TL;DR. “Screen time” is one number standing in for very different activities, and on its own it predicts little about how well people think. Across 355,358 adolescents, digital technology use explained at most 0.4% of the variation in well-being. The claim that a phone on the desk drains your mind has shrunk under replication: the largest meta-analysis found no significant effect on cognition (d = -0.02), another found a small working-memory cost only (d = -0.20), and the “heavy multitaskers are more distractible” effect fell to near zero after bias correction. What does show up is context. In PISA 2022, students distracted by classmates’ devices scored 15 points lower in maths, about three-quarters of a year of learning, while moderate use was linked to higher scores. For adults over 50, technology users had lower odds of cognitive impairment (OR = 0.42). We put these results on one scale and joined them with the 2025 American Time Use Survey: 3.2 hours a day of leisure screens against 16 minutes of reading. The rule that follows: stop counting hours and manage interruption, content, displacement and the hour before sleep.
Why “screen time” is the wrong unit
The public argument about screens usually starts from a total: hours per day. That number mixes a video call, a spreadsheet, a documentary, a feed and a game into one figure, then asks whether “it” is bad for the brain. It is not surprising that the answer keeps coming back weak.
The clearest demonstration came from Amy Orben and Andrew Przybylski in 2019. Large social datasets contain so many variables that researchers can analyse them in thousands of defensible ways, and that flexibility lets small effects look significant and produces conflicting results. So they ran the analysis every defensible way at once, a method called specification curve analysis, across three datasets totalling 355,358 adolescents. The association between technology use and well-being was negative but small, explaining at most 0.4% of the variation, which they judged too small to warrant policy change.
That study was about well-being, not cognition, but the lesson transfers: an effect can be real and significant in a big sample and still say almost nothing about any individual’s day. The useful question is not “are screens bad?” but “which screen behaviour, in which situation, moves which outcome, by how much?”
The effect-size ladder: the evidence on one scale
Studies here report results in different currencies: variance explained, standardised differences, test points, odds ratios. Table 1 lists what each source reported. Table 2 converts them to one scale.
Table 1. What the main studies report (verified data)
| Study | Design and sample | Exposure | Reported result |
|---|---|---|---|
| Orben and Przybylski (2019) | Specification curves, 3 datasets, n = 355,358 adolescents | Technology use (outcome: well-being) | Negative, at most 0.4% of variance |
| Ward et al. (2017) | 2 experiments, final samples of 520 and 275 | Own phone nearby vs in another room | Reduced working memory capacity and fluid intelligence |
| Parry (2023) | Meta-analysis, k = 56, n = 7,093 | Phone present vs absent | Working memory d = -0.20 (significant); sustained attention d = -0.14 (not significant) |
| Hartanto et al. (2024) | Meta-analysis, 166 effect sizes, 33 studies, n = 4,368 | Phone present vs absent | All cognitive outcomes d = -0.02, 95% CI -0.06 to 0.01 |
| Wiradhany and Nieuwenstein (2017) | 2 replications plus meta-analysis of 39 effects | Heavy vs light media multitasking | Distractibility d = 0.17, falling to 0.01-0.07 after bias correction |
| Walsh et al. (2018), ABCD study | Cross-sectional, n = 4,524 children aged 9-10 | Meeting the 2-hour recreational screen limit | Global cognition 4.25-5.15 points higher (sample SD 16.8) |
| OECD PISA 2022 | 15-year-olds, OECD average | Distracted by classmates’ devices in maths | 15 points lower (socio-economically adjusted) |
| OECD PISA 2022 | Same | Up to 1 hour of leisure device use at school vs none | 10 points higher (adjusted); over 1 hour linked to lower scores |
| Benge and Scullin (2025) | Meta-analysis, 57 studies, 411,430 adults over 50 | Using digital technology | Cognitive impairment OR = 0.42; decline HR = 0.74 |
Table 2. The same results on a common scale
| Result | Standardised difference (d) | Correlation (r) | Variance explained | Reading |
|---|---|---|---|---|
| Phone presence, all outcomes (Hartanto) | 0.02 | 0.010 | 0.01% | Indistinguishable from zero |
| Media multitasking, bias-corrected (Wiradhany) | 0.01-0.07 | 0.005-0.035 | 0.00-0.12% | Near zero |
| Technology use and adolescent well-being, upper bound (Orben) | 0.13 | 0.063 | 0.40% | Very small |
| Media multitasking, uncorrected (Wiradhany) | 0.17 | 0.085 | 0.72% | Small, mostly a small-study artefact |
| Phone presence, working memory only (Parry) | 0.20 | 0.100 | 0.99% | Small, the one effect that survives |
| Meeting the 2-hour screen limit, children (Walsh) | 0.25-0.31 | 0.125-0.152 | 1.6-2.3% | Small to moderate, cross-sectional |
| Technology use, adults over 50 (Benge) | 0.48 (95% CI 0.36-0.58) | 0.233 | 5.4% | Moderate, in the protective direction |
| Distracted by classmates’ devices (PISA) | Not converted | Not converted | Not converted | Gap of 0.75 years of learning |
| Up to 1 hour leisure use at school vs none (PISA) | Not converted | Not converted | Not converted | 0.5 years of learning, favouring moderate use |
Table 2: CEOtudent calculation from the sources in Table 1. Conversions: r = d / √(d² + 4), d = 2r / √(1 – r²), d = ln(OR) × √3 / π for the odds ratio, the ABCD sample SD of 16.8 for Walsh et al., and the OECD benchmark that 20 PISA points equal the average annual pace of learning. Absolute sizes; direction in the last column. Approximations that assume roughly equal groups, not re-analyses.
