The Wheel of Misfortune
In 1974, Amos Tversky and Daniel Kahneman did something beautifully devious. They brought research subjects into a room with a large wheel of fortune — the carnival kind, with numbers painted around its rim — and asked them to spin it. The wheel was rigged to stop on either 10 or 65. Then they asked a question that had nothing whatsoever to do with the wheel: What percentage of African countries are members of the United Nations?1
Think about this for a moment. You just spun a random wheel. You know it's random. Nobody told you the wheel had anything to do with Africa, or the United Nations, or geopolitics of any kind. And yet:
Subjects who saw the wheel land on 10 guessed, on average, 25%. Subjects who saw it land on 65 guessed 45%. A meaningless, obviously random number moved their answers by twenty percentage points.
This is anchoring. And it is, I think, one of the most unsettling results in all of psychology — not because the effect is large (it is), and not because it replicates (it does, beautifully), but because it tells us something we really don't want to hear about how our minds work. We don't reason from first principles. We reason from whatever number is lying around.
Insufficient Adjustment
Here is the mathematical intuition. When you're asked to estimate something uncertain — the population of Chicago, the boiling point of ethanol, the length of the Nile — your brain doesn't open a fresh spreadsheet. It grabs the nearest number and adjusts. The problem is that adjustment is effortful, and we're cognitively lazy creatures. We adjust upward or downward from whatever starting value we have, but we stop too soon. Always too soon.
Think of it as a kind of mental friction. Each step away from the anchor requires work. And at some point your brain says, "Eh, that feels about right," and you stop. The technical term is insufficient adjustment — you moved in the right direction, sure, but not far enough.2
The scary part? Even transparently absurd anchors work. Researchers have asked people, "Is the Mississippi River longer or shorter than 5,000 miles?" (It's 2,320 miles.) People anchored on 5,000 give dramatically higher estimates than people asked the same question with an anchor of 500 miles. They know the Mississippi isn't 5,000 miles long. They reject the anchor explicitly. And it shifts their answer anyway.3
Try It Yourself
I can tell you about anchoring all day. But the really convincing thing is to feel it happen to you. Below, you'll see a random number followed by an estimation question. Just type your best guess. After five questions, I'll show you what the anchors did to your brain.
The Anchoring Experiment
You'll see a random number, then an estimation question. Answer honestly — don't try to "beat" the anchor.
The Price of Everything
If anchoring only worked in psychology labs, it would be a cute finding and nothing more. But anchoring operates everywhere there's a number and a decision — which is to say, everywhere.
Listing prices are anchors. When a seller lists a house at $650,000, every subsequent negotiation orbits that number. Buyers' agents know this; sellers' agents know this even better. One study found that for every $1,000 increase in the listing price, the final sale price rose by roughly $400 — even when the house was objectively overpriced. The listing price isn't an invitation to negotiate. It's a gravitational field.4
Prosecutors anchor judges. In 2006, Birte Englich and colleagues showed that even experienced German judges were anchored by the prosecutor's sentencing demand. That's not so surprising — prosecutors have relevant information. What is surprising is the follow-up: they had judges roll dice before sentencing, and the dice outcomes shifted their sentences. The judges who rolled high numbers gave longer sentences. Dice.5
Wine lists are engineered anchors. That $300 bottle at the top of the list isn't there because the restaurant expects you to order it. It's there because after seeing $300, an $80 bottle looks like a sensible, moderate choice. Without the $300 anchor, you'd have ordered the $40 bottle. The decoy does its job without anyone ever buying it.
Why Experts Aren't Immune
You might be thinking: Okay, but I'm not a random undergraduate in a psych lab. I know about anchoring. Surely that protects me?
It doesn't. Not really.
Real estate agents who were shown a listing price and asked to estimate the home's value were just as anchored as regular folks — they just denied it more vehemently.6 Experienced judges, as we saw, were anchored by dice. Physicians shown an anchor before making a diagnosis adjust from it just as reliably as patients do. Knowing about anchoring does not stop anchoring — it merely changes how you feel about being anchored.
This is what makes it different from many cognitive biases. Most biases yield somewhat to expertise and deliberation. Anchoring doesn't particularly care about your credentials. The mechanism runs beneath the level where your expertise operates. Your analytical mind can check an answer's reasonableness, but it can't undo the anchor's effect on where you started checking.
The Negotiation Game
Anchoring's most consequential playground is negotiation — and specifically, salary negotiation. Research consistently shows that the first offer in a negotiation acts as a powerful anchor: whoever names a number first sets the gravitational center around which the final agreement orbits.7
Try this: negotiate a salary below. You'll play three rounds with different strategies — letting the employer go first, going first with a moderate ask, and going first with an aggressive ask. Watch how your first offer reshapes the entire conversation.
Salary Negotiation Simulator
Negotiate a starting salary. The employer's budget is hidden. Try different strategies and see how anchoring changes the outcome.
