Chapter 07 · Field guide

How Do You Read a Study Cited in an X Post?

Short answer

Find the actual paper, then check five things: whether it was peer reviewed, whether it was an experiment or an observation, who or what was studied, how big the effect was, and whether later work agrees.

Key takeaways

  • Read the paper, or at least its abstract, instead of stopping at the press release. Press releases and posts tend to drop the caveats.
  • Preprints are useful and fast, but servers such as medRxiv say plainly that they have not been peer reviewed.
  • Observational studies find associations. Only well-designed experiments, usually randomized, can show cause with confidence.
  • A statistically significant result can still be tiny. Look for the size of the effect as well as the word significant.

"Scientists prove dark chocolate makes you smarter." The post has 60,000 likes and a stock photo of a chocolate bar. Somewhere behind it there may be a real study. It might have tested 30 college students for two weeks on a memory game. It might have tested rats. It might not exist at all. Finding out which takes about five minutes and a handful of questions.

How do you find the actual study?

Follow links until you reach the paper itself, or search its title, DOI or lead author, and treat a post with no traceable study as unverified.

Most viral science posts are three steps from the source: a post quotes a news article, which quotes a university press release, which summarizes a paper. Each step trims nuance. Look for clues along the way: a journal name, an author, a date, or a DOI (a digital object identifier, which looks like 10.1234/abcd.5678). Paste a DOI after https://doi.org/ to reach the paper.

If you only have a title, search it in quotes, or try Google Scholar, which often links to a free copy. If none of this turns anything up, write that in the "original found" field of the claim-check worksheet. "I could not find the study" is a real result.

Has it been peer reviewed, or is it a preprint?

A preprint is a paper shared before independent experts have reviewed it, so treat its findings as early and open to change.

Preprint servers such as medRxiv and arXiv speed up science and are used by serious researchers. They also say clearly that the work has not been peer reviewed and should not guide health decisions on its own. Peer review filters out a lot, though it guarantees nothing. Reviewers check whether methods and conclusions make sense, but they rarely redo the analysis.

A press release deserves even more caution. It is written to get attention, and research on science news has found that exaggerations in news stories often start in the press release.

Was it an experiment or an observation?

Observational studies can show that two things go together, while randomized experiments are needed to show that one causes the other.

Suppose a study finds that people who eat dark chocolate score higher on memory tests. That is an association. Chocolate eaters might also sleep more, earn more or exercise more. Any of these confounders could explain the gap. A randomized trial, where a coin flip decides who gets chocolate and who gets a placebo, removes most of those differences.

Study type What it can show Common catch
Cell or lab study A possible mechanism Cells in a dish are not people
Animal study Effects in mice, rats or other animals Many results do not carry over to humans
Observational study An association in real populations Confounders can explain the link
Randomized controlled trial Cause and effect, if well run Often small or short
Systematic review or meta-analysis The overall weight of many studies Only as good as the studies included

If the post says "proves" and the study is observational, the post is overstating it.

Who or what was actually studied?

Check whether the subjects were people, animals or cells, how many there were, and whether they resemble you.

A study of 24 elite cyclists tells you about elite cyclists. A study of mice fed a dose equal to 40 chocolate bars a day tells you about mice on an unusual diet. Small samples swing by chance; a result in 30 people is much less settled than one in 30,000. Look for:

  • the number of participants, usually in the abstract
  • how long the study lasted
  • the age, health and background of participants
  • whether there was a comparison or control group

How big was the effect?

A result can be statistically significant and still too small to matter, so look for the size of the effect in plain units.

"Statistically significant" means the result was unlikely to arise by chance under the study's assumptions. It does not mean large. With enough participants, a memory score improvement of half a point out of 100 can be significant. Ask what the change was in real terms: points, minutes, kilograms or cases per 1,000 people. If the paper reports a relative risk, the percentage check will convert it to absolute numbers. The chapter on misleading statistics explains why that matters.

Does the rest of the evidence agree?

One study is a data point, so check whether systematic reviews or later studies support it, and whether it has been corrected or retracted.

Science moves by accumulation. A single surprising finding is interesting; it becomes reliable when other teams repeat it. Systematic reviews, such as those published by Cochrane, gather every relevant study on a question and weigh them together. If a review exists, it usually beats any one paper.

Last, check that the study still stands. Retracted papers keep circulating long after a journal pulls them. Search the title in the Retraction Watch database and look for a notice on the journal page. Then ask one question of the post itself: does it describe the study the way the study describes itself? If not, you have your answer, whatever the research found.

Frequently asked questions

How do I find the study behind a post?

Look for a DOI, a journal name or the lead author in the post or linked article. Search the paper's title in quotes, or paste the DOI after https://doi.org/. Google Scholar often shows a free version if the journal page is paywalled.

Is a peer-reviewed study always right?

No. Peer review catches many problems but not all, and reviewers rarely rerun the analysis. Treat a single peer-reviewed study as evidence, and give more weight to systematic reviews that combine many studies.

Why do mouse studies get so much attention?

They often report striking results and make good headlines. Many treatments that work in mice or in cells fail in humans, so a post that skips the word mice is overstating the finding.

How can I tell if a study was retracted?

Check the journal page for a retraction notice and search the title in the Retraction Watch database. Retracted papers keep circulating on social media for years.

Sources

  1. medRxiv, About (preprints have not been peer reviewed)
  2. Retraction Watch Database
  3. Cochrane, systematic reviews of health evidence
  4. DOI Foundation, Resolve a DOI
  5. Google Scholar

Examples in this chapter are invented for teaching. Facts about platforms were checked in November 2026 and can change. See how we research.