Is attention a trainable skill?

Partly, and much less than the advice implies. Training improves the thing you trained. Whether it improves anything else, which is the entire point of calling attention a skill, is close to unsupported. The strongest review of the field examined every study the brain-training industry cites in its own defence and found little evidence that training carries over to distantly related tasks or to everyday performance.

This matters because "train your attention" is becoming standard advice for surviving a world of infinite output, and the evidence underneath it is thinner than the confidence attached to it.

The transfer problem, which is the whole problem

Simons and colleagues did something unusual. Rather than review the literature at large, they took every peer-reviewed study cited by commercial brain-training companies as evidence that their products work, and applied pre-specified best-practice standards to each.

Their conclusion, in their words: extensive evidence that these interventions improve performance on the trained tasks, less evidence for closely related tasks, and little evidence that training enhances performance on distantly related tasks or improves everyday cognitive performance. Not one of the studies met all the standards.

Near transfer holds up. Far transfer is what any claim to have trained your attention actually requires. That is the part that keeps failing. Get better at the exercise and you have got better at the exercise.

Meditation does something, smaller and narrower than advertised

The most serious case for trainable attention runs through mindfulness, and Verhaeghen's three meta-analyses are the place to look: 109 effect sizes from 40 intervention studies, 59 from 18 studies of long-term meditators, 197 from 28 trait studies.

The effects are real and modest, on both counts. Hedges g of 0.29 for interventions and 0.32 for long-term practice, which is small to moderate. They concentrate in inhibition and executive control rather than in sustained attention, which is the thing people mean when they say their attention is going.

One limit worth holding. The published breakdown does not separate active controls from passive ones. Where a study compares meditation against doing nothing, some of the effect is likely to be expectation rather than mechanism, and this analysis cannot rule that out.

The famous finding did not replicate

Much of the public argument rests on a 2009 result suggesting heavy media multitaskers are worse at filtering distraction. It has been tested properly.

Wiradhany and Nieuwenstein ran two direct replications, fourteen tests at reasonable statistical power, plus a meta-analysis of 39 effect sizes. Only five of the fourteen showed the effect, and only two survived a more conservative analysis. The meta-analytic association became non-significant once small-study effects were corrected for. The authors say they question whether the association exists.

That is one specific effect rather than the whole question. It does not show multitasking is costless. It does mean the most-cited empirical anchor for the attention-collapse argument is not load-bearing.

What does hold up: attention residue

Leroy's work is narrower and sturdier. Two laboratory experiments show that people struggle to move attention away from an unfinished task, and that performance on the next task suffers as a result. Time pressure on the first task helps, by forcing closure.

The useful reading is that the lever is completion rather than willpower. Attention is released by finishing something, or by parking it deliberately, not by resolving to concentrate harder. That is a design instruction about how work is sequenced, and it survives scrutiny better than anything in the training literature.

Numbers to stop repeating

Two figures circulate constantly: that the average attention span on a screen has fallen to around 47 seconds, and that it takes 23 minutes to refocus after an interruption.

Neither could be traced, in preparing this page, to a peer-reviewed paper reporting that number with a stated methodology. They come from trade press and book promotion. They may well be approximately right. They are not currently checkable, and this research does not repeat figures it cannot check.

There is also a peer-reviewed interruption study pointing the other way, finding that people compensate for interruption by working faster, with no measured loss of output quality, at the cost of higher stress and frustration. That is a materially different claim from attention collapse, and the better evidenced of the two.

What this means for AI

If attention cannot reliably be trained upward, the useful move is to stop treating it as a personal capacity to be improved and start treating it as a resource to be allocated and protected by design.

Which is the aviation answer. The sterile cockpit rule does not ask anyone to concentrate. It removes what would prevent concentration, during a window defined in advance, and places the obligation on the organisation rather than the individual. See what professions can learn from aviation.

It also reframes what AI does to attention. The interesting question is not whether tools shorten your attention span, which is poorly evidenced, but whether they remove the occasions on which sustained attention was previously required. A capability that is never called upon is not being trained, whatever anyone's span is.

What to do, given all that

• Do not buy attention training. The far-transfer evidence is not there, and the industry's own cited studies are what the strongest review examined.
• Finish things, or park them explicitly. Residue is the best-evidenced effect here, and completion is what releases attention.
• Protect windows by rule rather than by intention. Define the period, remove the interruptions, and put the obligation on the system rather than on willpower.
• Watch the occasions, not the span. If nothing in a week required forty unbroken minutes of thinking, that is the finding, regardless of how focused you felt.
• Meditate if you want to, with accurate expectations. Small effects, concentrated in inhibition, possibly partly expectation. Worth doing, not a solution to this problem.

Related SuperSkills research

On the mechanism of effortful practice, desirable difficulty and cognitive load. On what happens when practice stops, can you regain a skill you have lost and capability debt. On protected attention as a rule, what professions can learn from aviation. On offloading generally, cognitive offloading.

Key research and primary sources

• Simons, D. J. et al. (2016). Do Brain-Training Programs Work? Psychological Science in the Public Interest, 17(3).
• Verhaeghen, P. (2021). Mindfulness as Attention Training. Mindfulness, 12(3).
• Wiradhany, W. and Nieuwenstein, M. R. (2017). Cognitive Control in Media Multitaskers: Two Replication Studies and a Meta-Analysis. Attention, Perception and Psychophysics, 79(8).
• Leroy, S. (2009). Why Is It So Hard to Do My Work? Organizational Behavior and Human Decision Processes, 109(2).

About this research

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026) and founder of The SuperSkills Intelligence Company. Two widely repeated figures about attention span are named here and not used, because their methodology could not be verified. The mindfulness effect is reported at the size the meta-analysis states rather than the size it is usually quoted at. Reviewed quarterly.