In 1974, Amos Tversky and Daniel Kahneman ran an experiment that should have been impossible. They wheeled a large numbered wheel — the kind you might find on a game show — into a psychology laboratory at Hebrew University of Jerusalem. Before asking subjects any question, they spun the wheel. The wheel was rigged: it could only stop at 10 or 65. Everyone in the room knew it was rigged. The experimenters told them directly. Then came the question: what percentage of African countries are members of the United Nations?
The subjects who watched the wheel land on 10 gave estimates averaging 25 percent. The subjects who watched it land on 65 gave estimates averaging 45 percent. A gap of 20 percentage points — produced by a number that was explicitly, visibly, demonstrably meaningless. The correct answer at the time was approximately 35 percent, but that is almost incidental. What the experiment exposed was something more fundamental: human judgment does not start from nothing. It starts from wherever it happens to be standing, and then moves — but never far enough.
Tversky and Kahneman published this finding in Science in 1974, in a paper titled "Judgment Under Uncertainty: Heuristics and Biases" (Vol. 185, Issue 4157, pp. 1124–1131). It became one of the most cited papers in the history of cognitive psychology. The phenomenon they named was the anchoring-and-adjustment heuristic: a two-stage cognitive process in which the mind first fixes on an initial value — the anchor — and then attempts to adjust from it toward a more accurate estimate. The problem, documented across five decades of subsequent research, is that the adjustment is almost always insufficient. People stop too soon, leaving their estimates too close to the anchor.
What Anchoring and Adjustment Is
Anchoring and adjustment is a cognitive heuristic — a mental shortcut — in which a person generates a numerical estimate by starting from an initial reference value (the anchor) and incrementally shifting that value toward what seems more plausible. The adjustment terminates when the person reaches a value that no longer appears obviously wrong, rather than continuing until the most accurate value is reached.
The result is systematic under-adjustment: final estimates are predictably biased toward whatever starting point was used, regardless of whether that starting point had any legitimate informational relationship to the question being answered.
Anchoring and Adjustment vs. Reference Point Independence
The concept becomes sharper when contrasted with its logical opposite: the normative ideal of reference point independence, in which a rational agent's estimate of a quantity would not depend on what number they happened to encounter before reasoning about it.
| Dimension | Anchoring and Adjustment | Reference Point Independence |
|---|---|---|
| Starting point | Depends on an initial anchor, whether relevant or arbitrary | Derives from first principles and available evidence only |
| Adjustment process | Incremental, stops when estimate feels plausible | Not applicable — estimate is built directly from evidence |
| Sensitivity to irrelevant numbers | High — arbitrary numbers shift final estimates measurably | Zero — irrelevant numbers carry no weight |
| Sensitivity to anchor magnitude | Monotonic — higher anchors produce higher estimates, lower anchors produce lower estimates | Not applicable |
| Expert immunity | Partial — experts show reduced but not eliminated anchoring | Full immunity is the normative ideal |
| Computational cost | Low — heuristic requires minimal deliberation | High — requires exhaustive evidence integration |
| Conditions of greatest distortion | Uncertainty, time pressure, high cognitive load, unfamiliar domains | N/A |
| Debiasing potential | Limited — awareness reduces but does not eliminate the effect | N/A — the ideal requires no debiasing |
The gap between these two columns is the gap between how human judgment actually works and how it would work if cognition were costless and unbounded. That gap is not a malfunction. It is a feature of a system optimized for speed under resource constraints. The problem is that the speed comes at the cost of accuracy in ways that are predictable and exploitable.
The Cognitive Mechanics: How Adjustment Actually Works
The Original Insufficient Adjustment Account
Tversky and Kahneman's original 1974 account described anchoring and adjustment as a sequential process. The mind selects a starting point — either an externally provided number or a self-generated initial estimate — and then mentally "slides" the estimate in the appropriate direction. Adjustment continues until the person reaches a value that no longer seems clearly wrong, at which point they stop.