Three things stand out. The effects that made headlines sit near the bottom of the ladder once better data arrives. Apart from the cross-sectional children’s 2-hour threshold, the larger effects are not about hours: they are about disruption at a specific moment (a classroom full of devices) and, for older adults, engagement in the protective direction. And the one effect that survived, a small working-memory cost when a phone is present, concerns attention during demanding tasks, which is exactly where knowledge workers should look.
The phone on the desk: what survived replication
In 2017, Adrian Ward and colleagues proposed the “brain drain” hypothesis: that the mere presence of your own smartphone occupies limited cognitive resources even when you are not using it. In two experiments, with final samples of 520 and 275 participants, they reported that phone presence reduced available working memory capacity and fluid intelligence, with the largest costs for the most phone-dependent participants.
The follow-up literature has been less kind. Douglas Parry’s meta-analysis found one significant pooled effect across cognitive functions, working memory at d = -0.20; sustained attention, inhibitory control, cognitive flexibility and fluid intelligence all produced null summary effects, and most studies had very low statistical power. The largest synthesis, by Andree Hartanto and colleagues in 2024, pooled 166 effect sizes and found an overall effect of d = -0.02, not significant and not moderated by how dependent people were on their phones. The authors concluded there is little reason at present to think complete isolation from smartphones at work would improve productivity and performance.
The honest reading is narrow. A silent phone in view probably costs little on most tasks. For the hardest thinking, holding several ideas in mind at once, moving it out of sight is a cheap hedge against the one surviving effect. It is not a cure for a scattered day, which is usually caused by what the phone does when it lights up, not by the fact that it exists.
Multitasking: the trait claim collapsed, the moment-to-moment cost did not
A second famous claim held that habitual media multitaskers become worse at filtering distraction. Wubbo Wiradhany and Mark Nieuwenstein ran two replications with an average replication power of 0.81: of 14 tests, only five showed the expected effect, and only two survived a more conservative Bayesian analysis. Their meta-analysis of 39 effect sizes found a weak association (d = 0.17) that became non-significant after correcting for small-study effects, at 0.01 to 0.07. They concluded there is reason to question whether the association exists at all in laboratory tasks.
So the evidence does not show that heavy multitasking permanently damages focus. It is consistent with something simpler and more controllable: switching during a task costs you at that moment. PISA 2022 points the same way. Students were less likely to report distraction when they switched off notifications from social networks and apps during class, did not have devices open for notes or searches, and did not feel pressured to answer messages. The lever is the interruption pattern, not the lifetime total. Our piece on attention residue explains why a switch keeps costing you after you return.
The adolescent data behind the headlines
Most alarm about screens and thinking comes from studies of young people. Two of the most cited sources are more nuanced than the headlines built on them.
The ABCD study. Walsh and colleagues analysed 4,524 US children aged 9 to 10. Only 36.6% met the recommendation of no more than 2 hours of recreational screen time a day. Children meeting the screen recommendation, alone or with the sleep recommendation, scored 4.25 to 5.15 points higher on global cognition (sample SD 16.8) than children meeting no recommendations. Two cautions: the data are cross-sectional, so family routines could drive both screen habits and scores; and the paper’s abstract and results text assign the two values to the two groups in opposite order, which is why we give a range. The authors raised the possibility that screen use beyond 2 hours weakens the benefit of sleep, which points to timing and displacement rather than screens as such.
