The study closest to the 23-minute refocus claim timed something else

A coworker leans over your desk with a quick question. Two minutes later you turn back to your spreadsheet, and a line you’ve seen on a dozen productivity blogs comes to mind: it takes 23 minutes to regain your focus after an interruption.

That number is everywhere, yet the published study that comes closest to it measured something else. That 2005 study followed 24 people at one company and timed how long they took to return to an interrupted project, usually after handling other work in between.

The real research is more useful than the myth. In a controlled test, interrupted people worked faster, not slower, and paid for it in stress. On hospital wards, each interruption came with more medication errors. And over a whole career, jobs that make people think hard are linked with sharper thinking in old age.

The study closest to the 23-minute claim

In 2005, Gloria Mark, Victor M. Gonzalez and Justin Harris presented a paper called “No Task Left Behind?” at CHI, a major computing conference. They shadowed 24 information workers at one company: seven managers, nine analysts and eight developers. Over more than 700 hours, they logged every shift between what the authors called working spheres, meaning the projects or themes that make up a job.

Three findings matter here:

• People spent an average of 11 minutes and 4 seconds in one working sphere before switching or being interrupted.
• About 77% of interrupted work was picked up again the same day.
• When it was, the average wait was 25 minutes and 26 seconds, with an average of 2.26 other projects handled in between. Work interrupted by someone else was resumed after about 22 minutes and 37 seconds; work people broke off themselves, after about 29 minutes. That difference fell just short of statistical significance.

That 22 minutes and 37 seconds rounds to 23, the nearest number in this research to the popular claim. But it measures something else.

First, the clock measured when people came back to a project, not when their concentration recovered. During those minutes they were doing other work. Second, the spread was enormous. The standard deviation, a measure of how much individual cases varied, was 54 minutes and 48 seconds, more than twice the average. Some returns took moments and others took hours, so the average hides a very wide range.

A 2008 paper that is often cited for the statistic, by Mark with Daniela Gudith and Ulrich Klocke, doesn’t contain it at all.

What interruptions cost, from an inbox to a hospital ward

That 2008 study is still worth knowing, because it found something the myth hides. Forty-eight people, most of them German university students with an average age of 26, played a human resources manager answering a batch of emails. In some rounds, a phone call or instant message interrupted them about every two minutes.

Interrupted people finished faster: about 20.3 and 20.6 minutes in the two interruption conditions, compared with 22.8 minutes without interruptions. Their emails had no more errors and were no less polite. But on a standard workload questionnaire scored from 1 to 20, their stress rose from about 6.9 to more than 9. Frustration, time pressure and effort climbed too. The authors’ interpretation: people who expect interruptions speed up to compensate for time they expect to lose, and the strain is the price.

Stakes change the picture. In a 2010 study in Archives of Internal Medicine, Westbrook and colleagues observed 98 nurses giving 4,271 medications over 505 hours at two teaching hospitals in Sydney, Australia. Interruptions happened in about half of all administrations. Each interruption was associated with a 12.7% increase in clinical errors, such as a wrong dose, and a 12.1% increase in procedural failures. The estimated risk of a major error was 2.3% with no interruptions and 4.7% with four. Nurse experience offered no protection.

That study was observational, so it shows association rather than proof, but the link held regardless of hospital or nurse characteristics. The two findings don’t conflict. One is a short, low-stakes lab task measured by speed; the other is a high-stakes procedure where a single slip matters. Interruptions can leave routine output looking fine while making errors more likely in work that leaves no room for them.

How often work is broken up, and what each switch costs

Microsoft’s count of the “infinite workday”

In June 2025, Microsoft published figures from anonymized activity data on people using its Microsoft 365 software, gathered through February 2025 and excluding education and European Union customers, alongside a survey of 31,000 workers in 31 markets. Treat these as company data about its own product’s users, not peer-reviewed research.

The headline was that users were interrupted by a meeting, email or notification every two minutes during core work hours. The report’s methodology explains that the two-minute figure assumes an eight-hour day, while the often-repeated count of 275 interruptions a day covers a full 24 hours. The report also found 29% of active workers back in their inboxes by 10 p.m., meetings after 8 p.m. up 16% over the previous year, and an average of 58 messages per user outside core hours.

One figure was garbled almost immediately. The report says 40% of people who are online at 6 a.m. are reviewing email. A Microsoft Switzerland press release headlined it as 40% of employees checking email before 6 a.m., a much bigger claim.

Office design shifts interruptions rather than removing them. When two corporate headquarters moved to open-plan layouts, Ethan S. Bernstein and Stephen Turban of Harvard found that face-to-face interaction fell about 70% while email and instant messaging rose. They measured interaction, not concentration, so the study says where the interruptions went, not what they did to anyone’s focus.

