Minority influence

Why can one person who never wavers move a group that unanimously disagrees with them?

Everything we know about groups points the other way: weight of numbers wins, and the lone dissenter is the one who caves. Yet the direction of travel in a great many arguments is from the many toward the few.

So what does a minority have to work with? Not votes, or it would not be a minority. Not authority, or the question would not arise. Whatever it has must be a property of how it holds its position rather than of how many hold it. That narrows the search considerably.

Start from the other side's mechanism, because knowing what a majority does tells you what a minority cannot do. A majority works largely by making disagreement expensive: contradicting six people out loud costs something, and people pay by agreeing aloud while privately unmoved — the finding the conformity piece in this collection turns on, since most of Asch's conformers reported afterwards that they had seen the lines correctly all along.

In a minority's hands that lever vanishes. Agreeing with the one odd person out does not buy you acceptance; it costs you acceptance. So if a minority moves anyone, it is not by the majority's route, and its effects should show up somewhere else entirely — not in what people say in the room, but in what they think afterwards, alone. That is the whole argument, and it is testable: a majority should produce lots of public agreement and little private change; a minority the reverse.

What could produce private change? If everyone agrees with you and one person does not, you have a small anomaly to account for. The cheap explanations — they misheard, they are joking, they will fold — all predict the deviance goes away. When the same answer comes back trial after trial without hedging or drift, those explanations are used up, and what is left is the possibility that they are seeing something. To check that, you have to look at the thing itself.

So consistency is not stubbornness dressed up: it is the property that exhausts the social explanations and forces attention onto the content. That predicts something sharp — an inconsistent minority, saying the startling thing only sometimes, should do essentially nothing, because the first cheap explanation still works.

Moscovici's experiments in 1969 were built on exactly this. Groups of six judged unambiguously blue slides; two confederates called them green. Consistent confederates drew a small fraction of green responses from naive participants against a control rate near zero; inconsistent ones drew almost nothing. Then the part that matters more: asked afterwards, privately, to sort ambiguous blue-green chips, those exposed to the consistent minority shifted their threshold toward green further than their public answers had. The influence was larger where nobody was watching.

Be honest about that evidence. The public effect was small — a single-figure percentage of trials, and I am recalling rather than deriving it — and the blue-green paradigm comes from a period whose findings have often not survived replication. Nothing above needs Moscovici's figures. What it needs is the crossover: minorities beating majorities on private and indirect measures while losing badly on public ones. A meta-analysis by Wood and colleagues in 1994 reported that pattern, and the pattern rather than any one study is what is load-bearing.

Which gives a clean way to be wrong. If minority influence were simply weaker majority influence — smaller on every measure, public and private alike — then no second process is needed, and the single-process account that Latané and Wolf proposed, where influence is just a function of a source's number, closeness and strength, would explain everything more economically. Find that ordering and the conversion story collapses. It has not been found, but the debate between one process and two is genuinely unresolved, and the honest position is that the crossover is better established than any explanation of it.

Three refinements connect the pieces.

First, this is not the same phenomenon as Asch's ally. A single confederate who broke the unanimous majority cut conformity to a small fraction of its level — but that ally released people who already privately agreed. Releasing agreement is not creating it. Minority influence is the harder claim: that the few can move people who genuinely started on the other side. The two mechanisms even have opposite requirements, since the ally works by being visible in the room and the minority works by being thought about after leaving it.

Second, consistency is not rigidity, and here the literature is least settled. Work in Mugny's tradition found that a minority perceived as dogmatic gets written off as a personality rather than a position, while one firm about the claim and flexible about everything else does better. That fits the mechanism — rigidity restores a cheap social explanation — but it is a smaller and more contested body of work than the consistency finding itself.

Third, a warning the mechanism supplies about itself. Nothing here makes a persistent minority right. The process that converts is attention forced onto content, which favours whatever survives examination; but it is triggered by unwavering deviance, which any determined person can supply. The same route serves the dissenting scientist and the crank, and from inside the room they are hard to tell apart.

• Whether a minority's convert becomes a second visible dissenter, and how that changes the group's arithmetic.
• Why influence that arrives without a remembered source is so hard to attribute, and what that does to studying it.

Every term the collection defines is gathered in the glossary.