Social media and IQ: what the studies measure, and what they do not

0 Quick Answer

No study has shown that social media use lowers measured intelligence, the best designed longitudinal test found no effect of online socializing on children's IQ over two years, and the harms that are documented run through sleep and displaced activity rather than through any direct erosion of ability. Sauce, Liebherr, Judd and Klingberg analysed 9,855 children from the Adolescent Brain Cognitive Development study in Scientific Reports in 2022, with measures of intelligence at ages 9 to 10 and again two years later, controlling for polygenic scores for cognition and for socioeconomic status. At baseline, time spent socializing online correlated negatively with intelligence at r = 0.10, and time spent watching videos at r = 0.12. After two years, socializing had no effect on the change in intelligence, gaming had a positive effect with a standardized coefficient of 0.17, and watching also came out positive, a result that did not survive a post hoc control for parental education. The baseline correlations are what a cross-sectional study sees; the longitudinal result is what remains once the children are followed.

The two most cited claims about phones and attention fare worse. Ward and colleagues reported in 2017 that the mere presence of a participant's own smartphone on the desk reduced working memory capacity and fluid intelligence scores. Ruiz Pardo and Minda's pre-registered direct replication in 2022 found no difference between phone locations on either task, and Hartanto and colleagues' four level meta-analysis of 166 effect sizes from 33 studies and 4,368 participants put the overall effect at d = 0.02 in the harmful direction, not significantly different from zero. Orben and Przybylski's specification curve analysis of 355,358 adolescents found that digital technology use explained at most 0.4 percent of the variation in well-being, which is the outcome most of the public debate is actually about.

What is documented is narrower and more useful. Carter and colleagues' meta-analysis of 20 studies covering 125,198 children found that using a device at bedtime roughly doubled the odds of inadequate sleep, and sleep loss lowers next day performance on exactly the tasks an IQ test contains. Walsh and colleagues' ABCD cross section of 4,524 children found that meeting the two hour recreational screen limit and the sleep recommendation together was associated with about five points more on a cognition composite scaled like an IQ. In adults over 50 the direction reverses, with regular internet use associated with better cognitive trajectories and lower dementia incidence in large cohorts. The rest of this page takes the studies by design.

1 What a Study of Social Media and IQ Would Have to Measure

The phrase social media and IQ names an exposure that is rarely measured and an outcome that is almost never measured, and most of the disagreement in this literature is disagreement about what was substituted for each. The exposure is usually hours of screen time reported by a parent or by the person, which mixes video, games, messaging and social platforms, and which correlates only modestly with logged use. The ABCD study separates watching, socializing and gaming; most studies do not. The outcome is usually not an IQ at all but a developmental screening questionnaire, a school grade, a single attention task, or a self-report of well-being, and the page on what IQ measures explains why a Full Scale score from a normed battery is a different object from any of those. When a headline reports that screens lower intelligence, the first question is which of these stood in for intelligence.

The design question matters as much. A cross-sectional correlation between screen time and a test score cannot separate three explanations: that screens lower scores, that children with lower scores are given or choose more screens, and that a third factor, such as household income, parental education or genetic propensity, produces both. Madigan and colleagues' cross-lagged panel study of 2,441 Canadian children was designed to test the direction, and the ABCD analysis was designed to remove the genetic and socioeconomic confounders with polygenic scores and family measures. Experiments settle direction but usually measure a task on one afternoon rather than an ability, and their results have to replicate before they mean anything, which is the subject of a later section.

The page on reliability and validity sets out what a measure has to show before a difference in it can be interpreted, and the same standard applies to the exposure side. Self-reported screen time has known low agreement with device logs, and the categories that matter for cognition, active messaging against passive scrolling, short video against long, are seldom recorded. That is why the strongest statements in this literature are about broad screen time and sleep, where the measurement is crude but the designs are large, and why the specific claim about social media platforms and intelligence remains, on the published evidence, untested in either direction at the level of a normed IQ.

