Notifications are the easy part

A considered reading of the research on interruption and the phone — the recovery figure that appears in no paper, the randomised trial that disabled notifications for a week and measured nothing, and what is left of the advice once you count who is actually doing the interrupting.

---

There is a settings screen on every phone that promises to fix this. Notifications, off. It is the first recommendation in almost every article about focus, the opening move of every productivity system, and the thing people mean when they say they have taken their attention back. It is free, it takes about ninety seconds, and it is reversible, which is a rare combination in behaviour change and part of why the advice travels so well.

The reasoning behind it is sound as far as it goes. Interruptions carry a real cost, that cost has been measured carefully for decades in settings where the stakes are considerably higher than an unanswered email, and a phone is a device purpose-built to interrupt. If the interruption is the problem and the phone is the source, removing the source should remove the problem.

The difficulty is what happened when somebody finally ran that as an experiment. In 2025, a preregistered randomised controlled trial put two hundred and five people through a week with push notifications disabled, logged their actual phone behaviour rather than asking them about it, and found that checking frequency did not change, screen time did not change, and the only reliable effects were that participants felt their use had become more deliberate and that they became more anxious about what they were missing.

That result is not an outlier. It is what most of the literature predicts once you look at the question underneath the advice — not does an interruption cost something, which is settled, but who is sending the interruptions. On the best available measurement, taken from first-person video of people going about ordinary days, the answer is that eighty-nine per cent of the time it is you.

What follows separates the interruption from the notification, because they are not the same thing and the evidence behind them is not the same strength. One is among the better-established findings in applied psychology. The other is a much smaller effect that has been asked to carry a great deal of weight.

The number that is not in any paper

Start with the figure everyone quotes. It takes twenty-three minutes and fifteen seconds to recover from an interruption. The number appears in business journalism, in productivity books, in corporate training material, and in the marketing copy of roughly every focus application ever shipped. It is almost always attributed to Gloria Mark, a computer scientist at the University of California, Irvine, who has spent more than two decades instrumenting how knowledge workers actually spend attention.

The attribution is half right. Mark's work is real, careful, and central to this subject. The number is not in it.

The study usually cited is Mark, Gonzalez and Harris's 2005 paper on fragmented work, which followed twenty-four information workers at a financial services firm through more than seven hundred hours of observation, timing every activity to the second (Mark, Gonzalez, & Harris, 2005). It found that people spent an average of eleven minutes and four seconds on a coherent unit of work before switching or being interrupted, that 57.1 per cent of those units were interrupted, and that 77.2 per cent of interrupted work was resumed the same day. When work was resumed on the same day, it took an average of twenty-five minutes and twenty-six seconds, and people passed through an average of 2.26 other activities on the way back.

Twenty-five minutes and twenty-six seconds. Not twenty-three fifteen. And the paper says something about that number that the popular version never carries: the standard deviation was fifty-four minutes and forty-eight seconds, which is larger than the mean. The distribution is enormously wide. Treating it as a fixed tax on every interruption — that ping just cost you twenty-three minutes — is a misreading of a long-tailed distribution as a constant.

The other paper frequently cited for the figure, Mark, Gudith and Klocke's 2008 experiment with forty-eight participants, does not contain it either, and reports something closer to the opposite of what it is usually enlisted to support. Interrupted tasks were completed in less time than uninterrupted ones, with no measurable difference in quality. The cost was not in the clock. It was in what people reported afterwards: significantly more stress, more frustration, more time pressure and more effort (Mark, Gudith, & Klocke, 2008). Interruption did not slow the work down. It made the work worse to do.

This is not a pedantic complaint about a rounded number. The twenty-three-minute figure is the single most load-bearing statistic in public discussion of attention, and it is a statistic with no study behind it, misattributed to a researcher whose actual findings are more interesting and less convenient. It is worth holding onto as a warning about how this literature travels: the version that circulates is consistently more dramatic, more precise-sounding, and more actionable than the version that was measured.

