Editors pick science stories by weighing novelty, human stakes, and surprise, but the filter leaks. A 2021 analysis in PLOS Biology of 1,535 news stories found that 69.5 percent of headlines about mouse studies left the mice out. The wording of a paper’s own title, it turns out, does much of the choosing.
Science wants detail and precision, while journalism wants brevity and stories right now. Wikipedia’s science journalism entry frames that daily tension in one line each: science values facts, numbers, and being right; journalism values immediacy, words, and being right now. The site’s science news coverage works inside that gap. Choosing what to cover is an editorial act, not a mechanical one.
Slow, unglamorous beats lose most often. Ocean heat content, a climate number that quietly keeps rising, rarely delivers the surprise a news meeting wants. Flashy lab results win the slot instead. Two recent studies put measurable shape on that trade.
Where do the selection instincts come from?
The trade’s habits are old. Wikipedia’s entry credits James Crowther, hired as The Manchester Guardian’s scientific correspondent in 1928, with shaping the role, and its account of the field’s aims still lists the tug between being right and being right now. The entry’s own history starts earlier still: almanacs carried natural history notes for centuries, and an 1818 Bengali monthly called Digdarshan printed science items. Classic news values sit closer to the journalism side of that list. Surprise, scale, and human stakes decide most mornings. We covered a connected angle in Science news for students: how a youth newsroom adapts research.
Where do headlines get their claims?
Often from the paper itself. In the PLOS Biology study, researchers sorted 623 Alzheimer’s papers into titles that named mice and titles that did not, then tracked the news coverage each drew. Among papers without mice in the title, only 10.4 percent of news stories mentioned mice in headlines, against 46.2 percent for titles that did. The 2021 paper’s conclusion is blunt: headline writers tend to follow the author’s lead.
Stranger still, understated titles seemed to travel better. Papers that kept mice out of their titles drew 3.9 news stories each on average, against 3.0 for titles that named the animal. A quiet title read as more newsworthy, which the authors did not expect. The finding is observational, so it shows a pattern, not a cause.
Can training change the odds?
A 2026 experiment in PNAS Nexus suggests part of the answer. Researchers gave a short video on research checkpoints, covering funding, sample, statistics, causal claims, and visuals, to 260 German journalists in a survey experiment. Treated journalists wrote correct headlines for 64 percent of study-based stories, against 36 percent for the control group. A 28-percentage-point gain is large for one short video. An exploratory follow-up of 268 real articles hinted at fewer factual mistakes after training.
Where the evidence comes from
The PLOS Biology analysis tracked coverage after the fact, so a hidden third cause, such as journal press practices, cannot be ruled out. The authors did check journal guidelines and found only 39 of 156 journals capped title length. The PNAS Nexus result is experimental, with random assignment and 260 participants, but it tests one training video, not a whole newsroom. Neither study offers a formula, and the German sample may not travel. The samples stay visible here: 623 papers, 1,535 news stories, 260 journalists. Each is a small slice of a big trade, and the authors say so.
Conclusion
How science news is chosen has no secret list behind it. Newsrooms pick with old news values, then inherit framing from the papers and titles in front of them. Mouse studies shed their mice, and careful numbers struggle for airtime. The 2021 and 2026 studies point to a modest fix: train the filter, and question the title. For readers, the practical habit is to check what a headline left out, species and sample first.