The critical flaw in this process is the stopping criterion. People stop when they reach the edge of a plausible range, not when they reach the most likely value. Because the plausible range is itself anchored — its near boundary is closer to the anchor than its far boundary — early stopping systematically produces estimates that are biased toward the anchor. The adjustment was, in Tversky and Kahneman's phrase, "anchored in" rather than calibrated to the true value.
The Confirmatory Hypothesis Testing Account
In 1997, Fritz Strack and Thomas Mussweiler published a second account of the mechanism in the Journal of Personality and Social Psychology (Vol. 73, No. 3, pp. 437–446), titled "Explaining the Enigmatic Anchoring Effect: Mechanisms of Selective Accessibility." Their argument challenged and extended the insufficient adjustment model.
Strack and Mussweiler proposed that when an anchor is presented, the mind does not merely use it as a numerical starting point. Instead, it engages in confirmatory hypothesis testing: the anchor is treated as a hypothesis ("Is the true value approximately this?"), and the mind searches memory for evidence consistent with that hypothesis. This selective accessibility of anchor-consistent information temporarily biases the evidence pool from which the final estimate is drawn. The estimate shifts toward the anchor not only because adjustment is lazy, but because the information retrieved during the anchoring phase was pre-filtered to support the anchor.
They demonstrated this with a clever experiment: subjects were asked to compare the Mississippi River to an anchor value (either very short or very long), then asked to estimate the river's length. Subjects exposed to a "long" anchor retrieved anchor-consistent facts (the river traverses much of North America, drains a vast basin) and gave longer estimates. Subjects exposed to a "short" anchor retrieved different facts (the navigable section, the urban portion) and gave shorter estimates. The same river, the same subjects, different information pulled from long-term memory — all because the anchor primed confirmatory search.
This mechanism explains why anchoring persists even when adjustment is motivated: the deck was stacked during the retrieval phase, before conscious adjustment began.
The Epley and Gilovich Distinction: External vs. Self-Generated Anchors
Nicholas Epley and Thomas Gilovich, in a series of studies published in 2001 (Journal of Personality and Social Psychology, Vol. 80, No. 5, pp. 668–680) and 2006 (Psychological Science, Vol. 17, No. 4, pp. 311–318), introduced a critical distinction that the prior literature had conflated.
External anchors — numbers provided by an experimenter, a price tag, a first offer — operate primarily through selective accessibility. The mind searches for anchor-consistent information and adjusts from the contaminated starting point. Self-generated anchors — numbers that participants produce themselves as initial approximations before refining — operate primarily through insufficient adjustment. When you try to estimate how many gallons of water a bathtub holds by starting from your gut reaction and adjusting, the adjustment itself is the bottleneck, not selective retrieval.
This distinction matters practically. Epley and Gilovich showed that self-generated anchoring is more sensitive to motivational factors. When participants were given strong incentives for accuracy, or when they were told to keep adjusting even after reaching a first estimate, the anchoring effect from self-generated anchors was substantially reduced. This reduction did not occur with external anchors. The implication: insufficient adjustment, in the self-generation case, is at least partially about insufficient motivation to continue the effortful adjustment process — not just an automatic retrieval contamination.
In their 2006 paper, Epley and Gilovich also showed that self-generated adjustment typically begins at a value the person knows to be wrong (a boundary value, like zero, or a qualitatively different category) and moves toward plausibility. This produces characteristic under-adjustment because people interpret "plausible" as "no longer obviously wrong" rather than "most likely correct."
The Role of Cognitive Load
Ariely, Loewenstein, and Prelec's 2003 study in the Quarterly Journal of Economics (Vol. 118, No. 1, pp. 73–105), "Coherent Arbitrariness," demonstrated a related and troubling pattern: even arbitrary personal numbers (specifically, the last two digits of Social Security Numbers) served as powerful anchors for willingness-to-pay estimates for unrelated consumer goods. High-digit participants bid systematically more than low-digit participants, with a correlation of approximately 0.5 between the arbitrary number and the bid.