What an interruption actually costs

Having deflated the number, it is necessary to be fair to the finding, because the underlying effect is not in doubt. Interruption is one of the more robustly demonstrated costs in applied cognition, and the laboratory work is unusually clean.

The sharpest demonstration is Altmann, Trafton and Hambrick's 2014 experiments, which put three hundred undergraduates through a task requiring them to keep their place in a sequence of steps, then interrupted them briefly and without warning. Interruptions averaging 2.8 seconds — roughly the time it takes to perform one step — doubled the rate of sequence errors on the trials that followed. Interruptions averaging 4.4 seconds tripled them (Altmann, Trafton, & Hambrick, 2014).

The critical detail is what did not change. Errors unrelated to sequence position showed no interruption effect at all. Attention did not globally collapse; people were not generally worse at the task. What broke was specifically the memory for where they were, which is the fragile part of any procedure performed in order. A few seconds was enough to lose the place, and losing the place was enough to produce an error that looked, from the outside, like carelessness.

Sophie Leroy's work identifies a complementary mechanism. When people switch tasks, some portion of attention stays behind on what they were doing, and performance on the new task suffers until the residue dissipates — particularly when the previous task was left unfinished (Leroy, 2009). An interruption, by definition, leaves something unfinished.

Mark, Gudith and Klocke's finding that interrupted work finished faster fits this better than it first appears. Their participants were answering emails under interruption from a supervisor; the interrupted groups compensated by working more quickly and writing less, arriving at the same quality in less time. Nothing was lost that a stopwatch could see. What the workload measures picked up was the price of that compensation — the stress, the frustration, the time pressure, the effort (Mark, Gudith, & Klocke, 2008). An interruption is not necessarily a hole in the day. It can be a surcharge on the same day, paid in a currency that productivity metrics do not record.

Where this matters most is not the office. Westbrook and colleagues observed ninety-eight nurses across six wards at two Sydney teaching hospitals, watching 4,271 medication administrations to 720 patients over 505 hours, and matched what they saw against the patients' charts. Each interruption was associated with a 12.1 per cent increase in procedural failures and a 12.7 per cent increase in clinical errors. Administrations with no interruptions carried a clinical error rate of 25.3 per cent; with three interruptions the rate was 38.9 per cent. Severity scaled the same way: the estimated risk of a major error rose from 2.3 per cent with no interruptions to 4.7 per cent with four. Nurse experience provided no protection (Westbrook, Woods, Rob, Dunsmuir, & Day, 2010).

So the interruption literature stands. It stands most strongly for sequential, procedural work where losing your place has consequences, and the mechanism is specific — place-keeping failure, not a general draining of attention. Hold that shape in mind, because the next question is whether a phone notification is an interruption of that kind, and how much of one.

The notification experiments, read closely

The foundational study here is Stothart, Mitchum and Yehnert's 2015 paper, which did something genuinely clever. Participants performed a sustained attention task while receiving calls or texts on their own phones, which they did not interact with at any point during the task. They could not answer. They simply received the notification and kept working.

Performance degraded anyway. The authors report the disruption as comparable in magnitude to what is seen when people actively take a call or read a message (Stothart, Mitchum, & Yehnert, 2015). The cost was imposed by the signal, not by the response to it, which is a striking result and deserves its influence. It has been cited several hundred times.

The mechanism the authors propose is worth noting, because it is more modest than the way the finding is usually reported. They do not argue that the notification seizes attention directly. They argue that it prompts task-irrelevant thought — wondering who it was, whether it matters, whether to deal with it later — and that mind-wandering of this kind is already known to damage performance. On that account the alert is not an interruption so much as the seed of one, and the interruption proper is something the participant then performs internally. This will turn out to be the shape of the whole subject.