Switch costs measured in the lab

Every switch between tasks costs a little time. The American Psychological Association’s explainer on multitasking, published in 2006, says each switch can cost as little as a few tenths of a second, and that those fractions add up when people switch back and forth repeatedly.

The 2001 study it describes, by Rubinstein, Meyer and Evans, ran four experiments; the first author worked at the Federal Aviation Administration’s technical center. Switching costs grew as the rules of a task got more complex, shrank when a cue signaled which task came next, and were larger when switching into an unfamiliar task than into a familiar one.

The “40 percent” figure

Many articles say multitasking wastes 40% of your productivity. The trail leads to the same 2006 APA page, which reports that researcher David Meyer has said brief mental blocks from shifting between tasks “can cost as much as 40 percent of someone’s productive time.” That is a researcher’s estimate of an upper limit, reported secondhand, with no study attached on the page. The measured finding is narrower: switch costs are real, and they grow with complexity and unfamiliarity.

Too much at once: cognitive load

Cognitive load theory helps explain why some interruptions hurt more than others. A 2014 guide for medical educators by Young and colleagues in Medical Teacher describes working memory, the mental workspace that holds whatever you’re actively using, as able to process only a limited number of pieces of information at a time.

The theory divides mental load three ways. Intrinsic load is the difficulty built into the task. Extraneous load is effort spent on things that don’t serve the task, such as a confusing form. Germane load is effort that builds understanding. When the total exceeds working memory’s capacity, performance and learning suffer, and tasks with many pieces that must be handled together, such as a complex diagnosis, are where that ceiling is easiest to hit.

The guide adds a twist, the expertise reversal effect: methods that help beginners can interfere once people become skilled, so a step-by-step worked example that steadies a trainee can get in a veteran’s way.

Demanding work and the aging brain

Here the story turns. Interruptions are a short-term cost, but cognitively demanding work itself is linked with better thinking decades later.

• Seven cohorts, three regions (BMJ, 2021). Kivimäki and colleagues pooled 107,896 people in the UK, elsewhere in Europe and the US, followed for 13.7 to 30.1 years, during which 1,143 developed dementia. Dementia occurred at 4.8 cases per 10,000 person-years among people with highly stimulating jobs and 7.3 among those with low stimulation. The researchers adjusted for education, smoking, heavy drinking, inactivity, job strain, obesity, high blood pressure and diabetes, plus heart disease, stroke or diabetes arising before any dementia diagnosis. The risk was still about 18% lower (a hazard ratio of 0.82).
• Careers tracked from 30 to 65 (Neurology, 2024). Edwin and colleagues traced the occupations of 7,003 older Norwegians in the HUNT4 70+ study, rating how routine each person’s work was from age 30 to 65. Compared with the least routine careers, the most routine carried a 74% higher relative risk of mild cognitive impairment and a 37% higher relative risk of dementia, after accounting for age, sex and education. When the team also adjusted for income and health, the impairment link held, but the dementia estimate’s range then included no difference at all.
• Before and after retirement (Journal of Occupational Health Psychology, 2014). Fisher and colleagues used the national Health and Retirement Study, following 4,182 Americans from 1992 to 2010. Jobs with higher mental demands were linked with better cognitive functioning before retirement and a slower rate of decline after it, with education, income, demographics and health taken into account.

All three studies are observational. Nobody can randomly assign careers, so people who start out with sharper thinking may be more likely to land and keep complex jobs, and adjustments for education and income can’t capture everything. The fair verdict is a consistent association across several countries, not proof that a harder job prevents dementia. The BMJ team also found lower levels of three blood proteins tied to dementia risk among people in stimulating jobs, a clue about mechanism rather than an outcome anyone would notice.

Note what these studies measured: how mentally demanding a job was, not how often its days were broken up. Demanding work and a fragmented day are different things, and the long-term association attaches to the first. The wider evidence on what raises dementia risk is laid out in the library’s article on the risk factors people rarely guess.

Right-to-disconnect rules, and why the US has none

Several governments now regulate after-hours contact. The rules differ more than headlines suggest:

Place	In force	Who is covered	What the rule does
France	January 1, 2017	Companies with 50 or more employees	Requires negotiation with employee representatives over disconnecting; if they can’t agree, the employer issues a charter that includes training on reasonable use of digital tools
Australia	August 26, 2024, for employers with 15 or more staff; August 26, 2025, for smaller businesses	Employees of covered employers	Gives employees a right to refuse to monitor, read or respond to contact outside working hours, unless refusing is unreasonable
Ontario, Canada	First deadline June 2, 2022	Employers with 25 or more employees on January 1	Requires a written policy on disconnecting; the employer decides what it says

Australia’s Fair Work Ombudsman sums up its rule in one sentence: “Employees have the right to refuse contact outside their working hours unless that refusal is unreasonable.” Whether a refusal is unreasonable depends on the reason for the contact, how it’s made and how disruptive it is, whether the employee is paid to be available, the employee’s role, and personal circumstances such as caring duties. Refusing is always unreasonable if the law requires the contact. Nothing in the rule bars employers from making contact in the first place.

