A simple number with a complicated origin
Studies of media use, mood and sleep often rely on a single figure: how much time someone spends on screens. It sounds straightforward, but the number depends heavily on how it was gathered. Two studies of the same group could report different amounts of screen time simply because one asked questions and the other read a device.
Asking people: self-report
The most common method is a questionnaire. Participants may be asked how many hours they used their phone yesterday, on a typical weekday, or in a typical week. Self-report is cheap, works for large samples and can ask about specific activities.
It also has well-known weaknesses.
- Recall error. Short, frequent, unplanned checks of a phone are easy to forget, so people often misjudge a total built from many small moments.
- Averaging. “A typical day” requires mentally blending very different days.
- Social desirability. People may round down, or up, depending on what they think is expected of them. The effect of social desirability on honest answers is a general issue in survey work.
- Question wording. Asking about “social media” versus “your phone” versus “the internet” can produce different totals, a pattern described under question wording.
More broadly, this is an instance of the gap between self-reported behaviour and what can be observed. The error is not always in one direction, and it can differ between groups, which makes comparisons trickier.
Reading the device: logs and trackers
Many phones and apps record usage, and some research projects ask participants to share this data or install monitoring software. Logs avoid memory problems and can show timing, such as late-night use.
They have limits of their own.
- Device is not person. A tablet or computer shared among family members records the use of whoever picked it up.
- Gaps in coverage. A log from one phone does not capture a laptop, a games console or a television.
- Activity is blurred. Screen-on time may include a video playing in the background, or a screen left on while someone does something else.
- Different definitions. Operating systems and apps may count “use” in different ways, so figures from different tools are not always comparable.
- Opt-in bias. People who agree to be tracked may differ from those who decline, an example of selection effects.
Platform records, meanwhile, show what happens on one service only, and the blind spots of platform data apply here too.
Why the two often disagree
When researchers compare self-report with logs for the same people, the two measures often differ noticeably, and the differences vary from person to person. Typically the correlation between them is positive but far from perfect. Some people overestimate, some underestimate, and the pattern can depend on the type of activity and the person’s age.
That matters because a finding such as “heavier use goes with lower mood” might look different depending on which measure was used. If self-reports are the measure, the link could partly reflect how people who feel low describe their own habits, not only what they actually did. Researchers sometimes call this a reporting bias, and it is hard to rule out without objective data.
Other ways to count
- Experience sampling: participants are prompted at random moments to say what they are doing right now, reducing reliance on memory.
- Time diaries: people log activities across a day. Time use surveys apply this approach to everyday life, and diaries can reveal what screens displace.
- Direct observation or screen recording: very detailed, but expensive, intrusive and limited to small groups.
Each has a trade-off between accuracy, cost, sample size and how much it alters behaviour. Being watched can change what people do.
What minutes cannot say
Even a perfect count of minutes is a thin measure. An hour video-calling relatives and an hour scrolling an anonymous feed are logged in the same way, yet they are very different experiences. Researchers increasingly try to separate type of use, timing, and the person’s circumstances, but this makes studies smaller and harder to compare. Total screen time is also a rough label that lumps work, school, entertainment and communication together.
For this reason, a claim built only on hours should be read cautiously, particularly when the article does not say how the hours were obtained. Useful questions are: was the figure self-reported or recorded, over how many days, on which devices, and for which people?
Takeaway for readers
Screen time is best treated as an estimate with a margin of uncertainty that depends on the method. Neither self-report nor logs is a perfect truth, and they are often most informative when used together. This article is a general explanation of measurement and is not advice about anyone’s own habits or health.