Crossed arms are the classic example. Popular guides read them as a sign of resistance. But people also cross their arms when they're cold, tired, comfortable or simply have nowhere else to put their hands. A single gesture, read without context, is closer to a horoscope than a measurement. Here's what research can and can't support.
The 93% number and where it came from
unsupported popular claimThe claim that communication is 7% words, 38% tone and 55% body language traces back to two 1967 studies by Albert Mehrabian. Participants judged how a speaker felt from single words, such as "dear" or "terrible," spoken in a tone or paired with a facial expression that sometimes contradicted the word. The findings concerned feelings and attitudes in mismatched messages, not everyday conversation. Mehrabian himself has said the formula shouldn't be applied to communication in general. If it could, you'd understand most of a foreign-language film with the sound off.
A horse that read people, not numbers
classic investigationIn early 1900s Berlin, a horse called Clever Hans appeared to answer arithmetic questions by tapping his hoof. In 1907, psychologist Oskar Pfungst showed that Hans was responding to tiny, unintentional cues from questioners, such as a slight change in posture as the right number approached. When the questioner didn't know the answer, Hans failed. The real lesson is about observers: we leak cues without meaning to, which is why modern studies use blind and double-blind designs.
Lie detection is barely better than a coin flip
"Liars avoid eye contact and fidget."
Shared by TV dramas, interview tips and countless viral threads.
Weak cues, modest accuracy
A 2006 meta-analysis by Charles Bond and Bella DePaulo, combining hundreds of studies, found people judged truth and lies correctly about 54% of the time. A 2003 review of cues to deception led by DePaulo found most body signals, including gaze aversion, were weakly related to lying or not at all.
Institutions have tested this too. In 2013, the US Government Accountability Office reported that the Transportation Security Administration's behaviour-detection programme lacked scientific evidence that it could identify threats from behavioural indicators. A 2003 National Research Council review raised serious doubts about the polygraph for security screening.
Study limit: many lab lies are low-stakes; high-stakes lies are harder to study ethically
Power poses and the replication test
partly replicatedA 2010 study by Dana Carney, Amy Cuddy and Andy Yap reported that holding expansive "power poses" for a couple of minutes changed hormone levels and risk-taking. A larger 2015 replication led by Eva Ranehill found no hormonal effects. In 2016, Carney stated that she no longer believed the effects were real. Later analyses suggest poses may modestly affect how powerful people say they feel, which is a much smaller claim than changing your biology.
Gestures change meaning across places
everyday demonstrationThe thumbs-up and the "OK" hand sign are friendly in many countries but considered rude in some regions. In parts of Bulgaria, head movements for "yes" and "no" are often described as reversed relative to much of Europe. Even eye contact norms vary between cultures and between individuals, including many autistic people, for whom avoiding eye contact says nothing about honesty.
Words, tone and gesture work together. Pulling one out and reading it alone loses most of the message.
Observation prompt
Watch a few minutes of a conversation in a café or on a talk show with the sound off. Guess the mood, then turn the sound on. Notice how often the words change your reading, and how often your guess was simply too confident.
What to take away
Non-verbal signals carry real information about emotion and emphasis, especially when they change against someone's usual behaviour. What they don't offer is a reliable dictionary. Treat any guide that promises to expose liars or read minds through posture as entertainment, and look for baseline, context and words before drawing conclusions.