Your critical thinking gym
Home
The Commandments
Join the Club
Your critical thinking gym
● Cognitive bias
← Named and defined by psychologists Amos Tversky and Daniel Kahneman, 1972, in their paper on subjective probability judgments.
Judging how likely something is by how much it resembles your mental picture of that category, instead of by the actual odds.
Sometimes shortened to representativeness bias. Not a synonym for stereotyping, base rate neglect, or the gambler's fallacy, each a specific offshoot of this heuristic, not another name for it (Sections 03 and 09).
The tendency to judge how likely something is to belong to a category, or how likely an event is to happen, by how closely it matches your mental prototype of that category or event, rather than by the actual statistical odds.
Plain version
Your brain has a stock "picture" of what things in a given category usually look like: what a nurse looks like, what a scam looks like, what a winning lottery sequence looks like. When something in front of you matches that picture closely, it feels probable. When it doesn't match, it feels unlikely. The catch: how well something matches your mental picture and how likely it actually is are two completely different questions, and this heuristic quietly treats them as the same one.
Aliases
Representativeness bias is used interchangeably with representativeness heuristic. No other common aliases.
Related, but not the same thing
Base rate neglect, the gambler's fallacy, and the conjunction fallacy are all specific things that can happen because of this heuristic, not other names for it. Representativeness heuristic is the general mechanism (judging by resemblance); the other three are particular failures it can produce. The exact distinctions are spelled out in Section 09.
Scope
Hard to catch in the moment. A good resemblance-match doesn't register as a shortcut, it registers as recognition, the same feeling as correctly identifying a friend's face in a crowd. There's an easy, reasonable-sounding story available afterward ("well, it did fit"), and getting the pattern-match "click" delivers a small, real payoff of feeling like you've understood something quickly. Nothing about it feels like a guess.
Why your brain does this
Sorting things into categories by how much they resemble a known example is one of the fastest, cheapest ways to make a decision with incomplete information, and for most of human history it worked well enough to matter. If something rustling in the grass moves and sounds like past predators, treating it as probably dangerous, without stopping to calculate an exact statistical likelihood, is the difference between reacting in time and not. Resemblance-based judgment is a genuinely useful shortcut for a mind that has to act fast and can't run the numbers on everything. This isn't a broken brain. It's a fast one, built for a world where a close-enough match to a known pattern usually was a meaningful signal.
Where it misfires
The shortcut breaks down specifically when a vivid, resemblance-based impression is pitted against numerical information the brain treats as boring background noise, especially base rates (how common something actually is overall) and sample size (how much data the impression is even based on). Told a detailed, specific-sounding description, people will confidently rate it as fitting a narrow, more "colorful" category over a broader, blander one, even when the broader category is, mathematically, far more likely to be true, and even when adding an extra plausible-sounding detail to a description makes it feel more probable when logically it can only be equally or less likely (the conjunction fallacy, Section 09).
The live academic debate, briefly
Gerd Gigerenzer's ecological-rationality research found that reframing these same probability problems using natural frequencies ("40 out of 1,000 people," instead of "4%") sharply reduces the error, arguing the mind is actually well-adapted to reason about frequencies encountered directly in experience, and it's the abstract percentage format itself that's unnatural. The more traditional Tversky and Kahneman "heuristics and biases" framing treats the underlying substitution, resemblance standing in for probability, as a specific, provable reasoning error regardless of format. This page doesn't take a side; both views agree that resemblance and real-world frequency are different things, and confusing them is where the trouble starts.
How it activates, step by step
1. You're given a description, an impression, or a case, and asked (or ask yourself) how likely it is to belong to a category, or how likely an outcome is.
2. Your mind compares it to a stored prototype: what a "typical" member of that category looks, sounds, or acts like.
3. It registers how closely the two match, not how common the category actually is, and not how much real evidence you have.
4. That match score gets reported back as if it were the answer to the original probability question, with base rates and sample size quietly dropped from the calculation.
Representativeness is one of the easiest biases to weaponize, because triggering it doesn't require presenting false information, only presenting a resemblance and letting the audience's own brain fill in the probability judgment.
This is the most direct exploitation on the list: a communicator paints a group with a small set of exaggerated, easily-matched traits, then relies on the audience's representativeness heuristic to treat any individual who fits those traits as probably embodying the whole caricature, base rates and individual variation ignored entirely.
Choosing a spokesperson who visibly resembles the audience's mental prototype of a "typical successful user," a "regular parent," or a "credible expert" borrows the same match-and-conclude shortcut: the messenger's resemblance to the category does the persuading, separate from any actual evidence about the product or claim.
Pairing a claim or product with imagery that resembles trusted authority, a lab coat, an official-looking seal, a flag, activates the same resemblance-to-category shortcut in reverse: it isn't evidence of legitimacy, but it looks enough like legitimacy's usual packaging to be processed as if it were.
Placing a person or group next to an unrelated bad actor, visually or rhetorically, lets the audience's own category-matching do the smearing: "stands near a bad example" gets processed as "probably resembles that category," with no actual claim of wrongdoing ever stated outright.
Outside politics
Packaging design leans on this constantly: rustic fonts, kraft-paper textures, and hand-drawn illustrations are used to make a mass-produced product resemble the prototype of something small-batch and artisanal, and health-food marketing uses clinical-looking fonts, white space, and lab-style imagery to make a product resemble the prototype of something scientifically vetted, independent of what's actually inside.