Ontario’s own guide says its law doesn’t require employers to create a new right to disconnect. France’s 2016 labor law made disconnection a subject of mandatory bargaining rather than a fixed limit. All three are rules about workplace expectations, not tested health measures, and their effect on concentration has not been established.

The United States has no federal right to disconnect, and a September 2025 review by the law firm Akerman found no state or local law in effect either. Bills keep appearing. California’s AB 2751, which would have let workers ignore non-emergency messages outside agreed hours, was held in committee on May 16, 2024, and died there. Vermont’s H.263 was referred to a House committee in February 2025, and New Jersey’s A3416 was pre-filed for the 2026 session. The legislative records for both showed them at an early stage.

Federal workplace health guidance stops short of limits. The National Institute for Occupational Safety and Health’s Total Worker Health framework lists work arrangements and work organization among its priority areas, but it’s a framework rather than a rule, and its overview doesn’t address cognitive workload specifically.

Structuring demanding work, graded by the evidence

This guide is built from the studies above, and each item says how strong its support is.

• Batch email instead of living in the inbox. Tested in one field experiment. In a 2015 study in Computers in Human Behavior, Kostadin Kushlev and Elizabeth W. Dunn of the University of British Columbia had 124 adults spend one week checking email three times a day with alerts off and another week checking as often as they liked. Daily stress was lower in the limited week, and so was tension during important activities; overall stress for the week didn’t change. The authors called the work exploratory and warned that jobs needing constant updates could suffer. Verdict: preliminary.
• Protect blocks of time for your most complex work. This follows from lab findings, not workplace trials: switch costs grow with complexity, and working memory has a hard ceiling. A policy of protected focus blocks hasn’t been tested in the research described here. Verdict: a reasonable inference, untested.
• Leave yourself a cue when you stop. In the 2001 experiments, cues about which task came next shrank switch costs. Jotting a one-line note about your next step before you answer an interruption extends that idea; it wasn’t what the study tested.
• Shield safety-critical steps first. The clearest evidence of harm comes from medication rounds, where errors climbed with each interruption. If your work has steps where one slip matters, those are the ones to protect. Verdict: observational evidence of harm; specific protective methods weren’t tested in these sources.
• Strip out extraneous load. Clear templates, fewer tools and instructions that don’t need decoding reduce the part of mental effort that adds nothing. The support comes from cognitive load theory in education research rather than from workplace trials.

When it is more than a busy job

Workload explains a lot of scattered thinking, but not all of it. These signs point beyond the job:

• Trouble concentrating that persists on weekends and vacations, or keeps getting worse.
• Concentration problems that arrive with low mood or loss of interest most of the day, nearly every day, for two weeks or more. On the National Institute of Mental Health’s list of depression symptoms, trouble concentrating, remembering or deciding sits alongside low mood. If you’re having thoughts of suicide, call or text 988 to reach the 988 Suicide & Crisis Lifeline; for a life-threatening emergency, call 911.
• New forgetfulness that worries you or the people close to you. Memory changes and when they call for assessment are covered elsewhere in the library.
• A change that began around the time you started a new medication. Whether a drug is behind it is a question for the prescriber.
• Exhaustion and cynicism about the job, which belong to burnout research more than to cognitive load.

Whether persistent concentration trouble reflects workload, depression, a sleep disorder or a medication effect is a decision for a clinician who can evaluate you, not for a time-management system.

The number to replace the 23 minutes

If you want a statistic about interruptions, use one the evidence supports. In a careful lab test, interrupted people sped up, kept their quality and reported about a third more stress. On hospital wards, errors rose with every interruption. The cost of a break in your work isn’t a fixed 23 minutes; it scales with how complex the task is and how much is riding on it. So rank your work by those two things, guard the top of the list from interruptions first, and batch the messages that can wait.

