Myth 12 · Myths
The 7% rule that its own author says is misused
Trainers often say only 7% of a message comes from words, 38% from tone and 55% from body language, but the studies behind it used single words about feelings, and the researcher says it does not apply to everyday talk.
- Albert Mehrabian's 1967 studies
- University of California, Los Angeles, United States
- 1967
- 3 min read
On this page
Before you readWhat would you do?
A trainer tells your team that words carry only 7% of meaning, so prompts to AI tools "barely matter" as long as you sound confident. What is the best response?
Pick an answer now. The file tells you whether you were right.
01The file in one minute
1The myth
Only 7% of any message comes from words, 38% from tone and 55% from body language.
2The studies
Two small 1967 experiments used single words spoken with a tone of liking or dislike, plus photos of faces.
3The limit
Mehrabian says the numbers apply only when people judge feelings or attitudes, not everyday communication.
4The lesson for AI
A chatbot reads your words, so the wording of a prompt carries most of the meaning.
02Timeline
1967
Mehrabian and colleagues publish two small studies about words, tone and faces.
1971
Mehrabian's book Silent Messages discusses the findings for a general audience.
Later decades
The 7-38-55 numbers spread through business and presentation courses.
03The myth
In 1967 the psychologist Albert Mehrabian and colleagues ran two small experiments. Volunteers heard single words, such as "dear" or "terrible", spoken in a tone that showed liking, dislike or neutrality, and in one study also saw photos of faces. They then judged how the speaker felt about the listener.
When words, tone and face disagreed, people relied more on tone and face. Combining the results, Mehrabian suggested weights of about 7% for words, 38% for voice and 55% for facial expression, for judging feelings and attitudes in this kind of setting.
Over the years the numbers spread through presentation courses and business books as a general rule of communication. Mehrabian has said that his findings do not apply unless a speaker is talking about their feelings or attitudes.
04What the evidence says
The experiments were narrow: single words, small groups of university students, and a question about liking. They say little about a lecture, a contract or directions to the station, where the words carry most of the meaning. If words were worth only 7%, you could follow a film in a language you do not know.
What the research does suggest is useful: when someone's tone or face clashes with their words, listeners tend to believe the tone and face. That matters for apologies, praise and bad news.
The AI angle: chatbots work almost entirely with words, and they do very well with them. The words in your prompt carry most of what the AI understands, so clear words matter more than ever.
The idea behind this caseLesson: The four-part prompt →
05How to spot the next one
- Check the original study before repeating a precise percentage about people.
- Ask what the study measured: single words about feelings, or real conversations.
- When your tone and words might clash, as with bad news, take care with both.
- In written messages and AI prompts, put the important details in clear words.
07 — Sources
- Decoding of inconsistent communications · Journal of Personality and Social Psychology · 1967
- Inference of attitudes from nonverbal communication in two channels · Journal of Consulting Psychology · 1967
- Albert Mehrabian · Wikipedia · 2024