It has also held up reasonably well in kind, which is more than can be said for some findings in this area. Kaminske and colleagues ran 105 participants through a Stroop task across five conditions and found notifications reliably increased completion time, regardless of whether the phone was the participant's own and regardless of task difficulty (Kaminske, Brown, Aylward, & Haller, 2022). Nason and Wilbiks reproduced the overall effect on sustained attention and added a wrinkle: visual notifications produced more errors than auditory ones (Nason & Wilbiks, 2025). Upshaw and colleagues measured the electrophysiology and found slower responses on trials paired with smartphone sounds — but also a larger N2 component, indicating that participants were recruiting more cognitive control rather than less (Upshaw, Stevens, Ganis, & Zabelina, 2022). That is a real complication for the simple story. The brain is not being passively hijacked; it is spending something to stay on task.

The most useful recent work puts a duration on the effect. Fournier and colleagues had participants perform a Stroop task while receiving smartphone-style notifications, tracking pupil dilation alongside response times. The disruption was a transient slowdown lasting roughly seven seconds, scaled by how relevant the notification seemed, and — importantly — predicted by how frequently a person interacted with their phone rather than by how long they spent on it (Fournier et al., 2026).

Seven seconds. Doubled sequence errors in Altmann's task. These are real effects, measured properly. They are also effects of a particular shape: brief, acute, occurring on tasks specifically engineered to detect momentary lapses, delivered at unpredictable intervals to people who are not permitted to resolve them. Under those conditions, an interruption you cannot act on is close to a worst case.

None of which tells you what happens across an ordinary day, where the interruption is usually resolvable in three seconds, where the task is rarely a sustained-attention paradigm, and where — as the next sections establish — the notification is not where most of the interruptions are coming from.

A cautionary detour: the phone that just sits there

Before going further, a case study in how this literature can mislead, because it bears directly on how much confidence to place in the next step.

In 2017, Ward, Duke, Gneezy and Bos published what became one of the most widely repeated findings about phones and cognition. Participants completed working-memory and fluid-intelligence measures with their own silenced phone either on the desk, in a bag, or in another room entirely. Performance improved as the phone got further away. They called it brain drain: the mere presence of the device consuming attentional resources through the effort of ignoring it (Ward, Duke, Gneezy, & Bos, 2017).

It is a clean design with an arresting conclusion, and it was reported everywhere. This publication has cited it before, describing the replication picture as mixed. That description now needs updating, and in one direction.

Ruiz Pardo and Minda ran a preregistered direct replication of the key experiment, with the full six-condition design, and did not reproduce the gradient (Ruiz Pardo & Minda, 2022). Two meta-analyses have since assessed the whole body of work. Böttger, Poschik and Zierer pooled twenty-two studies yielding forty-three effects and found a significant effect on memory (g = −0.23) but nothing on attention (g = −0.07) or general cognitive performance (g = 0.10) (Böttger, Poschik, & Zierer, 2023). Parry's systematic review and six meta-analyses, covering fifty-six effects and 7,093 participants, found exactly one significant pooled effect — on working memory capacity — with everything else null, alongside substantial methodological heterogeneity and generally poor statistical power (Parry, 2024).

Two further experiments point the same way. Kaminske and colleagues, whose notification effect was solid, specifically did not find that mere presence contributed to distraction. Christodoulou and Roussos randomised just over 140 young adults to a visible phone on the desk or no phone at all, measured attention with the Attention Network Test, and found no difference in either response time or errors (Christodoulou & Roussos, 2025).

The most-shared claim in this field is the one that has survived scrutiny worst. That is the relevant lesson, and it is not an argument that phones are harmless. It is an argument for asking, of any confident statement here, how it performed when somebody tried to repeat it.

Who is interrupting whom

Now the question that reframes the whole subject.

Maxi Heitmayer and Saadi Lahlou fitted thirty-seven people with first-person wearable cameras and recorded them through ordinary life — work, commuting, meals, social time — producing over two hundred hours of video containing 1,130 distinct smartphone interactions. They then sat participants down in front of their own footage and asked them to account for what they had done. It is the closest thing the literature has to a direct observation of how phones actually enter a day.

Eighty-nine per cent of those interactions were initiated by the user. Eleven per cent were triggered by a notification (Heitmayer & Lahlou, 2021). The rhythm they describe is roughly one minute of phone every five minutes — people reaching for the screen about twelve times an hour, almost always without being asked to.

