9 min readResearch
How Long Does It Take to Refocus After an Interruption?
The famous answer is 23 minutes and 15 seconds. I traced the figure back to its source, and what the research measured is stranger and more useful than the soundbite.
Twenty-three minutes and 15 seconds. That is the number the internet quotes for how long it takes to refocus after an interruption, and it comes from real research. Gloria Mark, a professor of informatics at UC Irvine, spent years shadowing office workers with a stopwatch, logging their task switches to the second. In a 2006 Gallup Business Journal interview, she reported that interrupted work was resumed, on average, in 23 minutes and 15 seconds.
So the short answer is: about 23 minutes, and the figure has a legitimate pedigree. The longer answer is more interesting. The number measures something different from what most people assume, it appears in an interview rather than any of Mark's published papers, and the studies behind it contain findings far more uncomfortable than "interruptions are bad." For a start: you cause about half of them yourself.
I quote this research for a living, so I went and read the original papers. This is where the number comes from, what the studies measured, and what the soundbite leaves out.
Where the 23-minute figure comes from
The figure traces to that 2006 Gallup interview. Mark described a field study in which researchers shadowed 36 managers, financial analysts, software developers, engineers and project leaders for three working days each, recording the time of each event to the second. Her words: "Most interrupted work was resumed on the same day - 81.9 percent - and it was resumed, on average, in 23 minutes and 15 seconds."
One catch. That exact figure appears in none of her peer-reviewed papers. A software engineer traced the citation trail in 2023, checked the five academic papers most often cited as the source, and found the number in none of them. The closest published figure sits in Mark's 2005 CHI paper with Victor Gonzalez and Justin Harris, "No Task Left Behind? Examining the Nature of Fragmented Work", built on more than 700 hours of second-by-second observation of 24 information workers. In that dataset, interrupted work that got picked up again the same day was resumed after an average of 25 minutes and 26 seconds.
So the honest version is a range: somewhere between 23 and 25 minutes, depending on which of Mark's own datasets you read. The 15-seconds precision that makes the soundbite feel scientific is the least reliable part of it. The research itself holds up. The decimal-point theater does not.
What "refocus" meant in the study
The image the statistic conjures is a worker staring at their screen for 23 minutes, waiting for focus to reboot. That is a misreading of what Mark measured.
Her team measured resumption lag: the gap between an interruption and the moment you return to the interrupted task. And that gap is filled with other work. In the 2005 paper, workers passed through an average of 2.26 other tasks before coming back to the one they got pulled from. You get interrupted on the report, the interruption drags you into email, email surfaces a question from a colleague, the question turns into a ten-minute detour, and 25 minutes later you find your way back to the report. The 23 minutes is a measure of how long your original task sat abandoned while you bounced through everything in between.
The surrounding numbers from the same study describe the terrain:
11 minutes and 4 seconds: the average time workers spent in one "working sphere" (a coherent project or task) before switching or being interrupted
57%: the share of working spheres that got interrupted rather than completed
77%: the share of interrupted work resumed the same day, which means roughly a quarter of interrupted tasks did not come back at all that day
2.26: the average number of intervening tasks before returning to interrupted work
If you recognize this pattern, where the enemy is other legitimate work rather than TikTok, I wrote about it at length in the fragmentation problem.
You cause about half the interruptions yourself
This is the finding the soundbite buries, and to me it is the most important one in the whole body of research.
In the Gallup interview, Mark reported that people interrupted themselves about 44% of the time. The rest came from outside: phone calls, colleagues, notifications. The 2005 paper puts internal interruptions at 52% of the total. Managers skewed external (59% of their interruptions came from other people), while analysts and developers interrupted themselves about as often as anyone else interrupted them.
No colleague appeared at your desk. No notification fired. You stopped mid-task, of your own volition, and went somewhere else. Sometimes that self-interruption is useful, a deliberate break to let a hard problem incubate. Much of the time it is the familiar reflex: the task got uncomfortable, and a new tab was two keystrokes away.
This changes the shape of the solution. Notification blockers and do-not-disturb modes address the external half of the problem. The internal half walks straight past them, because the interruption originates in you.
Interrupted workers finish faster, and pay for it
In 2008, Mark ran a lab experiment with Daniela Gudith and Ulrich Klocke, published as "The Cost of Interrupted Work: More Speed and Stress". Forty-eight participants answered batches of emails while the researchers interrupted them roughly every two minutes by phone or instant message.
The result surprised everyone, including the researchers. Interrupted participants finished faster: 20.3 minutes on the task, against 22.8 minutes for the uninterrupted group, with no difference in error rates. People compensate. They sense the time being taken from them, so they work at a higher clip and write shorter, more clipped replies.
