Why Sad Posts Go Viral Less Often
A famous experiment said that getting people's bodies going makes them share. It failed in 2024. The newspaper pattern behind sad posts going viral less often still stands, though its explanation is less certain.
Key takeaways
- Jonah Berger and Katherine Milkman (2012) found that, among every New York Times article over three months, those evoking awe, anger or anxiety were emailed on more than those evoking sadness.
- Berger's 2011 jogging experiment, which tied sharing to arousal alone, did not hold in two larger 2024 replications led by Skyler Prowten.
- For publishers, the test is what a piece leaves readers wanting to do: content that leaves them moved and still is less likely to travel.
A researcher once asked students to jog on the spot for a minute, then offered them a news article and asked whether they'd email it to someone. Most of the joggers said yes. Far fewer of the students who had sat still did. It became a widely cited explanation of why some posts go viral and sad ones go viral less often: get people's bodies going and they pass things on.
More than a decade later the jogging result failed twice. The explanation built on top of it did not fail with it, and the reason why is the most useful thing in this story.
2011, the jogging experiment
Jonah Berger published the experiment in Psychological Science with forty participants. The idea was clean. If physiological arousal, the state that readies a body to act, makes people share, then arousal produced by anything at all should do it, even arousal that has nothing to do with the article.
Berger tells the story at book length in Contagious, in the chapter on emotion. There it carries the causal half of the argument. As the book reports it, about three in four joggers chose to email the article, against roughly one in three of those who had sat still.
A small study with a striking result and a simple mechanism is exactly the kind that travels. It was easy to retell, easy to remember and easy to turn into advice: make your audience feel something strong enough to move.
Forty people is a small sample for a result that large. With so few participants, a handful of unusual individuals can produce a dramatic difference by chance, and the first published version of an effect is often its biggest. None of that means the result was wrong. It means the result was a reason to look harder rather than a fact to build on. It also tends to be the kind that needs repeating before anyone leans on it.
2012, one newspaper's articles
A year later Berger and Katherine Milkman published a very different kind of evidence in the Journal of Marketing Research. They took every article the New York Times published over three months, scored each for the emotions it evoked, and tracked which ones readers emailed on.
| Emotion an article evoked | Feels | Activation | Shared |
|---|---|---|---|
| Awe | Good | High | More |
| Anger | Bad | High | More |
| Anxiety | Bad | High | More |
| Sadness | Bad | Low | Less |
Positive articles were shared more than negative ones. Within the negative emotions, though, the activating ones went viral more readily and sadness less so. The abstract reports that the pattern held after allowing for how surprising, interesting or useful an article was and how prominently it was featured, and the paper adds experiments that its authors read as showing that the emotions in the content drive sharing.
The outcome was the newspaper's own record of which articles readers emailed on, which is what makes the design unusual. Rather than asking people in a laboratory what they might share, it looked at what readers actually sent to someone else, article by article, over a whole season of news.
This is a sturdier kind of evidence than the jogging study. Nothing was chosen to fit a hypothesis, and the outcome was real behaviour by real readers. It has its own limits. It covers one newspaper's email button, before ranking algorithms sat between readers and what they share, and the emotions were judged from the articles rather than asked of the people who sent them.
2024, two failed replications
Skyler Prowten and colleagues went back to the jogging study. They ran it twice, with more than a hundred people the first time and more than a hundred and fifty the second, kept the same jogging manipulation and updated the outcome to sharing on social media. Neither run found any effect of arousal on willingness to share.
A replication is simply the same study run again, ideally with more people, to see whether the original result appears a second time. Larger samples matter because they make a chance result less likely to masquerade as a real one. Two failures with groups roughly three and four times the size of the original are hard to dismiss as bad luck.
Their conclusion is careful: the results cast doubt on the idea that physiological arousal, on its own and apart from other factors, changes what people decide to share. That is a narrower target than it first appears. The jogging study separated arousal from content on purpose. The newspaper study never did. Content that makes you angry or awed is activating and also about something, and the replications say nothing about that combination.
So the chronology leaves a division, and it is worth stating carefully. Arousal cut loose from what you're reading is the part that didn't survive. The newspaper pattern, that content evoking high-arousal emotions went viral more readily than sad content, still stands as an observation. What is less certain now is the reason behind it, because the cleanest evidence that arousal itself does the work is the evidence that failed.
What survives for publishers
Start with the practical test. Read a piece back and ask what it leaves a reader wanting to do. An article that leaves people moved and still is, on this evidence, less likely to go viral than one that leaves them alarmed, indignant or astonished, however good it is.
That has an awkward edge. Anger and anxiety spread more in the newspaper data. It's not an argument for outrage. A finding about what spreads says nothing about what deserves to, and content built to provoke can travel widely while costing its publisher the trust it was meant to build.
