The problem: attention wanders off constantly — roughly a third of your waking life — and occasionally it does so at a moment that costs a life. The instinctive fix is to demand more focus, more discipline, more vigilance.
- That fix is aimed at the wrong thing. Most wandering is harmless. The damage comes from how deep the mind goes and when — not how often it leaves.
- Automation is not the culprit on the flight deck, despite everyone assuming it is. Idle hands, being the monitoring pilot, and things going well are what send attention away.
- Concentrating harder doesn’t work — and there’s a specific reason. Pilots showed no fatigue-style decline in watching over time. The mechanism everyone tries to fight isn’t the one that’s failing.
- Brain-monitoring headbands can’t do what they claim. When properly tested against fake feedback, the real thing performs the same. And the popular claim about tracking your “default mode network” is physically unsupportable at that sensor count.
- What does work is unglamorous: sleep, real exercise, giving hands and voice something to do, designing tasks so the drift has somewhere harmless to land — and, when you catch yourself circling, changing the question instead of fighting the thought.
You are driving a road you know well. You arrive, park, and realise you remember nothing of the last ten minutes. You did not drive badly. You stayed in lane, you stopped at the lights. Something in you was doing the driving. It just wasn’t the part of you that was there.
Or you read a page of a report and absorb none of it. You walk into a room and can’t recall why. You make a cup of tea, get interrupted, and come back an hour later to a dry tea bag in an empty cup.
This is not a defect. It is what a normal mind does. Depending on how you measure it, somewhere between a fifth and a half of your waking life is spent thinking about something other than what you are doing — call it a third as a working figure.
So the obvious conclusion is: wander less, focus more.
The obvious conclusion is wrong, and the evidence for why is unusually clear.
The first surprise: most wandering costs nothing
Researchers once sampled a few thousand people at random moments through the day and asked what they were doing, what they were thinking about, and how happy they felt. The headline that went around the world was that a wandering mind is an unhappy mind.
Look at the breakdown, though, and the headline falls apart.
Roughly four in ten wandering thoughts go somewhere pleasant. About three in ten are neutral. The remainder are unpleasant — and that last slice is carrying the whole effect.
The same pattern shows up somewhere far more consequential. In a study of nearly a thousand drivers injured badly enough to reach a hospital emergency department, those whose minds had been deeply, disruptively elsewhere in the moments before the crash were about twice as likely to be the one at fault.
In that same study, ordinary casual mind-wandering showed no meaningful link to crash responsibility at all — and just over half of all the drivers reported some.
It was not whether the mind had left. It was how far it had gone, and when.
There is also a decent case that the arrow points the other way from how it’s usually drawn. When researchers measured mood and wandering fifteen minutes apart, low mood predicted later wandering, but wandering did not predict later low mood. The bad mood came first and pulled attention inward. Which means “catch yourself drifting and snap out of it” is treating a symptom.
Your mind leaves constantly, and most of the time that’s fine — pleasant wandering costs nothing at all. The harm sits in a minority of episodes that are deep, unpleasant, and badly timed. And low mood seems to cause the wandering rather than the other way round.
What happens when the stakes are real
Aviation is the best place to look at this, for an unglamorous reason: somebody actually measured it. Researchers put experienced Boeing 747 pilots into a full-motion simulator, flew them a demanding arrival, and interrupted them every two minutes to ask what they had just been thinking about.
Before the results, the backdrop. A landmark accident review looked at major airline accidents where crew actions were a factor. In more than eight out of ten of them, somebody had failed to monitor properly or failed to speak up about something they’d seen. Not a lack of skill. Not a lack of knowledge. A lapse in watching.
So what did the simulator find?
1. Automation was not the villain
This is the opposite of what almost everyone assumes, including me before I read the papers. Pilots’ minds wandered slightly more with less automation engaged — and the difference was too small to mean anything. A second study, measuring the monitoring pilot instead, found the same nothing.
What the researchers concluded is more interesting than “automation makes you drift.” Pilots’ thoughts wandered not when a particular level of automation was used, but when whatever they were using was working.
2. Idle hands were the villain
When pilots were actively doing something with their hands — setting up the panel, working the flight computer — almost none of their thoughts were off-task. When they weren’t touching anything, a quarter were.