2 The Largest Longitudinal Test: ABCD and the 2022 Analysis

The one study that measured intelligence properly, followed a large sample and controlled for the confounders that cross-sectional work cannot remove found that socializing online did nothing to children's IQ over two years, and its authors expected otherwise. The ABCD study recruited children at 21 sites across the United States and administered the NIH Toolbox cognition battery alongside other measures; the 2022 analysis built an intelligence factor from those tasks at baseline and at the two year follow up. Screen time was reported in three categories: watching television and videos, socializing through social media, video chat and texting, and gaming. Polygenic scores for cognition stood in for genetic differences, and household income and parental education for socioeconomic status.

At baseline the picture looked like every cross-sectional study: watching correlated at r = 0.12 and socializing at r = 0.10 with lower intelligence, and gaming did not correlate at all. Over two years the picture changed. Gaming predicted gains in intelligence with a standardized coefficient of 0.17, which on a scale with a standard deviation of 15 is about two and a half points, by our arithmetic. Watching also predicted gains, at 0.12, though that effect disappeared when parental education replaced the socioeconomic index in a post hoc check. Socializing predicted nothing. The authors read the gaming result as consistent with experimental evidence of cognitive benefits from video games, which the page on video games and IQ reviews, and read the whole pattern as consistent with the malleability of measured ability that the Flynn effect and cognitive training studies also show.

Three limits belong beside the result. The children were 9 to 10 at baseline and 11 to 12 at follow up, which is before the ages at which social media use is heaviest, and the socializing category bundled messaging and video chat with platforms. Two years is a short window for an effect that critics describe as cumulative. And the study is a single analysis of a single cohort, however large. What it establishes is narrower than a verdict: in the largest sample with the best controls, the negative baseline correlation between online socializing and intelligence was a product of who socializes online rather than of socializing, because it did not predict any change. The page on what heritability actually means explains why controlling for polygenic scores changes the reading of a correlation, and this study is the clearest example in the screen literature of that change.

3 The Cross-Sectional Baselines: Screen Time, Sleep and Toolbox Scores

Cross-sectional studies of screen time and cognition find consistent small negative associations, and the best of them measure the thing that the longitudinal work then shows to be selection, so they are worth reading for what they measure rather than for what they are quoted as proving. Walsh and colleagues used the first ABCD release, 4,524 children aged 8 to 11, to test the Canadian 24 hour movement guidelines: at least 60 minutes of physical activity, two hours or less of recreational screen time, and nine to eleven hours of sleep. Only 37 percent of the children met the screen limit, 51 percent met the sleep recommendation, 18 percent met the activity recommendation and 5 percent met all three. Each additional recommendation met was associated with 1.44 points more on the NIH Toolbox global cognition composite, which is scaled with a mean of 100 and a standard deviation of 15; meeting the screen limit alone was associated with 4.25 points more, and meeting the screen limit and the sleep recommendation together with 5.15 points more, in models adjusted for the usual covariates. Physical activity, by contrast, showed no independent association, and the page on exercise and IQ discusses why that result surprised the authors.

Madigan and colleagues followed 2,441 mothers and children in Calgary at 24, 36 and 60 months with a random intercepts cross-lagged model, which separates a child's stable level from changes within the child. Higher screen time at 24 months predicted poorer performance on the Ages and Stages developmental questionnaire at 36 months, and higher screen time at 36 months predicted poorer performance at 60 months, with standardized coefficients of 0.06 and 0.08; the reverse paths, from development to later screen time, were not significant. Takahashi and colleagues' Japanese cohort reported in JAMA Pediatrics in 2023 that screen time at age one was associated with communication and problem solving delays at ages two and four. Those are the strongest findings in the literature for young children, and they are findings about developmental screening questionnaires in toddlers, not about the IQ of anyone old enough to hold an account on a social platform.

The gap between the two age groups is the point. In toddlers, screen time displaces the interaction through which language and problem solving develop, and the effects, though small, run in the expected direction. In the 9 to 12 year olds of ABCD, the baseline association is the same size and sign, and then the longitudinal analysis with genetic controls finds it does not predict change. The page on whether IQ changes with age explains why developmental measures at two and intelligence measures at ten are not the same construct, and why an association at one age cannot be carried to another.

