Association is not causation: reading a study

An association means that two measured things are connected in the data. On its own, it does not establish that one causes the other. When reading an observational study, check who took part, what was measured, when events happened and which alternative explanations were considered. This is a guide to reading research, rather than changing a child's care.

What was observed, and what was assigned?

In observational research, investigators examine data without assigning participants the intervention being studied. They may use questionnaires or existing records, for example. NIH explains different types of clinical studies, including the value of observational approaches for questions that cannot simply be investigated through an experiment.

In an interventional study, the intervention is assigned according to a protocol. However, not every interventional study uses random allocation. If a report calls something an experiment, it is still worth checking how participants were allocated and what was compared. The label alone does not tell you everything about the method.

Three picture cards: rain, umbrellas and puddles

Imagine a fictional classroom exercise. Street photographs show umbrellas and puddles appearing together. Their shared appearance would not justify concluding that opening umbrellas creates puddles. Rain could explain both observations. The illustration above refers to this simple example, rather than reporting an actual scientific finding.

In a real study, identifying another relevant factor can be much harder. The CDC Field Epidemiology Manual describes confounding, where another factor distorts the association being investigated. A useful reading question is therefore not only “Are these things connected?” but also “What else could affect this comparison?”

Rain is easy to picture in our card exercise. In real data, some important circumstances may not have been measured. Saying that a result was adjusted for certain factors does not automatically remove every alternative explanation. Look for the factors the researchers actually included, alongside the limitations they acknowledge. A long adjustment list is something to read, rather than a guarantee to accept without question.

The order of events matters too

Consider another invented example: do library visits lead to an interest in books? It is also possible that an existing interest encourages families to visit more often. If we only have answers collected at one moment, we cannot simply invent the sequence. This example contains no real family data and makes no claim about the effect of library visits.

The US National Library of Medicine discusses time order and alternative explanations when assessing causation. For an initial reading, look for whether researchers followed people over time or asked about past events later. Knowing which event happened first matters, but that fact alone does not settle the entire causal question.

Turn a headline into a question, not an instruction

When you read “X is associated with Y”, open the original paper and find the participants and measured outcome. A questionnaire score, a recorded diagnosis and a family's reported experience are not interchangeable measures. A study of adults in one country may not answer a question about a child elsewhere.

Try a brief note in this form: “In this group, the authors examined these two things. They reported this association and these limitations.” If you cannot find one detail, leave a question mark rather than filling the gap with “proved”. Your note is a way of following the meaning, not an expert assessment of the paper's validity.

Notice changes in wording between the paper and a social post. “Associated with” can become “leads to” in a short caption, although those statements make different claims. Keep the original wording beside your note. You do not have to resolve the whole question before deciding that the stronger caption needs more support.

What to save with the link

Save the paper's title, date, population and one question in your own words. Distinguish the date of a news story from the date of the study. If the paper is a preprint, record that status too. Publication status and research design describe different things; one does not automatically tell you the other.

For a number, look at the group behind it. Our guide to 1 in 31 and 1 in 127 explains why similar-looking statistics can describe different populations. For a summary of several studies, see systematic reviews and meta-analysis. These are separate checks that the word “association” cannot answer on its own.

Do not throw observational research away

“An association alone is not enough” does not mean “this study is worthless”. Observational evidence can provide important knowledge and contribute to investigating causes. Methods, assumptions and the wider body of evidence matter. Some questions cannot ethically or practically be studied by randomly assigning people to a particular exposure or experience.

It would therefore be misleading to dismiss every observational study, just as it would be misleading to announce that a single association has settled a causal question. A parent does not need to recreate an epidemiological analysis at the kitchen table. A useful first step is noticing what the paper claims, what remains uncertain and what a headline has added.

Frequently asked questions

Does the word association mean the research is wrong?

No. It describes a relationship in data. The questions are how the relationship was examined and what can reasonably be concluded. Researchers can accurately report an association without claiming they have established its cause.

Does a very large sample prove causation?

Not by itself. More participants do not automatically fix measurement problems, participant selection or alternative explanations. You still need to understand how the data were collected and compared.

Should I change something immediately after reading a headline?

This article does not recommend treatment or changes to care. If a report relates to a specific decision for your child, save the original source and your question for an appropriate professional.

Space for a story as well

This evergreen explanation uses sources checked on 17 October 2026. It is not a report of a new discovery. Away from research headlines, explore our own TAMITOS book and current edition information for shared story time. A book about understanding differences does not replace research appraisal or professional care.