And the cleanest version of the same point: simply being designated the monitoring pilot roughly doubled the wandering. Not because those pilots were doing less — measured activity was almost identical. From outside the cockpit you could barely tell the two roles apart. Inside their heads they were in completely different places.
3. Things going well was a risk factor
When pilots made the required crossing restriction, they reported far more wandering than when they missed it. Success frees up attention, and freed attention goes somewhere.
Which is the profile of a great many incidents: a perfectly normal flight that quietly stopped being watched.
Blaming the autopilot is comfortable and wrong. What pulls a pilot’s mind away is having nothing to do with their hands, being cast as the watcher rather than the doer, and the flight going smoothly. The quiet, well-behaved leg is the dangerous one.
How often the watching actually failed
In the second study, pilots acting as monitoring pilot had to make a standard altitude call as the aircraft descended through each assigned level. A routine, drilled, unremarkable task.
- One in four of those calls was missed.
- Only one pilot out of sixteen made every single one.
- Across the sessions, pilots spent something like two-fifths of their available monitoring time thinking about something else.
- When they were purely monitoring, they got the call right most of the time. When they’d picked up some other cockpit task, they got it wrong more often than right.
These were pilots averaging close to eighteen thousand hours, in a sterile-cockpit-enforced simulator, being watched by a researcher. The authors’ conclusion was blunt: monitoring misses, even among very experienced pilots, are in no way rare.
There was no fatigue-style decline over time. Pilots were no worse at the third callout than the first. In laboratory studies of sustained watching, performance normally starts sagging within about fifteen minutes — and these sessions were longer than that. So the absence is meaningful. Attentional “running down” explained none of the missed calls.
Why not? Because a flight deck never gives you the long uninterrupted watch that the laboratory assumes. The watching is constantly broken up. And the two things pilots did instead of sustaining a long vigil — picking up other tasks, and letting the mind drift — each produced their own misses.
So “concentrate harder for longer” is aimed at a mechanism that isn’t the one failing. As the researchers put it: diligent monitoring, mixing in other activities, and taking mental breaks all lead to lapses.
The encouraging part: pilots already schedule it
Minds don’t wander at random. They wander into gaps.
During the approach phase, where air traffic control naturally creates pauses, pilots put the overwhelming majority of their off-task thinking into those pauses — and their callout performance was fine. During the arrival phase, where one call runs straight into the next with no breathing room, most of their wandering landed at inconvenient moments. That is where the misses came from.
In other words, people are already doing crude attentional budgeting. It works when the task structure gives them somewhere to put the drift. It fails when it doesn’t.
Which is exactly the pattern from the driving data: depth and timing, not presence.
Even very experienced crews miss routine checks about a quarter of the time — and not because they run out of concentration. There’s no evidence of attention “running down.” People simply drift, or fill the gap with another job. The good news: they mostly drift into the pauses, when the task structure gives them pauses.
Two failure modes worth naming
Missing the alarm
In one study, pilots in a motion simulator were given a landing-gear failure with a loud alarm — comfortably louder than the ambient cockpit noise, the sort of thing you cannot imagine missing.
With nothing else going on, not one pilot missed it. Introduce a destabilising windshear at the same moment, and almost four in ten did not react at all.
This is not deafness and it is not carelessness. Under heavy load the sound reaches the ear and does not reach awareness. It’s the same attentional bottleneck as mind-wandering, pointed outward instead of inward.
Forgetting to do the thing you already decided to do
Someone went through a large sample of confidential incident reports filed by airline pilots and pulled out every one involving a memory failure. Nearly every single one was the same type: not forgetting something learned, but forgetting to carry out an intention — a task deferred, an interruption absorbed, a step that never got resumed.
The classic cases are the flap and slat omissions. One well-known accident came down to a crew not working the taxi checklist that would have caught the configuration — with investigators noting the captain had started several conversations that had nothing to do with the job after pushback.
In another accident, the flaps were called. Prompted for flaps, the first officer answered with the setting and confirmed a green light — while the flaps were still up.
A callout is not a check. Under the right conditions a callout becomes a habit that runs perfectly well without the perception it is supposed to verify. The words come out. The looking doesn’t happen. That is a wandering mind wearing the uniform of a procedure.
