Social & Critical Psychology · Simplified

Two Reasons to Agree

People conform is not a finding. It is the beginning of a question about which of two quite different things is happening.

The task, and what it is not. Six people look at a fictional display and say how many marks it contains. The classic studies used line lengths and points of light; nothing here reproduces their stimuli, their procedure or their data. The percentages come from a small two-route model whose parameters are printed below. They are chosen to put the classic conditions in roughly the right order of magnitude, which is a teaching decision and not a replication.

Step 1 of 2

First, a prediction

A participant sits with five others who all give the same obviously wrong answer aloud. The display is easy to judge. On what share of those rounds does the participant go along with them?

Step 2 of 2

Six conditions, one at a time

Run whichever you like. Each one stays in the table so you can compare them, and each is reported as two separate routes rather than as one rate. Watch which conditions move which route.

Key terms

Normative influence: Agreeing because of what saying otherwise would cost you in front of these people. It needs an audience.
Informational influence: Agreeing because other people are evidence about a thing you cannot judge on your own. It does not need an audience at all.
Unanimity: Whether anybody else breaks with the majority. It works as a step rather than a headcount: one person is a great deal more than a fifth of five.
Identification: Whether the participant regards these people as the sort of people they are. The classic two-route account has no term for it.

Two bars per condition. The upper bar is the normative route and the lower is the informational one, each labelled with its value, so the two are never told apart by colour alone.


Condition	Normative route	Informational route	Agrees on

What this shows

The manipulation that separates them

Key idea: Taking away the audience is the manipulation that tells the two routes apart, because only one of them needs one. On the easy display, going private removes about twenty points of agreement, and almost all of what is left is the small informational component. On the ambiguous display it removes about eleven, because most of the agreement there was people using each other as evidence, and that does not care who is watching. So the same intervention, a private response, is highly effective in one room and half as effective in the other, and you cannot know which room you are in from the rate alone. Two further things the table shows. Unanimity works as a step: one person breaking ranks changes far more than the arithmetic of four against one suggests, because what has changed is whether another answer is sayable. And a group the participant does not regard as their sort of people exerts much less pull than five unanimous voices should, which the classic two-route account has no term for at all.

This is a teaching model, not a replication. The parameters are printed so they can be argued with, they are chosen to put the classic conditions in roughly the right order of magnitude, and no figure here is an estimate of anything. Three cautions about the underlying research rather than the model. The classic findings are of their time and their place: rates vary substantially across decades and across societies, and the studies that produced them would not be run in the same way today. Conformity is not a defect being measured, and going along with a group that knows more than you is often the sensible thing to do; the ambiguous condition here is a case where the agreement may well be correct. And a group average conceals a great deal, since a rate of a third is compatible with most participants never agreeing once and a minority agreeing almost always.

The longer version lets you move each parameter continuously rather than choosing conditions, adds a participant task, and simulates a hundred individuals so you can see the distribution behind the group rate. It is at Conformity Under Context in the main collection.