Two ways of knowing what people do
There are broadly two ways to learn about behaviour. One is to ask: people tell a researcher how often they exercise, how much they read, what they eat or how they spend their evenings. The other is to observe: a device logs screen use, a till records a purchase, a turnstile counts entries, or a researcher watches and notes down what happens.
Trend stories often mix the two without saying so. A claim that “people are reading less” might rest on a questionnaire in which people estimate their reading, or on records of book sales or library loans. The difference is not a technicality. It can change the conclusion.
Why self-reports and records often differ
Researchers have long noted that stated behaviour and recorded behaviour do not always match. Several causes are typically involved.
Memory. Recalling how many times you did something last month is difficult, especially for routine activities that leave no distinct memory. People often reconstruct an answer from a general impression, and that impression is easily skewed by recent or unusual events.
Rounding and estimation. Asked how long they spend on an activity, people tend to give round numbers or familiar figures. Estimating quantities across a long period is imprecise, and errors are not always evenly spread in both directions.
Wanting to look good. Some behaviours are seen as virtuous and others as embarrassing. People may over-report the first kind and under-report the second, even in anonymous surveys. This tendency, known as social desirability, is a recognised issue for topics like diet, exercise, reading and time spent on devices.
Different definitions. A person who says they “cook at home most nights” may count reheating something as cooking. A question about “exercise” may be read to include walking or not. The respondent’s private definition may differ from the researcher’s.
Self-image. People often answer in line with the kind of person they believe themselves to be, which can differ from what they actually do on a particular day.
When the gap tends to be larger or smaller
The size of the gap is not constant. It is often smaller for rare, memorable events, such as a holiday abroad or a major purchase, and larger for frequent, low-attention habits. It is often larger for topics with a moral tone. It also depends on how the question is asked: shorter recall windows, specific time frames and concrete wording can help, whereas vague wording like “regularly” invites loose answers. The effect of phrasing is covered under question wording.
Evidence on screen use is a well-known example. Researchers comparing self-estimates with device logs have generally found that people are imperfect at estimating their own use, which is why discussions of screen time often distinguish between the two kinds of measure. Similarly, reading habits are especially prone to flattering answers because reading carries a positive image.
Observed data has its own problems
It would be wrong to conclude that records are simply the truth and surveys the error. Observed data has limits of its own.
- Coverage. A log shows activity on one device, app or store, and not elsewhere. A person who reads on a borrowed device or buys with cash leaves a gap.
- Meaning. A record shows an event, not why it occurred. A device may be on while the person is not looking at it.
- Who is observed. People who agree to install trackers, or who use a loyalty card, may not be typical.
- Awareness. Knowing they are watched can change how people behave, at least for a while.
- Access. Records held by companies are often available only as summaries, so outsiders cannot check how they were produced, as discussed under transaction records and platform data.
So each approach is partial. Self-reports reach private habits and reasons but are fuzzy. Records are precise about what they capture but narrow and silent about motive.
Why the gap matters for trend claims
A trend is a change over time, and the two kinds of evidence can disagree about change as well as level. If the social pressure around a behaviour grows, people may report it more or less than before even if their conduct is steady. For instance, a rise in claimed healthy eating might reflect a stronger sense that it is the right answer, not a change in what is eaten. In the other direction, if recall is steadily more difficult because a habit has become automatic, reports may drift without any change in behaviour.
This is one reason claims about trends in food, reading or leisure should be read carefully when they rest only on questionnaires. A figure showing what share say they do something is evidence about answers. Treating it as evidence about actions needs an extra step that may or may not be justified.
Making sensible use of both
Combining the two often works best. Observed records can anchor the quantity, while surveys explain context and motives. Where a diary-style approach is used, such as in time use surveys, respondents record activities close to when they happen, which may reduce some recall problems, though not the wish to look good.
When reading a claim, it is useful to ask whether the behaviour was reported or recorded, over what period, with what definition, and whether another source points the same way. Agreement between a self-report and an independent record is reassuring. Disagreement is not a failure; it is information about how people understand and describe their own lives.