Nothing Slipped Today. You Are Wrecked Anyway, and That Is the Finding.

Mind  ·  Can’t Focus at Work, Vol. 3  ·  The Reading Room  ·  9 min read  ·  What a 2008 experiment measured about interrupted work, why the famous 23-minute figure has no paper behind it, and what to count instead of tasks

Everything went out. Nothing was late. You cannot explain to anyone, including yourself, why you are sitting in the car at six not turning the key.

You were interrupted eleven times before lunch. You know this because you counted, once, on a Wednesday, out of a vague need for evidence. And the evidence went nowhere, because the deliverables all landed and the deck was finished and you answered every one of those eleven people, which meant there was nothing to point at. What follows is a reading of published research on that day; we are not clinicians, the findings belong to Gloria Mark and her co-authors, and only the plain-English version is ours.

That is the specific trap of constant interruptions at work when you are the competent one. The damage does not show up in the place everybody looks.

The short answer: in a controlled experiment, interrupted people did not take longer. They finished faster, at the same quality — and reported significantly higher stress, frustration, time pressure, effort and mental workload than people who were left alone.¹ The interruptions did not cost you the work. They were paid for out of you, which is why the output looks fine and you do not.

There is a name worth having for what you handed over, because you will want to point at something: the invisible bill. Every time you sped up to protect a deadline, you paid a small amount of effort that nothing recorded. The work is the receipt. The bill is not on it.

What the study actually measured

Gloria Mark, Daniela Gudith and Ulrich Klocke ran the experiment and published it at CHI in 2008 under a title that gives away the result: The Cost of Interrupted Work: More Speed and Stress.¹

Forty-eight people were given office work to do. Some were left alone. Some were interrupted by phone or instant message with content related to the task; some were interrupted with content about something else entirely. Afterwards, everyone reported on how the work had felt.

The headline sentence from the paper is worth reading exactly as written: “Surprisingly, people completed interrupted tasks in less time with no difference in quality.”¹

They compensated. Interrupted, they worked faster, and the work came out just as good. Which is the finding almost nobody quotes, because it is not a satisfying story about how interruptions ruin your day, and because it describes something most competent women recognise instantly and have never once been given credit for.

Then the second sentence: “Working faster with interruptions has its cost: people in the interrupted conditions experienced a higher workload, more stress, higher frustration, more time pressure, and effort.”¹

Definition: the cost of an interruption at work is not usually measured in lost minutes. It is absorbed as compensatory effort — you speed up to protect the deadline — and shows up as stress, frustration and workload rather than as anything late.¹

About the 23 minutes everybody quotes

You will have seen it. It is on posters. It takes 23 minutes and 15 seconds to return to a task after an interruption.

That number is not in the 2008 paper. The paper does not examine recovery time between finishing an interruption and returning to the original task, and it never mentions the figure 23.² The number appears to have entered circulation through interviews and press coverage rather than a printed result, and traces back to no locatable primary source. Where published work has estimated resumption lag, the figures reported have generally been in the range of eleven to sixteen minutes.²

This matters for you rather than for the record. The 23-minute framing tells you your day is being shredded and the arithmetic proves it — and then the arithmetic does not match your experience, because your day is not visibly shredded. Everything went out. So you conclude the shortfall must be you.

The real finding fits what actually happened far better. Nothing was late because you absorbed it. The bill arrived separately, at six, in the car.

If the day only holds together because you absorb the difference, the Quiet Audit is ten private minutes on how much you are absorbing. Start here.

Why the compensation is invisible to everyone including you

Speeding up to protect a deadline does not announce itself. There is no moment where you decide to do it. You simply notice, around four, that you have not stood up since eleven, and that the sandwich is still in its packet.

And the feedback you get is about the output, because the output is the only visible thing. The deck was good. The client was happy. Nobody sees the compression, so nobody comments on it, so the only person in a position to notice is you — and you are the person who has been trained for thirty years to read “it all got done” as evidence that the day was fine.

