First, a judgement
You are about to judge how far a faint marker moves in a dark field. There is no ruler, no reference object and no correct answer. After five rounds alone you will see three other people's estimates before each of the next five, and then you will be alone again.
The laboratory
Key terms
- Ambiguity
- Having no way to check an answer. It is what makes this situation different from one where the group is plainly wrong.
- Convergence
- Estimates moving towards one another over successive rounds.
- Norm
- A shared sense of what a reasonable answer looks like, which can outlast the group that produced it.
Fifteen rounds, no right answer
Self-paced. Nothing is timed, nothing is scored, and you can type any number you like - including the same one every round.
- 1-5. Alone
- 6-10. With three others
- 11-15. Alone again
- + Three other groups
Your fifteen rounds
Four groups, one display
| Group | Mean first estimate | Norm by round 10 | Spread within the group |
|---|
The historical experiment, described separately
In the 1930s Muzafer Sherif used the autokinetic effect - a stationary point of light in a completely dark room appears to move, because the eye has no reference against which to stabilise it - to create judgements that could not be checked. Participants gave estimates of how far the light had travelled, first alone, then in small groups where estimates were spoken aloud, and in some conditions alone again afterwards.
What he reported was not that people copied one another. Alone, individuals settled into their own characteristic range. In groups, those ranges converged towards a common one, and different groups converged on different values. When participants returned alone, they carried the group's range with them. Sherif read this as the formation of a frame of reference rather than as compliance: nobody had been asked to agree, nobody was watching, and the frame outlived the situation that produced it.
This activity borrows that logic and none of its materials. There is no dark room, no point of light and no autokinetic effect here; the display is a drawing and the other participants are generated by a documented model. Nothing on this page is a replication, and no number produced here corresponds to anything Sherif reported.
What this demonstrates
Convergence under ambiguity is not caving in
Nobody in this laboratory is asked to agree, nobody sees your answer, and there is no cost to differing. What the three other estimates provide is not pressure but information - the only information available about a judgement that has no answer. Taking it into account is reasonable. That is why this phenomenon is usually treated as the clearest case of informational influence, and why it sits so awkwardly beside the classic conformity studies, where the judgement was easy and the pressure was the point.
A frame outlives the situation that produced it
The interesting round is round 11. The others are gone, nobody is watching, and most people's estimates stay much closer to the group's range than to their own opening one. Whatever happened in rounds six to ten was not a decision to go along with anybody: it changed what a plausible answer looks like. That is the difference between complying with a demand and acquiring a definition of the situation.
Different groups, different worlds
The three other groups saw exactly the same display and arrived somewhere else. Nothing distinguishes their norms except who happened to be in the room and what they happened to say first. Once a norm exists it feels like a property of the display, which is the whole trouble: shared frames present themselves as facts about the world rather than as the residue of a particular set of conversations.
Where this goes next
The mechanism scales badly and interestingly. Later work has traced such frames across generations of participants, with each new arrival inheriting a norm from people who inherited it from people who invented it. The norm survives long after everyone present at the start has left. That is a laboratory model of tradition, and it is worth asking what it does and does not capture about the real thing.
For teaching elsewhere: take this activity as one self-contained block of HTML, on the clipboard or as a file. Either way it is styled so that it will not disturb the page you put it into.