Why text messages get misunderstood more than you think
A 2005 study found senders are reliably overconfident that their tone will land, that it is worse in writing than in speech, and that knowing someone well does not help.
When you write a message you hear it in your head in the tone you meant, and that private soundtrack is not attached to the message when it leaves. A 2005 study by Justin Kruger, Nicholas Epley, Jason Parker and Zhi-Wen Ng found that senders are systematically overconfident that their tone will land, that this happens more over writing than over speech, and that knowing the other person well does not fix it.
The interesting part is not that messages get misread. Everyone knows that. It is that people are reliably wrong about how often it happens to them, and that the reason is mechanical rather than careless.
You hear your own tone. Nobody else does.
The British Psychological Society’s research digest puts the mechanism in one sentence: as we compose a message we read it to ourselves silently with our intended tone, forgetting as we do that the recipient might not read it that way at all.
That is the whole trick. The sender is reading a message with a soundtrack. The recipient is reading the same words with no soundtrack, or worse, with their own. NYU Stern’s summary of the research makes the same point about what writing strips out: a message carries no body language, no facial expression and no vocal inflection, whatever emoticons the sender adds.
So the gap is not a failure of effort or of goodwill. It is a missing channel, and the sender is the one person who cannot notice it is missing, because their own head keeps supplying it.
What Kruger and colleagues tested in 2005
The study is Egocentrism Over E-Mail: Can We Communicate as Well as We Think?, published in the Journal of Personality and Social Psychology. NYU Stern’s own account of it says the authors started from the premise that social judgement is inherently egocentric, then tested how confident senders were that messages carrying sarcasm, seriousness, anger, sadness or humour would be understood as intended.
Two results came out of it. Misunderstandings were common, including between people who already knew each other. And senders were largely unaware of that, typically overconfident that their intent had been grasped. The research digest adds that the overconfidence showed up more for written messages than for spoken communication, and that it persisted whether or not sender and recipient knew each other well.
Worth stating plainly: this was email in the mid 2000s, among study participants, not messaging apps in 2026 among your own contacts. The mechanism it describes is general, the specific setting was not. Nothing here is a measurement of your conversations.
As the researchers put it in the digest’s closing quotation, if understanding another person were only a matter of turning sentences into ideas there would be no room for misunderstanding, and since it is not, there is.
Both sides are overconfident, not just the sender
This is the part that usually gets dropped when the study is repeated secondhand. The research digest records that recipients also overestimated their own ability to read the intended tone of a message.
So a misread message often involves two people who are each sure the exchange was clear, in opposite directions. That is a different problem from one careless writer, and it is why going back over an argument line by line so rarely settles who was being unfair.
It also connects to something strange about how we judge descriptions of ourselves. Why personality descriptions feel accurate covers the other side of that coin, which is how easily a general statement feels precisely aimed once it is pointed at you.
The one thing that measurably helped
The digest reports a tested intervention rather than a piece of advice. When participants were asked to read a sarcastic message out loud to themselves in a neutral tone before sending it, they became better than usual at judging how unlikely the recipient was to catch the sarcasm.
Saying the words without your own tone appears to break the loop, because it stops your head from supplying the soundtrack. The reverse also showed up: in another of the studies, senders who watched a video of a joke being performed before emailing it became more likely to overestimate how funny the recipient would find it. Fresh exposure to the intended delivery made the gap wider.
This was measured on sarcasm and humour under study conditions. It is not a technique for resolving a disagreement, and nothing in the research says it is.
Why an export cannot tell you how you landed
This is the honest limit of reading back over your own conversations, and it applies to any tool that does it, this one included.
A chat export contains what was written. It does not contain how any of it was received, because that never got typed. So an analysis of an export can describe how your messages read to an outside party, which is genuinely useful and is close to the thing the 2005 study says you cannot see for yourself. What it cannot do is report how the other person actually took them.
Portraitor reads one participant in a conversation and returns inferences about how that person comes across, each with an estimate of confidence attached. Analyse my WhatsApp chat with AI describes what that looks like, and what a chat analysis cannot tell you is the list of things it does not reach. The received half of the conversation is on that list and always will be.
If what you want is the plain version of how your own messages read, what your texts sound like to everyone else covers it without the study, and politeness theory explains why a polite word can be the thing that lands badly.
Frequently asked questions
Frequently asked questions
Who found that people misread text messages?
Justin Kruger, Nicholas Epley, Jason Parker and Zhi-Wen Ng, in a 2005 paper in the Journal of Personality and Social Psychology titled Egocentrism Over E-Mail: Can We Communicate as Well as We Think?
Does it help if the other person knows you well?
The research digest reports that the overconfidence persisted whether or not sender and recipient knew each other well, and that misunderstandings were common even between people who already knew each other.
Do emoji fix the problem?
NYU Stern’s account of the research says a written message carries no body language, facial expression or vocal inflection whatever emoticons the sender inserts. The study consulted here does not report emoji as a remedy.
Is there anything that was shown to help?
One thing, under study conditions. Participants who read a sarcastic message aloud to themselves in a neutral tone before sending it became better at judging how unlikely the recipient was to catch the sarcasm.
Can analysing a chat export tell me how my messages were received?
No. An export holds what was written, not how it was read, so no analysis of one can report the other person’s reaction. It can describe how the messages read to an outside party, which is a different question.
A figure often quoted alongside this one is the claim that the words themselves carry only a small fraction of a message. The 7-38-55 rule traces that number back to the two studies it came from.
You cannot hear your own messages the way anyone else does. That is the finding, and it is the reason reading them back from outside your own head is worth anything at all.
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 →