What this demonstrated was not merely that anchors work, but that the cognitive system cannot reliably distinguish informative from uninformative anchors at the point of encoding. The anchor enters the estimation process before deliberate assessment of its relevance can exclude it.
Under high cognitive load — time pressure, concurrent tasks, emotional arousal — this pre-filtering failure worsens. Studies by Wilson, Houston, Etling, and Brekke (1996, Journal of Personality and Social Psychology, Vol. 71, No. 4) showed that distracted participants exhibited stronger anchoring effects than undistracted ones. Adjustment is an effortful process; anything that reduces available cognitive resources reduces adjustment, leaving estimates closer to the anchor.
Intellectual Lineage
Pre-1974 Precursors
The conceptual roots of anchoring and adjustment run earlier than 1974. In the late 1950s, Muzafer Sherif's work on the autokinetic effect — in which participants' judgments of a stationary light's apparent movement converged on group norms — demonstrated that numerical judgments anchor on socially provided reference points. Sherif was not studying heuristics; he was studying conformity. But the structure of the effect is recognizable: an initial value, provided externally, pulls subsequent estimates toward itself.
Helson's Adaptation Level Theory (1964) proposed that all judgments are made relative to an adaptation level — a weighted average of recent stimuli — and that departures from this level are systematically underestimated. This is not identical to anchoring-and-adjustment, but it prefigured the key insight: cognition is reference-relative, not absolute.
Tversky and Kahneman's Heuristics and Biases Program
The 1974 Science paper was part of a broader intellectual project that Tversky and Kahneman called the "heuristics and biases" research program. The paper described three heuristics — representativeness, availability, and anchoring-and-adjustment — each of which produced characteristic and predictable errors. The program's central claim was that human judgment under uncertainty is not random noise, but systematically biased in directions determined by cognitive shortcuts.
This framing was deliberately provocative. It challenged both the Bayesian model of rational belief updating and the prior consensus among economists that departures from rationality were small, unsystematic, and self-correcting. The heuristics and biases program argued instead that the biases are large, consistent, and robust to expertise, motivation, and stakes.
Daniel Kahneman received the Nobel Prize in Economic Sciences in 2002 (Tversky had died in 1996) in part for this body of work. The citation specifically mentioned the demonstration that human judgment departs systematically from rational models in predictable ways.
Post-1974 Theoretical Development
Through the 1980s, the anchoring-and-adjustment heuristic was treated primarily as a single mechanism. The Strack and Mussweiler (1997) paper was the first major challenge, arguing for a selective accessibility account that applied primarily to externally provided anchors. Their follow-up work with Mussweiler and Strack (1999, Personality and Social Psychology Bulletin, Vol. 25, No. 1, pp. 71–79) elaborated the confirmatory hypothesis testing mechanism and showed that the semantic overlap between anchor and target moderated effect size: anchors that shared conceptual features with the target produced stronger effects than semantically distant anchors.
Epley and Gilovich's 2001 and 2006 papers then rehabilitated the original insufficient adjustment account, but restricted it to self-generated anchors and showed that it was motivationally malleable in ways that external-anchor effects were not. By the mid-2000s, the field had arrived at a two-process model: selective accessibility for external anchors, insufficient adjustment for internal ones, with overlapping influence in many real-world cases.
Chapman and Johnson's 1999 chapter "Anchoring, Activation, and the Construction of Values" (in Elicitation of Preferences, edited by Kahneman and Tversky, Cambridge University Press) provided a synthesis, arguing that both mechanisms operate through a common deeper process: anchors activate information that becomes disproportionately accessible in subsequent judgment, whether through confirmatory retrieval or through the simple cognitive inertia of self-generated adjustment.