So what do we do about it?
Here is the uncomfortable part. The standard countermeasures — the sterile flight deck rule, monitoring callouts, crew resource management — are sensible, widely adopted, and built on hard-won accident experience. But almost none of them has published evidence that it actually reduces monitoring failures.
That is not a claim that they don’t work. It’s a claim that nobody has demonstrated that they do. The industry’s own monitoring guidance says as much about itself — its authors noted openly that they’d prioritised getting the material out over including the human-factors science behind it. And when researchers reviewed decades of crew resource management studies, their conclusion was that they could not determine whether it had affected safety outcomes at all.
A study of planned rest on long-haul flights gave each pilot in turn a forty-minute rest opportunity during quiet cruise, finishing an hour before descent.
- Pilots took it up almost every time, and fell asleep within a few minutes.
- A forty-minute slot produced around 25 minutes of real sleep.
- In the final ninety minutes of flight — descent and landing, the part that matters — microsleep events dropped by roughly two-thirds compared with the crews who hadn’t rested.
Set that against what chronic short sleep does. A fortnight of six hours or less a night produces impairment comparable to a couple of nights without any sleep at all — while people’s sense of how sleepy they are plateaus. At their worst measured performance, subjects reported feeling only slightly tired.
People do not know how impaired they are. That is the operationally dangerous part.
Attention lapses are normal, frequent, and largely immune to being willed away. The standard procedural defences are sensible but mostly unproven. The one intervention with hard evidence behind it is rest. Which naturally raises the question everyone is currently being sold an answer to.
Can’t we just monitor the brain and fix it?
This is where a lot of money is currently going, so it’s worth being precise.
The technology is called neurofeedback, and the idea is genuinely elegant. Put sensors on the scalp, measure electrical activity, calculate a number, and show that number back to the person in real time — as a tone, a bar, a game. Reward the number when it moves the right way, and see whether people learn to move it deliberately.
The underlying learning mechanism is real, and it’s the same one that trains animals and people to do almost anything: reward. There’s a beautiful piece of neuroscience behind it. Brain cells that release dopamine fire when something turns out better than expected, go quiet when it turns out exactly as expected, and dip below baseline when an expected reward fails to arrive. Dopamine isn’t the pleasure chemical, as it’s usually described — it’s a signal about surprise in value. That’s why these systems deliberately keep you succeeding most but not all of the time. A reward you fully expect teaches you nothing.
Brain scans confirm that the reward and learning machinery genuinely switches on during neurofeedback. People really are learning something.
The trouble is what happens next.
Problem one: the trials don’t hold up
When you test neurofeedback properly — half the people get feedback from their own brain, half get convincing fake feedback, and nobody involved knows which — the advantage of the real thing disappears. This has now happened across attention, sleep, anxiety, mood and performance training, in trial after trial, including in healthy people trying to improve concentration.
People do improve. They improve just as much with the fake version. Which tells you the working ingredient is the thirty-odd sessions of structured attention practice, coaching and expectation — not the brainwaves.
Problem two: the headband can’t see what it claims to see
You will read that consumer devices track your “default mode network” — the brain system associated with mind-wandering, self-reflection, remembering your past and imagining your future.
That network is a real and important thing. But it was defined using MRI scanners, which measure blood flow. EEG measures electricity. They are different physical signals, related only indirectly. There is no brainwave frequency that is the default mode network.
Worse, the network’s two main hubs sit deep on the inner wall between the two halves of the brain — several centimetres in, and about the least accessible tissue there is from a sensor stuck on your forehead. And working out which bit of brain produced a given scalp reading is a mathematically under-determined problem: you need to sample the electrical field densely across the whole head to even attempt it. Research systems use sixty-four, a hundred and twenty-eight, or two hundred and fifty-six sensors, plus an individual scan of your head. A consumer headband has between two and eight.
Studies that recorded EEG and MRI simultaneously to find the network’s electrical signature have reported, variously: that it tracks alpha waves, that it doesn’t, that it tracks one slice of beta waves, that it tracks the opposite way in an adjacent slice of the same band, that it tracks frontal theta, that it doesn’t — and, in one careful study, that it has no EEG correlate at all, even though the network was plainly visible in the MRI from the same session.
