Does a mood spread through a chat? What the study showed
The most cited evidence that feeling travels through written words is a 2014 Facebook experiment. It is worth reading closely, including the editorial notice attached to it.
The best known experimental evidence is Kramer, Guillory and Hancock’s 2014 study, which altered the amount of emotional content in the Facebook News Feed of 689,003 people and reported that reducing positive expressions led people to post less positively and more negatively. It measured a feed of public posts, not a conversation, and the journal later published an editorial expression of concern about it.
The claim that a mood spreads through written words gets repeated as though it were settled. It rests heavily on one experiment.
That experiment is real, it is large, and the gap between what it tested and what people say it proved is wide enough to be worth laying out.
The 2014 experiment, and what it actually changed
Experimental evidence of massive-scale emotional contagion through social networks was published in the Proceedings of the National Academy of Sciences on 17 June 2014 by Adam Kramer of Facebook’s Core Data Science Team, Jamie Guillory of the University of California, San Francisco, and Jeffrey Hancock of Cornell.
The significance statement puts the scale at 689,003 people. What was manipulated is stated in the abstract: the researchers tested whether emotional contagion occurs outside in-person interaction by reducing the amount of emotional content in the News Feed.
So the intervention was subtraction rather than addition. Nothing was written to anyone. Posts that already existed were shown or not shown.
A feed of posts is not a conversation between two people
This is the distinction that disappears when the study is summarised. A News Feed is a stream of public and semi public posts from a wide circle, consumed in one direction. A chat is an exchange between people who know each other, where both sides write.
The authors are careful about what they take their result to show. Their framing is that emotions expressed by others on Facebook influence our own emotions, and that in-person interaction and nonverbal cues are not strictly necessary for contagion. That is a statement about exposure to expressed emotion, not about the dynamics of a two person thread.
Nothing in the design speaks to whether your friend’s bad week bleeds into your replies, which is the question most people have in mind when they repeat the finding.
What the abstract reports, and what it leaves out
The direction is clear. When positive expressions were reduced, people produced fewer positive posts and more negative posts; when negative expressions were reduced, the opposite pattern occurred.
The magnitude is not in the abstract. The published summary gives the direction of the effect and the size of the sample without stating how large the shift was, so this page will not quote a number for it. That is a real limitation on how much weight the study can carry in an argument, and it is the thing most summaries quietly skip.
The honest form of the finding is therefore narrow: with a very large sample, altering exposure to emotional posts moved the emotional tone of what people wrote, in the direction you would expect. How much it moved is not something the abstract supports anyone in saying.
The tendency of language to lean positive in the first place is useful background for reading any result of this shape.
The editorial expression of concern, stated plainly
On 22 July 2014 the Proceedings of the National Academy of Sciences published an editorial expression of concern relating to the paper, written by the journal’s editor in chief Inder M. Verma and carried in volume 111, issue 29, page 10779.
The notice is attached to the paper’s record and travels with it, so a citation of the study that does not mention it is an incomplete citation. This page states that the notice exists, who wrote it and where it appeared, and does not characterise its contents beyond that.
That is not a reason to discard the result. It is a reason to cite the paper with the notice attached, which is what anyone quoting it ought to be doing.
Why your own export cannot answer this about you
The question people actually want answered is personal: does this person’s mood get into mine. A chat export cannot settle it, and it is worth being precise about why rather than waving at complexity.
An export is observational. Two people whose messages move together may be influencing each other, responding to the same events, or simply being two people who were already alike. The 2014 study could separate those because it intervened; nothing read after the fact can.
There is a second reason, specific to how a tool like this works. Portraitor’s home page says you choose one person to produce a portrait of, that their messages are read, and that everyone else stays context. A reading of one side’s writing is not a measurement of transfer between two people, whatever the two sets of messages look like side by side. The wider list of what a reading cannot reach covers the rest of that boundary, the research on misreading tone covers the nearest thing it can speak to, and the home page describes the method itself.
Frequently asked questions
Frequently asked questions
Is there real evidence that emotions spread through text?
There is one large experiment. Kramer, Guillory and Hancock altered emotional content in the Facebook News Feed of 689,003 people in 2014 and reported that reducing positive expressions led to fewer positive and more negative posts, with the opposite pattern when negative expressions were reduced.
How big was the effect?
The published abstract does not state the magnitude. It reports the direction of the change and the size of the sample, so no figure for the size of the shift is quoted here.
Did the study look at private messages?
No. The manipulation was to the News Feed, a stream of posts, rather than to conversations between two people.
Is there anything attached to the paper I should know about?
Yes. The journal published an editorial expression of concern on 22 July 2014, written by its editor in chief and carried at volume 111, issue 29, page 10779. It travels with the paper’s record.
One experiment, a very large sample, a direction but no published magnitude, and an editorial notice attached. That is the whole of the evidence, and it is less than the claim it gets used for.
A mood is not the only thing that moves between two people in a thread. Telling someone your good news has a name of its own in close relationships research, and capitalization in relationships: replies to good news reports what the 2004 study measured about the reply that came back.
Your conversations, read the way you would read someone else’s.
Portraitor takes a chat export you already have, anonymises it in your browser, and returns a structured reading of how you come across. Identifying details are removed in your browser before anything is sent.
See how it works →- Kramer, A. D. I., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences, 111(24), 8788 to 8790.
- Verma, I. M. (2014). Editorial expression of concern. Proceedings of the National Academy of Sciences, 111(29), 10779.
- Portraitor, home page.