The meeting ends a minute early, which everyone takes as a good sign. The decision felt clear. Heads nodded. Someone said "great, we're aligned," and nobody objected. Then, over the following week, the engineer builds the small version, the designer plans for the big one, and the project manager tells a stakeholder the whole thing was postponed. Nobody lied. Nobody zoned out. Everyone simply left the same room carrying a different meeting.
It's tempting to file this under carelessness — someone wasn't listening, someone didn't take notes. But the research points somewhere less flattering and more interesting: misalignment like this is manufactured by two well-documented perceptual biases working in tandem. Speakers systematically overestimate how clearly they communicated. Listeners systematically overestimate how well they understood. And a meeting, it turns out, is close to a perfect habitat for both.
A Song Everyone Can Hear but You
In 1990, a Stanford psychology graduate student named Elizabeth Newton ran what has become one of the most quietly devastating experiments about communication. She paired people up as "tappers" and "listeners." Tappers picked a well-known song — something like "Happy Birthday" — and tapped its rhythm on a table. Listeners had to name the tune.
Before the tapping began, Newton asked the tappers to predict how often listeners would succeed. They guessed about half the time. The actual success rate was roughly 2.5 percent — three songs out of 120.
The gap makes sense once you sit in each chair. When you tap a song, you hear the melody playing in your head; the taps feel rich, expressive, unmistakable. The listener hears knuckles on wood. The tapper cannot un-hear the music, and so cannot imagine how impoverished the signal really is.
Everyone who speaks in a meeting is a tapper. You arrive with the full arrangement in your head — the constraints you've been wrestling with, the Slack thread from Tuesday, the three alternatives you already considered and discarded. Your words are the taps. What lands on the other side of the table is often just rhythm.
The Illusion of Transparency
Psychologists Thomas Gilovich, Kenneth Savitsky, and Victoria Medvec gave the speaker's half of this problem its formal name in 1998: the illusion of transparency. Across a series of studies, they found that people consistently overestimate how visible their internal states are to others. Participants who told lies believed observers would catch them far more often than observers actually did; people drinking a foul-tasting sample believed their disgust was written on their faces when it mostly wasn't.
The mechanism is a kind of anchoring. Your own experience — what you meant, what you felt, what you intended — is vivid and immediate, so you start there and adjust for the fact that others can't read your mind. But the adjustment is chronically insufficient. You know what you meant, so it feels like you said it.
The listener's side has a mirror-image flaw. The feeling of understanding is generated cheaply, from fluency: if the sentences parse and nothing snags, the brain registers comprehension. But fluency is not fidelity. Boaz Keysar and Anne Henly demonstrated this directly in 2002 — speakers who uttered ambiguous sentences overestimated how often listeners recovered their intended meaning, and listeners who picked the wrong interpretation almost never realized another reading existed. Both sides walk away satisfied. Both sides are wrong about the same sentence.
The Curse of Knowledge Does the Rest
There's a third gear in this machine. In 1989, economists Colin Camerer, George Loewenstein, and Martin Weber coined the term "curse of knowledge" for a stubborn asymmetry: once you know something, you cannot fully simulate the mind of someone who doesn't. Their subjects — traders with extra information — couldn't help acting as if others shared it, even when money was on the line.
This is why experts skip steps when they explain things, and it's why the person proposing a decision in a meeting is uniquely unqualified to judge whether the decision was stated clearly. They've spent hours, maybe weeks, inside the problem. The compressed shorthand that comes out — "let's go with the lighter approach for Q3" — feels precise to them because their knowledge silently fills in every gap. To everyone else, it's a Rorschach blot. Which approach was "the lighter" one? Lighter in scope or in staffing? Is Q3 the deadline or the start?
Ambiguity is invisible from the inside. That's the whole trick of it.
Why Meetings Are the Perfect Habitat
The psycholinguist Herbert Clark spent much of his career studying what he called grounding: the process by which conversation partners accumulate evidence that they've actually understood each other, not just that words were delivered. In a one-on-one conversation, grounding is cheap — you get real-time questions, puzzled looks, restatements.
In a meeting, grounding quietly collapses. One person speaks; eight people listen; and each listener assumes that if their interpretation were wrong, surely someone else would have said so. Nods circulate as social lubricant, not as comprehension receipts. Asking "wait — what exactly did we just decide?" feels like confessing you weren't paying attention, so the question that would save the week goes unasked, and the silence gets read as consensus.
Add one more structural fact: meeting decisions are usually spoken, not written. Speech is forgiving of ambiguity in a way writing is not. A vague sentence can be said aloud and nodded through; the same sentence typed out and read back tends to expose its own holes. Most meetings end before that exposure ever gets a chance to happen.
Closing the Loop
Fields where misunderstanding kills people stopped trusting the feeling of clarity a long time ago. Aviation and medicine both run on closed-loop communication: an instruction isn't considered delivered until the receiver has said it back in their own words and the sender has confirmed the read-back. Not because pilots and nurses are careless — because those professions accepted, institutionally, that transmission is not reception.
A few translations for the conference room:
State decisions in executable language. A decision is stated clearly when someone who wasn't in the room could act on it. That means no pronouns without antecedents, no "soon," no "the lighter approach" — a named owner, a dated deadline, a described scope. If the sentence can't pass that test, the decision isn't finished; it's still a vibe.
Ask for a read-back, not for questions. "Any questions?" invites silence, and silence tells you nothing. Instead, ask the person who will do the work: "So what's the first thing you'll do?" Medicine calls this teach-back, and it works because it samples the listener's actual model instead of their feeling of fluency.
Write the sentence while everyone is still in the room. Writing forces disambiguation that speech lets slide. Thirty seconds of typing the agreed sentence where everyone can see it will surface more misalignment than thirty minutes of confident discussion.
Distrust smooth agreement. Genuine alignment usually has a little friction in it — a clarifying question, a "just to confirm." When agreement arrives with zero resistance, the illusion of transparency is often doing the agreeing for you.
None of this is bureaucracy. It's one sentence, said back and written down — a seatbelt that costs half a minute.
The Version That Survives
The deeper lesson of the tappers and listeners is that no amount of goodwill fixes this. You cannot try your way out of the curse of knowledge, because the bias lives precisely in the part of your mind that checks whether you were clear. The only reliable fix is an external artifact — a written version of the decision that exists outside anyone's head, that every attendee can read and dispute while disputing is still cheap. That's the gap Meeting Mortem was built for: it turns the fog at the end of a meeting into a concrete record of what was decided, who owns it, and by when — the read-back your team never quite gets around to doing out loud. If your meetings keep ending in agreement and unraveling by Thursday, you can see how it works at meetingmortem.lumenlabs.works.