"A Study Shows..." — How to Decode Research Journalism

A research paper is published. A university press office reads it and writes a press release summarising the findings. A journalist reads the press release — rarely the paper itself — and writes a news article. A sub-editor reads the article and writes a headline. A social media editor selects the most compelling sentence from the article and turns it into a shareable post.

At each of those five steps, information is simplified, abbreviated, and made more striking. The full picture narrows. Caveats disappear. 'May be associated with' becomes 'causes'. 'In a small sample of 47 patients' becomes 'scientists discover'. 'Under certain conditions in a laboratory mouse model' becomes 'breakthrough in human health'.

This is not primarily a story of deliberate deception. It is a story of structural incentives and the cumulative effect of small compressions. Each person in the chain is doing their job: the press office is making the research sound important enough to attract media attention; the journalist is writing something readable within a tight deadline; the sub-editor is writing a headline that will compete for clicks among hundreds of others. Each compression is individually defensible. The cumulative effect can be a headline that bears only a distant relationship to what the original paper actually found.

The Sumner et al. finding

In 2014, a team led by Petroc Sumner at Cardiff University published a study in the BMJ that quantified exactly how much exaggeration enters the science news chain, and where it enters. The researchers tracked 462 press releases from twenty UK universities and matched them to associated news stories and academic papers.

Their findings were striking. Forty per cent of press releases contained exaggerated advice (directional claims about what the findings implied people should do), 33% contained exaggerated causal claims (correlation reported as causation), and 36% contained exaggerated inference to humans from animal research. When press releases contained exaggeration, news stories were substantially more likely to contain the same exaggeration — 58%, 81%, and 86% respectively. When press releases did not contain exaggeration, news story exaggeration rates dropped to 17%, 18%, and 10% (Sumner et al., 2014, p. g7015).

The implication was important: the primary source of exaggeration in science journalism was not journalists but academic press releases. Universities were, in seeking media coverage for their research, introducing distortions that then propagated into news coverage.

What this means for readers

The chain of simplification means that headlines about scientific research carry a known, documented distortion risk — even when every person in the chain is acting in good faith. The farther the headline is from the original paper, the more likely it is that caveats have been dropped, effect sizes have been inflated, and the scope of the findings has been overstated.

Recognising this is not a counsel to dismiss science journalism — much of it is valuable and accurate. It is a reason to treat individual research headlines with proportionate caution, and to understand what questions to ask before updating beliefs based on a 'study shows' claim.