninterrupted blocks, a pause before transitions. The unreliable lever is training yourself to be someone who handles interruption well, which is the one the productivity market mostly sells.

Editorial Comment

MindHeaven® makes no claim that any product reduces switch costs or improves task-switching performance. None of the studies in this series tested a supplement, and we are not aware of evidence that any compound in our formulations affects these measures.

We publish a series ending on a limitation because that is the honest shape of the evidence. The finding that practice improves the trained task without transferring elsewhere is inconvenient for a large industry, and it is the same standard we apply on the pages describing our own compositions.