Four Case Studies Across Domains
Case Study 1: Real Estate Appraisal — Northcraft and Neale, Tucson, 1987
Gregory Northcraft and Margaret Neale, then at the University of Arizona, recruited participants and provided them with detailed property information packets for actual residential properties in Tucson, including square footage, room count, condition, and recent comparable sales data. The packets varied in one respect: the listing price, which was set at approximately $119,900, $129,900, $139,900, or $149,900 for the same property.
Two populations were tested: undergraduate students and professional real estate agents with active listings and appraisal experience. Appraised values from student participants ranged across the four conditions by approximately $22,000 — a direct tracking of the anchor. Professional agents' estimates ranged by approximately $14,000 — smaller, but still substantial. The study, published in Organizational Behavior and Human Decision Processes (Vol. 39, No. 1, pp. 84–97), showed a correlation of 0.41 between listing price and agent appraisal.
Critically, post-experiment interviews revealed that agents almost uniformly denied relying on the listing price. They cited their professional reliance on comparable sales and objective property features. The numerical data contradicted their self-report. The anchor had operated below the level of conscious awareness even in the domain professionals claimed as their own.
Case Study 2: Clinical Diagnosis — Chapman and Johnson on Anchoring in Medical Judgment
Gretchen Chapman and Eric Johnson's work on anchoring in clinical contexts demonstrated that diagnostic anchors — an initial hypothesis about a patient's condition — produce under-adjustment even when contradictory evidence is presented. Physicians who received an initial diagnostic suggestion at the start of a case file showed reduced sensitivity to subsequently presented evidence that contradicted that diagnosis.
This pattern, sometimes called diagnostic anchoring in the clinical literature, was documented in emergency medicine settings where initial triage assessments (e.g., "probable cardiac event") influenced subsequent interpretation of diagnostic test results. Physicians adjusted their diagnostic probabilities less than Bayesian updating would require, producing a characteristic bias toward the initial working hypothesis. Chapman and Johnson argued that this is the confirmatory hypothesis testing mechanism applied to clinical reasoning: the initial diagnosis primes retrieval of diagnosis-consistent information, reducing the functional weight assigned to contradictory findings.
Case Study 3: Legal Sentencing — Englich, Mussweiler, and Strack, Germany, 2006
Birte Englich of the University of Wurzburg, collaborating with Thomas Mussweiler and Fritz Strack, published a series of studies in 2006 in Psychological Science (Vol. 17, No. 7, pp. 611–617) examining anchoring in judicial sentencing. The design was direct: experienced German judges were given a case file for a shoplifting offense and asked to recommend a sentence in months. Before deliberating, judges rolled a pair of dice that were, without their knowledge, loaded to produce either 3 or 9.
Judges who rolled 3 recommended sentences averaging approximately 5 months. Judges who rolled 9 recommended sentences averaging approximately 8 months. The dice roll — a transparently random outcome, bearing no legal or evidentiary relationship to the case — produced a 60 percent difference in sentencing recommendations between the two groups.
A second version of the experiment used a prosecutorial sentencing request as the anchor, varying the request from an implausibly short to an implausibly long sentence. Judges again adjusted insufficiently from the prosecutorial anchor. An explicit warning about anchoring, delivered to a subset of judges before they deliberated, reduced but did not eliminate the effect. The implication for legal systems that allow prosecutorial sentencing requests — or that expose judges to arbitrary numerical information before deliberation — is sobering.
Case Study 4: Investment and Valuation — Stock Price Anchoring
In equity valuation, the 52-week high price of a stock functions as a powerful anchor for investor price targets and sell-side analyst forecasts. George and Hwang (2004, Journal of Finance, Vol. 59, No. 5, pp. 2145–2176) documented that stocks trading near their 52-week high earned positive subsequent returns, while stocks far below their 52-week high earned negative returns — a pattern inconsistent with efficient markets but consistent with anchoring. Analysts and investors adjusted insufficiently from the historical high, treating prices near the high as "too expensive" and prices far below as "undervalued," even when the fundamental case was identical.