What increases it
Strong: vivid, specific, story-like descriptions. The more detailed and "novelistic" a description feels, the more strongly it triggers prototype-matching over base-rate thinking, this is the core mechanism behind the conjunction fallacy (Section 09).
Moderate to strong: lower numeracy. Individual-differences research finds that people who score lower on numeracy and cognitive-reflection measures are more likely to fall for representativeness-driven errors like the conjunction fallacy, though the relationship is a correlation, not a guarantee (Liberali et al., 2012).
Moderate: how the probability question is framed. Presenting the same problem as a natural frequency rather than an abstract percentage measurably reduces the error for most people (Section 08).
What does NOT predict it
Formal statistical training reduces but does not eliminate susceptibility. Tversky and Kahneman found graduate students trained in statistics still committed the conjunction fallacy on the Linda problem at high rates (Section 05); expertise narrows the gap, it doesn't close it.
State-dependent factors
Time pressure and mental workload increase reliance on fast, intuitive judgment generally, which favors resemblance-based answers over the slower, deliberate step of checking a base rate. No representativeness-specific stress study is cited here; this follows from the broader, well-established dual-process finding that cognitive load pushes judgment toward heuristic shortcuts across the board.
The pattern across all four: fictional students, real voters, and ordinary shoppers all substitute "does this match the picture in my head" for "what does the actual evidence or base rate say," and trained statisticians and everyday consumers do it at similar rates.
First named
Psychologists Amos Tversky and Daniel Kahneman named and defined the heuristic in "Subjective probability: A judgment of representativeness", Cognitive Psychology, 3(3), 430-454 (1972).
The landmark studies
The 1972 paper and its 1973 follow-up, "On the psychology of prediction" (Section 05's "Tom W" problem), established the core finding across a series of category-judgment tasks: given a resemblance-based impression, people systematically underweight or ignore how common the category actually is, even when the real base rate is stated to them directly in the same problem.
Since then
Tversky and Kahneman's 1983 "Linda problem" paper extended the finding into the conjunction fallacy (Section 05), showing the same substitution can produce a judgment that is not just imprecise but logically impossible. A 2025 pre-registered replication of nine of the original 1972 experiments successfully reproduced eight of the nine, indicating the core effect is well-replicated and not seriously in dispute (Mayiwar, Wan, Løhre & Feldman, Quarterly Journal of Experimental Psychology, 2025).
Cross-tradition note
See Section 02 for Gerd Gigerenzer's ecological-rationality framing and his natural-frequency research, the relevant non-English research tradition for this page's adaptive-origin section, alongside the Kahneman/Tversky error framing.
Score: 2 out of 5 (see the breakdown in the card after Section 01).
The in-the-moment question
"Am I judging this by how common it actually is, or just by how well it matches the picture in my head?" A useful follow-up: if you stripped away the vivid details and just asked how many people or cases in this whole category actually look like this, would the answer change?
Why this is hard to catch
A strong resemblance match doesn't feel like a shortcut, it feels exactly like ordinary, accurate recognition, the same feeling you get correctly spotting a friend's face in a crowd. There's an easy, complete-sounding cover story available afterward ("it just fit the pattern"), and the moment of recognition itself delivers a small, genuine payoff of feeling like you've quickly understood something. None of that registers as a warning sign in the moment.
Debiasing research note
The natural-frequency reframing is one of the more genuinely robust debiasing interventions in the literature, not just a plausible-sounding idea: Gigerenzer and Hoffrage's original demonstration has been replicated in professional decision-making contexts, including improving diagnostic reasoning among practicing managers (Natural frequencies facilitate diagnostic inferences of managers). Simple awareness training, just telling people the heuristic exists, has weaker and more mixed evidence for producing lasting behavior change; the format of the question seems to matter more than knowing the bias by name.
Commonly co-occurs with
Once a resemblance-based first impression forms, confirmation bias tends to take over from there, steering attention toward whatever further confirms the match and away from anything that would complicate it.
Often mistaken for
Not the same thing, though closely linked. Representativeness heuristic is the general mechanism: judging by resemblance to a prototype. Base rate neglect is one specific consequence of that mechanism: ignoring the actual numerical frequency of a category once a resemblance judgment has been made. Base rate neglect can occasionally happen through other routes too; representativeness is its most common cause, not its only one.
A narrow, specific case, not a synonym. The gambler's fallacy is representativeness applied to short sequences of independent random events specifically, expecting a small sample (a run of coin flips) to look as balanced as a very long one would. Representativeness heuristic is the broader phenomenon; the gambler's fallacy is one particular way it shows up, and only in that one context.
Not a separate Filter page yet, but worth naming precisely: the conjunction fallacy is the specific logical error (Section 05's Linda problem) where adding a plausible, resembling detail to a description makes a scenario feel more probable, even though it can only be equally or less probable than the description alone. It's a direct product of representativeness, not an independent bias.
Stereotyping (the propaganda technique, Section 03) is a deliberate communication tactic built to exploit this heuristic in an audience. Representativeness heuristic is the vulnerability inside any one person's thinking; stereotyping is the weapon a communicator aims at it. Same distinction as authority bias versus appeal to authority elsewhere on this site.
Compounding effect
Representativeness and base rate neglect frequently compound directly: a strong resemblance match first produces an inflated probability estimate, then the same match makes the actual base rate feel irrelevant or even wrong, "sure, but this case is different," closing the loop against correction from either direction.
Primary sources
Accessible reading