This is not a fringe result. It is consistent with everything that came before it. Mark's 2005 observational data found that 52 per cent of workplace interruptions were internal — self-initiated — against 48 per cent external. Of interrupted work that was resumed the same day, 90.1 per cent of the resumptions were initiated by the worker rather than prompted by anybody else (Mark, Gonzalez, & Harris, 2005). Oulasvirta and colleagues documented the mechanism a decade before the current discussion started, describing "checking habits" — brief, repetitive inspections of dynamic content, averaging around thirty-four sessions a day, reinforced by informational rewards available in under a second. Their participants described the pattern as an annoyance rather than an addiction, which is probably the more accurate word for it (Oulasvirta, Rattenbury, Ma, & Raita, 2012).

Two further findings sharpen the picture. Mark and colleagues logged computer activity in workplaces and found a median duration of focus on any one screen of forty seconds (Mark, Iqbal, Czerwinski, Johns, & Sano, 2016). And Andrews and colleagues compared what people estimated about their own phone use against two weeks of actual logs: estimates of duration were adequate, but estimates of the number of daily uses did not correlate with reality at all (Andrews, Ellis, Shaw, & Piwek, 2015).

That last one matters for everything that follows. People have some idea how long they are on their phones. They have no reliable idea how often they pick them up. Which means the behaviour most in need of changing is the one least available to introspection — and that any intervention evaluated by asking people how it went is measuring the wrong instrument.

The reinforcement structure behind that habit was the subject of the July essay on reward, and only one part of it needs restating here. Oulasvirta and colleagues' account of a checking habit is not about pleasure; it is about an informational reward that is quickly accessible — there almost the instant you look, every time, for almost no effort (Oulasvirta et al., 2012). Whatever is doing the maintaining, it is worth being clear about where it sits. Not in the notification. In what the screen shows when you look, which is the same whether the phone asked you to look or not.

For completeness: the notification load is real. Pielot, Church and de Oliveira logged a week of notifications for fifteen people and found an average of 63.5 per day, mostly from messengers and email (Pielot, Church, & de Oliveira, 2014). Sixty-three interruptions a day is not nothing. But against roughly a dozen self-initiated pick-ups an hour, it is the smaller stream.

The trial that turned the notifications off

Which brings us to the experiment this essay opened with.

Dekker, Baumgartner, Sumter and Ohme ran a preregistered randomised controlled trial with 205 participants, assigning some of them to disable push notifications for one week. Critically, they did not rely on self-report for the behavioural outcomes: phone use was objectively logged, with daily mobile diaries capturing the subjective side separately.

The intervention did not affect smartphone behaviour. Not checking frequency, not screen time. The null was not explained away by individual differences either — the pattern did not depend on participants' trait fear of missing out. It also produced no effect on perceived control over phone use, on perceived overuse, on smartphone vigilance, on productivity, or on smartphone-related distraction.

Two things did move. Perceived checking-habit strength decreased: participants experienced their own use as more intentional, even though the logs said the amount had not changed. And fear of missing out increased. The authors' summary is blunt — disabling notifications "results in drawbacks rather than improvements in digital well-being" — and their conclusion is that the findings challenge the assumption that notifications play a prominent role in driving smartphone use at all (Dekker, Baumgartner, Sumter, & Ohme, 2025).

This is one trial, one week, 205 people. It should not be treated as the final word, and a week is short enough that a habit could plausibly outlast the intervention and then reassert itself. But it is the only properly controlled test of the most widely given piece of attention advice in circulation, it was preregistered, and it measured behaviour instead of asking about it. On the standard this publication applies elsewhere — that a well-measured null outweighs a pile of self-reported positives — it carries considerable weight.

It also explains a pattern in the older literature that looked odd at the time. Iqbal and Horvitz asked twenty information workers to turn off email notifications for a week. Some checked email more frequently as a result. Every participant turned notifications back on when the study ended (Iqbal & Horvitz, 2010). Pielot and Rello's twenty-four-hour deprivation study found the same undertow: twelve of their thirty participants reported checking their devices more often than usual during the day without alerts, one of them leaving the screen on so as not to miss anything (Pielot & Rello, 2017).

