The 7-38-55 Rule Does Not Say What You Think

You have seen the chart. A circle split three ways, usually in a sales deck or a communication training slide. Words 7 percent. Tone 38 percent. Body language 55 percent. Underneath it, some version of the line "people decide whether they like you in this ratio."

Almost every part of that is wrong, and the way it is wrong is more interesting than the correction.

Where the numbers come from

Albert Mehrabian, a psychologist at the University of California, published two studies in 1967. The one the chart is credited to is Mehrabian and Ferris, "Inference of Attitudes from Nonverbal Communication in Two Channels", in the Journal of Consulting Psychology.

I read it. Here is what is actually in it.

The word everyone deletes

Mehrabian's own website still carries the equation. It does not say what the chart says. Verbatim:

"Total Liking = 7% Verbal Liking + 38% Vocal Liking + 55% Facial Liking"

The word "Liking" appears three times. Every popular retelling deletes it, and that deletion is the whole myth. It was never a formula for communication. It is a formula for working out whether somebody likes you.

He states the limit himself, on the same page: "Please note that this and other equations regarding relative importance of verbal and nonverbal messages were derived from experiments dealing with communications of feelings and attitudes (i.e., like-dislike). Unless a communicator is talking about their feelings or attitudes, these equations are not applicable."

Read that last sentence twice. Unless the person is talking about their feelings, the rule does not apply. Which covers almost everything anybody says at work.

The chart as it is usually drawn

Facial expression, as claimed

55%

Tone of voice, as claimed

38%

The words, as claimed

7%

Now here is what the experiment actually did.

The words were never tested

One word was used. "Maybe", chosen because it rated as neutral. It was used in all thirty six conditions, unchanged.

So the 7 percent figure cannot have come from the study it is always credited to, because that study never varied the words at all.

There was no video, no body, no gesture and no posture either. The visual material was still black and white photographs of heads.

62women took part, across three separate tasks, all of them undergraduates doing it for introductory psychology course credit. The headline result rests on twenty of them.

Subjects were shown a photograph of a woman and a recording of her saying "maybe", and asked to guess how she felt about a third person who did not exist.

What it actually found

A two term equation with no verbal term in it. Facial expression carried roughly three to two the weight of tone. These are the only variance figures in the paper:

Facial expression

41.4%

Tone of voice

19.3%

Sixty point seven percent between them, and no figure at all for words.

The three famous numbers appear exactly once, in the closing discussion, as a suggestion for a model somebody ought to go and test. The paper's own words: "It is suggested that the combined effect of simultaneous verbal, vocal, and facial attitude communications is a weighted sum of their independent effects, with the coefficients of .07, .38, and .55, respectively." The very next sentence asks for someone to test it.

Nobody ever ran the experiment the chart describes. The split is stitched together from two separate studies, one comparing words against tone, one comparing face against tone. Neither compared all three at once.

The study was designed to test the double bind theory of schizophrenia.

He has been saying this for thirty years

Mehrabian, in correspondence: "My findings are often misquoted ... Clearly, it is absurd to imply or suggest that the verbal portion of all communication constitutes only 7% of the message. Suppose I want to tell you that the eraser you are looking for is in the second right-hand drawer of my desk in my third floor office. How could anyone contend that the verbal part of this message is only 7% of the message?"

Nobody who repeats it has read it

A communication scholar ran a content analysis of the first hundred Google results for "communication 93 percent nonverbal". Of the 79 usable pages:

• 75 passed the numbers on with no qualification at all
• 4 questioned them
• 1 described the finding correctly

Only 16 of the 79 named Mehrabian. 19 gave no source whatsoever.

So what is actually true

Here is the part that matters for a practice, and it points the same way, for better reasons.

Being heard beats being read, and it is not a small effect. When evaluators heard a candidate's pitch instead of reading the identical words, they rated the person's intellect higher and were more likely to want to hire them. Among 39 professional recruiters from Fortune 500 firms the gap was enormous.

The same paper tested video, on purpose, to see whether seeing a person adds anything on top of hearing them. It did not. The authors: "simply adding more individuating information about a candidate through visual cues, such as physical appearance and nonverbal mannerisms, had no measurable impact on evaluations of the candidate's mind."

Two other labs found the same null. Voice beat text every time. Video added nothing on top of voice.

A smile is audible. When neutral sentences were given a "smile" acoustic treatment, listeners heard it, and the muscles in their own faces moved in response, measured electrically.

And the one that should change how you follow up:

34times more likely to say yes when asked face to face rather than over email, in a study where each requester approached ten people. The requesters predicted no difference at all.

What this means for your practice

Three things, and the third is the useful one.

Do not quote the 7-38-55 rule. Anybody can look it up in ninety seconds and find Mehrabian himself telling them it does not mean that.

The honest version is narrower and still worth knowing. When what somebody says and how they say it disagree, people believe the delivery. That is why a paragraph reading "caring, compassionate team" does almost nothing, and why hearing a doctor say one sentence does a lot.

Buy video for the voice, not for the face. The evidence that hearing somebody beats reading them is strong and has been replicated. The evidence that adding a picture on top of the voice helps is not there. Video is simply the only way to get a real voice onto a web page, and that is a good enough reason on its own.