How We Stopped Paying Attention

In 2004, a researcher at UC Irvine named Gloria Mark started following office workers around with a stopwatch, timing how long each of them stayed on one screen before switching to another. The answer was two and a half minutes. She has been running versions of that measurement ever since. It is now about 47 seconds.1

That number gets quoted a lot, almost always in service of a story you have already heard: phones broke our brains, nobody can concentrate, the species is cooked. The number is real. The story around it is mostly wrong, and the way it's wrong turns out to matter, because the honest version points somewhere far more useful than despair.

So we went through the actual literature: the observational studies that produced the 47 seconds, the meta-analyses that have quietly dismantled half the popular claims about multitasking, and the randomized controlled trials that tested whether any of this can be reversed. What follows is what the evidence supports, chart by chart, every figure linked to its source. Fair warning on two counts. Some of the most-repeated statistics in this field were invented, including the most famous one of all. And the most alarming interpretation of the rest, the one where your capacity for attention has been physically degraded, is the interpretation the data supports least.

01 - The measurementForty-seven seconds

Start with what was actually measured, because it is narrower and stranger than the headline suggests.

Mark's method is not a lab test. She instruments real people doing real work: software that logs every window and application switch, second by second, for days at a time. In 2004 the average duration of attention on a single screen was about 150 seconds. By 2012 the same measurement gave 75 seconds. In her 2016 study with Microsoft researchers, 40 information workers logged over two full work weeks, the median duration of focus on one screen was 40 seconds.2 Her pooled estimate across the more recent studies, the one that made the rounds, is 47 seconds.1

Chart 1 · The shrinking window

How long we stay on one screen

Average duration of attention on a single computer screen before switching, from studies that logged real workers in real jobs. These are separate studies, not one continuous instrument, which is why the line is not perfectly smooth.

Sources: Gloria Mark's research program as reported in Attention Span (2023) and summarized by the American Psychological Association; the 2016 point is the published median from Mark et al., CHI 2016. Populations and logging methods differ between waves; the 2016 figure is a median, the others are averages.

Two things about that chart deserve more skepticism than they usually get. The samples are information workers at desks, not humanity. And the 2004 and 2012 figures come from different studies than the 2016 one, with different logging setups, so this is a stitched series rather than one instrument tracking one thing. We think the direction is solid and the precision is not, which is roughly how everything in this field works.

But notice what the measurement actually is. It is not a test of how long you can concentrate. Nobody strapped these workers into a chair and asked them to hold focus. It is a record of how long they chose to stay on one thing when a hundred other things were one click away. That distinction is the entire article.

02 - The mythThe goldfish that never existed

Before going further we have to clear out the most famous statistic in this entire subject, because it is fabricated, and because you have almost certainly seen it presented as fact.

This matters beyond pedantry. That fake number became the load-bearing justification for a decade of making everything shorter: shorter videos, shorter paragraphs, shorter everything, on the theory that the audience had been neurologically downgraded. And underneath it sits a claim nobody has ever demonstrated, which is that the human capacity for sustained attention has actually declined. There is no good evidence that it has.

Attention didn't shrink. The number of things bidding for it exploded.

03 - The environmentWhat actually changed

If capacity didn't collapse, something else did. The measurable change is not in the person; it's in the volume of interruption the person is standing in.

Take the same logging studies. Across 32 information workers tracked for five working days, the average participant checked email 74 times a day and visited Facebook 21 times a day.4 The typical Facebook visit lasted about 18 seconds and the typical email visit about 32 seconds.4 These are not deep dives. They are the behavioral equivalent of a nervous tic.

For teenagers the number is worse, and better documented. In 2023, Common Sense Media installed tracking software on the phones of 203 American 11-to-17-year-olds for a week and simply counted. The median was 237 notifications a day. About a quarter of them landed during school hours. Some participants received more than 4,500 in a single day.5

Chart 2 · The interruption load

A typical teenager's day, in notifications

Median notifications received per day by 11-to-17-year-olds, from a week of on-device tracking. The school-hours and overnight slices are shares of that median.

Source: Common Sense Media, Constant Companion: A Week in the Life of a Young Person's Smartphone Use (2023), n=203, one week of tracked device data. The school-hours and overnight bars are our arithmetic on the report's stated percentages of the median.

Each of those interruptions has a tail. The most-quoted figure in this field is that it takes about 23 minutes to return to an interrupted task, and it comes from Mark's own work, although we should flag that the precise "23 minutes and 15 seconds" version circulating online is not a sentence you will find in the 2008 conference paper it is usually pinned to.6 The figure Mark states in her later peer-reviewed work is "on average around 23 minutes."4 Treat it as an order of magnitude, not a stopwatch reading.

The cleanest evidence that the environment is doing the damage, rather than the people, comes from taking the environment away. In 2012, Mark and colleagues cut email off entirely for 13 information workers for five working days, with heart-rate monitors attached throughout. Without email, people switched windows about half as often, held focus longer, and showed more variable heart rates. With email, they sat in a flatter, more constant state consistent with sustained low-grade alertness.7 It is a tiny study. It is also very hard to explain by saying the workers' brains changed over five days.

04 - The steelmanYour brain is probably fine

Here is where this investigation has to turn on its own side of the argument, because several of the scariest findings in the attention literature have not held up, and you rarely hear about it.

Claim one: heavy media multitaskers have damaged cognitive control. This is the famous 2009 PNAS study by Ophir, Nass and Wagner, which found that people who juggle many media streams were more susceptible to distraction.8 It launched a thousand headlines. It has also failed to replicate repeatedly: the change-detection result was not reproduced in at least three subsequent studies, the task-switching result was not reproduced in four, and one study found the opposite, with heavy multitaskers switching better.9 A meta-analysis pooling the evidence found the overall association between media multitasking and cognitive control to be small.9

Claim two: the mere presence of your phone drains your mind. The 2017 "brain drain" experiments found that having your phone on the desk, even face-down and silent, reduced working-memory performance compared to leaving it in another room.10 A 2023 meta-analysis gathered 22 studies and 43 effects and found the pooled effect real but small, Hedges' g = 0.14. More awkwardly for the popular version: the memory effect reached significance, but the effect on attention specifically did not, and the result was clear only in Asian samples, not North American or European ones.11

Claim three: you can train your way out of this with an app. The most thorough review ever conducted on brain-training programs, a 180-page assessment by seven researchers for Psychological Science in the Public Interest, concluded that these programs reliably make you better at the trained task, somewhat better at very similar tasks, and produce little credible evidence of transfer to everyday cognition.12 Getting good at the game makes you good at the game.

Chart 3 · The size of things

Everything in this field is a small effect

Standardized effect sizes from meta-analyses across the attention literature: acute exercise on executive function, mindfulness training on attention, paper over screens for reading comprehension, and a phone's mere presence on cognition. By the conventional benchmark, 0.2 is small and 0.5 is medium. Nothing here is close to the language used to describe it in headlines.

Sources: Ludyga et al. (2016); Verhaeghen (2021); Delgado et al. (2018); Böttger et al. (2023). These come from different designs and outcome measures and are not strictly comparable; they are shown together to convey scale, not to rank interventions.

So the steelman is strong: your attention hardware has not been demonstrably damaged, the multitasking-brain-rot literature is shakier than its press, and every effect in this field is modest. If you want to believe nothing is wrong, there is real evidence for you.

We don't believe it, for one reason. The 47 seconds is not an effect size in a lab. It is a description of how people now actually live, and it did not require anyone's brain to change.

05 - The mechanismYou are losing an auction, not a race

Put the pieces together and the picture resolves. Attention is a finite resource being allocated, continuously, between whatever is available. For most of human history the competing bidders were few, slow, and often worse than the thing in front of you. What changed over twenty years is not the size of the resource but the quality, speed and availability of the competition.

This is why the "capacity" framing fails and the environment framing works. A person who cannot hold a screen for more than 47 seconds at work will, that same evening, watch television for three hours, or read a novel until 2am on holiday, or spend six hours on a video game without a single voluntary switch. The machinery is intact. It is being outbid, in real time, several hundred times a day, by things engineered by very well-funded people to win exactly that auction.