4 The Smartphone in the Room: An Effect That Did Not Survive Replication

The most quoted experiment about phones and thinking reported that a silent smartphone on the desk lowered working memory and fluid intelligence, and the pre-registered replication and the meta-analysis that followed found nothing. Ward, Duke, Gneezy and Bos published Brain Drain in the Journal of the Association for Consumer Research in 2017. In two experiments with undergraduates, participants left their own phone on the desk, in a pocket or bag, or in another room, and then completed an automated operation span task, a measure of working memory capacity, and in the first experiment a set of Raven's matrices items. Participants whose phones were on the desk scored lower on working memory than those whose phones were in another room, and the authors proposed that the phone's presence consumed attentional resources even when unused. The finding fitted the mood of the time and travelled far.

Ruiz Pardo and Minda ran a pre-registered direct replication of the second experiment, assigning phone location and whether the phone was on or off across six conditions, with the same operation span task and a go no-go task of response inhibition, and reported in Acta Psychologica in 2022 that the effect did not replicate: there was no difference between phone locations on either task. Hartanto, Lua, Kasturiratna, Koh, Tng, Kaur and colleagues then pooled the whole literature in Technology, Mind, and Behavior in 2024, 166 effect sizes from 53 samples in 33 studies with 4,368 participants, across working memory, sustained attention, content retention, fluid intelligence and other outcomes. The pooled effect was d = 0.02 in the harmful direction with a confidence interval from 0.06 to 0.01 on the other side of zero, and none of the methodological moderators they examined was significant.

The page on working memory tests explains what an operation span task measures and why it is sensitive to distraction when distraction is real, and the page on matrix reasoning describes the Raven's format used in the first experiment. The replication history is not a reason to keep a phone on the desk during a timed test; it is a reason not to cite the 2017 result as evidence that phones reduce intelligence. A person taking any timed assessment, including an ACIS session, should still remove the phone, because notifications interrupt and interruption is a documented cost even if mere presence is not.

5 Media Multitasking and Attention: Real Differences, Unknown Direction

People who habitually consume several media streams at once perform differently on some attention tasks, and after fifteen years of research nobody has shown whether the habit causes the difference or the difference the habit. Ophir, Nass and Wagner's study of cognitive control in media multitaskers in PNAS in 2009 classified students by a media multitasking index and found that heavy multitaskers were worse at filtering irrelevant stimuli from memory and at switching between tasks, the opposite of what a practice account would predict. Uncapher and Wagner's review of the decade that followed, also in PNAS, found that heavy media multitasking was associated with poorer performance on tasks of working memory and sustained attention in a number of studies, that the effects were modest, that several studies found no association, and that virtually all of the evidence was cross-sectional, so that the direction remained open.

Firth and colleagues' review of the online brain in World Psychiatry in 2019 reached a similar position across attention, memory and social cognition: the internet may be altering attentional capacities and the way memory is used, the evidence is suggestive rather than settled, and the neuroimaging findings that exist are correlational. The page on the six cognitive domains explains that attention and processing speed are measured by separate indices in a battery, and the page on processing speed describes what those tasks require. A difference in a filtering task among heavy multitaskers is a difference on a component, and no study in this literature has reported a difference in Full Scale IQ between heavy and light media multitaskers.

What the multitasking literature does support is a practical statement about performance on the day. Attempting to attend to two streams at once degrades performance on both in every experimental paradigm that has tested it, which is why timed tests are administered in silence. Whether a person who does this habitually has changed their capacity, or simply has a preference that a task of the moment reveals, is a question that the cross-sectional designs cannot answer and the longitudinal designs have not been run to answer.

6 Sleep: The Pathway That Is Documented

If social media affects cognitive performance in the general population, the route for which the evidence is strongest is sleep, and that route has been measured at every step. Carter, Rees, Hale, Bhattacharjee and Paradkar's meta-analysis in JAMA Pediatrics in 2016 covered 20 studies of 125,198 children aged 6 to 19 with a mean age of 14.5. Using a portable device at bedtime was associated with inadequate sleep quantity at an odds ratio of 2.17, poor sleep quality at 1.46 and excessive daytime sleepiness at 2.72. Merely having access to a device in the bedroom, without reported use, carried odds ratios of 1.79, 1.53 and 2.27 for the same outcomes. The authors described the association as strong and consistent, with heterogeneity across studies but no study pointing the other way.

