The short version
A group is close, capable, and under pressure. Agreement starts to feel like the thing they’re producing. Doubts get softened, then held back, then — this is the part that does the damage — stop being formed at all, because a mind that has learned there’s no use raising something learns not to notice it.
The group doesn’t decide to ignore the risk. It arranges not to hear about it.
Where the name comes from
The word was coined twice. William H. Whyte had it first, in Fortune in March 1952, for the corporate cult of togetherness — “a rationalized conformity, an open, articulate philosophy which holds that group values are not only expedient but right and good as well.” Irving Janis arrived at it again twenty years later, apparently on his own — he never cites Whyte — building it on the pattern of Orwell’s newspeak. Janis’s is the meaning that stuck: Psychology Today in 1971, then Victims of Groupthink the following year. He defined it as what happens when the members’ drive for unanimity overrides their motivation to realistically appraise the alternatives, and he built the case out of American foreign-policy disasters — the Bay of Pigs above all, where a room of unusually intelligent people signed off on a plan that several of them privately thought was madness.
Janis listed eight symptoms. The ones you can actually spot in a meeting are these four:
- An illusion of unanimity — silence read as consent
- Self-censorship — the doubt you decide isn’t worth raising
- Direct pressure on the one person who does raise it
- Mindguards — someone quietly keeping inconvenient information away from the group
Janis’s other four are about how the group sees itself and the world: invulnerable, inherently decent, warnings explained away rather than reckoned with, and opponents too stupid or evil to matter.
Why it happens
Cohesion is the cause, which is what makes this so awkward — the groups most vulnerable are the ones that like and respect each other. A group of strangers argues freely because none of them has anything to protect. A close team has something to lose by arguing, and it isn’t the argument.
The mechanism is a slow one. First the objection gets softened, so as not to sound like an attack on colleagues. Then it gets held back for a better moment that never arrives. Then, for some, the person stops having it — because a mind that has repeatedly learned an objection has nowhere to go stops generating the objection. By the time the decision is made the doubts have been set aside rather than answered, and the silence gets read as consent — which is exactly why the group is so certain.
Two things make it worse, and both look like good leadership. A leader who states their preference early converts every subsequent disagreement into a disagreement with the boss. And time pressure, which everyone treats as a reason to consult fewer people, removes the slack in which a doubt would have been raised.
This is not groupthink
- A group that agrees because the answer is obvious. Unanimity isn’t a symptom of anything. Sometimes the room is right, quickly, and there was nothing to debate.
- People you disagree with agreeing with each other. This is the commonest misuse of the word by a wide margin. “Groupthink” has become a way of dismissing a consensus without engaging it, which is itself an unearned dismissal. The concept is about the process inside a group, not about whether their conclusion happens to be wrong.
- Deference to genuine expertise. A room going along with the one person who actually knows is functioning correctly.
- The Abilene paradox. Related but distinct, and worth keeping apart: in Abilene there is no dissent to suppress, because every member privately agrees and each has misread the others. Groupthink is a group with real doubts that it declines to surface.
The test is about the record, not the outcome: when did somebody last say something in this room that the room didn’t want to hear, and what happened to them afterwards?
How to head it off
Janis proposed structural fixes rather than exhortations to be braver, and the structural ones are the ones that work:
- The leader speaks last, and says so. Withholding your preference costs nothing and removes the largest single source of self-censorship in the room.
- Assign the objection to a person. A named devil’s advocate can argue against the plan without it being read as disloyalty, because it’s their job. Rotate it, or it becomes one person’s personality.
- Split into independent subgroups and let them work the problem separately before comparing. Two rooms rarely converge on the same blind spot.
- Hold a second meeting after everyone has agreed, explicitly for doubts that surfaced overnight. Janis called this the second-chance meeting, and it works because it makes raising something later legitimate rather than obstructive.
- Bring in somebody with no stake. Cohesion is the fuel; an outsider has none of it.
The single most useful habit is smaller than any of those. When a big decision passes with no objection at all, treat the silence as information rather than as agreement, and go and ask two people privately what they didn’t say.