Keep this life skill for later
Sign in to save
← All life skills

Learning & thinking · a life skill

Deliberate Source Checking

trace it upstream

A claim reaches you after several retellings: a paper, then a press release, then a news story, then a post. Each one makes it shorter and surer. Following it back to where it started takes about fifteen minutes, and the version you find there is usually smaller, and more interesting, than the one you saw.

  • Moderate to start
  • High impact
  • Learn it in 2-4 weeks
  • One claim, followed to its source

Words that tend to go missing on the way to you

  • in mice
  • may
  • was linked to
  • in 30 people
  • early results
  • compared with
  • slightly
one claim, four retellings
A claim traced upstream from a post to the study it came from Four documents stacked from the source at the top to the reader at the bottom. The study says compound X was associated with longer life in thirty mice. The press release says compound X may slow ageing, dropping the mice. The news story says a pill could slow ageing. The post says scientists find an anti-ageing pill. Arrows run up the left side from the post to the study: that is the reader going upstream. Between the documents, notes say what each retelling lost: “in 30 mice”, “compound”, then “could”. At the bottom, what the reader would now say: one study linked compound X to longer life in mice, and nobody has tested it in people. the study · source Compound X was associated with longer life in 30 mice [lost: “in 30 mice”, “associated”] [lost: “compound”: now a pill] [lost: “could”] the post Scientists find anti-ageing pill you met it here ↑ copy these words upstream “One study linked compound X to longer life in mice. Nobody has tried it in people.”

An illustrative claim, not a real study. Read it bottom to top: each retelling sounds surer, and each one leaves out something the source said.

For the next claim that matters to you

Follow it back.

Pick one claim you'd act on or pass on: a health tip, a statistic, a quote. Then walk it back, one retelling at a time.

  1. Copy the exact words

    Paste the claim as you saw it, with where and when. Memory rounds a claim up, and you'll want the exact wording to compare against later.

  2. Find who said it first

    Follow the links in the story. Look for “a study published in…” or a press release. Search a distinctive phrase in quotation marks, like "slow ageing" compound, and look for the earliest date.

  3. Read the original's own words

    For a study, read the abstract: who or what was studied, how many, and compared with what. For a quote, read the whole paragraph around it. For a number, find the table it came from.

  4. Mark what each retelling lost

    Hedges like “may” and “linked to”. The subjects: mice, cells, a small group. The size of the effect. The comparison. Usually each step drops one.

  5. Say it the way the source would

    Write the claim in one sentence the original would agree with. If you pass it on, pass on that, with a link to the source.

Compare two versions

What went missing?

Paste the claim you saw and the sentence it came from. The checker lists the hedging words the source used and the version you saw dropped, and the strong words it added. It's filled in with the example above.

Hedges the source used, and the version you saw dropped

    Strong words the version you saw added

      It only matches a list of words, so it can't tell you what the study did. Read the source for that. Nothing you paste leaves this page.

      Where the exaggeration starts

      Often before the news.

      Advice to readers

      Release exaggerated58%

      Release didn't17%

      Cause from a correlation

      Release exaggerated81%

      Release didn't18%

      People from animal research

      Release exaggerated86%

      Release didn't10%

      Share of news stories that exaggerated, by whether their press release had. Sumner et al., BMJ, 2014.

      Petroc Sumner and colleagues traced 462 health press releases from 20 leading UK universities, the papers behind them, and 668 news stories built on them. The releases themselves already overstated things often: 40% gave more direct advice than the paper did, 33% turned a correlation into a cause, and 36% drew conclusions about people from animal work.

      When the release exaggerated, the news usually did too. When it didn't, the news mostly didn't either. The study is observational, so it can't prove the releases caused it. But it does say where to look: blaming the reporter stops one step short of the source.

      The idea

      Why it works.

      Every retelling is a summary, and summaries lose the hedges first. A careful sentence about thirty mice and a possible link becomes a headline, because a headline has room for a verb and a noun. Nobody needs to lie for the claim to grow; it only needs to be retold a few times.

      So the version closest to the source is almost always the most accurate one you can find, and often the only one with the details you need to judge it: who was studied, how many, and compared with what.

      On the trail

      Habits that help.

      • “A study found” with no link is a reason to look. A story that names the journal and the date is easier to check, and usually more careful.
      • Search the quote in quotation marks. Then look for the earliest result, not the first one.
      • Read the methods line first. Who, how many, compared with what. The abstract usually says it in one sentence.
      • A cold trail is an answer. “I couldn't find where this came from” is worth saying out loud.

      Go deeper

      Keep going.

      Keep this going in the app.

      HourLife turns these ideas into a daily system — goals, habits, focus, and reflection in one private place.

      Download on the App Store Download on the Mac App Store See everything HourLife does