The second step, from short sleep to cognitive performance, is one of the oldest findings in experimental psychology, and the page on sleep and IQ reviews it: restricted sleep lowers vigilance, working memory and reasoning on the following day, with effects that accumulate across nights and that people underestimate in themselves. The ABCD cross section is the two steps joined in one sample, where meeting the sleep recommendation and the screen limit together carried the largest association with cognition of any combination tested. None of this requires that the content of a platform does anything to the brain; it requires only that a device in the bedroom keeps a child awake, which is what the odds ratios show.

The distinction matters because it points to the intervention. A pathway through sleep is interrupted by moving the device, not by changing the app, and the Carter meta-analysis found that access alone, not just use, carried risk. The page on how long an IQ test takes notes that a full battery is demanding enough that the night before matters, and the same applies to a school day. What the evidence does not show is any effect of bedtime device use on ability as opposed to performance; the studies measure the next morning, not the next decade.

7 Displacement: Television, Games and the Academic Record

The largest meta-analysis of screens and school performance found that total screen time was not associated with grades, while television and video games were, which is the signature of displacement rather than of any general effect of screens on ability. Adelantado-Renau and colleagues reviewed 58 cross-sectional studies of 480,479 participants aged 4 to 18 from 23 countries in JAMA Pediatrics in 2019, with 30 studies and 106,653 participants in the meta-analysis. Overall screen media time was not associated with academic performance, with an effect size of 0.29 and a confidence interval that crossed zero. Television viewing was inversely associated with composite scores at 0.19, with language at 0.18 and with mathematics at 0.25, and video game playing with composite scores at 0.15, with both effects concentrated among adolescents rather than younger children. Computer, internet and mobile phone use, the categories closest to social media, did not show consistent associations.

Grades are not IQ, but they are closely related to it, as Deary and colleagues' finding of a latent correlation of 0.81 between general ability at eleven and examination results at sixteen shows, and the page on IQ and academic achievement sets out. A displacement effect on grades runs through homework and reading time, and the page on reading and IQ reviews the evidence that the amount a child reads has effects on verbal ability that persist. A screen that replaces an hour of reading has a cost that shows up in vocabulary; a screen that replaces an hour of television does not, and the crude category of screen time cannot tell the two apart, which is why the pooled estimate for total screen time is null.

The same review is a caution against the opposite claim. Video games showed a small negative association with grades among adolescents in this cross-sectional literature, while the longitudinal ABCD analysis found gaming predicting gains in intelligence, and both can be true: a game that displaces homework can lower a grade while exercising the fluid abilities a matrices test rewards. The page on fluid versus crystallized intelligence explains why an activity can affect the two sides of ability differently, and why a single number for the effect of screens is not a coherent quantity.

8 Memory Offloading: The Google Effect and Its Replication

The claim that having information available online changes how memory works came from a 2011 experiment that a high powered replication could not reproduce, and the surviving literature describes a small and context dependent effect on what people choose to remember rather than any loss of the capacity to remember. Sparrow, Liu and Wegner's Google Effects on Memory in Science reported that people primed with difficult questions showed slower colour naming for computer related words, that people who expected information to be saved recalled it less well, and that they remembered where to find it better than what it was. Camerer and colleagues' replication project in Nature Human Behaviour in 2018 re-ran 21 social science experiments published in Nature and Science between 2010 and 2015 with samples about five times larger than the originals, and the Google Stroop effect was among the findings that did not replicate.

The distinction that survives is between what a person stores and what a person can store. Choosing not to memorize something that will be available later is a strategy, and a rational one; it changes the contents of memory, not the capacity that a memory test measures. No study has reported lower working memory capacity, lower digit span or lower delayed recall in people who use search engines or social platforms heavily, as opposed to people who choose not to encode what they expect to find again. The page on IQ and artificial intelligence takes up the same question for the tools that have replaced search since 2022, and the answer is the same for lack of evidence in either direction.

