Why is science reporting so often wrong?
Because a chain of incentives between the laboratory and the reader systematically amplifies certainty and strips caveats — and the errors are introduced at several points, not only in newsrooms.
Where distortion enters:
The press release. Research has found that much exaggeration in science news is already present in the university press release, not added by journalists. A widely cited BMJ study found that where press releases contained exaggerated claims about causation or human relevance, the resulting news coverage usually reproduced them — and where releases were accurate, coverage largely was too.
Universities compete for profile and funding, and press offices are incentivised accordingly.
Correlation reported as causation. The most common single error. Observational studies find associations; headlines state that one thing causes another.
Animal and cell studies presented as human findings. "Cures cancer" frequently means it killed cells in a dish or shrank tumours in mice. Most such results never translate.
Effect sizes omitted. A relative risk increase of 50% sounds alarming; if the absolute risk moves from 2 in 10,000 to 3, it is not. Relative risk without absolute risk is the single most misleading common practice.
Single studies reported as settled. Science advances through accumulated evidence, and any individual study — particularly a small or surprising one — is provisional. Coverage rarely conveys where a result sits in the literature.
Sample size and preprint status frequently unmentioned.
False balance. Presenting a fringe position alongside an overwhelming consensus as though they are comparable, which distorted climate coverage for years.
Structural causes: shrinking specialist science desks, general reporters covering technical material on deadline, and headlines written by someone who did not read the paper.
How to read science news: find the original paper; check whether it was humans, animals or cells; look for absolute numbers; note the sample size; and treat a single striking finding as interesting rather than established.