There is someone on your team you have quietly written off. You wouldn't put it that way. You'd say they're not really engaged, or hard to read, or that they don't bring much to the room. The evidence is a handful of Thursdays: the meeting where they said nothing, the one where they pushed back on a plan everyone else liked, the one where their camera stayed off and their replies came a beat too late.

Here is the uncomfortable part. If someone asked you to explain your own worst hour in that same room — the day you were curt, or silent, or dug in on something you didn't even care about — you would not reach for a description of your character. You would reach for a description of your day. You'd hadn't slept. Your kid was sick. You'd just come out of a call where you got blindsided. You know, with total clarity, that your behavior in that meeting was not the same thing as who you are.

You extend that understanding to yourself automatically. You almost never extend it to them.

The oldest finding in social psychology, and the one we keep forgetting

In 1967, Edward Jones and Victor Harris ran a study that should have permanently changed how we read each other. They had participants read essays either supporting or criticizing Fidel Castro, then guess the essay writer's real opinion. Reasonable enough — until they told participants that the writers had been assigned their position. No choice at all. A coin flip decided who praised Castro and who condemned him.

Participants still believed the pro-Castro writers liked Castro more. They could see the situation. They knew it fully explained the behavior. And they reached past it anyway, straight for the person underneath.

Lee Ross later named this the fundamental attribution error: our tendency to explain other people's behavior by their disposition — their character, their motives, who they are — while systematically discounting the situation they were standing in. Some researchers prefer correspondence bias, because it names the mechanism more precisely. We assume behavior corresponds to the person. The quiet one is a quiet person. The blunt one is a blunt person. The one who missed the deadline is unreliable.

Ross and his colleagues ran a second study in 1977 that lands even harder in a meeting context. They set up a mock quiz game: one participant was randomly assigned to write hard questions, another to answer them. Everyone watching knew the roles were assigned by chance. Everyone knew the questioner got to pick trivia from the corners of their own knowledge. And observers still rated the questioners as substantially more knowledgeable than the contestants.

Read that again with a conference room in mind. The person who runs the agenda gets to choose the terrain. They pick the topics they know cold. They ask, they don't answer. And everyone in the room — including the person answering — walks out with a slightly upgraded sense of how smart the questioner is.

Meetings are not neutral instruments for measuring people. They are quiz games with assigned roles.

The mechanism: you're too busy to see the situation

Why does this happen even when the situation is staring us in the face? Daniel Gilbert and his colleagues found the machinery in 1988. Attribution appears to work in two steps. The first is fast and nearly automatic: we characterize the behavior in terms of the person. Silent → disengaged. The second step is slow and effortful: we correct for the circumstances. Although — she was in back-to-back calls all morning and this topic isn't hers.

The correction requires cognitive resources. In Gilbert's experiments, observers who were made "cognitively busy" — asked to hold something else in mind while watching — performed the first step just fine and largely failed the second. They saw a nervous person, not a person being asked nerve-wracking questions.

Now consider what your brain is actually doing in a meeting. Tracking the agenda. Rehearsing the thing you want to say. Watching your own face in the corner of the screen. Reading the room. Wondering whether that Slack message needs a reply. You are, by any technical definition, cognitively busy for the entire hour. The correction step never runs. What you're left with is step one, unedited, and step one is a character judgment.

This is why meetings — the single richest source of interpersonal data most of us have about our colleagues — are also the least trustworthy. We form our most durable impressions of each other in precisely the conditions that guarantee those impressions are wrong.

The camera is pointed the wrong way

In 1973, Michael Storms did something clever. He had two people hold a conversation while two others observed. Then he showed each participant a videotape — but from the opposite angle. Actors watched themselves. Observers watched from the actor's point of view.

The attributions flipped. When people saw themselves on tape, they made more dispositional attributions about their own behavior. Perspective, quite literally, drove the judgment. You explain yourself by the situation because the situation is what you can see; you are looking outward at a room full of pressures. You explain others by their character because they are what you can see.

A caveat worth keeping, because the honest version of this science is more interesting than the tidy one: Bertram Malle's 2006 meta-analysis found the classic actor–observer asymmetry is much weaker on average than textbooks imply. It shows up strongly under specific conditions — and one of the most reliable is negative events. When something goes badly, we excuse ourselves and indict each other. Which is to say: the bias is small when nothing's at stake and largest exactly when it matters. In the meeting where the plan fell apart. In the retro after the launch slipped.

What this costs you

Every meeting produces two outputs. The first is the decision, and everyone tracks it. The second is a set of quiet revisions to how each person in the room is seen — who was sharp, who was checked out, who was difficult. Nobody tracks that one, and it compounds.

It compounds into who gets invited to the next conversation. Who gets the visible project. Who gets described, in a calibration meeting eighteen months from now, with a word that traces back to a Thursday when they had the flu.

The cruelty of correspondence bias isn't that it's harsh. It's that it's confident. It doesn't feel like a guess. It feels like knowing someone.

Your next moves

  • Before your next meeting, write down one line about your own situation — "third call in a row, didn't prep, low sleep." Then, after the meeting, look at it. Whatever you'd count as an excuse for yourself is data you owe everyone else in the room.
  • Use consensus information as a circuit-breaker. Harold Kelley's covariation model says we discount dispositional causes when behavior is common. So ask literally: do most people go quiet in this meeting? If yes, the meeting is the cause, not the person. If half your team goes silent at the same agenda item, you don't have half a team of passive people.
  • Steal Storms's camera trick. Before judging someone's behavior, describe the meeting from where they were sitting: what they knew, who outranked them, what they'd been asked to defend. Do it in one written sentence, not in your head — the head version skips the effortful step.
  • Rotate who asks the questions. If one person always sets the agenda and probes, that person will accumulate an unearned reputation for being the smartest in the room, and everyone else will accumulate the opposite. Swap the role monthly. Watch what happens to who "seems senior."
  • When you catch yourself using a trait word — lazy, defensive, checked out, difficult — force a rewrite into a behavior plus a context. "He pushed back twice on a spec he found out about ten minutes earlier." Trait words end inquiry. Behavior-in-context restarts it.

The thing worth building

None of this works from memory, because memory is where the bias lives. The correction step needs something outside your head — a written trace of what actually happened in the room, what the constraints were, who was set up to shine and who was set up to answer. That's the whole reason meeting retrospectives exist, and the whole reason most of them fail: they capture decisions and lose people.

MeetingMortem is built on the idea that a meeting deserves an autopsy, and that the autopsy should notice more than what got decided — who spoke, who was crowded out, what the room's shape was doing to the people inside it. If you'd like to see your meetings from the other end of Storms's camera, you can find it at meetingmortem.lumenlabs.works.

And if you never install it, keep the one idea anyway: the person you've written off had a situation. So did you. Only one of you got the benefit of the doubt.