Albert Mehrabian is a psychologist who has spent a substantial part of his career trying to correct a misuse of his own findings, without much success.
The two studies were run in 1967 at UCLA. In the first, with Morton Wiener, subjects heard a single recorded word — words like honey, dear, brute, terrible — spoken in a tone that either matched or contradicted the word's meaning, and judged the speaker's attitude toward them. In the second, with Susan Ferris, subjects saw photographs of faces paired with a single spoken word, again with matched or mismatched valence.
The question in both was narrow: when the channels disagree, which one does the listener believe? Mehrabian combined the results into a weighting: 7 percent verbal, 38 percent vocal, 55 percent facial.
Those numbers now appear on slides in presentation-skills courses worldwide, glossed as: only 7 percent of communication is the words.
The claim is absurd on ten seconds' inspection, and the counterexample is available to anyone. If words carried 7 percent of meaning, you could follow 93 percent of a lecture in a language you do not speak. You could get the gist of a phone call from the tone alone. Radio would be nearly useless. The figures cannot possibly describe communication in general, and they were never claimed to.
What they describe is a single, specific, artificial task: the resolution of inconsistency between channels, using single words, when judging a speaker's feelings toward the listener. Mehrabian has stated this in print repeatedly, in terms as close to exasperated as academic prose permits — that the equation was derived from experiments on communications of feelings and attitudes, that it is not applicable unless the communicator is talking about their feelings or attitudes, and that he is uncomfortable with how it has been used.
The methodological limits are worse than the scope limits. Single words, not sentences. A small number of subjects, all women in the first study. Photographs, not moving faces. No conversation. No context. Two studies, arithmetically combined into a formula that was never intended to be a formula.
The reason this matters for a guide about influence is what the myth licenses.
If words are 7 percent, then reading bodies is where the real information is, and an entire industry follows: body-language experts on television analyzing politicians' hand gestures, negotiation courses built around posture, statement analysis vendors, and interview training that instructs people to distrust what a candidate says in favor of how they say it. The 7-38-55 figure is the founding citation for the proposition that observation beats listening.
And the evidence in the rest of this part says the opposite. Nonverbal reading of deception runs at 54 percent. The content of an account — its verifiable details, its internal consistency, its resistance to unexpected questions — is where the discriminating information actually lives, which is the entire finding of Part 4.
The myth has it exactly backwards, and it has been repeated so often that it is now easier to process than the correction. Which is chapter 25 operating on a claim about how communication works.
See Defense: reading the original study