Attention residue: the hidden cost of switching tasks

When you jump from one task to another, part of your mind stays behind. The lab numbers on that leftover attention, and the one-minute habit that cuts it, matter more than any productivity slogan.

Rows of telephone operators at switchboards in San Francisco in 1918
Photo: AnonymousUnknown author, Public domain, via Wikimedia Commons

Attention residue is the share of your attention that stays on a previous task after you switch to a new one, and it lowers performance on the new task. Sophie Leroy described it in 2009: people who left a task unfinished, or finished it without time pressure, kept thinking about it and did worse on the next one. Writing a one-minute "ready-to-resume" note before switching reduced the residue in her 2018 experiments.

Key takeaways
  • Attention residue (Leroy 2009): thoughts about the last task keep using capacity the next task needs.
  • 57% of work segments were interrupted in 700+ hours of office observation (Mark 2005).
  • The "23 minutes to refocus" line is a misreading: about 25 minutes is how long interrupted work waited, mostly spent on other tasks.
  • Interrupted people worked faster, but stress rose from 6.9 to about 9.5 on a 20-point scale within 20 minutes (Mark 2008).
  • A one-minute ready-to-resume note is the fix with experimental support (Leroy and Glomb 2018).

Switching is not free even when it feels instant. Lab studies since the 1990s measure a switch cost in time, the extra milliseconds the brain needs to reconfigure for a new rule.

Field studies of office work by Gloria Mark at the University of California, Irvine, measure the bigger cost: fragmented days, interrupted work that waits a long time to be resumed, and higher stress. This post puts the three layers together, separates the numbers from the slogans (including the famous "23 minutes"), and ends with what actually helps.

57%of work segments interrupted, in 700+ hours of office observation (Mark 2005)
25 min 26 saverage wait before interrupted work was resumed the same day
6.9 → 9.5stress score (1-20) after just 20 minutes of interrupted email work (Mark 2008)
~47 saverage time on one screen before switching, down from about 2.5 min in 2004

What is attention residue?

Attention residue is leftover thinking about Task A that occupies the mind while you work on Task B. Sophie Leroy, an organisational psychologist, published the concept in 2009 in Organizational Behavior and Human Decision Processes, with two experiments in which participants moved from a first task (such as a word puzzle) to a second (such as evaluating a job candidate's résumé) under different deadline conditions.

Two findings stood out:

  • Unfinished work sticks. Leaving the first task unfinished made it harder to move attention away, and performance on the second task suffered.
  • Finishing is not enough. People who completed the first task without time pressure still carried residue, apparently because they kept thinking about how well they had done. Finishing under a tight deadline helped people close the first task mentally and do better on the next one.
Task A left open, or no deadline Task B residue zone switch time Attention residue: thoughts about A occupy the first part of B
A schematic of Leroy's idea. The shaded part of Task B is where thoughts about Task A compete for attention; a ready-to-resume note shrinks it.

The residue is not just about incomplete work; it is about open loops of any kind.

The story

Between September 2006 and March 2008, observers followed 98 nurses on six wards of two major teaching hospitals as they prepared and gave medicines: 4,271 administrations to 720 patients, over 505 hours. Johanna Westbrook's team at the University of Sydney logged every interruption and checked each dose for mistakes.

Interruptions were the norm, in 53.1% of administrations. Each one was linked to a 12.1% rise in procedural failures, such as not checking a patient's identity or not recording the dose, and a 12.7% rise in clinical errors, such as the wrong drug or the wrong dose. The estimated risk of a major error was 2.3% with no interruptions and 4.7% with four. Reporting in Archives of Internal Medicine in April 2010, the authors pointed out that returning to a disrupted task first requires finishing the interrupting one. The hands move on; part of the mind is still on the last job.

ScienceDaily: Nurses who are interrupted while giving medication make more errors (report on Westbrook et al., Archives of Internal Medicine, 2010)
98 nursesobserved on six hospital wards
53.1%of medicine rounds were interrupted
12.1%more procedural failures per interruption
2.3% → 4.7%estimated major-error risk, none vs four interruptions

That is the link to working memory. With only about four chunks of active capacity, as covered in why you can hold only about four things, every thought still parked on Task A is a slot Task B cannot use.

How big is the switch cost in the lab?

In the lab the cost is measured in fractions of a second per switch, and it grows with task complexity. Joshua Rubinstein, David Meyer and Jeffrey Evans ran four experiments, published in 2001 in the Journal of Experimental Psychology: Human Perception and Performance, in which people alternated between tasks or repeated one task. Two experiments used classification rules for geometric shapes; two used arithmetic with different operations.

Switching was always slower than repeating. The cost rose when the rules were more complex and fell when a visual cue announced which task came next, and those two effects added up rather than interacting.

The authors read this as two separate stages of executive control: goal shifting (deciding "now I do task B") and rule activation (loading B's rules and switching off A's). Rule activation took longer when moving from a familiar task to an unfamiliar one than the other way round, which is why returning to hard, half-learned work after easy messages feels so sticky.

Factor tested (Rubinstein et al. 2001)Effect on switch costWhat it means at work
Rule complexity higherCost increasesSwitching into analysis or writing costs more than into routine admin
Cue tells you the next taskCost decreasesA written next step reduces reload time
Familiar to unfamiliar taskRule activation slowerGoing from WhatsApp back to new material is the expensive direction
Complexity and cuing togetherEffects add upTwo separate stages: goal shift, then rule loading
What happens in one task switch Task A running (its rules loaded) Stage 1: goal shifting "Now I do task B" faster when a cue announces the next task Stage 2: rule activation load B's rules, switch off A's slower for complex or unfamiliar rules Task B running
The two-stage model of executive control from Rubinstein, Meyer and Evans (2001). Cues speed stage 1; complex rules slow stage 2, and the two effects add up.
Executive Control of Cognitive Processes in Task SwitchingJoshua S. Rubinstein, David E. Meyer, Jeffrey E. Evans · Journal of Experimental Psychology: Human Perception and Performance 27(4) · 2001

Four experiments compared alternating between tasks with repeating one task, varying task familiarity, rule complexity and the presence of visual cues. Alternation produced time costs that increased with rule complexity and decreased with cuing, and the effects were additive, supporting a two-stage model of goal shifting and rule activation. Switching from familiar to unfamiliar tasks took longer to activate rules than the reverse. The limit: the costs are measured on simple, rapid lab tasks, so the size of the cost for real knowledge work has to come from field studies.

Read the paper

What do interruptions cost in real office work?

They fragment the day and delay the return to important work by tens of minutes. In a 2005 field study, Gloria Mark, Victor Gonzalez and Justin Harris shadowed 24 information workers at one US company for about three days each, over 700 hours of timed observation. They grouped activity into "working spheres", the projects people were actually trying to advance, and found that 57% of working spheres were interrupted.

A telephone operator in a blue uniform plugging a cord into a switchboard in the Tokyo Telephone Exchange, with rows of operators behind her
An operator at the Tokyo Telephone Exchange. Switchboard work was built around constant interruption; most knowledge work today has inherited the pattern without the training.Photo: NTT Public Corporation, Public domain, via Wikimedia Commons

Where "23 minutes" really comes from

Of interrupted work, 77.2% was resumed the same day. When it was, the average wait was 25 minutes 26 seconds, with a very wide spread (standard deviation of almost 55 minutes), and people passed through an average of 2.26 other working spheres before returning.

This is where the popular claim that it takes "23 minutes to refocus" comes from, and it is a misreading. The figure is the time until people came back to the interrupted task, mostly spent on other work, not the time their brain needed to regain focus.

The real cost is that important work waits behind a queue of other things.

Not all work was equal. Central working spheres, the core of someone's job, were resumed the same day 82.0% of the time; peripheral ones only 43.8%. Collocated workers worked longer before switching but were interrupted more.

No Task Left Behind? Examining the Nature of Fragmented WorkGloria Mark, Victor M. Gonzalez, Justin Harris · Proceedings of CHI 2005

Detailed observation of 24 information workers, averaging 25 hours 42 minutes of timed observation each. 57% of working spheres were interrupted; 77.2% of interrupted work was resumed the same day, after an average of 25 minutes 26 seconds and 2.26 intervening activities. Central work was about twice as likely to be resumed the same day as peripheral work (82.0% vs 43.8%). The limit: a small sample at one US company in 2004, before smartphones and chat apps, so today's fragmentation is likely worse, not better.

Read the paper (free)

Do interruptions make you slower, or just more stressed?

