Does Screen Time Really Cause Depression? What 21,797 Swedish Twins Reveal

The question behind the headlines

A popular claim sounds straightforward: more screen time makes young people depressed. Yet a correlation between two behaviours can emerge for several reasons. Does digital entertainment change mental health? Can someone already struggling spend more time online? Or do family experiences and inherited traits influence both?

The Swedish CATSS research was designed to examine those alternatives. It did not ask whether all digital technology is good or bad, and it did not test a phone-ban intervention. It examined patterns of recreational screen use and internalizing problems—depression- and anxiety-related symptoms—over adolescence and into early adulthood.

How the Swedish twin study worked

Tong Gong and colleagues analysed data from 21,797 participants in the Child and Adolescent Twin Study in Sweden. Parents reported some screen activities around age nine; participants subsequently reported screen-use duration at ages 15 and 18. Mental-health questionnaires were available at ages nine, 15, 18 and 24. Subsets of participants could also be linked to Swedish registers for clinical depression, anxiety diagnoses and dispensed medication.

The researchers separated relatively passive entertainment, such as watching television or videos, from interactive activities involving gaming, chatting, blogging or browsing. Heavy use was defined as more than six hours per day in the adolescent screen-time categories—not a universally validated medical threshold.

They examined simultaneous associations, longitudinal associations and later clinical outcomes. Critically, they then compared twins within the same family when the twins differed in their reported screen use. This design is unusually informative because many family conditions that are hard to measure in an ordinary survey are shared within a twin pair.

The numbers behind a long follow-up

The headline sample of 21,797 does not mean that every participant answered every question at every age. Of the twins included at the age-nine stage, 16,027 completed the age-15 assessment, 13,050 completed the age-18 assessment and 5,772 completed questionnaires at age 24. In other words, later observations came from progressively smaller subsets of the original cohort.

That matters when interpreting longitudinal conclusions. Attrition can make later responses less representative if continued participation is related to wellbeing, family circumstances or technology habits. Researchers used statistical models suited to their available data, but no analytical method can transform missing responses into directly observed experiences.

Clinical follow-up was also a different data source from symptom questionnaires. Registers included diagnosed depressive or anxiety disorders and dispensed medication, capturing people who accessed care or received recorded treatment. Those measures provide valuable corroboration, but do not include every person living with symptoms or tell us exactly what caused a diagnosis.

The first results looked worrying

In ordinary population models, higher adolescent recreational screen use was associated with more internalizing symptoms and with subsequent recorded depression and anxiety outcomes. Associations were often stronger during mid- and late adolescence and among girls. Some analyses indicated that heavy interactive screen use at age 15 was associated with roughly two- to threefold higher subsequent clinical risks.

These are relative associations from particular models, not a statement that two or three out of every group of young people will become ill. The study does not establish that all interactive activities have the same impact; gaming, chatting and browsing were grouped in ways that cannot isolate modern apps or particular content.

What changed when the researchers compared twins?

The most revealing stage of the investigation was the co-twin comparison. If one twin reports more recreational screen use than the other, does that twin also show higher symptom scores or clinical risk once the shared family background is taken into account?

The associations became substantially smaller within discordant twin pairs. After correction for multiple statistical testing, they were no longer statistically significant. The researchers interpreted this as evidence that shared familial factors—including genetic predispositions and shared childhood environment—may explain an important part of the population-level association.

That is not equivalent to proving that screens never influence well-being. Within-pair estimates can be imprecise, individuals within a pair still have different friends and experiences, and measured screen time is an imperfect proxy for what actually happens online. A result that no longer passes a statistical threshold is not proof of an exact zero effect.

Why the twin method is valuable—and what it cannot remove

An ordinary observational comparison might find that adolescents spending six hours online differ from those using screens for two hours. The groups may also differ in inherited behavioural tendencies, household rules, parental education, pre-existing mental-health symptoms and many other characteristics. Statistical adjustment helps only for factors that are measured adequately.

A co-twin design asks a different question. If two people raised in one family reported different screen habits, how closely did their mental-health outcomes differ? Identical twins share effectively all inherited DNA, while fraternal twins share about half on average. Both typically share many aspects of childhood family life. The CATSS investigators drew on these comparisons to reduce the influence of shared family and inherited characteristics.