Remove the signal and the habit does not stop. It goes looking.

What silence actually buys

That is not the whole story, and the rest of it is more sympathetic to the advice — with a consistent pattern in which measures move and which do not.

Kushlev, Proulx and Dunn put 221 participants through two contrasting weeks: one maximising phone interruptions, with alerts on and the device within reach and sight, and one minimising them, with alerts off and the phone away. Participants reported higher levels of inattention and hyperactivity during the alerts-on week, and higher inattention in turn predicted lower productivity and lower psychological well-being (Kushlev, Proulx, & Dunn, 2016). It is a substantial study with a genuine within-person design. It is also entirely self-report, and the comparison is not against normal use but against a deliberately maximised condition.

Pielot and Rello's participants, through a full day without notifications, reported feeling less distracted and more productive — with the authors themselves noting that the effect sizes suggested low practical significance — while also feeling unable to be as responsive as expected, which made some of them anxious, and feeling less connected to their social group. Two-thirds intended to change how they managed notifications afterwards; two years later, half were still doing so (Pielot & Rello, 2017).

The most instructive intervention study is Fitz and colleagues' randomised field experiment. Two hundred and thirty-seven participants were assigned to receive notifications as usual, batched hourly, batched three times a day, or not at all. The three-times-daily group came out ahead on nearly everything measured: more attentive, in a better mood, less stressed, more in control of their phones. Hourly batching was indistinguishable from the control. And the group that received no notifications at all got few of the benefits while reporting higher anxiety and higher phone-related fear of missing out (Fitz et al., 2019). The paper lost 31 per cent of its sample to attrition, which is worth knowing, and the outcomes are self-reported.

Wasmuth and colleagues ran a smaller field experiment giving participants specific instructions for reducing phone-related distraction. Hyperactivity symptoms fell. Inattention symptoms did not, and neither did working-memory accuracy (Wasmuth et al., 2022).

Lay these side by side and a shape emerges. Every study in which people are asked how the week felt finds improvement. The study that logged what people did found nothing. The study that measured working memory found nothing. This is the same measurement problem that ran through the August essay on sleep, arriving from a different direction: the subjective instruments move, the objective ones mostly do not, and the advice in circulation is built almost entirely on the former.

There is a reading of this that is not dismissive. Feeling less harassed by your own device is worth something on its own terms, and Dekker's finding that use came to feel more intentional without becoming less frequent is a real effect, not a placebo to be sneered at. But it should be described accurately. Turning notifications off changes your relationship to the phone. It does not appear to change how much you use it.

The part that is not about attention

There is a finding that recurs in every deprivation study here and is almost never included in the advice, and it is the most consistent result in this entire literature.

Pielot and Rello's participants, freed from notifications for a day, reported feeling less distracted — and also reported that they could no longer be as responsive as they were expected to be, which made some of them anxious. They felt less connected to their social group. One participant left their screen on so as not to miss a friend's message, because the friend would be annoyed. In Fitz's trial, the group receiving no notifications at all showed higher anxiety and higher phone-related fear of missing out. In Dekker's randomised trial — where nothing else moved at all — fear of missing out went up. Iqbal and Horvitz's information workers all turned their alerts back on at the end of the week.

Four studies, four designs, fifteen years apart, one result. Whatever else a notification does, it also discharges an obligation. It tells you that you have not missed anything, and it does so continuously, which is why removing it produces a specific and unpleasant sensation that is not about concentration at all.

This reframes what the advice is asking. "Turn off your notifications" sounds like a proposal about attention. Measured behaviourally, it is a proposal about availability: you are being asked to become less reachable, and to absorb the social cost of that, in exchange for a focus benefit that the one properly controlled trial could not detect. Some people will take that trade and be glad of it. It should at least be described as a trade.