Similarly, initial public offering (IPO) pricing has been shown to anchor on prior comparable offering prices. Ljungqvist and Wilhelm (2003, Journal of Finance, Vol. 58, No. 2, pp. 723–761) found that IPO offer prices are systematically anchored to the prices of recent comparable IPOs, producing clusters of over- and underpricing as the anchor moves. The adjustment from comparable pricing is insufficient to produce market-clearing prices, contributing to the well-documented IPO underpricing phenomenon.
Empirical Research Findings
Effect Sizes and Robustness
Anchoring effects are among the most replicable findings in behavioral science. A 2006 meta-analysis by Furnham and Boo (Journal of Behavioral Decision Making, Vol. 24, No. 1, pp. 35–65, 2011) examined 40 studies and found a mean effect size of d = 0.97 — large by conventional standards — with minimal publication bias. Unlike many behavioral science findings that weakened or collapsed in replication attempts during the 2010s, anchoring has consistently replicated across laboratories, cultures, participant populations, and experimental paradigms.
Effect sizes are moderated by several factors. Domain familiarity reduces but does not eliminate anchoring — the correlation between expertise and anchoring resistance is consistently negative but small. The plausibility of the anchor matters: implausible anchors (Gandhi died at 140) produce somewhat smaller effects than moderately plausible ones, because subjects discount extreme anchors more aggressively. However, even implausible anchors produce statistically significant anchoring effects, suggesting that partial discounting of an extreme anchor still leaves substantial residual influence.
Debiasing Attempts and Their Limits
Attempts to reduce anchoring through training, incentives, and awareness have produced a consistent finding: partial success, not elimination. Wilson, Houston, Etling, and Brekke (1996) showed that asking participants to consider reasons why the anchor might be wrong — "consider the opposite" instructions — reduced anchoring substantially, more than simply being warned about it. The mechanism appears to be that considering opposing reasons manually generates anchor-inconsistent information, partially counteracting the selective accessibility process.
Epley and Gilovich (2001) showed that increasing incentives for accuracy reduced self-generated anchor effects but not external anchor effects, consistent with the claim that self-generated adjustment is motivationally malleable while external anchor effects operate through more automatic selective retrieval.
Critically, professional expertise does not confer immunity. Northcraft and Neale (1987) with real estate agents, Englich and colleagues (2006) with judges, and multiple studies of financial analysts have all shown that domain experts exhibit anchoring effects that are statistically significant even when reduced relative to novices. The Dunning-Kruger-adjacent assumption that expertise eliminates cognitive bias has found no support in the anchoring literature specifically.
Cross-Cultural Evidence
Studies across the United States, Germany, China, and Israel have found anchoring effects of comparable magnitude, suggesting that the mechanism is not culturally specific. Chaxel, Russo, and Kerimi (2013, Journal of Behavioral Decision Making, Vol. 26, No. 4, pp. 323–334) demonstrated anchoring in choice contexts beyond numerical estimation, including categorical decisions, showing that the phenomenon generalizes beyond the original experimental paradigm. The robustness across cultural and methodological variations is one reason the 1974 Tversky and Kahneman paper retained its influence even as other findings from that era have faced replication challenges.
Limits, Nuances, and Boundary Conditions
When Anchors Lose Their Power
Chapman and Johnson's 1999 synthesis identified several conditions under which anchoring effects are attenuated:
First, when the target judgment is unambiguously determined by available information — when there is effectively only one correct answer and participants know the formula for reaching it — anchors have little room to operate. The mechanism requires uncertainty. In the absence of uncertainty, adjustment reaches the correct value regardless of starting point.
Second, when subjects have strong, accessible prior beliefs about the correct value — genuine domain expertise combined with relevant experience — anchors are resisted more effectively. The expert's pre-existing knowledge competes with the anchor for influence over the final estimate.