Importantly, some of the study's later within-pair analyses combined identical and fraternal twins because complete longitudinal measurements within pairs were less common. These pooled comparisons do not completely control inherited DNA in the way an identical-twin-only contrast can. Nor can any co-twin approach automatically control differences in friends, school experiences, bullying, stressful events or what happens online.

Understanding the evidence threshold

Researchers run many models when studying several ages, outcomes, screen types, weekdays and weekends. The more statistical tests are performed, the greater the chance that an apparently significant finding will occur by coincidence. Multiple-testing correction requires stronger evidence before a result is called statistically significant.

In the published CATSS analysis, several within-pair estimates remained above zero, but their uncertainty was greater and they did not survive the study's multiple-testing threshold. That is stronger grounds for caution than simply saying the relationship 'disappeared'. It also means we should not selectively highlight one uncorrected positive estimate while ignoring the wider set of tests.

The practical distinction is between the direction of an estimate, its uncertainty and whether the data support a clear causal conclusion. A large population-level association can shrink when shared family characteristics are accounted for, yet still leave open whether certain individuals, kinds of online activity or vulnerable groups experience a direct effect. The published paper also states that its analyses were not preregistered. This is a transparency limitation to consider when interpreting a large set of exploratory statistical comparisons.

Can mental health affect screen use too?

Direction matters. The published study also explored whether symptoms at one age preceded later screen use, alongside the reverse direction. Some analyses suggested bidirectional patterns for interactive activity and internalizing symptoms.

For a young person, time online may be recreation, social contact, avoidance, a coping strategy, or several things at once. Without experimentally changing exposure and measuring outcomes, those possibilities cannot be cleanly separated. The published study should not be transformed into a simple statement that one side necessarily causes the other.

Why one published version matters

An earlier preprint of this project described its findings in more causal-sounding terms. The version published in Nature Human Behaviour on 8 October 2026 is the appropriate source for readers. Its central conclusion is more cautious: significant population associations were largely attenuated by co-twin comparisons and did not survive correction for multiple testing.

This is a useful example of how science can change while a paper moves through review. At TwinPare Research, conclusions should follow the definitive publication and its actual methods—not the most dramatic phrasing available in an earlier draft.

Time online is not the same as the experience online

One hour spent talking with a trusted friend, one hour of gaming with classmates and one hour of viewing distressing content are not necessarily interchangeable exposures. The study's broad passive and interactive categories are helpful for investigating historical patterns but cannot tell us precisely which kinds of content, social comparison, harassment or supportive interaction mattered to each participant.

Technology also changed during the CATSS study. Data collection began in 2004, before smartphones and algorithmic social feeds became routine parts of adolescent life. A result about the activities measured in those birth cohorts therefore cannot be copied directly into a claim about every social platform used by teenagers in 2026.

The next scientific questions are about context and susceptibility: which online experiences coincide with worsening symptoms, how sleep and social relationships interact with use, and whether carefully designed interventions actually change outcomes. Those questions require more than a count of minutes on a screen.

What should families and readers take away?

The study does not justify ignoring distress, nor does it validate one universal daily screen-time limit. A more useful conversation may ask when screen use is displacing sleep, physical activity, schoolwork or supportive relationships—and whether a young person is experiencing harm, bullying or persistent low mood.

Because the study is observational, it cannot tell parents exactly how much time a particular child should spend on a phone. Household routines, age, individual needs and the type of digital activity matter. Young people with persistent depression or anxiety symptoms deserve appropriate professional support regardless of their screen habits.

The research is based on Swedish twins from specific birth cohorts and self- or parent-reported behaviours. Technology, social platforms and the surrounding social context change quickly. Findings should not automatically be applied to every country, age group or new online platform.

The TwinPare perspective: similar genes, different lives

Twin research is valuable not because identical twins always get the same outcome, but because their shared starting point lets us test explanations that ordinary comparisons cannot easily untangle. Seeing an association weaken within families reminds us that human health is not a simple scorecard of one habit.

There is also an important limit to the TwinPare Health & Fitness connection: wearable sleep, activity and heart-rate trends may help users reflect on everyday routines where supported, but they do not measure the cause of depression, diagnose anxiety or establish a person's risk from screen time. Research should spark better questions, not unsupported individual predictions.

The most defensible conclusion is therefore neither 'screens cause depression' nor 'screens are harmless'. Young people's experiences with screens and mental health are linked in population data, while family and genetic contexts seem to account for much of that link. Understanding why—and for whom digital life can become harmful—remains an active scientific question.