A wheel of fortune stood at the front of the room, numbered 0 to 100. It was rigged to stop at 10 or at 65. Subjects watched it spin, then answered two questions: is the percentage of African nations in the United Nations higher or lower than the number you just saw, and what do you think it actually is?
The group that saw 10 guessed 25 percent on average. The group that saw 65 guessed 45 percent.
Amos Tversky and Daniel Kahneman published this in Science in 1974, in the paper that introduced heuristics and biases as a research program, and the wheel is the detail that makes the finding unarguable. Nobody in that room believed the wheel knew anything about the United Nations. They watched it spin. The number was visibly, demonstrably random, and it moved their judgments by twenty percentage points.
Anchoring and adjustment describes what appears to happen. Faced with a quantity you do not know, you start from whatever number is available and adjust away from it — and the adjustment stops too early, typically at the first value that seems plausible rather than at the value you would have reached with no anchor at all. Later work has complicated the account: with informative anchors, the effect seems to run partly through selective retrieval, where the anchor makes anchor-consistent evidence easier to recall. The competing explanations matter less here than the two robust findings that follow from all of them.
First, anchoring survives full knowledge of the anchor's irrelevance. Telling people the number is random does not protect them. Neither does paying them for accuracy, in most studies. Second, it works on experts in their own domain. Real estate agents shown identical properties with different list prices produced different appraisals, and denied that the list price had influenced them. German judges who rolled dice before recommending a sentence recommended longer sentences after high rolls. The professional judgment is not an alternative to the anchor; the anchor gets into the professional judgment.
This is why the first number in any negotiation is contested so hard by people who do their negotiating professionally. An opening offer is not an estimate of value or a starting position in a dance. It is an attempt to install a reference point in your head before you have generated one of your own, and it works even when you know exactly what is being done, because knowing does not switch the mechanism off. The list price, the manufacturer's suggested retail price, the crossed-out original price beside the sale price, the suggested donation amounts, the plaintiff's damage demand, the prosecutor's charge before plea bargaining — all anchors, all installed early.
Two countermeasures have some empirical support and both are more work than they sound. The first is to generate your own number before exposure: decide what a fair price or reasonable settlement is, in writing, before you hear theirs. An anchor already in place is much harder to displace than an empty slot is to fill. The second is the consider-the-opposite strategy, which asks you to actively generate reasons the anchor is too high, or too low, since the bias runs partly on one-sided retrieval.
What does not work is confidence in your own immunity. That is the one thing the judges, the agents, and the students watching the wheel all had.
Defense: Consider the Opposite: Disarming Anchors in Negotiation