None of this means every activating post will go viral. Most content of any kind reaches few people, and timing, audience and luck all play their part. What the newspaper data offer is a tilt in the odds. Across thousands of articles, the ones that activated readers were more likely to go viral than the ones that left them still, and over enough posts a tilt is worth having.
Awe is the better lesson for anyone who wants work to go viral without leaning on anger. It is activating and positive, and it went viral more readily in the 2012 data. A charity appeal makes the point. The same child can be shown so that you feel sorrow, or so that you feel indignant that the situation exists. Only the second is activating. Whether the sorrowful version raises more money from people who do see it is a separate question that sharing data cannot settle.
It also helps to be honest about what you are measuring. Shares, donations, sign-ups and brand recall are different outcomes, and content that wins on one can lose on another. A sad appeal that goes viral less often may still persuade the people who see it. A campaign that only counts shares will quietly push every brief towards anger and alarm.
A related study adds a caution about time. In 2011 Berger and Eric Schwartz found that interesting products drew an initial burst of conversation that did not last, while products people were reminded of by their everyday surroundings kept being discussed. That research is about products rather than emotions, but its warning carries over: what sparks a burst of sharing and what keeps something talked about over months are not the same thing.
The honest summary of thirteen years is that sad posts go viral less often than activating ones, that the pattern tracks activation as well as pleasantness, and that the cleanest experiment for why was the one that broke. For anyone judged on reach, test an activating version against a calm one before launch. The research gives a direction. Only your own audience can give you the size, and a simple comparison of two versions of the same piece will tell you more than any general rule about emotion.
Common questions
Why don't sad posts go viral as often?
In the newspaper study, sadness was the one low-activation emotion, and articles that evoked it were emailed on less often than articles evoking awe, anger or anxiety. Sad content still spreads; it simply went viral less often on average.
Does positive content spread more than negative?
In the newspaper data, yes, on average. But activation mattered too: anger and anxiety, both unpleasant, were shared more than sadness. Pleasantness and activation work together rather than one replacing the other.
Did the jogging study fail?
Its result did not hold in two larger replications published in 2024, which used social media sharing as the outcome. Those replications tested arousal produced by exercise, not arousal produced by the content itself, so they do not test the newspaper pattern directly, though they weaken the clearest evidence that arousal itself is the cause.
Sources
Berger, J., & Milkman, K. L. (2012). 'What makes online content viral?' Journal of Marketing Research, 49(2), 192-205. All New York Times articles published over a three-month period; positive content more viral than negative, but virality partially driven by physiological arousal; high-arousal positive (awe) and negative (anger, anxiety) content more viral; low-arousal deactivating content (sadness) less viral. The abstract reports the results hold when controlling for how surprising, interesting or practically useful content is and how prominently it was featured, and that experimental results demonstrate a causal effect of specific emotions driven by the level of activation induced. Abstract read at OpenAlex 2026-10-06. Verified at the published record 2026-10-06. doi:10.1509/jmr.10.0353
Berger, J. (2011). 'Arousal increases social transmission of information.' Psychological Science, 22(7), 891-893. doi:10.1177/0956797611413294. Forty participants; those made physiologically aroused by jogging on the spot, for sixty seconds as Contagious reports it, were more willing to email a news article. DOI confirmed at the Association for Psychological Science journal page, 2026-10-06; sample size as reported in the 2024 replication.
Berger, J. (2013). Contagious: Why Things Catch On. Simon & Schuster. Ch. 3, Emotion, pp. 93-124. Read in full by this publication, 2026-09-02.
Prowten, S. D., Walker, E., London, B., Pearce, E. N., Napoli, A., Chenevert, B., Clevenger, C. T., & Smith, A. R. (2024). 'Does physiological arousal increase social transmission of information? Two replications of Berger (2011).' Psychological Science, 35(9), 1025-1034. Two replications, N = 111 and N = 160, with the sharing measure updated to social media, finding no effect of incidental arousal on willingness to share; the original, Berger (2011), Psychological Science 22(7), 891-893, had N = 40. Abstract checked at OpenAlex 2026-10-06. doi:10.1177/09567976241257255
Berger, J., & Schwartz, E. M. (2011). 'What drives immediate and ongoing word of mouth?' Journal of Marketing Research, 48(5), 869-880. Dataset of everyday conversations covering more than 300 products, a large field experiment across several cities, and a controlled laboratory experiment with real conversations. More interesting products received more immediate word of mouth but, contrary to intuition, not more ongoing word of mouth over multiple months or overall; products cued more by the environment or more publicly visible received more word of mouth both immediately and over time. Verified at the published record 2026-10-01. doi:10.1509/jmkr.48.5.869