This is where the shame enters, and it enters through a gap in the evidence. There is no artefact. A late deliverable is a thing you can hold up. Two hundred small compensations are not, which is why “I should have handled that better” arrives so easily and finds nothing to argue with.

It is also why the after-hours message costs more than its content suggests. A 6:58 p.m. “quick question” is not one minute of reading. It is one more thing to compensate for, arriving at the hour when there is nothing left to compensate with.

Five Things That Look Like This and Are Not This

Everything above protects the research. This part protects you, and it is the section most productivity writing skips, because it involves conceding that the workload might not be the whole story.

Five ordinary conditions produce being wrecked at six on a day when nothing slipped. Each is worth raising rather than assuming.

• Low ferritin, with or without anaemia. Common in the years before the final period. The tell is breathlessness on stairs and a tiredness that a good night does not touch.
• An underactive thyroid. Fatigue with cold intolerance, dry skin, weight change and slowed thinking. One blood test settles it.
• Obstructive sleep apnoea. Waking unrefreshed after a full night is the signature, and somebody else has usually noticed the breathing first.
• The perimenopausal sleep pattern. Fragmented second half of the night, most nights, with the exhaustion arriving before the working day rather than at the end of it.
• Depression. The distinguishing feature is not how heavy it feels but what happens on a clear weekend: depletion lifts a little, anhedonia does not.

None of these rules the invisible bill in, and none of them rules anything else out. They stack, which is why the same Wednesday costs far more with untreated apnoea than without it. These are questions worth putting to a clinician, not a list anyone can clear themselves on.

Who Else Has Measured This

One experiment is never enough, and the gap in this one is real: 48 people with a mean age of 26 doing a short controlled task is not a Wednesday with eleven live threads. Three other groups have measured the same mechanism from different directions.

Prof. Sophie Leroy (Faculty · ResearchGate · LinkedIn), now at the University of Washington Bothell, named attention residue in 2009: part of your attention stays on the previous task after you switch, and performance on the new one suffers accordingly.³ That is the mechanism sitting underneath the compensating.

Dr. Mary Czerwinski (Microsoft Research · Google Scholar · Wikipedia) and colleagues at Microsoft Research ran a diary study of real task switching and interruption in real offices, rather than in a laboratory, and found the same pattern of fragmentation and returned-to work.&sup4;

Erik Altmann (MSU faculty · Google Scholar) and J. Gregory Trafton (Site · Google Scholar · LinkedIn) built the memory-for-goals account of why resuming an interrupted task costs anything at all: the goal has to be reactivated, and reactivation takes time and effort that nothing on the deliverable records.&sup5;

A laboratory experiment, a field diary study and a computational model, arriving at the same place. The cost is real, it is paid in effort, and it does not show up in the output.

The work is the receipt. The bill is not on it.

What This Actually Changes

Not the workload. Nothing here removes a single message from your inbox, and pretending otherwise would be the same trick the 23-minute poster plays.

What it changes is the sentence you have been saying in the car at six, which is usually some version of everything got done, so why am I like this. The research answers that exactly. Everything got done because you were like this. The compensating is the mechanism by which the day held together, and the state you are in at six is what it cost.

That reframe matters because it moves the evidence. For years the only thing you could point at was output, and output kept saying the day was fine. The invisible bill is the other column, and once you start recording it you have something to bring to a conversation that is not a feeling.

It also predicts something specific. If the cost is compensatory effort rather than lost time, then the fix is not to work faster or to prepare harder. It is to reduce how often you are asked to absorb the difference — which is a conversation about the channel, not about your resilience.

What to Say, and What You Will Probably Hear Back

One sentence is rarely enough, because there is a standard reply waiting for it. Two versions, depending on which room you are in. Both short enough to read off a phone.

If what you want to take into either conversation is a clear account of what the weeks have actually contained, rather than one remembered Wednesday, the Quiet Audit is ten private minutes on exactly that. No programme, no advice, nobody sees your answers. Start here.

What to count instead of tasks

The measurement you are using is broken. “Did everything get done?” will return yes on the days that are hurting you most, because getting everything done is the mechanism of the harm.