Then the bill arrives. On the workload measures, interrupted participants scored significantly higher on stress, frustration, time pressure and effort, and those differences showed up after only about 20 minutes of interrupted work. Scale that across an eight-hour day of Slack pings and self-inflicted tab detours and you get a familiar creature: the day where you got things done on paper and still ended up wrung out, snappish, and sure that you worked twice as hard as your output suggests.
So interruptions cost less time than the 23-minute soundbite implies, and more of everything else. The tax gets paid in cortisol instead of minutes.
Attention residue: the part of you that stays behind
The stress finding connects to a second body of research. In 2009, organizational behavior researcher Sophie Leroy published "Why Is It So Hard to Do My Work?" in Organizational Behavior and Human Decision Processes. Her experiments showed that when you switch from Task A to Task B, part of your attention stays with Task A. She named it attention residue.
Leroy found the residue is thickest under exact office conditions: when Task A is unfinished and time-pressured, your brain keeps processing it while you try to work on Task B, and your performance on Task B drops. You are at the meeting, and a background thread is still chewing on the half-written email. You return to the report after "two minutes" in Slack, and the Slack conversation rides along with you.
So minutes alone understate the refocus tax. Even after you return to the task, you work at reduced capacity until the residue clears, and if you switch often enough it never clears. You spend the afternoon operating through a film of half-dismissed thoughts. The context switching research puts numbers on how much capability that film eats.
The 47-second baseline
One more layer, because the interruption research lands on top of an attention environment that has degraded since the original studies. Mark kept measuring after 2005, and she compiled the trajectory in her 2023 book Attention Span. In 2004, her team's screen-recording data showed people averaged about two and a half minutes of attention on any single screen before switching. A few years later it was around 75 seconds. In her most recent measurements, the average is 47 seconds, a figure covered by the University of California when the book came out.
Read those two numbers together. If refocusing after an interruption takes 20-something minutes, and the average stretch of screen attention is 47 seconds, the arithmetic stops working. Nobody pays a full refocus tax after each switch, because at that switching rate there would be no work left. Instead, most knowledge workers never reach full focus in the first place. They operate in a shallow state where switches are cheap because there is so little depth to lose. The 23-minute figure describes the cost of losing real focus. The 47-second figure suggests real focus has become the rare condition. More numbers on that decline live in my roundup of digital distraction statistics.
The honest answer
Pull it together and the answer to the headline question looks like this:
About 23 to 25 minutes to return to interrupted office work, per Mark's field studies, with roughly two other tasks handled in between
A quarter of interrupted tasks not resumed at all that day
Elevated stress, frustration and effort even when total time recovers, per the 2008 experiment
Attention residue degrading the next task after the switch, per Leroy
All of it sitting on a 47-second baseline of screen attention
The exact number is messier than the soundbite. The tax is not. Every version of the data agrees that once an interruption lands, you pay: in delay, in stress, in residue, or in all three. Which points at the only strategy that survives contact with the research. You cannot make the tax cheaper. Refocusing costs what it costs, because that is how human attention works. The winning move is to not get interrupted, and since half of interruptions are self-inflicted, that means building an environment that catches your own hand on the door. That principle drives my whole deep work setup.
Why I built Hugo around prevention
This research is the reason Hugo works the way it does. I built a focus app for the Mac, and I built it to stop interruptions before the tax comes due, including the ones you start yourself.
You open a session and tell Hugo what you're working on. From then on, its AI evaluates each new tab and app against that goal, in real time. The 44% problem, the self-interruption, is the case Hugo was designed for: the moment your hand drifts to a new tab mid-task, Hugo makes the call before the detour starts, and if something looks off-task it asks you to justify how it serves the goal you set. Most of the time, having to type a justification is the moment you realize there isn't one. That is the interruption dying at second zero, before the 23 minutes, the residue, or the stress ever begin. And because it covers desktop apps as well as the browser, the escape hatches that a blocklist leaves open are handled too.
Honesty clause: Hugo cannot stop a colleague tapping your shoulder or a phone ringing on your desk. External, human interruptions remain your calendar's problem. But on a Mac, the digital surface is where nearly all self-interruption happens, and self-interruption is half the tax. Hugo handles that half so you can lock in and crank through the work you planned.
Download Hugo and see what a session without the tax feels like.
Ready to stop losing time to distractions?
Hugo uses AI to keep you on task. No blocklist required. Set your goal, and Hugo evaluates every tab in real time.
Frequently Asked Questions
AI-Powered Focus
The focus app that thinks for you.
Hugo sits between you and distraction. It hides your apps, locks down your browser, and uses AI to silently decide if what you're opening is actually work - so you never have to burn willpower again.