In the best-known experiment, interrupted people actually finished faster, and paid for it in stress. Mark, Daniela Gudith and Ulrich Klocke had 48 participants, mostly German university students, play a human-resources manager answering 12 emails per block. In the baseline block there were no interruptions; in two other blocks a "supervisor" called or messaged every two minutes with questions on the same topic or an unrelated one, such as how many hot dogs to order for 240 employees.

48participants answering 12 emails
22.77 minemail task with no interruptions
~20.4 minemail task with interruptions: faster
6.9 → 9.5stress score on a 20-point scale

Time on the email task itself was 22.77 minutes without interruptions and about 20.3 to 20.6 minutes with them. Errors and politeness did not change; the interrupted emails were simply shorter. But after only about 20 minutes of interrupted work, people reported significantly more stress, frustration, time pressure and effort, and the most mental workload when interruptions came from an unrelated topic. People compensate for interruptions by working faster, and the bill arrives as strain.

Self-rated workload after 20 minutes of email work, by interruption type
See the data as a table
No interruptions (score (1-20))Same-topic interruptions (score (1-20))Different-topic interruptions (score (1-20))
Mental workload10.0210.8311.5
Stress6.929.469.13
Frustration4.736.636.48
Time pressure11.0212.6912.17
Effort9.511.0411.52

Source: Mark, Gudith & Klocke, CHI 2008, Table 3 · Modified NASA-TLX, 48 participants. All differences across conditions significant.

The Cost of Interrupted Work: More Speed and StressGloria Mark, Daniela Gudith, Ulrich Klocke · Proceedings of CHI 2008

A lab simulation of office email work with 48 participants (15 men, 33 women, mean age 26). Each person answered 12 emails under three conditions: no interruptions, interruptions every two minutes on the same topic, or on a different topic, delivered by phone or instant message. Interrupted participants completed the task faster (about 20.3 to 20.6 minutes vs 22.77) with no loss in accuracy or politeness, but wrote shorter emails and reported higher stress (about 9.1 to 9.5 vs 6.9 on a 20-point scale), frustration, time pressure and effort. The limit: a short, simple task with mostly student participants; the authors note they could not test whether the stress accumulates over a full working day.

Read the paper (free)

Are heavy multitaskers better at switching?

No, the evidence points the other way. Eyal Ophir, Clifford Nass and Anthony Wagner at Stanford built a media multitasking index and compared heavy and light media multitaskers on standard tests of cognitive control in a 2009 PNAS paper.

Heavy multitaskers were more distracted by irrelevant stimuli in their surroundings and by irrelevant items in memory, and, surprisingly, did worse on a task-switching test, likely because they were worse at filtering out the task set they were supposed to drop. Practice at switching did not make them better switchers; it seemed to make them leakier.

That study is correlational, so it cannot say whether multitasking causes weaker filtering or whether people who filter poorly drift towards multitasking. Either way, the self-image of the skilled multitasker does not survive the data. For how this plays out with phones and studying, see why you can scroll but can't study, and for the reward-learning side of the pull, how dopamine builds habits.

Telephone operators in headsets seated in a row at a switchboard in Anamosa, Iowa, in 1915
Telephone operators at a switchboard. Each incoming call meant dropping one connection task and loading another, the same goal-shift and rule-load cycle the lab studies measure.Photo: Donated by Anamosa Library and Learning Center, CC BY-SA 4.0, via Wikimedia Commons

Why does switching feel so tiring?

Because each switch asks the executive system to do two jobs, shifting the goal and loading new rules, while residue from the last task occupies limited working memory. Mark's 2008 data show the subjective side: within 20 minutes, interrupted work feels more effortful and stressful even when output looks the same. Over a day of hundreds of switches, that becomes the familiar end-of-day exhaustion with little to show for it.

Attention spans on screens

Screen behaviour has moved in the wrong direction. In a June 2026 interview on TED's ReThinking podcast, Mark gave her team's measurements of time on one screen before switching:

  • 2003-04: about 2.5 minutes;
  • 2012: about 75 seconds;
  • later studies: roughly 44 to 50 seconds, averaging about 47 seconds.
2.5 minon one screen, 2003-04
75 son one screen, 2012
47 saverage in later studies

She also argued against cold-turkey digital detoxes, comparing them to crash diets, and favoured lasting changes to daily routines.