PISA 2022. The OECD data on 15-year-olds is often summarised as “phones lower scores”. The report is more specific. About 30% of students said they got distracted using devices in most or every maths lesson, and 25% were distracted by other students’ devices; those distracted by classmates’ devices in at least some lessons scored 15 points lower, after socio-economic adjustment. Yet up to an hour a day of device use for learning at school was linked to 14 points higher than none, and up to an hour of leisure use to 10 points higher; use beyond an hour went with lower scores. The OECD frames this as the “Goldilocks hypothesis”: moderate use is not intrinsically harmful, overuse and misuse are. One more figure matters for adults too: 45% of students felt nervous or anxious without their devices nearby, and that feeling was linked to lower life satisfaction and lower maths scores. The relationship with the device carries the signal, not the minutes.
Adults and ageing: the “digital dementia” fear meets the data
For adults, the popular worry is that a lifetime of screens erodes the mind. Jared Benge and Michael Scullin tested this “digital dementia” hypothesis in a 2025 meta-analysis of 57 poolable studies covering 411,430 adults over 50, average baseline age 68.7. Technology use was associated with lower odds of cognitive impairment (OR = 0.42, 95% CI 0.35 to 0.52) and slower cognitive decline over time (HR = 0.74). The effects held after accounting for demographic, socioeconomic, health and cognitive reserve proxies, and in the highest-quality studies alone.
On our scale that is the largest effect on the ladder, pointing the opposite way from the moral panic. It is still observational: people whose cognition is slipping may simply stop using complex technology, and the authors call for tests of causality in both directions. The fair conclusion is not that screens protect the brain, but that actively using digital tools is not the cognitive threat it is often presented as.
How much screen time adults actually have: the 2025 time-use data
The American Time Use Survey, released by the US Bureau of Labor Statistics in June 2026, is the most recent official picture of leisure time. It counts primary activities only, so screens at work or used while doing something else are excluded. We combined watching TV with playing games or using a computer for leisure (a category that includes social media and video games, and also board games) and compared the total with reading.
Table 3. Leisure screen time vs reading, US 2025 (hours per day unless stated)
| Group | Leisure screen time | Share of all leisure | Reading | Screen minutes per reading minute |
|---|---|---|---|---|
| All people 15 and over | 3.23 (194 min) | 62.6% | 16 min | 12.0 |
| Ages 20 to 24 | 3.15 (189 min) | 67.5% | 6 min | 31.5 |
| Ages 25 to 34 | 2.80 (168 min) | 62.2% | 10 min | 16.5 |
| Ages 35 to 44 | 2.20 (132 min) | 56.6% | 13 min | 10.0 |
| Ages 45 to 54 | 2.65 (159 min) | 61.6% | 13 min | 12.6 |
| Ages 55 to 64 | 3.32 (199 min) | 63.7% | 13 min | 15.8 |
| Ages 65 to 74 | 4.63 (278 min) | 67.1% | 28 min | 10.1 |
| Full-time workers | 2.39 (143 min) | 59.9% | 10 min | 14.1 |
| Advanced degree holders (25 and over) | 2.35 (141 min) | 53.5% | 26 min | 5.3 |
CEOtudent calculation from BLS American Time Use Survey 2025, Table 11A: watching TV plus playing games and computer use for leisure; reading; total leisure and sports.
Every group exceeds the 2-hour threshold used for children in the ABCD analysis, and screens take between about half and two-thirds of leisure time. The lowest screen total, ages 35 to 44, coincides with the least leisure overall, a reminder that time pressure, not willpower, sets most totals. The mix is also shifting: BLS reports games and computer use rose from 25 to 37 minutes a day between 2015 and 2025, a 48% increase.
The ratio column is where the cognitive question lives. A full-time worker spends about 14 minutes on leisure screens for every minute of leisure reading. The research above does not say those screen minutes are damaging, but it does say minutes come from somewhere. If you value the sustained attention reading trains, watch your reading total, not your screen total. Our piece on deep reading as a competitive advantage covers what that practice builds.
The four questions that matter more than your total
Table 4. Screen-time decisions the evidence supports
| Question | What the evidence shows | Decision rule |
|---|---|---|
| 1. Interruption: does the screen break into demanding work? | Trait multitasking harm fell to near zero; distraction and notifications track lower performance; a small working-memory cost of phone presence survives | Shield blocks of hard work from alerts; phone out of sight for the hardest tasks; skip total isolation for routine work |
| 2. Content: what are you doing on it? | Learning use and moderate leisure use were linked to higher PISA scores, heavier leisure to lower; for adults over 50, use goes with lower odds of impairment | Sort screen time into create, learn, connect and scroll; cut scroll first, not everything |
| 3. Displacement: what does it replace? | Screens take 62.6% of leisure for people 15 and over; reading averages 16 minutes a day, 10 for full-time workers | Schedule the activity you want more of before setting any screen limit |
| 4. Timing: does it eat into sleep? | In ABCD, meeting both screen and sleep recommendations went with better cognition; the authors flag screens weakening sleep’s benefit | Fix a device cut-off before bed; rank it above daytime minutes |
CEOtudent editorial framework, derived from the sources in Tables 1 to 3.
