Apparently, it takes 23 minutes and 15 seconds to refocus after an interruption.
At least, that’s the rule according to wellbeing programmes, slide decks I’ve seen, the occasional HR newsletter and, perhaps, presentations you’ve sat through yourself.
Fast Company ran the 23 minute version in 2013. The Washington Post covered it in 2015. Today it resides in the blogs of time management and workplace analytics tools as seemingly accepted fact about the average person. For real precision, the 15 seconds gets added too.
I had it in mind when working in the emergency management sector recently. The broad context was cognitive performance degrading under pressure, and I made a couple of passing comments about multitasking, attention switching and attention residue. As I got part-way into the bit, I realised that someone might bring up the 23 minute rule.
Nobody did. Awesome.
It isn’t a thing, actually
It’s not in the published research.
The figure traces back to a small number of interviews with Gloria Mark and a quotation in the Wall Street Journal. From there it slid effortlessly into business media. Then it found a life of its own, bred, and proliferated into slide decks, webinars, newsletters and organisational wisdom. Maybe educational wisdom too?
Mark is a serious researcher with strong research chops and published the book ‘Attention Span’ in 2023, to sit alongside her several papers. In 2005 she and colleagues shadowed 24 information workers for more than 700 hours. Using a stopwatch, researchers found people spent 11 minutes on a unit of work before they switched or faced interruptions. Just over 57% of those work pieces were interrupted, and roughly half the interruptions were self-initiated. When the interrupted work was picked up later the same day, which happened 77% of the time, the average gap before it came back around was 25 minutes 26 seconds. During that gap, people deliberately worked on an average of 2.26 other things.
That time is a measure of time in queue, or how long a task waits in the pile, NOT how long a brain takes to come online after an interruption. It is a scheduling observation that shouldn’t be quoted as a cognitive claim.
Consequently, it shouldn’t be used that way in how organisations think about interruption. I write that because organisations have based real decisions on the number. They probably still do.
Here’s how it goes
The logic runs thus: every interruption burns 23 minutes of real cognitive capacity. A normal day holds dozens of interruptions, including meetings. We are therefore losing hours of output, so we should do everything we can to protect the calendar, by way of no-meeting days, focus blocks, notification bans and so on. The entire response works on the premise that interruption is a tax on output.
When tested, though, it didn’t hold. Mark, Gudith and Klocke ran a simulated email task with 48 [participants in 2008. Some were interrupted every two minutes, and some were not.
Given what you know of the tax, what would you expect?
Turns out, the interrupted participants finished faster (!) around 20 minutes versus 22.8, with no significant difference in errors or quality. What changed, was super revealing. Self-reported stress ran from 9.13 to 9.46 out of a possible 20, compared to 6.92 for the uninterrupted group, or about a third higher. Frustration, effort, and, surprise surprise, time pressure, were all significantly higher too.
People absorbed the interruption and compensated with effort. They worked harder and faster, and paid the tax in strain, not in output.
That last bit should change how we frame this discussion within our organisations. If you sell interruption as a productivity intervention, you’ll see the output is fine. In my opinion, the better answer is a sustainable load case. People are compensating, they are being literally and figuratively taxed, and the tax has a cost. The costs are sick leave, exhaustion, burnout, and turnover. Throughput numbers will hold. At least in the short term.
Attention residue
Let me pivot here to Sophie Leroy. Her attention residue research is the mechanism most often invoked here. Her 2018 work with Theresa Glomb, across four studies, found that people can’t disengage from an interrupted task when they anticipate having to rush back to it. It’s that residue that then degrades performance on the new task. The intervention that removed the penalty was small. It was a brief written note on where you had got to and how you would pick it up again.
Leroy’s own 2020 review with Schmidt and Madjar is a useful read here too. It integrates the interruption and multitasking literatures and names the inconsistencies. It’s not as crude as “switching costs you” but more accurately “switching costs you when you leave things unfinished and expect to have to rush back to them”. (Hopefully you’ll see how that could apply in an emergency management context and why I mentioned it!). A closing ritual removes the cost. (if you can smell Zei
In a field study, Van Zoonen and Scharp published a daily-diary experiment in 2025, using 50 employees at a Dutch technology consultancy over four workdays.The intervention group wrote a brief ready-to-resume plan each time they were interrupted. Attention residue dropped and the daily link between residue and exhaustion weakened.
Interestingly, the intervention itself also raised overall exhaustion over the four days. The authors put this down to the effort of learning a new habit. If implementing this, you can expect a hockey stick shaped curve. Things may get a little worse before they get a lot better. It’s helpful to identify that up front.
The takeaway
So what now? Well for one, stop with the 23 minute thing if your organisation does it. Better, correct it if you hear it.
The closing ritual is helpful and better than a calendar rule.
Watch for anticipated time pressure rather than the number of interruptions. The residue came from expecting to have to rush. If you know you’ll get the time back, you’ll handle it differently from someone who knows they won’t.
Argue from the position of load, not productivity.
For transparency, a few notes. The studies were small (and why I included the numbers) and before Slack, most smartphones, video calls etc, so something using these interruptions would be interesting. For balance, some recent research has found that socially driven interruptions can raise job satisfaction, and that emotionally-valenced interruptions can lengthen the time taken to resume while improving accuracy afterwards. Time to resume and performance aren’t the same thing.
Photo by Markus Winkler on Unsplash