Popular claimWhat the data saySource
"It takes 23 minutes to refocus after an interruption"Interrupted work was resumed after about 25 minutes on average, mostly spent on 2.26 other tasks; it is a delay to return, not a refocus timeMark et al. 2005
"Multitaskers get good at switching"Heavy media multitaskers did worse on a task-switching testOphir et al. 2009
"Interruptions make you slower"In a 20-minute email task, people got faster but more stressedMark et al. 2008
"Finishing a task frees your mind"Finishing without time pressure still left residueLeroy 2009

How many switches is your day making? Use the calculator

Use this to check what you just read against your own day. Put in your screen hours and how long you usually stay on one window; the last two lines apply Mark's 2005 average wait to the interruptions you count.

Your screen switches and waiting work

What reduces attention residue?

A short written plan for where you will resume is the best-tested fix. In a 2018 paper in Organization Science, Leroy and Theresa Glomb ran four studies, including a survey of 202 professionals and lab studies with 66 and 44 students.

When people expected to return to an interrupted task under time pressure, residue rose and their work on the interrupting task suffered. Writing a brief ready-to-resume plan before switching cut the residue, and those participants made better hiring decisions and remembered more from the résumés they reviewed. The University of Washington's summary says even a minute's work will do.

  1. Write where you stopped. One line on the current state of Task A.
  2. Write where you will restart. The exact next action, so the goal-shift stage has a cue.
  3. Note what is unresolved. Open questions go on paper instead of staying in working memory.
  4. Batch the shallow work. Group messages and calls into fixed windows instead of answering every ping.
  5. Protect central work. Mark's data show central tasks get resumed; peripheral ones drift. Put the important block first.
Artist's impressionBefore I switchStopped at:intro done, methods half writtenRestart with:write the sample-size sentenceStill open:which table to cite?
Artist's impression of a ready-to-resume note: three short lines, written in under a minute before you leave the task.

A full routine built on these ideas is in a deep work focus routine. For the wider Manthan series on attention and the mind, start at Manthan.

Questions people ask

Is attention residue the same as multitasking?

No. Multitasking is trying to do two things at once; attention residue is what happens after a switch, when thoughts about the previous task linger. You can have residue even if you never multitask, simply by moving between tasks without closing the first one mentally.

Does it really take 23 minutes to get back on task?

That figure is a misreading. Mark's 2005 field study found interrupted work was resumed after about 25 minutes on average, but people spent that time on two or more other tasks. It measures how long important work waits, not how long the brain takes to refocus.

Why can I focus for hours on games but not on study?

Games give frequent, unpredictable rewards and clear next actions, so there is little residue and strong pull. Study often starts with an open, vague task and a slow payoff. Making the next action explicit and removing competing cues helps more than willpower.

Are some interruptions less harmful?

In Mark's 2008 lab study, interruptions on the same topic and on a different topic both raised stress; mental workload was highest for unrelated interruptions. Rubinstein's data add that a clear cue for what comes next reduces the switching cost.

Sources

  1. Leroy S. 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 (2009)
  2. Leroy S, Glomb TM. Tasks interrupted: how anticipating time pressure on resumption of an interrupted task causes attention residue. Organization Science 29(3):380-397 (2018)
  3. Rubinstein JS, Meyer DE, Evans JE. Executive control of cognitive processes in task switching. J Exp Psychol: Human Perception and Performance 27(4):763-797 (2001)
  4. Mark G, Gonzalez VM, Harris J. No task left behind? Examining the nature of fragmented work. CHI 2005 (2005)
  5. Mark G, Gudith D, Klocke U. The cost of interrupted work: more speed and stress. CHI 2008 (2008)
  6. Ophir E, Nass C, Wagner AD. Cognitive control in media multitaskers. PNAS 106(37):15583-15587 (2009)
  7. strategy+business: summary of Leroy (2009) on attention residue
  8. University of Washington News: Task interrupted? A plan for returning helps you move on (2018)
  9. TED ReThinking with Adam Grant: Why a digital detox won't reset your attention span, with Gloria Mark (2026)
  10. Westbrook J.I. et al.: Association of interruptions with an increased risk and severity of medication administration errors, Archives of Internal Medicine, as reported by ScienceDaily (98 nurses, 4,271 administrations, 12.1% and 12.7% per interruption) (2010)
attention residuetask switchingSophie LeroyGloria Markinterruptionsmultitaskingdeep workswitch cost
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Mindblog.

Keep reading in Mindblog

From the other blogs