Third, when the anchor is so extreme as to be treated as nonsensical — far outside any plausible range the subject can conceive — the anchor may be discarded entirely rather than used as a starting point. This is the "anchoring at the boundary" failure: an anchor of one trillion dollars for a bicycle does not produce estimates of several hundred million dollars; it is simply ignored. However, the threshold for "too extreme to matter" is itself context-dependent and higher than most people intuit.
The Assimilation-Contrast Distinction
Psychophysics literature has long distinguished between assimilation effects — where responses move toward a reference stimulus — and contrast effects — where responses move away from it. Anchoring is an assimilation effect. But under certain conditions, anchors can produce contrast: if the anchor is perceived as so extreme or obviously inappropriate that it triggers a correction in the opposite direction, the estimate may depart from the anchor in the direction away from it.
Mussweiler, Strack, and Pfeiffer (2000, Personality and Social Psychology Bulletin, Vol. 26, No. 9, pp. 1142–1150) demonstrated that forewarning subjects that they were about to be anchored produced contrast effects when subjects were highly motivated to overcome the bias — they overshot in the corrective direction. This has practical implications: telling people they are susceptible to anchoring without giving them effective tools to resist it may cause them to overcorrect, trading one distortion for another.
Anchoring Is Not the Same as Priming
A persistent confusion in both popular and academic treatments conflates anchoring with priming. Priming refers broadly to the influence of a prior stimulus on a subsequent response; anchoring is a specific application to numerical judgment involving a starting point and a directional adjustment process. Not all numerical priming is anchoring. A number that activates semantic associations without serving as an adjustment starting point is priming but not anchoring in the Tversky-Kahneman sense.
The Epley-Gilovich distinction between external and self-generated anchors complicates this further. External anchors may operate primarily through accessibility mechanisms that resemble priming. Self-generated anchors operate through insufficient effortful adjustment. Lumping both under "anchoring" risks obscuring the different conditions under which each is susceptible to debiasing.
The Insufficiency of "Knowing About It"
Perhaps the most practically significant limit is also the most frustrating: knowing that anchoring exists, knowing the mechanism, and knowing you are being anchored does not reliably allow you to escape it. Briefing on the phenomenon reduces the effect modestly. Deliberate counter-argumentation reduces it more substantially. But the residual effect remains statistically significant even in participants who are actively trying to resist it.
Kahneman's own account in Thinking, Fast and Slow (2011, Farrar, Straus and Giroux) acknowledges this flatly: after five decades of studying and writing about anchoring, he remains susceptible to it. The mechanism operates at a level of cognitive processing that deliberate metacognitive awareness cannot fully override. The practical implication for anyone making decisions under uncertainty — which is to say, anyone making decisions — is that process redesign, not awareness alone, is required for meaningful debiasing.
References
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131.
Strack, F., & Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility. Journal of Personality and Social Psychology, 73(3), 437–446.
Epley, N., & Gilovich, T. (2001). Putting adjustment back in the anchoring and adjustment heuristic: Differential processing of self-generated and experimenter-provided anchors. Journal of Personality and Social Psychology, 80(5), 668–680.
Epley, N., & Gilovich, T. (2006). The anchoring-and-adjustment heuristic: Why the adjustments are insufficient. Psychological Science, 17(4), 311–318.
Mussweiler, T., & Strack, F. (1999). Hypothesis-consistent testing and semantic priming in the anchoring paradigm: A selective accessibility model. Personality and Social Psychology Bulletin, 25(1), 71–79.
Northcraft, G. B., & Neale, M. A. (1987). Experts, amateurs, and real estate: An anchoring-and-adjustment perspective on property pricing decisions. Organizational Behavior and Human Decision Processes, 39(1), 84–97.
