The More a Psychology Study Goes Viral, the Less Likely It's True
June 21, 2026 · Keith Stanovich, How to Think Straight About Psychology~5 min read
That clever psychology finding you shared last year — the one with the perfect counterintuitive twist, the one your friends actually clicked — has worse odds than you'd hope. Researchers trained a machine-learning model to read more than 14,000 psychology papers published in top journals between 2000 and 2019 and score how likely each was to hold up. The pattern that fell out is uncomfortable: the more media coverage a study got, the less likely it was to replicate. The headline you remember was, on average, the weaker science.
Virality is a filter, and it filters for the wrong thing
It helps to see media attention not as a reward for good work but as a sieve. What makes a study shareable is a short list: it's surprising, it's clean, it turns on a single counterintuitive flip you can say in one sentence. Power poses make you bolder. Smiling makes you happier. Willpower runs out like a muscle. Those are the studies editors chase and readers forward. But the same traits that make a result spread are the ones that make it fragile — a single dramatic effect, found once, is exactly the kind that tends to shrink or vanish when someone runs it again. Spotlight and sturdiness are pulling in opposite directions. And it isn't that fame is rare and trustworthy: the analysis found highly cited papers were no more likely to replicate than the rest. Citations don't certify truth either.
What "replication" actually showed
This isn't one model's hunch. Back in 2015, a consortium of 270 researchers tried to redo 100 well-known psychology studies under the original conditions. Only about a third to forty percent produced the same significant result the second time. On average, the replicated effects came in at roughly half the size of the originals. Read that slowly: take a shelf of textbook findings, run them again carefully, and most of the drama thins out or disappears. Not because the first researchers lied — mostly they didn't — but because a single study is a noisy instrument, and a striking number from one run is often just the noise that happened to look like a signal.
A machine-learning analysis of 14,000+ psychology papers in top journals (2000–2019) found that the more media coverage a study received, the less likely it was to replicate; highly cited papers fared no better. Background: the Reproducibility Project: Psychology (2015) retested 100 studies and reproduced only about a third, with effect sizes roughly half the originals. Framework: Keith Stanovich, How to Think Straight About Psychology (converging evidence over a single dramatic study). Popular-science explainer; the link is correlational, not causal.
The unit of trust is the pile, not the headline
This is the spine of Keith Stanovich's How to Think Straight About Psychology, and it's the part worth tattooing somewhere. Real knowledge in psychology isn't built on the one brilliant experiment. It's built on converging evidence — many studies, different labs, different methods, slowly stacking until the same finding shows up no matter who looks. A claim earns trust by surviving repetition, not by being astonishing the first time. So the headline that screams about one study has picked the wrong unit. The right unit is boring on purpose: the slow accumulation, the replication you never read about, the effect that keeps showing up after the buzz has moved on.
Don't overcorrect into cynicism
Now the guardrail, because the easy misread is "so psychology is fake." It isn't, and the data say so. The relationship is correlational — coverage tracks with fragility; it doesn't prove the media broke anything. And there's genuinely good news buried in the same numbers: between 2010 and 2019, average replication scores rose. Preregistration, bigger samples, and a field that got publicly embarrassed and decided to fix it are working. So the lesson isn't to sneer at every study. It's to hold a single splashy result loosely, and to trust most the findings dull enough that nobody made them go viral.
What to do with the next viral headline
Here's the move, small enough to actually use. When a "new study finds…" lands in your feed, ask one question before you share: is this one study, or many? A lone, surprising result is a hypothesis wearing a headline's costume — interesting, worth watching, not yet something to rebuild your habits around. Wait for the words "replicated," "meta-analysis," "across multiple labs." Notice that the most trustworthy science usually arrives without fireworks, because by the time something is solid, it's no longer surprising. The thrill of a finding and its truth are, more often than we'd like, inversely related. Treat your own astonishment as a yellow light, not a green one.
A headline is one study in a costume. Truth is the boring pile of studies behind it.
The more a finding surprises you, the longer you should wait before you believe it.
Sources: a 20-year machine-learning analysis scoring the replicability of 14,000+ psychology papers in top journals (2000–2019), which found media coverage inversely related to replication likelihood and citations unrelated to it (reported via Big Think, "Psychology studies that go viral are likelier to be bogus"); the Reproducibility Project: Psychology (Open Science Collaboration, 2015), which replicated 100 studies with roughly a third reproducing and effect sizes about half the originals. Framework from Keith Stanovich, How to Think Straight About Psychology (falsifiability, converging evidence, why media spotlight ≠ truth). Popular-science reflection — the coverage–replication link is correlational, not causal; not professional or clinical advice.