You are not losing a race against your own mind. You are losing an auction against everyone who wants it.

It also explains the one place where a genuine capacity difference does show up: depth. A meta-analysis of 54 studies covering more than 170,000 readers found comprehension is reliably worse on screens than on paper, an effect concentrated in expository text and strongest when readers are under time pressure.13 Not because pixels are toxic, but because a screen is a place where you have been trained to skim, and where the next thing is always one gesture away. It is the same mechanism we found when we looked at what happened to reading in America: the habit didn't die of incapacity, it lost its slot.

06 - The reversalWhat the trials actually show

Now the useful part, and the reason the environment framing matters. If the problem were your brain, the outlook would be grim. Environments, unlike brains, can simply be changed. And when researchers change them, attention measurably comes back.

The strongest recent evidence is a 2025 randomized trial published in PNAS Nexus. Researchers recruited 467 adults and used a custom app to block mobile internet on their phones for two weeks, while leaving calls and texts working. Average daily phone use fell from 314 minutes to 161 minutes. Sustained attention, measured objectively with a gradual continuous performance test, improved by d_z = 0.24, which the authors note is roughly the magnitude of ten years of age-related decline, running the other way.14

Chart 4 · Two weeks without mobile internet

Daily phone use collapsed, then crept back

Average minutes of daily smartphone use before the block, during it, and after access was restored. The rebound is the interesting part: the behavior returns faster than the benefits fade.

Source: Castelo et al., PNAS Nexus (2025), n=467 enrolled, 313 completing all surveys.

What is striking about that trial is not the attention gain. It's that the attention gain was the smallest of the three things that improved.

Chart 5 · What two weeks bought

Attention improved least. Everything else improved more.

Within-person effect sizes from the same trial. The mental-health effect was, as the authors point out, larger than the meta-analytic effect of antidepressants.

Source: Castelo et al., PNAS Nexus (2025). Important caveats: only 25.5% of participants met the study's own compliance threshold, the sample was self-selected for wanting to cut down, and the authors explicitly note that expectancy effects could have contributed.

Those caveats are real and we would rather state them than bury them: a quarter compliance rate is low, and people who volunteer for a phone-blocking study know what result they are hoping for. But 91% of participants improved on at least one of the three outcomes, and the direction is consistent with everything else in this section.

The second-strongest lever is the one nobody wants to hear, because it is not about phones at all. In a laboratory-controlled experiment, 48 healthy adults were assigned to 8, 6 or 4 hours in bed for 14 consecutive nights, monitored continuously. Restriction to 6 hours or less produced attention deficits equivalent to up to two nights of total sleep deprivation. And the crucial finding, the one that should be on a poster in every office: their subjective sleepiness plateaued after a few days. By the end, with performance at its worst, the sleep-restricted subjects reported feeling only slightly sleepy.15

You are, in other words, a bad judge of your own attentional state. This is the single most important practical fact in the literature, and it applies to the phone as much as to sleep: the feeling of being fine is not evidence of being fine.

07 - The protocolWhat to actually do

Ranked by the strength of the evidence behind them, not by how good they sound. Each of these has a trial or a meta-analysis under it, and we've marked how solid each one is.

Notice what is missing from that list. There is no supplement, no app that rewires anything, no productivity system. The interventions that work are unglamorous and mostly consist of removing something.

08 - The conclusionThe good news is the boring news

The doom version of this story says your attention span has been permanently degraded by technology. Twenty years of research does not support that, and the most-cited evidence for it turns out to be a marketing statistic with no study underneath.

What the research does support is duller and much more actionable. Your capacity is intact. What collapsed is the market position of anything slow. Forty-seven seconds is not a measure of what your mind can do; it is a measure of how thoroughly the alternatives have been optimized, and how little friction stands between you and them. The 2025 trial is the proof: change the friction for two weeks and a decade of apparent age-related decline reverses, without anyone getting smarter or more disciplined.

That is the whole finding. Attention is not a virtue you were issued in short supply. It is a resource you are auctioning off, several hundred times a day, in increments of about 47 seconds, usually without noticing, to whoever bid last.

The price is negotiable. You just have to stop taking the highest bid by default.