It also explains why batching works where silence does not. Three scheduled deliveries a day preserve the reassurance — nothing has been missed for long — while removing the unpredictability. The intervention that performed best is the one that left the social function intact and changed only the timing (Fitz et al., 2019).

When you remove the phone entirely

One more result, because it closes the loop.

Heitmayer ran a within-participant study in which people completed two five-hour work sessions, video-recorded throughout, once with their phone at arm's reach and once with it on a table a metre and a half away. The manipulation worked exactly as intended on the phone: use fell from twenty-nine minutes to fifteen, and the number of separate interactions fell from 18.5 to 6.5.

Total time spent on leisure activities did not change. Computer leisure rose from twenty-seven minutes to one hour and six minutes (Heitmayer, 2025).

The sample is small — twenty-two participants — and it is a single study. But the direction is unsurprising given everything above. If the pull is originating in the person rather than the device, moving the device relocates the behaviour rather than removing it. The phone was never the only screen in the room.

What the evidence supports

Well supported. Interruptions impose real costs on sequential tasks, through a specific mechanism: loss of place rather than general attentional collapse. Interruptions of under three seconds double sequence errors (Altmann, Trafton, & Hambrick, 2014). In clinical work the costs are measurable in patient harm, rising with each additional interruption (Westbrook et al., 2010). Receiving a notification you cannot act on degrades laboratory task performance, on the order of seconds, and this has replicated across several paradigms (Stothart, Mitchum, & Yehnert, 2015; Kaminske et al., 2022; Nason & Wilbiks, 2025; Fournier et al., 2026).

Reasonably supported. Most phone pick-ups are self-initiated: eighty-nine per cent under direct observation (Heitmayer & Lahlou, 2021), with roughly half of all workplace interruptions internal in the earlier observational work (Mark, Gonzalez, & Harris, 2005). The checking pattern is habitual, frequent, and largely invisible to the person doing it — people cannot accurately estimate how often they pick up their phone (Oulasvirta et al., 2012; Andrews et al., 2015). Scheduled delivery outperforms both default notifications and total silence on how a fortnight feels (Fitz et al., 2019). Removing notifications reliably increases anxiety about responsiveness and fear of missing out — the most consistent finding across every deprivation study here (Pielot & Rello, 2017; Fitz et al., 2019; Dekker et al., 2025).

Weakly supported, or contradicted. That disabling notifications reduces phone use: contradicted by the only preregistered trial with objective logging, which found no change in checking frequency or screen time and an increase in fear of missing out (Dekker et al., 2025), and foreshadowed by earlier deprivation studies where people compensated by checking more (Iqbal & Horvitz, 2010; Pielot & Rello, 2017). That the mere presence of a phone drains cognition generally: one significant pooled effect on working memory across fifty-six effects and 7,093 participants, everything else null (Parry, 2024; Böttger, Poschik, & Zierer, 2023), with a preregistered replication failure (Ruiz Pardo & Minda, 2022) and two further nulls (Kaminske et al., 2022; Christodoulou & Roussos, 2025).

Not supported at all. That recovering from an interruption takes twenty-three minutes and fifteen seconds. No study reports this. The closest measured figure is twenty-five minutes and twenty-six seconds for same-day resumption of interrupted work, with a standard deviation larger than the mean (Mark, Gonzalez, & Harris, 2005), and the experimental follow-up found interrupted tasks completed faster (Mark, Gudith, & Klocke, 2008).

What to do, given the evidence

In descending order of how much weight the evidence behind them can bear:

1. Silence the phone for procedural work, and only make a rule about that. This is where the interruption effect is best established and where the consequences are worst: anything performed in sequence where losing your place produces an error. Hospitals have taken this seriously for fifteen years for good reason (Westbrook et al., 2010; Altmann, Trafton, & Hambrick, 2014). It is a narrow recommendation and it is the strongest one available.

2. Batch rather than abolish. Three scheduled deliveries a day beat both the default and total silence across nearly every measure in the one randomised comparison that tested all four (Fitz et al., 2019). Hourly batching did nothing. The interval matters more than the fact of batching, which suggests the mechanism is about predictability rather than volume.