Asking whether a headband can see the default mode network is a bit like asking whether a thermometer can see a rainbow. Both respond to something real. There are lawful relationships between them. You still can’t substitute one for the other.
To be fair to the technology: consumer EEG genuinely can pick up broad alertness and drowsiness, muscle tension, eye movement, and — in the better devices — decent automatic sleep staging. Those are real capabilities with real uses. “Tracks how alert you are” is defensible. “Measures your default mode network” is not.
Problem three: nobody has measured whether it helps
The most striking gap in this entire field: no neurofeedback study has ever used real-world experience sampling — the method that defined mind-wandering — as an outcome. Every study measures the brain signal itself, or performance on a task inside a scanner, or a questionnaire.
There is no evidence that neurofeedback reduces mind-wandering in daily life, because nobody has checked.
And when researchers ran the best-controlled test of the most direct target — training people to quieten the key hub of the default mode network, with a proper fake-feedback comparison — the fake feedback produced a similar effect, and there was no difference in how people said they felt.
One more thing worth saying: you wouldn’t want to switch it off
Even granting a device that worked perfectly, “suppress the wandering system” is a bad goal.
That network isn’t idling. It’s what you use to remember your own life, imagine your future, plan, follow a story, and work out what other people are thinking. In people whose brains are damaged in those regions, the result isn’t laser focus — it’s losing access to your own autobiography. Undemanding tasks that maximise mind-wandering improve creative problem-solving afterwards more than hard tasks, rest, or no break at all.
The thing worth changing was never the wandering. It’s the capture — the moment ordinary drifting tips into circling the same unpleasant thought, or drifts at exactly the wrong second.
The idea is sound and the learning is real — but when you test it honestly, fake feedback works just as well, the sensors can’t see the brain system being advertised, and nobody has checked whether any of it changes anything in daily life. Save your money.
What actually helps
There is a large research literature comparing ways of handling difficult thoughts and feelings. Two of them are actively counterproductive.
Which gives a short and fairly practical list.
- Stop trying to wander less.It’s a third of everyone’s waking life, most of it is harmless or pleasant, and no study shows that reducing it improves anything. You’re aiming at the wrong target.
- Design for placement, not suppression.Pilots already park their wandering in the gaps — when the task structure gives them gaps. If you run operations, building predictable low-consequence windows into a procedure does more than telling people to concentrate. A procedure that demands an unbroken watch is fighting the hardware.
- Give hands and voice something to do.Almost no wandering with hands busy; a quarter with hands idle; double the wandering from the mere label of “monitoring pilot”. Make monitoring an active job with physical and verbal outputs, not a watching job. The mechanism is the action–feedback loop, not the motion itself: what binds attention is that the doing has to close on something you actually perceive — which is exactly why the flaps callout above failed, and why a hand on a switch is a proxy for engagement rather than the thing itself.
- Treat the smooth day as the risky one.Attention goes wandering precisely when things are going well. The flight, drive or shift that feels easy is the one that quietly stops being watched.
- Protect sleep, and don’t trust your own read on it.The single best-evidenced intervention here. And remember that your sense of how tired you are stops tracking how impaired you are.
- Get some real exercise.In a large multi-site trial, the benefit of exercise on mental health ran through a reduction in repetitive negative thinking — it works on this specific mechanism rather than around it. Thirty to sixty minutes, moderate or better, several times a week.
- When you catch yourself circling, change the question — don’t fight the thought.“Why am I like this?” is abstract and makes things worse. “What specifically happened, and what’s the next concrete step?” is the move that blunted the effect in controlled experiments. Highest value item on this list for most people.
- If the concrete question is too hard, step outside yourself.Think it through as if advising a friend, or in the third person. Reliable in the moment, costs nothing.
- Don’t suppress, and don’t dwell.The only two strategies with documented negative effects. Everything else is at worst neutral.
Ordinary mind-wandering, including the unpleasant slice of it, is normal. What’s worth taking seriously is thinking that is repetitive rather than moving, mostly about the past and about yourself, hard to interrupt, and present most days for weeks. That pattern is the one that predicts trouble rather than just reflecting a bad week — and it’s a conversation with a doctor, not a self-help project.
Discover more from Safety Matters Foundation
Subscribe to get the latest posts sent to your email.