Scale Distortion
There's a subtler flavor of anchoring that goes beyond shifting a number up or down. Sometimes an anchor doesn't just move your estimate — it warps the entire scale you're thinking on.
Ask someone, "How happy are you on a scale of 1 to 5?" and you get one distribution of answers. Ask them the same question on a scale of 1 to 10, and you don't just get the same answers doubled. The wider scale spreads people out differently. The endpoints anchor the meaning of the entire response space. A "3 out of 5" and a "6 out of 10" feel different even though they shouldn't.
This is why survey design is so much harder than it looks. The choice of scale endpoints isn't neutral — it's a choice about what counts as extreme. Rating systems create their own anchored reality.
The Geometry of Starting Points
Here's where the math gets interesting. There are actually two competing theories about why anchoring works, and they predict different things.
The first theory — Tversky and Kahneman's original — is anchoring-and-adjustment. You start at the anchor and adjust until you reach a value that seems plausible. But "seems plausible" is a fuzzy boundary, and you stop the moment you cross into the region of acceptable answers. Since you're approaching from one side, you stop at the near edge of that region rather than its center.
Picture it geometrically. The set of answers you'd consider "reasonable" for the population of Canada forms an interval — say, 25 million to 55 million. If you're adjusting down from an anchor of 85 million, you stop at 55 million, the first plausible number you hit. If you're adjusting up from 12 million, you stop at 25 million. Both answers are "reasonable," but they're on opposite edges of reasonable.
The second theory, proposed by Strack and Mussweiler, is more insidious: selective accessibility. The anchor doesn't just give you a starting point — it changes what information comes to mind. When asked "Is the population of Canada greater or less than 85 million?", your brain automatically starts testing the hypothesis that it is 85 million, and in doing so, activates knowledge consistent with a large population: Canada is geographically huge, it's a wealthy G7 nation, it has major cities. When the anchor is 12 million, different facts activate: it's mostly empty wilderness, the population clusters along the US border, winters are brutal.9
This is the deeper reason experts aren't immune. It's not that they fail to adjust far enough — it's that the anchor literally changes what they think of. The real estate agent anchored on $650K doesn't just start high and adjust down. She starts noticing the granite countertops and the south-facing windows. Anchored on $450K, she'd notice the water stain on the ceiling and the dated fixtures. The anchor doesn't just move the answer. It curates the evidence.
A Bayesian would say we should treat the anchor as zero-information and form our estimate from our prior knowledge alone. But Bayesian rationality requires knowing what information is and isn't relevant — and anchoring operates at the stage before that distinction gets made, at the level where your brain decides what to retrieve from memory. You can't rationally ignore a stimulus that has already reshaped the inputs to your reasoning.
De-Anchoring: Fighting the Gravity
So if knowing about anchoring doesn't fix it, what does? The research offers three imperfect but useful strategies:
Strategy 1: Consider the Opposite
When you catch yourself anchored, ask: What if the true answer were in the opposite direction? If you just saw a listing price of $650K, force yourself to generate reasons the house might be worth $450K. This doesn't eliminate anchoring, but it roughly halves its effect — which, given how strong anchoring is, qualifies as a major victory.
Strategy 2: Generate Multiple Anchors
One anchor dominates. Two anchors compete. If you're about to negotiate a salary, look up the 25th percentile, median, and 75th percentile for your role and location. Now your brain has three anchors pulling in different directions, and the result is closer to the truth than any single anchor would produce.
Strategy 3: Use Base Rates
Before encountering any anchor, do your homework. If you know the median home price in the neighborhood is $520K, the listing price of $650K has less gravitational pull. Base rates don't make you immune, but they give you a competing anchor that's actually informative rather than arbitrary.8
None of these are perfect. Anchoring is not the kind of bias you overcome through willpower or education. It's the kind you manage through process — by deliberately structuring your decision environment so that the first number you see is a number worth seeing.
Which brings us back to the wheel of fortune. Kahneman and Tversky weren't just showing us a quirk of cognition. They were showing us something about the geometry of thought. We don't start from zero and build up. We start from somewhere — and that somewhere, however arbitrary, shapes everywhere we end up.
There's a deep mathematical lesson here, one that goes beyond psychology. In optimization, your starting point determines which local minimum you find. In differential equations, initial conditions determine the trajectory forever after. In iterative algorithms, the seed value shapes the convergence. Our minds, it turns out, work the same way. We are initial-condition-dependent machines.
And this is both the bad news and the good news. The bad news: you cannot think about a number without being contaminated by it. The question isn't whether you're anchored — you're always anchored. But the good news is that anchors, like initial conditions, can be chosen. You can curate the first numbers you encounter. You can build checklists and base rates and decision procedures that plant good anchors before bad ones arrive. You can't escape the gravitational field, but you can choose which star to orbit.
The question isn't whether you're anchored. The question is: who chose your anchor?