Sources

1. Association for Computing Machinery (Proceedings of the SIGCHI Conference on Human Factors in Computing Systems). “Mark G, Gonzalez VM, Harris J. "No Task Left Behind? Examining the Nature of Fragmented Work." CHI 2005, Portland, Oregon, pp. 321-330. DOI: 10.1145/1054972.1055017.” doi.org
2. Association for Computing Machinery (Proceedings of the SIGCHI Conference on Human Factors in Computing Systems). “Mark G, Gudith D, Klocke U. "The Cost of Interrupted Work: More Speed and Stress." CHI 2008, Florence, Italy, pp. 107-110. DOI: 10.1145/1357054.1357072.” doi.org
3. Archives of Internal Medicine. “Westbrook JI, Woods A, Rob MI, Dunsmuir WT, Day RO. "Association of interruptions with an increased risk and severity of medication administration errors." Arch Intern Med, 2010. PubMed. DOI: 10.1001/archinternmed.2010.65.” pubmed.ncbi.nlm.nih.gov
4. Microsoft WorkLab. “Breaking down the infinite workday (Work Trend Index Special Report, June 17, 2025).” microsoft.com
5. Microsoft Switzerland. “New Microsoft Study Reveals the Rise of the Infinite Workday (press release, June 17, 2025).” news.microsoft.com
6. Philosophical Transactions of the Royal Society B. “Bernstein ES, Turban S. "The impact of the 'open' workspace on human collaboration." Philos Trans R Soc Lond B Biol Sci, 2018. PubMed. DOI: 10.1098/rstb.2017.0239.” pubmed.ncbi.nlm.nih.gov
7. American Psychological Association. “Multitasking: Switching costs (March 20, 2006).” apa.org
8. Journal of Experimental Psychology: Human Perception and Performance. “Rubinstein JS, Meyer DE, Evans JE. "Executive control of cognitive processes in task switching." J Exp Psychol Hum Percept Perform, 2001. PubMed. DOI: 10.1037//0096-1523.27.4.763.” pubmed.ncbi.nlm.nih.gov
9. Medical Teacher. “Young JQ, Van Merrienboer J, Durning S, Ten Cate O. "Cognitive Load Theory: implications for medical education: AMEE Guide No. 86." Med Teach, 2014. PubMed. DOI: 10.3109/0142159X.2014.889290.” pubmed.ncbi.nlm.nih.gov
10. BMJ. “Kivimäki M, Walker KA, Pentti J, et al. "Cognitive stimulation in the workplace, plasma proteins, and risk of dementia: three analyses of population cohort studies." BMJ, 2021. PubMed. DOI: 10.1136/bmj.n1804.” pubmed.ncbi.nlm.nih.gov
11. Neurology. “Edwin TH, Håberg AK, Zotcheva E, et al. "Trajectories of Occupational Cognitive Demands and Risk of Mild Cognitive Impairment and Dementia in Later Life: The HUNT4 70+ Study." Neurology, 2024. PubMed. DOI: 10.1212/WNL.0000000000209353.” pubmed.ncbi.nlm.nih.gov
12. Journal of Occupational Health Psychology. “Fisher GG, Stachowski A, Infurna FJ, Faul JD, Grosch J, Tetrick LE. "Mental work demands, retirement, and longitudinal trajectories of cognitive functioning." J Occup Health Psychol, 2014. PubMed. DOI: 10.1037/a0035724.” pubmed.ncbi.nlm.nih.gov
13. Fair Work Ombudsman (Australian Government). “New right to disconnect laws (last updated August 26, 2025).” fairwork.gov.au
14. Government of Ontario. “Written policy on disconnecting from work, Your guide to the Employment Standards Act (updated July 8, 2024).” ontario.ca
15. Actu-Juridique. “L'émergence du droit à la déconnexion en droit du travail (March 1, 2017).” actu-juridique.fr
16. Akerman LLP. “Yaqoob MA. The Right to Disconnect in the US: What Employers Need to Know About Emerging Proposals (September 10, 2025).” akerman.com
17. CalMatters Digital Democracy. “AB 2751: Employer communications during nonworking hours (2023-24 session; status: failed).” calmatters.digitaldemocracy.org
18. Vermont General Assembly. “H.263, An act relating to creating a right for employees to disconnect from work (bill record).” legislature.vermont.gov
19. New Jersey Legislature. “Assembly No. 3416 (222nd Legislature, pre-filed for the 2026 session).” pub.njleg.gov
20. National Institute for Occupational Safety and Health. “About the Total Worker Health Approach (last reviewed May 31, 2024).” cdc.gov
21. Computers in Human Behavior (Elsevier). “Kushlev K, Dunn EW. "Checking email less frequently reduces stress." Computers in Human Behavior, 2015;43:220-228. DOI: 10.1016/j.chb.2014.11.005.” doi.org
22. National Institute of Mental Health. “Depression (NIH Publication No. 24-MH-8079, revised 2024).” nimh.nih.gov