Ninety-three percent of communication is nonverbal. Seven percent is the actual words. You've probably heard some version of that claim — from a sales trainer, a dating coach, a LinkedIn post with a stock photo of two people mid-handshake. It has the specific, slightly-too-precise ring of a real finding, the kind of number that feels citable even when nobody citing it has ever read the study. Chase Hughes spends a page of his own profiling chapter taking that number apart — and then, in the very next paragraph, hands you a replacement number that's just as uncited.1
The claim traces back to Albert Mehrabian, a researcher at UCLA in the 1970s, whose work gets summarized — everywhere, endlessly — as "93% of communication is nonverbal, 7% is words." Hughes, along with a named cohort of body-language practitioners he cites (Mark Bowden, Scott Rouse, Greg Hartley, Tonya Reiman), rejects the popularized version outright, calling it "the body language myth." His counter-evidence is disarmingly simple: you're reading these words right now, and one hundred percent of what you're absorbing is language. Watch a film in a language you don't speak, and the plot mostly evaporates no matter how expressive the actors are. If 93% of communication really arrived through the body, subtitles wouldn't exist as an industry.1
Hughes adds a second, sharper jab: the very university that produced the study, UCLA, doesn't teach nonverbal communication as a meaningful component of its own psychology PhD program — he estimates roughly eight minutes of instruction time across the entire degree. If the 93% figure were something the field actually believed and organized around, it's a strange thing for the field's own flagship training pipeline to functionally ignore.1
Here's where it gets interesting. Having spent a page dismantling a specific, cited-sounding number for being unsupported and mismeasured, Hughes doesn't conclude "we don't actually know the ratio" — he immediately proposes a new one:
"For the purposes of this section, let's conservatively estimate that about 1/3 of communication is what's being said and 2/3 is nonverbal... For simplicity's sake, we will simply use the 2/3 rule."1
No study is cited for 2/3. No researcher's name attaches to it. It isn't measured, it's declared — "conservatively," which sounds like caution but is actually just a rhetorical softener for an equally invented figure. The critique of Mehrabian was a critique of an unsupported precise number masquerading as settled science. The 2/3 rule is exactly that same thing, just rounder and delivered with an apology attached.
Evidence: Hughes' rebuttal of the 93%-nonverbal claim is directionally reasonable and matches a broader, well-documented pattern of Mehrabian's original research (which actually measured congruence between tone, face, and words in narrow emotional-attitude contexts) being generalized far beyond what the original study supported.1 [PLAUSIBLE — needs corroboration] [CONTESTED] The replacement figure, however, carries none of that same scrutiny.
Tensions: This is a rare case where a single page of the book contains its own internal contradiction, fully visible without needing outside sources to surface it. Hughes correctly diagnoses the problem — a specific, authoritative-sounding percentage getting repeated by an entire industry with nobody checking the source — and then, one paragraph later, generates a new specific, authoritative-sounding percentage for his own readers to repeat. The book's broader "🚩 CONSENSUS MISREPRESENTATION" pattern (uncited "studies show" language used across dozens of other passages) is visible here in miniature, except this time it's aimed at correcting a different instance of the exact same habit.
Psychology — Anchoring and Adjustment. That page documents how a first number — even a number known to be random or irrelevant — disproportionately pulls subsequent judgment toward itself, because people adjust from an anchor rather than reasoning from scratch. The 93% figure functions as exactly this kind of anchor across the whole body-language training industry: even Hughes, actively debunking it, can't fully escape its gravitational pull — he doesn't propose "the ratio is unknown," he proposes a smaller number in the same direction, still overwhelmingly nonverbal, just less extreme. The insight the pairing produces: debunking a specific number is not the same operation as escaping its anchor. The 2/3 rule isn't an independent estimate; it's 93% adjusted downward by someone who still can't quite let go of "communication is mostly nonverbal" as a starting point, which is precisely the adjustment-insufficiency anchoring predicts.
Business — Hallucination and Confidence in Language Models. That page describes why language models generate confident, specific-sounding false claims: the training objective rewards producing answer-shaped text, never rewards saying "the actual number is unknown." Hughes, writing as a human author under a similar pressure — a chapter needs a working ratio to build the rest of its argument on, "I don't know" doesn't let the chapter continue — does structurally the same thing a hallucinating model does: generates a plausible, specific, confidently-delivered number to fill a gap where honestly nobody has one. The insight the pairing produces: confident false precision isn't a uniquely artificial-intelligence failure mode. It's what any system — human or machine — defaults to when its output format requires a definite answer and the training/incentive structure punishes visible uncertainty more than it punishes being wrong.
Sharpest implication: Disputing a statistic's citation trail and disputing its conclusion are two different acts, and this passage only does the first one while appearing to do both. Hughes never actually challenges the underlying claim that nonverbal channels dominate verbal ones — he keeps that thesis fully intact and just re-skins it with a rounder, uncited fraction. Which suggests the specific number was never the load-bearing part of the original myth. The load-bearing part was the authority that precision itself lends to a claim — and that authority survives the debunking completely undamaged, just wearing a different digit.
Generative Questions: