Explore journal

When a link is not a cause

A health headline can be accurate and still invite the wrong conclusion.

An unlabeled scatter plot in a research notebook beside everyday objects.
AI-generated illustration · illustrative scene, not a documentary photograph.

The verb that does too much

A headline reports that one habit is associated with a better health outcome. By the time the claim reaches a conversation, “associated with” has become “improves”, and an observation has quietly acquired the force of advice. No new evidence has appeared; the verb has simply done more work than the study necessarily permits.

Correlation describes a relationship between variables. Causation asserts that changing one factor brings about a change in another. Health researchers need evidence to move from the first claim to the second. Observational associations can be informative without settling that question.

A town invented for the example

Consider a wholly fictional survey in which people who visit a lakeside walking route report better wellbeing. We have no measurements beyond that invented association. It would be premature to conclude that the route itself produced the difference.

People with more free time might both visit the lake more often and experience less stress. People already feeling well might be more inclined to go out. Alternatively, walking there might genuinely help. Those explanations are compatible with the headline’s simple association, but they imply different stories.

The example contains neither actual findings nor a recommendation to change a health routine. Its purpose is to show why the direction of an explanation and the role of other factors cannot be read directly from two variables moving together.

Evidence has a design

The US National Library of Medicine explains the distinction between correlation and causality and the importance of study design. Randomised experiments can strengthen causal inference by helping separate an intervention from other differences between groups. They are not a magic exemption from questions about measurement, comparison or interpretation.

An observational study should not be dismissed merely for being observational. Some questions cannot readily be addressed by assigning people to experiences. The appropriate response is to ask what the design can establish and how other evidence bears on the proposed explanation.

Read the claim at its original strength

For an English reader, this is partly an exercise in preserving grammatical restraint. “Was associated with” does not mean “was proved to prevent”. “May contribute” does not mean “will cure”. Words that look cautious can carry the most important information in the sentence.

The opposite mistake is to treat uncertainty as evidence that nothing has been learned. A relationship worth investigating is still a result; it simply should not be promoted beyond its evidential rank.

Before passing on the next headline, try reconstructing its modest version: what was observed, among whom, and what remains unresolved? The aim is neither automatic belief nor automatic dismissal. It is an account whose confidence has not outrun its reasons.

Vocabulary & expressions

causal inference
inferência de causa e efeito

The design supports a stronger causal inference.

confounding factor
fator que pode distorcer uma relação observada

Free time is a possible confounding factor in the fictional example.

premature
prematuro

That conclusion would be premature.

restraint
contenção, cuidado

The wording shows appropriate restraint.

to bear on
ter relação com, ser relevante para

Other evidence may bear on the explanation.

Examples written by Better English Today.

02 · CHECK YOUR UNDERSTANDING

What stayed with you?

Choose one answer for each question. You can return to the text at any time.

1. What changes between the headline and the conversation?
2. Why invent several explanations for the lakeside association?
3. Does the article dismiss observational research?

Your turn to use it.

Rewrite an imaginary health headline at three levels of certainty: observation, possible explanation and established cause. Explain what evidence would be needed before using the strongest version.

The study observed… One explanation may be… A causal claim would require…

Try it aloud or in your notebook. This is an open response, not an automatically graded task.

Sources & further reading

Source: US National Library of Medicine · Finding and Using Health Statistics: correlation and causality

Original text, examples and activities by Better English Today. References support the factual background. Links do not imply partnership.