So change what you record. For one week, at the end of each day, write two things and nothing else:

• How many times you sped up to protect something. Not interruptions — compensations. The moment you thought “fine, I will just do this faster.”
• What time you last had four uninterrupted minutes. Not an hour. Four minutes.

Five days of that produces something a list of completed tasks never will: a record of the cost, in a form you can actually show someone. It also gives you a number for the conversation you have been putting off, because “I am finding it hard to concentrate” gets sympathy and no change, while “I compensated fourteen times on Tuesday to protect that deadline” is a fact about the workload rather than a fact about you.

Then take one interrupting channel and give it a shape. Not all of them — one. Messages answered on the half hour rather than on arrival. The point is not to become unreachable. It is to convert a stream of arrivals into two scheduled batches, so that the compensating happens on your clock instead of theirs.

Questions people ask about this

Do constant interruptions at work actually slow you down?
Usually not. In the 2008 CHI experiment, interrupted people completed the work in less time than people left alone — 20.31 and 20.60 minutes against 22.77 — with no measured difference in quality.¹ What rose instead was stress, frustration, time pressure, effort and mental workload. The interruption is paid for in effort rather than in minutes, which is why the day looks fine on paper and does not feel fine at six.

If the work still gets done, is this actually a problem?
The work getting done is the problem, in the sense that it removes the signal. Effort and stress that go unrecorded do not go unspent. The study measured the spend directly, which is precisely what a completed to-do list cannot do.¹

Am I just less resilient than I used to be?
The number of channels able to reach you has changed considerably more than you have. It is also worth separating this from genuine cognitive change — there is one specific question to ask before assuming the worst.

Why can I handle it fine some days and not others?
Because compensating draws on the same capacity as everything else you decide all day. By late afternoon there is less of it, which is why the identical interruption at 9:30 costs almost nothing and the same one at 4:40 feels like being asked to lift something.

Should I just get better at saying no to interruptions?
Most of the ones reaching you are not refusable and you know it. Shaping the channel is more achievable than refusing the people, and it does not require a conversation with anyone before it starts working.

Is the 23-minute figure completely wrong, then?
It is unsourced rather than disproved. Recovery time after an interruption is real and has been estimated in published work at somewhere around eleven to sixteen minutes. The specific “23 minutes 15 seconds” simply does not have a paper behind it.²

Next in the series: you cleared the calendar, closed the door, and got nothing done anyway. Jo Brennan on the afternoon that had no interruptions in it at all.

Where to Go Next

In order, and each for a reason.

1. If part of your attention is still on the last thing every time you start the next one, read attention residue and the sixty seconds that clear it — that is the mechanism underneath the compensating.
2. Then the afternoon that had no interruptions in it at all, because a cleared calendar is the obvious fix and it does not work on its own.
3. Then what happens when there is no artefact to point at, which is where the shame gets in.

Nothing was late. That was never the thing worth measuring.

References
1. Mark, G., Gudith, D., & Klocke, U. (2008). The Cost of Interrupted Work: More Speed and Stress. Proceedings of CHI 2008, 107–110. Read the paper
2. Analysis of the provenance of the “23 minutes 15 seconds” claim, tracing it to interviews and press coverage rather than to any printed result in the 2008 paper or its successors. Read the trace
3. Leroy, S. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. Organizational Behavior and Human Decision Processes, 109(2), 168–181. Read the paper
4. Czerwinski, M., Horvitz, E., & Wilhite, S. (2004). A diary study of task switching and interruptions. Proceedings of CHI 2004, 175–182. Read the paper
5. Altmann, E. M., & Trafton, J. G. (2002). Memory for goals: an activation-based model. Cognitive Science, 26(1), 39–83. Read the paper

On the researcher. Gloria Mark is a research scientist, not a clinician: her doctorate is in psychology and her chair is in informatics, and nothing in her published work is medical advice. Her own site lists the papers behind the book, which is where the figures in this piece were checked.

Written by Nora Whitfield for Blue Leaf Journal. Updated: 27 August 2026.