The Study About Sheep That Is Not About Sheep

Solomon Asch showed participants a line, then three comparison lines of obviously different lengths, and asked which matched. The task is so easy that control participants get it right over 99% of the time.

Each participant sat in a room with several others who, unknown to them, were confederates. On certain trials the confederates all gave the same wrong answer, confidently, before the real participant's turn.

About 75% of participants went along with the majority at least once. That is the number everybody quotes, and it is usually presented as a demonstration that people are sheep.

Here is the number that is not quoted. Across all critical trials, about two thirds of responses were correct and independent. And roughly a quarter of participants never conformed at all, on any trial.

So the finding is that social pressure measurably bends judgement, in a substantial minority of instances, on a task where the correct answer is unmistakable. That is remarkable enough. It is not the same as saying people abandon the evidence of their senses to fit in, which is what the pop version says.

The variations Asch ran are more interesting than the headline, and one of them is genuinely important.

Adding confederates raised conformity up to about three or four, and then made almost no difference. Seven people saying the wrong thing is not much worse than four.

But breaking the unanimity changed everything. A single confederate who gave the correct answer cut conformity by roughly three quarters. And (this is the crucial detail) a single confederate who gave a different wrong answer cut it almost as much.

What mattered was not support for the truth. It was that the majority was no longer unanimous. One person visibly declining to go along makes it possible for everyone else to decline too, regardless of what they decline in favour of.

The practical implication for any meeting you have ever sat in is direct. The first person to say "I'm not sure about this" is doing vastly more work than the second, third or fourth. And the cost of speaking first is borne entirely by them, which is why it so often does not happen.

Deutsch and Gerard added a distinction that explains the shape of the data. Informational influence is going along because you think the others are probably right: useful, rational, and stronger when the situation is ambiguous or unfamiliar. Normative influence is going along to be accepted, which operates when others can see what you said.

Let participants write their answers privately in the Asch paradigm and conformity drops sharply, because the normative component disappears. It does not drop to zero, because the informational component survives: if seven people say the line is shorter, maybe you are missing something.

Now the version of this that does the most damage in the world, and has the least name recognition.

Pluralistic ignorance is a state in which almost everyone in a group privately disagrees with a norm, publicly conforms to it, and reads everyone else's conformity as evidence that they genuinely believe it. The norm then enforces itself with nobody holding it.

Surveys of American college students found that most substantially overestimated how much their peers approved of heavy drinking, and adjusted their own behaviour toward that imagined norm. Similar gaps between private opinion and perceived consensus have been documented for attitudes to segregation in the American South, for gender norms in Saudi Arabia, and for a range of workplace practices.

The mechanism is self-sustaining and the intervention is unusually cheap: publish the actual distribution of private opinion. Studies that have simply told students what their peers really think have produced measurable drops in drinking, and a well-known Saudi study found that telling husbands how many other husbands privately supported women working substantially raised their wives' employment months later.

Finally, a caution about treating the Asch number as a fact about human beings. Bond and Smith's meta-analysis across seventeen countries found conformity rates varying substantially: higher in more collectivist societies, and declining in the United States across the decades since 1951.

That variation is not noise around a true value. It is the finding. What the paradigm measures is how a particular society at a particular moment weights group harmony against individual independence, and quoting one percentage as the human rate of conformity gets it backwards.