Chapman, G. B., & Johnson, E. J. (1999). Anchoring, activation, and the construction of values. In D. Kahneman & A. Tversky (Eds.), Choices, Values, and Frames (pp. 120–140). Cambridge University Press.
Ariely, D., Loewenstein, G., & Prelec, D. (2003). Coherent arbitrariness: Stable demand curves without stable preferences. Quarterly Journal of Economics, 118(1), 73–105.
Englich, B., Mussweiler, T., & Strack, F. (2006). Playing dice with criminal sentences: The influence of irrelevant anchors on experts' judicial decision making. Psychological Science, 17(7), 611–617.
Wilson, T. D., Houston, C. E., Etling, K. M., & Brekke, N. (1996). A new look at anchoring effects: Basic anchoring and its antecedents. Journal of Experimental Psychology: General, 125(4), 387–402.
George, J. M., & Hwang, C. Y. (2004). The 52-week high and momentum investing. Journal of Finance, 59(5), 2145–2176.
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
Frequently Asked Questions
What is anchoring and adjustment?
Anchoring and adjustment is a two-stage cognitive process in which people begin with an initial value — the anchor — and adjust their estimates from it to arrive at a final judgment. The systematic problem, identified by Tversky and Kahneman in their 1974 Science paper, is that adjustment is consistently insufficient: estimates remain biased toward the anchor even when the anchor is known to be arbitrary or irrelevant. The phenomenon explains why the first number encountered in a negotiation, appraisal, or estimation task exerts disproportionate influence on all subsequent reasoning.
What did the spinning wheel experiment find?
Tversky and Kahneman's 1974 study used a rigged wheel of fortune that was programmed to stop at either 10 or 65. Subjects watched the wheel spin and land, then were asked: 'What percentage of African countries are members of the United Nations?' Subjects who saw the wheel land on 65 gave a median estimate of 45%. Subjects who saw it land on 10 gave a median estimate of 25% — a 20-percentage-point difference driven entirely by an irrelevant random number. Subjects knew the wheel was random and the number had nothing to do with the question. The anchor influenced estimates regardless.
What is the selective accessibility model of anchoring?
Fritz Strack and Thomas Mussweiler's 1997 Journal of Personality and Social Psychology paper proposed that anchoring works through selective accessibility: when people consider whether the anchor value is a plausible answer, they activate anchor-consistent information in memory. This information then remains accessible when they make their final estimate, biasing it toward the anchor. Mussweiler and Strack tested this by showing that semantic priming after anchoring was specific to anchor-consistent concepts, confirming that the anchor had selectively activated related knowledge. The model explains why anchors work even when subjects are told to ignore them: the accessibility bias operates before the deliberate adjustment process begins.
How does anchoring affect legal sentencing?
Birte Englich, Thomas Mussweiler, and Fritz Strack's 2006 Personality and Social Psychology Bulletin study had experienced German judges read a case summary, then roll a pair of dice (manipulated to show either 1+2=3 or 4+5=9) before recommending a sentence. Judges who rolled the higher number recommended an average of 8 months; judges who rolled the lower number recommended an average of 5 months — a 50% difference driven by a pair of dice. The study replicated with legal prosecutor sentencing demands, showing that the number a prosecutor requests — itself potentially anchored on arbitrary factors — substantially influences judges' own independent sentencing recommendations.
Can anchoring be reduced?
Debiasing anchoring is difficult. Awareness of the anchor's irrelevance does not eliminate its effect — the Tversky and Kahneman subjects knew the wheel was random. Consider-the-opposite instructions (explicitly asking what reasons exist against the anchor) reduce the effect modestly. Epley and Gilovich's 2006 research found that self-generated anchors — where the subject produces their own starting estimate rather than receiving an external one — show more adjustment when motivation increases, suggesting that insufficient adjustment in self-generated cases reflects cognitive laziness that effort can partially overcome. External arbitrary anchors, however, appear to operate through accessibility mechanisms that are less responsive to deliberate correction.