3. Aim at the pick-up, not the alert. Nine in ten interactions are self-initiated (Heitmayer & Lahlou, 2021). Any intervention on notifications addresses the other one in ten by construction. What sustains a checking habit is a reward reliably available in under a second (Oulasvirta et al., 2012); the lever is what the pick-up gets you, not whether the phone announced itself. This is the attention-reclamation point in its narrowest form: rearrange the environment rather than resolving to try harder.

4. Expect displacement. Moving the phone away halved phone use and left total leisure time unchanged, with the difference reappearing on the computer (Heitmayer, 2025). Removing a device without replacing the activity relocates the behaviour. Plan for what the time is for.

5. Do not expect notification settings to change your screen time. They did not in the trial that measured it properly (Dekker et al., 2025). They may well change how deliberate your use feels, which is a legitimate thing to want, and it is worth wanting it under its own name rather than as a proxy for using the phone less.

6. If you do go silent, tell people. The cost that shows up most reliably when notifications are removed is not attentional but social — anxiety about responsiveness, and the sense of having dropped out of contact (Pielot & Rello, 2017; Fitz et al., 2019; Dekker et al., 2025). It is a manageable cost once it is named, and most of it is an assumption about what other people expect rather than anything they have said.

7. Discount the twenty-three-minute figure wherever you meet it. Its presence in an article is reasonable evidence that the author did not read the paper, which is useful information about everything else in it.

A note on certainty

The notification story is not wrong in the way the blue-light story was wrong. The mechanism is real and it replicates. An alert you cannot answer does cost you something, and on a task where place-keeping matters it can cost more than it looks like it should.

It is wrong about proportion. It took a small, well-measured, acute effect and made it the explanation for a pattern of behaviour that is overwhelmingly self-generated — then attached a remedy to it that, when tested, changed nothing except how people felt about what they were already doing. The most-repeated statistic in the field has no paper behind it. The most-cited laboratory finding about phone presence has not survived meta-analysis. And the one intervention everybody recommends has now been through a preregistered trial with objective measurement, and came out flat.

What survives is narrower and duller than the advice. Interruption is expensive where sequence matters, which is an argument for silence during specific work rather than as a way of life. The phone's pull does not depend on the phone doing anything — it is a habit of reaching, running at roughly a dozen times an hour, largely outside awareness and resistant to being switched off at the source. And an attention problem that originates in the person will follow the person to the next device.

The settings screen is the part of this that a toggle can reach. That is the reason it is the part everyone is told to fix, and it is not much of a reason. Nobody is pinging you most of the time. You are just looking.

