is 93% of communication really non-verbal? what Mehrabian actually tested

The study

In 1967 the psychologist Albert Mehrabian published two short papers that would later be stitched into one famous formula.

The first, written with Morton Wiener, appeared in the Journal of Personality and Social Psychology under the title "Decoding of inconsistent communications". The researchers chose nine single words after pretesting how positive or negative they sounded on paper. Three were positive (honey, thanks, dear), three neutral (maybe, really, oh) and three negative (don't, brute, terrible). Two speakers recorded each word in three tones of voice: one signalling liking, one neutral, one signalling dislike. In most combinations the word and the tone disagreed, for example "dear" said with obvious dislike.

Listeners heard these recordings and rated the speaker's attitude towards an imagined person being addressed. One group was told to attend only to the meaning of the word, another only to the tone, and a third to everything. When people attended to everything, tone had the stronger pull.

The second paper, with Susan Ferris, appeared in the Journal of Consulting Psychology. This time there was just one word: "maybe". Two speakers said it in a positive, neutral or negative tone, and the recordings were paired with photographs of faces showing positive, neutral or negative expressions. Listeners again estimated the speaker's attitude. The face had a larger effect than the voice.

What it found

From these two studies, Mehrabian proposed that when verbal, vocal and facial cues about an attitude are combined, the overall impression is a weighted sum with coefficients of .07, .38 and .55. That is the source of the 7-38-55 rule.

Notice how narrow the conditions were:

• Single words, not sentences, stories or arguments.
• Attitudes only: whether the speaker seemed to like or dislike someone.
• Mixed signals by design: in most trials, the word said one thing and the voice or face said another.
• No gestures, posture or context: just a recorded word, a tone and, in the second study, a still photograph.

A teaching summary of the studies hosted on a Florida International University faculty page puts it plainly: it is unclear exactly how the two separate results were combined into three precise percentages, and the popular reading that 93% of all meaning is non-verbal is not supported by experiments that used one-word messages in a laboratory.

Mehrabian has said the same thing. On his own website, alongside the equation, he explains that it was derived from experiments on the communication of feelings and attitudes, specifically like and dislike, and that unless a speaker is talking about their feelings or attitudes, "these equations are not applicable."

Did it hold up?

This is a case where the question is less "did the finding replicate?" and more "was the finding ever what people say it is?"

The narrow result, that a contradictory tone or facial expression can outweigh a single neutral or emotional word when people judge liking, is plausible and is not the problem. The problem is the leap from that small laboratory setup to all human communication. Writing in a 2011 Psychology Today column, the researcher Jeff Thompson pointed out that the formula concerns cases where verbal and non-verbal signals conflict, and that treating it as a universal rule is inaccurate, a limit Mehrabian agrees with.

Communication scholar David Lapakko, in a 2015 article in the Communication and Theater Association of Minnesota Journal, traced how the figures spread. He noted that the two studies had serious methodological limitations yet still appeared in many communication textbooks, and that the formula circulates widely online, often with little understanding of where it came from.

A 2020 explainer from the Donders Institute at Radboud University adds two further limits: the studies looked only at words, tone and face, ignoring gestures, posture and head movements, and they measured only attitudes in situations where signals contradicted each other. The relative weight of words and non-verbal cues depends on what is being communicated and where.

We could not find a modern, large-sample replication that tested the 7-38-55 split itself, so we make no claim either way about the exact numbers. The honest verdict: real study, misquoted claim.

How it gets misused

The figure usually appears in three ways:

1. "Words barely matter." A training slide tells sales teams or new managers that only 7% of their message is in the words. Taken seriously, that would mean you could follow a lecture in a language you do not speak just by watching the speaker. You cannot.
2. "Body language is everything." People spend hours on posture and hand positions and minutes on the actual structure of what they want to say.
3. "Trust the face, not the words." In the original studies, people judged a stranger's liking from one word. Real conversations carry far more context, and faces are easy to misread.

The real lesson from Mehrabian is narrower and more useful: when your words and your tone disagree about how you feel, people tend to believe the tone.

What it means for you

Think of a team lead in Pune telling her team, "I'm really pleased with this quarter," in a flat voice while scrolling her phone. The team will probably hear the flatness. That is the Mehrabian situation: feelings, mixed signals.

Now think of the same lead explaining a new approval process. Here, if the words are unclear, no amount of warm tone will rescue the message. Content carries it.

So there are two jobs:

• When you are conveying a feeling (thanks, concern, an apology, praise), make your voice and face match your words. If you cannot mean it, say less.
• When you are conveying information, get the words right first: a clear point, a reason, a next step. Then let your delivery support it.

"I want to say thank you properly. The client called this morning to say how smooth the handover was, and that was your work."

"Here's the change in one line: approvals now go to Ravi first, then to me. Starting Monday."

The first line is about a feeling, so it needs a warm, unhurried voice. The second is about information, so it needs to be clear before it needs to be charming.

If English is your second or third language, this is good news. You do not need a particular accent or perfect gestures to be understood. You need clear words, and a voice that agrees with them when you are talking about how you feel.

Sources

1. Albert Mehrabian and Morton Wiener, Decoding of Inconsistent Communications, Journal of Personality and Social Psychology, 6(1), 109 to 114, 1967. doi.org/10.1037/h0024532
2. Albert Mehrabian and Susan R. Ferris, Inference of Attitudes from Nonverbal Communication in Two Channels, Journal of Consulting Psychology, 31(3), 248 to 252, 1967. doi.org/10.1037/h0024648
3. Albert Mehrabian, official website, page on his book Silent Messages (1981), giving the equation and its stated limits. www.kaaj.com/psych/smorder.html
4. Florida International University faculty page, How Much Social Meaning Is Nonverbal? (research summary of both 1967 studies). faculty.fiu.edu/~brownj/Abstracts/Social_Meaning_Nonverbal.pdf
5. Jeff Thompson, Is Nonverbal Communication a Numbers Game?, Psychology Today, 30 September 2011. www.psychologytoday.com/us/blog/beyond-words/201109/is-nonverbal-communication-a-numbers-game
6. David Lapakko, Communication is 93% Nonverbal: An Urban Legend Proliferates, Communication and Theater Association of Minnesota Journal, 34, 2015. cornerstone.lib.mnsu.edu/ctamj/vol34/iss1/2
7. Marlijn ter Bekke, This is why "93% of communication is nonverbal" is a myth, Donders Wonders blog, Radboud University, 6 July 2020. blog.donders.ru.nl/this-is-why-93-of-communication-is-nonverbal-is-a-myth/?lang=en

Published by The Full Stop, the journal of Brightside English, in the The Experiments series. Every factual claim links to a source below; where accounts or findings are disputed, we say so. Sources checked on 26 Oct 2026. Reviewed by a Brightside senior coach. Corrections: see our editorial standard.