Is the 7-38-55 Rule True? The Experiment Behind '93% of Communication Is Nonverbal'
The short version: no, not as a rule for communication. The 7-38-55 figures come from two 1967 experiments in which students judged how much a speaker liked someone from single recorded words, such as 'maybe', said in different tones, and from photos of faces. When the word, the tone and the face disagreed, tone and face won. That's all the studies show. The researcher who ran them has said the formula doesn't apply unless someone is talking about feelings or attitudes. If 93% of meaning were nonverbal, you could follow most of a lecture in a language you don't speak.
The company has booked a communication workshop. The trainer opens with a slide: words 7%, tone of voice 38%, body language 55%. "Science," she says. "Only 7% of your message is what you actually say."
Kevin's boss takes this seriously. His next all-hands presentation has very few words on the slides and a great deal of walking about. Nobody remembers what it was about, which, according to the slide, is fine, because they got 93% of it.
Kevin wonders why the workshop came with a 40-page handbook.
Where the numbers come from
In 1967 a psychologist published two small experiments. In the first, students listened to single recorded words, some positive ("dear", "thanks"), some neutral ("maybe"), some negative ("terrible"), spoken in a positive, neutral or negative tone. They had to judge how the speaker felt about the person being addressed. When the word and the tone disagreed, the tone mostly won.
In the second experiment, one word, "maybe", was recorded in three tones and paired with photographs of faces showing liking, neutrality or dislike. Again students judged the speaker's attitude. The face counted for a bit more than the tone, roughly 3 to 2.
Putting the two results together gave a formula for how much each channel counts when judging liking: 0.07 for the word, 0.38 for the tone, 0.55 for the face. That's the whole origin of 7-38-55.
What the experiments did and didn't test
| In the experiments | In the workshop slide |
|---|---|
| Single words, no sentences | Any message, any length |
| Judging whether the speaker likes someone | Understanding what someone means |
| Word, tone and face deliberately contradicting each other | Normal conversation, where they usually agree |
| Photos of faces, not moving bodies | "Body language" |
| Small groups of university students | Everyone, everywhere |
The researcher himself has been clear about it: unless a speaker is talking about their feelings or attitudes, the equations don't apply. The finding is real but narrow. When someone says "fine" through clenched teeth, you believe the teeth. That's a useful thing to know, and it's a long way from "words don't matter".
Statistics for noobs: a result only travels as far as the test
Every study has a setting: who took part, what they did, what was measured. How far a result can be carried outside that setting is called external validity. A tightly controlled lab test can be perfectly correct about its own conditions and still say little about anything else.
Before you apply a number, ask three questions: who was tested, on what task, and does my situation look like theirs? Here the answers are: small groups of students, judging liking from one word, and no.
A quick test you can run
If 93% of meaning came from tone and body language, three things would follow. A phone call, with no face, would lose more than half of every message. You could follow most of a lecture in a language you don't speak, as long as the speaker was expressive. And email would be almost impossible. None of these is true, which is a decent sign that the rule has been stretched well past its evidence.
Where it costs you money
Market studies get stretched the same way. A pattern found in one country, in one decade, with one kind of stock, gets quoted as a law of investing. Before trusting it with your money, check what the original test covered, and whether anyone has tried it on other markets and later years.
The moral
Kevin's boss gave the next presentation in total silence, using only gestures and a confident expression. Afterwards, the team was asked what they had understood. Responses ranged from "a reorganisation" to "a fire drill". It was the annual budget. The trainer has been booked again, this time to explain the handbook.
What this can't tell you
- Tone and faces do matter. When words and tone clash, people lean on tone, and this is well supported. The problem is the universal percentages, not the idea that delivery counts.
- Small, old studies. The figures come from two experiments with small groups; the exact numbers wouldn't be expected to hold even for the same task.
- No single split exists. How much each channel counts depends on the topic, the people and the situation.
The workshop, the trainer and Kevin's boss are an illustration. The figures come from two experiments published in 1967, in the Journal of Personality and Social Psychology and the Journal of Consulting Psychology, and the researcher's own later comments on how they apply. Educational content, not career or financial advice.