A two-week protocol:
- Measure by category. For one normal week, sort your device activity reports into create, learn, connect and scroll. The split, not the total, is the decision input.
- Count interruptions. Note how often you break off a demanding task to check a device. The notification audit protocol gives a method for cutting them.
- Guard one block a day. Give 60 to 90 minutes of your hardest work alerts off and the phone out of sight.
- Protect the sleep boundary. Pick a fixed screen cut-off and keep it; our guide to sleep architecture and cognitive performance explains why the last hours matter.
- Add before you subtract. Put 20 minutes of reading or another chosen activity into the day first.
- Review. Compare guarded and unguarded work blocks; keep what helped.
This is where the CEO and the student meet. The CEO owns the decision about what attention is for and designs the environment instead of waiting for willpower or policy. The student keeps updating: the brain drain effect was a confident claim in 2017 and a much smaller one by 2024, and the multitasking effect largely disappeared under correction. Treat every study here as a hypothesis to test on your own work. For the tool-level side, see digital minimalism for AI power users.
FAQ
Does screen time lower cognitive performance in adults?
Not as a general effect of hours that current evidence can detect. The largest phone-presence meta-analysis found no significant effect (d = -0.02), and a 2025 meta-analysis of adults over 50 linked technology use to lower odds of impairment (OR = 0.42), though observationally. The supported cost is interruption during demanding tasks.
Is the “brain drain” effect of a nearby phone real?
Partly. Later syntheses found either a small working-memory effect only (d = -0.20) or no overall effect (d = -0.02). Keeping the phone out of sight for your hardest thinking is a cheap precaution, not a transformation.
What does PISA 2022 say about phones and learning?
Students distracted by classmates’ devices scored 15 points lower in maths, about three-quarters of a year of learning. But up to an hour of device use, for learning or leisure, was linked to higher scores than none. The OECD reads this as moderate use not being intrinsically harmful.
Is a full digital detox worth it?
The evidence gives little reason to expect cognitive gains from removing devices entirely. Targeted changes to interruptions, content, displacement and the time before sleep are better supported and easier to keep.
Sources
- Orben, A. and Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3(2), 173-182.
- Ward, A. F., Duke, K., Gneezy, A. and 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.
- Parry, D. A. (2023). Does the Mere Presence of a Smartphone Impact Cognitive Performance? A Meta-Analysis of the “Brain Drain Effect”. Media Psychology. Pooled estimates as reported in the author’s working-paper version (PsyArXiv, 2022).
- Hartanto, A., Lua, V. Y. Q., Kasturiratna, K. T. A. S., Koh, P. S., Tng, G. Y. Q., Kaur, M., Quek, F. Y. X., Chia, J. L. and Majeed, N. M. (2024). The Effect of Mere Presence of Smartphone on Cognitive Functions: A Four-Level Meta-Analysis. Technology, Mind, and Behavior, 5(1), 1-14.
- Wiradhany, W. and Nieuwenstein, M. R. (2017). Cognitive control in media multitaskers: Two replication studies and a meta-analysis. Attention, Perception, and Psychophysics, 79(8), 2620-2641.
- Walsh, J. J., Barnes, J. D., Cameron, J. D., Goldfield, G. S., Chaput, J.-P., Gunnell, K. E., Ledoux, A.-A., Zemek, R. L. and Tremblay, M. S. (2018). Associations between 24 hour movement behaviours and global cognition in US children: a cross-sectional observational study. The Lancet Child and Adolescent Health, 2(11), 783-791.
- OECD (2023). PISA 2022 Results (Volume II): Learning During and From Disruption. OECD Publishing, Paris. Chapters 3 and 5, Boxes II.5.1 and II.5.2.
- US Bureau of Labor Statistics (2026). American Time Use Survey, 2025 Results. News release and Table 11A, 2025 annual averages.
- Benge, J. F. and Scullin, M. K. (2025). A meta-analysis of technology use and cognitive aging. Nature Human Behaviour, 9(7), 1405-1419.
Table 1 reports verified results from the sources above. Tables 2 and 3 are CEOtudent calculations from those sources; the common-scale conversions and the screen-to-reading ratio are our analytical choices. Table 4 is a CEOtudent editorial framework.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:
