Why Do Coworkers Interrupt You? The Research
Because interrupting you is easier than waiting, and because much of what you blame on colleagues is not coming from colleagues. What the research measures, why the famous 23-minute figure is quoted wrong, and what actually reduces the number.
Coworkers interrupt you because asking is less effort than waiting, and because most of the interruptions you blame on colleagues are not coming from colleagues. The decision to approach happens before you know about it, and it is made from what is visible on your screen. That is why muted notifications, a closed chat status and a Do Not Disturb sign have not fixed it, and why an interruption that never happens feels so different from one you survive.
What follows is the measured answer, including a correction: the widely repeated 23 minutes and 15 seconds figure comes from an interview rather than from the paper people attribute it to, and the paper that does exist found something different.
Key takeaways
- The average work event lasts 3 minutes 5 seconds before a switch or an interruption, excluding formal meetings.
- 44% of interruptions are self-inflicted, and the average person touches 12.2 spheres of work per day.
- The 23-minute figure is misattributed: it is from a 2006 interview, while the peer-reviewed paper reports 25:26 and 2.3 intervening tasks.
- Interrupted work was finished faster in a 2008 lab study, with no quality loss, but with significantly more stress and effort.
- Cubicle workers get 29% more interruptions than people in private offices, and report 9% higher exhaustion.
What the numbers actually are
The most substantive work on this is observational research by Gloria Mark at the University of California, Irvine, published in "Too Many Interruptions at Work?" (Gallup Business Journal, 2006) and in the peer-reviewed No Task Left Behind? (Mark, González & Harris, CHI 2005). It did not ask people whether they were interrupted. It watched what they did.
Read the first two together and the pattern is clear: three minutes is barely long enough to get into a problem. The same research counted 12.2 different spheres of work per day, and found 81.9% of interrupted work resumed the same day — which sounds reassuring until you price the resumption at a quarter of an hour.
The 23-minute myth, corrected
Almost every article on interruptions opens with the claim that it takes 23 minutes and 15 seconds to get back to a task. That figure comes from the 2006 Gallup interview, a journalist's summary rather than a reviewed paper. The peer-reviewed version of the same research reports 25 minutes 26 seconds to return, with an average of 2.3 other tasks handled in between. That second number is the important one: you do not return to the interrupted task, you work your way back through two other things first.
The 2008 lab study, The Cost of Interrupted Work (Mark, Gudith & Klocke), does not contain the 23-minute number at all. People finished interrupted work faster and with no loss of quality, and reported significantly higher stress, frustration, time pressure and effort. The cost of an interruption is therefore not reliably a block of lost minutes. Sometimes it is a task completed sooner inside a workday that feels harder than it would have been.
A 2023 release from UC Irvine tracks the same behaviour forward: average on-screen attention fell to 47 seconds between 2016 and 2020, against roughly 2.5 minutes in 2003. Interruptions have not got louder. The gaps between them have got shorter.
Why finishing a task is not enough to let it go
Time is only part of the cost. The rest is that switching degrades the work you switch to, which is what Sophie Leroy calls attention residue (Organizational Behavior and Human Decision Processes 109(2), 168–181, 2009). When you move to a new task, part of your attention stays behind on the old one, and performance on the task you have just started suffers for it.
Her experiments add the finding that matters most here: finishing a task is not sufficient. What releases attention is closure, and what reliably creates closure is a deadline. Under time pressure, her participants performed better on the task that followed. "I'll just get back to it" therefore loses twice: the work keeps you attached, not just the file open.
Who interrupts you most
The largest single source is you. 44% of interruptions are self-inflicted: a device switch, a quick check of a message, a second application. Nobody had to walk anywhere for those, and no sign on your monitor would have prevented them.
So who interrupts you from outside? It depends on where you sit. UC Irvine found employees in cubicles receive 29% more interruptions than those in private offices, and that frequently interrupted workers report 9% higher exhaustion. Proximity is the driver, not seniority and not personality. The companion productivity-tool page covers the sound side of the same problem.
What actually reduces them, and where the gap is
There is a consistent gap between what people try and what the mechanism is. Most of what fails fails for one reason: it targets the digital layer while the interruption happens in the social one.
- Muted notifications and a closed chat status. Work for messages. Useless for the person already crossing the floor towards your chair.
- A Do Not Disturb sign on the monitor. The sign competes with everything else in the room, and the loudest competing signal is another person mid-sentence.
- Headphones. Reduce sound, not approach, and make you unable to hear the colleague who genuinely needs you.
- Asking people to respect your time. The right ask, and impossible to enforce on individuals. It works as a team norm and fails as a personal request.
Only two things reliably remove the approach: a visible signal that you are unavailable, legible without explanation, and a team-wide agreement about protected time, where the rule is already known. The second is stronger and worth arranging whenever you have the standing to. The gap is that the first is available to everyone, immediately and at no cost, and almost nobody uses it. Everyone has a strong existing template for what a live video call means, so it needs no negotiation and no follow-up.
What a colleague sees when they glance at your screen
Real people from licensed stock footage, no AI. Muted and looping, nothing recorded, nothing sent anywhere.
Start a meetingA visible unavailability signal, and its honest caveat
MeetingSimulator puts a video meeting grid on your screen, from one to sixteen participants, so the approach never gets made. Every participant is a real person filmed in licensed stock footage — nothing on screen is AI-generated, which matters practically, because synthetic faces give the illusion away in about two seconds. It is free, needs no account, and runs in the browser.
The caveat comes first rather than as a footnote: a colleague cannot tell this apart from deception. They will read a meeting on your screen the same way whether it is real or not, and you cannot correct the impression without explaining the mechanism, which is the one thing that defeats it. Where you can get a team norm agreed instead, that is the better route. The version that fails is the one that buys an hour and fills it with browsing.
Be clear about the limits: this removes other people's interruptions and does nothing about the 44% that are self-inflicted. It is not a substitute for a team-wide no-meeting block where one exists.
Working where nobody can reach you
If your interruptions are mostly people crossing the floor, the direct fix is to stop being on that floor. In a café or coworking space the person next to you has no idea who you work for, so the question is never asked.
The focus part is less obvious, and the evidence is mixed. A series of five experiments, summarised by David Burkus in Harvard Business Review, put the optimum for creative productivity near 70 decibels, roughly the murmur of a busy café, with both quieter (~50 dB) and louder (~85 dB) levels failing to help. That is a wide target, and an easy one to miss by accident.
Focus in cafés and coworking spaces covers the rest: which background sounds help, why music with lyrics is the worst option for language and memory work, and where the "optimal noise" theory stops holding.
Setting boundaries without being rude
A signal on your screen is a workaround for something colleagues cannot read. A boundary is the thing itself, and it works on the interruptions a screen cannot touch: the chat message, the standing question in the kitchen, the seven-attendee invite with no agenda.
Workplace boundaries without being rude covers the wording, and why a stated rule outperforms a stated preference almost every time. The research above is a reasonable argument for having the conversation at all: 9% higher exhaustion among frequently interrupted workers is measurable, and a boundary is cheaper to defend than exhaustion.
Questions people ask
Why do my coworkers keep interrupting me?
Gloria Mark's observational research at UC Irvine found the average work event lasts 3 minutes 5 seconds before a switch or interruption, and 44% of those switches are self-inflicted. People interrupt you for the same reason they check their phone: sitting still is harder than acting. The decision is social and made from what they can see on your screen.
Is it really 23 minutes and 15 seconds to refocus?
The 23 minutes 15 seconds figure comes from Gloria Mark's 2006 Gallup Business Journal interview, not from a peer-reviewed paper. The reviewed version, No Task Left Behind? (CHI 2005), reports 25 minutes 26 seconds with 2.3 intervening tasks. The 2008 lab study, The Cost of Interrupted Work, does not contain that number at all: interrupted work finished faster, with no quality loss, but with more stress and effort.
What is attention residue?
Attention residue is Sophie Leroy's term for the part of your attention that stays behind on a previous task after you switch, which lowers performance on the task you have just started. Her 2009 experiments found that finishing a task is not enough to release it. You need the closure a deadline creates, which is why time pressure improved results on the following task.
Who interrupts me the most?
Your own switching does, more than other people do: 44% of interruptions are self-inflicted, and the average person moves between 12.2 spheres of work per day. Then proximity. People in cubicles receive 29% more interruptions than those in private offices, and frequently interrupted workers report 9% higher exhaustion, according to UC Irvine.
What actually reduces interruptions at work?
Only two things reliably remove the approach: a visible signal that you are unavailable, and a team-wide agreement about protected time. Muted notifications, a closed chat status and a sign on the monitor all fail on the colleague already walking to your chair. A recurring team-wide no-meeting block beats any personal tool; your own screen is the fallback for the hours policy cannot cover.
Block the next approach
Open the tool, choose how many participants are in the call, and the grid takes over your screen while you work behind it. Free, no signup, no install.
Free forever No signup Real people, not AI Nothing leaves your device
Sources: Gloria Mark, UC Irvine — "Too Many Interruptions at Work?" (Gallup Business Journal, 2006); Mark, González & Harris, No Task Left Behind?, CHI 2005; Mark, Gudith & Klocke, The Cost of Interrupted Work, CHI 2008; UC Irvine, 2023; Sophie Leroy, Why is it so hard to do my work?, Organizational Behavior and Human Decision Processes 109(2), 2009; David Burkus, Harvard Business Review, on five experiments on optimal noise for creative work.