The replication history of this literature is itself a finding. Of the three most cited experimental results about digital technology and cognition, the smartphone presence effect and the Google Stroop effect failed direct replication and the media multitasking result has not been tested longitudinally. A reader who encounters a confident statement about what social media does to memory or attention should ask which study is meant, and then whether it replicated, and on the current record the answer to the second question is usually no.

9 Adults and Older Adults: The Direction Reverses

In adults over 50 the large cohorts point the other way, with regular internet use associated with better cognitive trajectories and lower dementia incidence, and the one randomized trial of social media training in older adults found a benefit on working memory updating. Kamin and Lang analysed 29,576 participants aged 50 to 100 across 14 countries in the Survey of Health, Ageing and Retirement in Europe, with measures in 2013 and 2015, and reported in the Journals of Gerontology that the cross-lagged effects ran in both directions but that internet use had the larger effect on subsequent cognitive functioning. Cho, Betensky and Chang followed 18,154 dementia free adults aged 50 to 64.9 in the Health and Retirement Study for up to 17 years and reported in the Journal of the American Geriatrics Society in 2023 that regular internet users had about half the hazard of dementia of non-regular users, a hazard ratio of 0.57 that held at 0.54 after adjustment for self-selection into use and at 0.62 after adjustment for signs of decline at baseline, with a U-shaped relationship to daily hours in which the lowest risk sat between a tenth of an hour and two hours a day.

Myhre, Mehl and Glisky assigned 41 healthy older adults to learn and use Facebook, to keep an online diary as an active control, or to a waiting list for eight weeks, and reported in the Journals of Gerontology that the Facebook group improved on a composite measure of updating, the executive function associated with complex working memory, while the controls did not, with no differential change on the other cognitive measures. The sample is small and the result is a pilot, but it is the only randomized test of social media use on a cognitive outcome in any age group, and it went the way the cohort studies go.

The reversal is not a paradox. In late adulthood the alternative to online activity is often less stimulation and less social contact, and both are risk factors for decline that the page on IQ and longevity and the page on whether IQ changes with age discuss. In childhood the alternative is sleep, reading and play. The same exposure sits in a different opportunity cost at different ages, which is one more reason a single sign for the effect of social media on cognition is not a meaningful quantity, and why the cohort evidence for adults should not be read as licence any more than the toddler evidence should be read as alarm.

10 Well-Being Is Not Ability, and the Flynn Reversal Predates the Feed

Two arguments are routinely borrowed to support the claim that social media lowers intelligence, one about adolescent well-being and one about falling test scores, and neither is about social media and IQ. The well-being literature is the one with the large effects in public debate and the small effects in the data. Orben and Przybylski's specification curve analysis across three datasets and 355,358 adolescents found a negative association between digital technology use and well-being that explained at most 0.4 percent of the variation, and their time use diary study of 17,247 adolescents in Ireland, the United States and the United Kingdom, using diary measures rather than recall, found little evidence of substantial negative associations at any time of day, including before bed. The page on IQ and mental health explains why well-being and measured ability are different constructs with different correlates, and the page on anxiety and IQ examines the one route by which mood does affect a test score, which is performance on the day.

The falling scores argument rests on the reversal of the Flynn effect observed in several countries, and its timing rules social media out as a cause. Bratsberg and Rogeberg's analysis of Norwegian conscription data in PNAS in 2018 covered men born from 1962 to 1991 and showed that the rise, the turning point and the subsequent decline could all be recovered from variation within families, which points to environmental causes that vary between brothers and rules out the genetic and between family explanations. The turning point falls among cohorts born in the mid 1970s, who sat their conscription tests around 1993, before any social platform existed, and the youngest cohort, born in 1991, was tested around 2009 to 2010. The page on the Flynn effect and the page on average IQ by generation set out what the reversal does and does not show, and it does not show anything about a technology that arrived after it began.

The honest summary is that the two strongest public arguments for a cognitive cost of social media each rest on a dataset that measures something else. That is not evidence of safety, and the toddler and sleep findings above are evidence of specific costs. It is evidence that the sentence "social media lowers IQ" has not been tested by the studies most often cited for it.

11 The Evidence by Design

Read by design rather than by headline, the literature sorts into one longitudinal study with proper controls, several large cross sections, a set of experiments with a poor replication record, and cohort studies of adults that point the other way, and the column that matters is what each measured as the outcome.

The column to read is the outcome. One study measured intelligence longitudinally and found no effect of socializing. The cross sections measured composites and questionnaires and found small associations that the longitudinal design attributes to selection. The experiments measured tasks on the day and did not replicate. The sleep studies measured sleep, and the effect is there. The adult studies measured cognition over years, and the association is favourable.

12 What the Evidence Supports, Stated Narrowly

Stated as narrowly as the designs allow, the evidence supports six sentences about social media and IQ, and none of them is the one in the headlines. First, in the only large longitudinal study that measured intelligence with a normed battery and controlled for genetic and socioeconomic confounders, time spent socializing online had no effect on children's intelligence over two years, and the negative cross-sectional correlation at baseline reflected who socializes online rather than what socializing does. Second, screen time in toddlers is associated with small delays on developmental questionnaires in designs that establish direction, and those findings concern children years younger than any platform's users. Third, the mere presence of a smartphone does not measurably impair cognitive performance, on a pre-registered replication and a meta-analysis of 4,368 participants, and the 2017 result should no longer be cited. Fourth, bedtime device use and even bedroom access roughly double the odds of inadequate sleep in children, and short sleep lowers next day performance on the tasks an IQ test contains, which is the documented pathway from screens to scores. Fifth, total screen time is not associated with school grades, while television and video games carry small negative associations among adolescents, consistent with displacement of homework and reading. Sixth, in adults over 50, regular internet use is associated with better cognitive trajectories and about half the hazard of dementia in large cohorts, and a small randomized trial of learning a social platform improved working memory updating.

What the evidence does not support is any claim that social media lowers, or raises, measured intelligence in adolescents or adults, because no study has tested that claim with a normed outcome over a period long enough to matter. It does not support the transfer of the well-being literature, where the effects are real but tiny, to ability. It does not support the attribution of the Flynn reversal to platforms that did not exist when the reversal began. And it does not support the reading of any attention task result as a change in capacity, because the designs that could distinguish capacity from habit have not been run.

The narrow statement points to two practical facts and one open question. The facts are that a device in the bedroom costs sleep and that sleep costs performance, and that habitual multitasking degrades performance on the task in front of a person. The open question is whether years of a particular pattern of use change what a battery measures, and the honest answer on September 19, 2026 is that nobody has measured it. The page on whether you can improve your IQ reviews what does move scores, and the list is short and does not include either abstaining from or indulging in any platform.

13 What This Does Not Say About You

No finding on this page describes an individual, and the ones that come closest describe the night before a test rather than the person taking it. A correlation of 0.10 between online socializing and intelligence at baseline in ABCD means that knowing a child's socializing hours tells you almost nothing about their score, and the longitudinal result means it tells you nothing about how the score will change. An odds ratio of 2.17 for inadequate sleep with bedtime device use describes a population of 125,198 children and says nothing about whether a particular reader slept last night. A null meta-analysis of smartphone presence says that the average effect across 4,368 people is indistinguishable from zero, and an individual who finds a phone on the desk distracting is not contradicted by it; they are simply not the average.

For a person who wants to know what their own profile looks like, the evidence points to conditions of measurement rather than to history of use. An ACIS session scores twenty timed subtests across six domains against an adult reference frame of 3,243 records aged 16 to 90, described in the technical manual, and prints the standard error of every index; the domains include working memory, attention and processing speed, which are the ones the multitasking and sleep literatures concern. The practical implications of this page for that session are the ones the preparation page already gives: sleep the night before, remove the phone, and take the subtests without other streams open, because the time limits are part of what is measured. A reader who does those things has removed every documented pathway on this page from their own score.

What remains is the question the studies have not answered, whether a decade of a particular pattern of use has changed a person's capacity, and no test can answer it for an individual either, because there is no measurement of the same person before the decade began. What a normed battery can do is state where the person stands now, on a stated scale with a stated error, and the page on score versus percentile explains how to read that statement. The national conversation about social media is about populations and about well-being; a score is about one person and about ability, and the two should not be confused in either direction.

14 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. Effect sizes, sample sizes and confidence intervals are quoted from the abstracts and results of the papers as published; the conversion of a standardized coefficient to IQ points is our arithmetic and is labelled as such where it appears.

• Sauce B, Liebherr M, Judd N and Klingberg T. The impact of digital media on children's intelligence while controlling for genetic differences in cognition and socioeconomic background. Scientific Reports, 2022, volume 12, article 7720.
• 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. Associations between 24 hour movement behaviours and global cognition in US children: a cross-sectional observational study. The Lancet Child and Adolescent Health, 2018, volume 2, issue 11, pages 783 to 791.
• Madigan S, Browne D, Racine N, Mori C and Tough S. Association between screen time and children's performance on a developmental screening test. JAMA Pediatrics, 2019, volume 173, issue 3, pages 244 to 250.
• Takahashi I and colleagues. Screen time at age 1 year and communication and problem-solving developmental delay at 2 and 4 years. JAMA Pediatrics, 2023, volume 177, issue 10, page 1039.
• Ward A F, Duke K, Gneezy A and Bos M W. Brain drain: The mere presence of one's own smartphone reduces available cognitive capacity. Journal of the Association for Consumer Research, 2017, volume 2, issue 2, pages 140 to 154.
• Ruiz Pardo A C and Minda J P. Reexamining the "brain drain" effect: A replication of Ward et al. (2017). Acta Psychologica, 2022, volume 230, article 103717.
• Hartanto A, Lua V Y Q, Kasturiratna K T A S, Koh J R, Tng G Y Q, Kaur M and colleagues. The effect of mere presence of smartphone on cognitive functions: A four-level meta-analysis. Technology, Mind, and Behavior, 2024, volume 5, issue 1, pages 1 to 14.
• Ophir E, Nass C and Wagner A D. Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences, 2009, volume 106, issue 37, pages 15583 to 15587.
• Uncapher M R and Wagner A D. Minds and brains of media multitaskers: Current findings and future directions. Proceedings of the National Academy of Sciences, 2018, volume 115, issue 40, pages 9889 to 9896.
• Firth J, Torous J, Stubbs B, Firth J A, Steiner G Z, Smith L, Alvarez-Jimenez M, Gleeson J, Vancampfort D, Armitage C J and Sarris J. The "online brain": how the Internet may be changing our cognition. World Psychiatry, 2019, volume 18, issue 2, pages 119 to 129.
• Carter B, Rees P, Hale L, Bhattacharjee D and Paradkar M S. Association between portable screen-based media device access or use and sleep outcomes: A systematic review and meta-analysis. JAMA Pediatrics, 2016, volume 170, issue 12, pages 1202 to 1208.
• Adelantado-Renau M, Moliner-Urdiales D, Cavero-Redondo I, Beltran-Valls M R, Martínez-Vizcaíno V and Álvarez-Bueno C. Association between screen media use and academic performance among children and adolescents: A systematic review and meta-analysis. JAMA Pediatrics, 2019, volume 173, issue 11, pages 1058 to 1067.
• Sparrow B, Liu J and Wegner D M. Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 2011, volume 333, issue 6043, pages 776 to 778.
• Camerer C F and colleagues. Evaluating the replicability of social science experiments in Nature and Science between 2010 and 2015. Nature Human Behaviour, 2018, volume 2, issue 9, pages 637 to 644.
• Orben A and Przybylski A K. The association between adolescent well-being and digital technology use. Nature Human Behaviour, 2019, volume 3, issue 2, pages 173 to 182.
• Cho G, Betensky R A and Chang V W. Internet usage and the prospective risk of dementia: A population-based cohort study. Journal of the American Geriatrics Society, 2023, volume 71, issue 8, pages 2419 to 2429.

15 Frequently Asked Questions