---

References

• Altmann, E. M., Trafton, J. G., & Hambrick, D. Z. (2014). Momentary interruptions can derail the train of thought. Journal of Experimental Psychology: General, 143(1), 215–226.
• Andrews, S., Ellis, D. A., Shaw, H., & Piwek, L. (2015). Beyond self-report: Tools to compare estimated and real-world smartphone use. PLOS ONE, 10(10), e0139004.
• Böttger, T., Poschik, M., & Zierer, K. (2023). Does the brain drain effect really exist? A meta-analysis. Behavioral Sciences, 13(9), 751.
• Christodoulou, A., & Roussos, P. (2025). "Phone in the room, mind on the roam": Investigating the impact of mobile phone presence on distraction. European Journal of Investigation in Health, Psychology and Education, 15(5), 74.
• Dekker, C. A., Baumgartner, S. E., Sumter, S. R., & Ohme, J. (2025). Beyond the buzz: Investigating the effects of a notification-disabling intervention on smartphone behavior and digital well-being. Media Psychology, 28(1), 162–188.
• Fitz, N., Kushlev, K., Jagannathan, R., Lewis, T., Paliwal, D., & Ariely, D. (2019). Batching smartphone notifications can improve well-being. Computers in Human Behavior, 101, 84–94.
• Fournier, H., Fournel, A., Osiurak, F., Koenig, O., Pâris, F., Gaujoux, V., & Ringeval, F. (2026). Attention hijacked: How social media notifications disrupt cognitive processing. Computers in Human Behavior, 179, 108926.
• Heitmayer, M. (2025). When the phone's away, people use their computer to play: Distance to the smartphone reduces device usage but not overall distraction and task fragmentation during work. Frontiers in Computer Science, 7, 1422244.
• Heitmayer, M., & Lahlou, S. (2021). Why are smartphones disruptive? An empirical study of smartphone use in real-life contexts. Computers in Human Behavior, 116, 106637.
• Iqbal, S. T., & Horvitz, E. (2010). Notifications and awareness: A field study of alert usage and preferences. Proceedings of the 2010 ACM Conference on Computer Supported Cooperative Work, 27–30.
• Kaminske, A., Brown, A., Aylward, A., & Haller, M. (2022). Cell phone notifications harm attention: An exploration of the factors that contribute to distraction. European Journal of Educational Research, 11(3), 1487–1494.
• Kushlev, K., Proulx, J., & Dunn, E. W. (2016). "Silence your phones": Smartphone notifications increase inattention and hyperactivity symptoms. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 1011–1020.
• 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.
• Mark, G., Gonzalez, V. M., & Harris, J. (2005). No task left behind? Examining the nature of fragmented work. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 321–330.
• Mark, G., Gudith, D., & Klocke, U. (2008). The cost of interrupted work: More speed and stress. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 107–110.
• Mark, G., Iqbal, S. T., Czerwinski, M., Johns, P., & Sano, A. (2016). Neurotics can't focus: An in situ study of online multitasking in the workplace. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 1739–1744.
• Nason, K., & Wilbiks, J. M. P. (2025). Call me maybe: Effects of notification modality on visual sustained attention. Multisensory Research, 38(1–3), 61–75.
• Oulasvirta, A., Rattenbury, T., Ma, L., & Raita, E. (2012). Habits make smartphone use more pervasive. Personal and Ubiquitous Computing, 16(1), 105–114.
• Parry, D. A. (2024). Does the mere presence of a smartphone impact cognitive performance? A meta-analysis of the "brain drain effect". Media Psychology, 27(5), 737–762.
• Pielot, M., Church, K., & de Oliveira, R. (2014). An in-situ study of mobile phone notifications. Proceedings of the 16th International Conference on Human-Computer Interaction with Mobile Devices & Services, 233–242.
• Pielot, M., & Rello, L. (2017). Productive, anxious, lonely: 24 hours without push notifications. Proceedings of the 19th International Conference on Human-Computer Interaction with Mobile Devices and Services, 1–11.
• Ruiz Pardo, A. C., & Minda, J. P. (2022). Reexamining the "brain drain" effect: A replication of Ward et al. (2017). Acta Psychologica, 230, 103717.
• Stothart, C., Mitchum, A., & Yehnert, C. (2015). The attentional cost of receiving a cell phone notification. Journal of Experimental Psychology: Human Perception and Performance, 41(4), 893–897.
• Upshaw, J. D., Stevens, C. E., Ganis, G., & Zabelina, D. L. (2022). The hidden cost of a smartphone: The effects of smartphone notifications on cognitive control from a behavioral and electrophysiological perspective. PLOS ONE, 17(11), e0277220.
• Ward, A. F., Duke, K., Gneezy, A., & Bos, M. W. (2017). Brain drain: The mere presence of one's own smartphone reduces available cognitive capacity. Journal of the Association for Consumer Research, 2(2), 140–154.
• Wasmuth, J. M., Reinhard, I., Hill, H., Alpers, G. W., Shevchenko, Y., Kiefer, F., & Leménager, T. (2022). A smarter way to use your smartphone: An intervention to limit smartphone-related distractions reduces hyperactivity but not inattention symptoms. European Addiction Research, 28(4), 255–266.
• Westbrook, J. I., Woods, A., Rob, M. I., Dunsmuir, W. T. M., & Day, R. O. (2010). Association of interruptions with an increased risk and severity of medication administration errors. Archives of Internal Medicine, 170(8), 683–690.