There is no subject in this magazine's territory where the gap between the confidence of public discussion and the state of the evidence is wider. Screens are discussed as a moral question, in the vocabulary of temptation and discipline, and the research is recruited afterwards to support positions already held.
Treated as an evidence question it looks quite different. Less alarming in some respects, more troubling in others, and much less settled than either side would like.
The measure is the first problem
Nearly all the large studies use screen time: hours per day, usually reported by a parent or by the young person.
Consider what that number combines. A video call with a grandparent. A film watched with a parent who talks about it. A game requiring planning. An endless feed. Homework. Each of these is a different activity with different demands, and screen time treats them as interchangeable.
No other area would tolerate this. Nobody studies the effects of paper time, aggregating novels, tax forms and cigarette packets. Yet a great many published conclusions rest on exactly that aggregation.
Self report makes it worse. People are poor at estimating their own device use, and parent estimates of children's use correlate imperfectly with logged data. The measurement error is substantial and not random.
The effects are smaller than the coverage
When large datasets are analysed carefully, associations between screen time and wellbeing or attainment are consistently found to be small. Work using specification curve methods, which run every defensible version of an analysis rather than one chosen version, has repeatedly shown that the reported association depends heavily on analytical choices, and that the typical effect sits close to trivial in size.
Small does not mean zero, and at population scale a small effect can matter. But small also means that a headline describing a crisis is not describing what the study found. For why this distinction is so consistently lost, see effect size in plain words.
Which way does the arrow point
The second structural problem is direction. A child who is struggling may use screens more as a consequence. A child with fewer alternatives available may use screens more because of the alternatives, not instead of them.
Longitudinal designs help but do not settle it, and the more careful studies that look at within person change over time tend to find smaller effects than cross sectional comparisons between children.
Randomised experiments could settle it and are largely infeasible for meaningful durations. This is a chronic condition in research on children's lives, and it means the evidence base will remain partly inferential no matter how much data accumulates. We describe the general problem in correlation and cause.
The one strong finding
There is a genuinely robust result in this area, and it is about the youngest children.
Toddlers learn substantially less from video than from a person doing the same thing in front of them. This has been demonstrated many times, across tasks, and it is the reason guidance from bodies including the Royal College of Paediatrics and Child Health treats very early video differently from later use.
It also has a known exception that clarifies the mechanism: when the video is interactive, as in a live call where the person responds to the child, the gap narrows considerably. The operative variable appears to be contingency, the same thing that runs through conversation, not vocabulary. Video without response is a poor teacher not because it is a screen but because it does not answer.
Displacement is the real mechanism
If you want one framing that fits the evidence better than the others, it is displacement.
Time spent on a device is time not spent on something else. Where that something else was sleep, physical activity, or interaction with a person, the consequences of losing it are reasonably well established on their own terms. Where it was a third repeat of a television programme, the loss is negligible.
This framing has the advantage of being answerable at the level of a family without any moral content. The question is not whether screens are bad. It is what this use is instead of, and that is a question a household can actually consider.
Why official guidance is cautious and vague
Readers often find institutional guidance on screens frustratingly non committal. The vagueness is a fair reflection of the evidence rather than a failure of nerve.
Bodies that reviewed the literature carefully have generally concluded that the evidence does not support specific hour thresholds, and have declined to invent them. Where numerical limits are issued, they are usually acknowledged as pragmatic rather than derived. A guideline that admits its own basis is doing better science than one that supplies a confident number, even though the confident number is what everybody wants.
What a reader can reasonably conclude
That screen time as a single quantity is a poor unit of analysis and should be treated with suspicion whenever it appears in a headline. That the strong finding concerns very young children and video without response. That displacement is a better lens than harm. That the reported effects on older children's wellbeing are small and contested, which is neither a reassurance nor an alarm but is what the evidence says.
And that the moral framing, whichever direction it runs in, is doing something other than describing children. This magazine has no position on how much television anybody's child watches, and no evidence that would justify one.