Where it sits
We estimate by starting somewhere and adjusting — and we reliably stop adjusting too early.
Tell it apart
| Aspect | Anchoring | Framing effect |
|---|---|---|
| Operates on | Numbers | Descriptions |
| Changes | Your estimate | Your choice |
| Needs relevance | No | No |
| Defence | Pre-commit a figure | Restate both ways |
Anchoring is about a reference point — usually a starting figure, and it pulls your estimate toward it. Framing is verbal — the same fact described as '90% survival' or '10% mortality' changes which option you pick. Anchoring shifts how much; framing shifts whether.
What it doesn't mean
Experts are anchored too, if only slightly. In a 2025 study, scientists assessing papers they had previously cited shifted their quality judgments a little with a random number printed on the questionnaire — while the real citation counts, the anchor you would expect to bite, did nothing.
It doesn't. The original anchor came from a wheel of fortune spun in front of the participants, who had every reason to treat the result as chance — it was in fact rigged to stop on 10 or 65 — and it still moved the median estimate by 20 percentage points.
What it really means
The detail that makes the original anchoring experiment unforgettable is that the participants watched the number appear in front of them, as far as they knew by chance. Tversky and Kahneman spun a wheel of fortune — quietly rigged to stop on 10 or 65 — let it land, and then asked what percentage of African countries belonged to the UN. People who had just seen the wheel stop on 10 gave a median answer of 25%. People who had seen it stop on 65 said 45%.
Nobody thought the wheel knew anything about Africa. The anchor didn't need to be credible, relevant, or even plausible — it only needed to be first.
The mechanism is in the name Tversky and Kahneman gave it: anchoring and adjustment. Asked to estimate something uncertain, we start from whatever number is available and move away from it until the answer feels acceptable. The flaw isn't the starting point; it's that we stop adjusting far too early. Which is why the same paper's other demonstration works: given five seconds, people estimate 8×7×6×5×4×3×2×1 at around 2,250, and the identical digits ascending at 512. The real answer is 40,320, and the only thing that changed was which digit you saw first.
Where it comes from
Old English ancor, from Latin ancora
Anchor is Old English ancor, from Latin ancora, from Greek ankyra; the psychological sense comes from Tversky and Kahneman's anchoring-and-adjustment account (1974). The image holds: the first figure drops an anchor, and your adjustment can only swing so far before the chain pulls tight.
- 1974
Tversky and Kahneman describe 'anchoring and adjustment'. A wheel of fortune was spun in front of participants to produce a number; they then estimated what percentage of African countries are in the UN. Those who saw 10 gave a median of 25%. Those who saw 65 gave 45%.
- 1974
In the same paper, two groups had five seconds to estimate a product. Shown 8×7×6×5×4×3×2×1 the median guess was 2,250; shown the same digits ascending, 512. Both are 40,320.
In real life
- A jacket is tagged was £240, now £120. Whether £120 is a fair price for that jacket never quite gets asked; the £240 has already set the scale.
- In a salary negotiation, the first figure on the table, whoever says it, tends to pull the final number towards itself.
- Asked how many jellybeans are in a jar right after someone else guessed 1,000, people's guesses cluster nearer 1,000 than they would have otherwise, even though the first guess was just a guess.
Questions
Q01What is anchoring bias?
Letting an initial number set the range you argue inside. A flat listed at £400,000 makes £385,000 feel like a win — even if comparable flats on the same street go for £350,000. The asking price was never evidence, but it became the reference point.
Q02Why does it work even with random numbers?
Because we estimate by starting somewhere and adjusting, and the adjustment stops too soon. Participants watched a wheel of fortune land on 10 or 65 and still let it move their answers by 20 points — knowing it was random didn't help.
Q03How do you defend against it?
Generate your own number before you see theirs, and write it down. Decide what a used car is worth from listings and its service history before opening the seller's ad. A figure you committed to in advance is much harder to reel in.
Sources
- What Is the Anchoring Bias? — Simply Psychology.
- Are quality assessments in science affected by anchoring effects? (PLOS ONE) — PMC.
- anchor (etymology) — Wiktionary.