The Filter

The Filter Your critical thinking gym
Join the Club
Home
Propaganda Techniques
Cognitive Biases
The Commandments Join the Club

● Cognitive bias

Base Rate Neglect

No Greek or Latin root, just a description of what happens. Named following psychologists Daniel Kahneman and Amos Tversky's 1970s studies on statistical prediction.

Ignoring how common something actually is in favor of a specific story that feels like it fits, even when the story is a far worse guide than the plain statistic sitting right next to it.

Also called: base rate fallacy, base rate bias. In legal settings this same error, applied to statistical test evidence, is sometimes called the prosecutor's fallacy or defense attorney's fallacy (Section 01). Not the same as the gambler's fallacy (Section 09).

Judgment & Decision-Making Not Enough Meaning Universal: risk, forecasting, snap judgments about people

The tendency to underweight or ignore the underlying statistical prevalence (the base rate) of an event or category, in favor of specific, individuating information about the case at hand, even when that individuating information is weaker or less reliable evidence than the base rate itself.

Plain version

Your brain would rather run with a good, specific-feeling story than a boring statistic about a whole population, even in the exact situations where the statistic is the far more trustworthy guide.

The classic demonstration: told that a group is 70% lawyers and 30% engineers, then given a personality sketch that sounds a bit like a stereotypical engineer, most people guess "engineer," largely ignoring that lawyers outnumbered engineers more than two to one in the room to begin with.

Aliases

Base rate fallacy and base rate bias are used interchangeably with base rate neglect in most research. In law, when the same error shows up in how jurors or lawyers interpret statistical test evidence (like a DNA match probability), it's sometimes labeled the prosecutor's fallacy or defense attorney's fallacy, useful courtroom-specific names, but the underlying mechanism is identical to the general bias, not a distinct one.

Related biases

Not the same as the gambler's fallacy: both involve misjudging probability, but they're different mistakes about different kinds of numbers. Full distinction in Section 09.

Scope

Judgment shortcut Social-perception distortion

Gets confused with

Gambler's fallacy: precise distinction in Section 09.

Self-check score 2 / 5

Slightly easier to catch than the hardest biases on this site, but still low. There's almost no internal alarm: using the vivid, specific detail in front of you doesn't feel like skipping a step, it feels like paying attention to the more relevant thing. The cover story afterward is airtight ("the description just fit"), and there's a mild but real payoff, pattern-matching to a story feels like sharper thinking than looking up a boring number.

Feels like something's happening?
2 / 5
No dramatic internal signal, just a quiet sense that this description is telling you what you need to know.
Easy to explain away after?
5 / 5
"The description just fit" is a completely reasonable-sounding account of what happened, whether you were right or wrong.
Feels good to believe?
3 / 5
Moderate. Pattern-matching to a vivid story feels like sharper thinking than looking up a boring number.

Why your brain does this

For almost all of human history, the probability judgments that mattered weren't handed to you as clean population statistics, they came from direct experience with specific, concrete cases you could see, hear, or remember. A brain built to quickly match a new case against a familiar pattern ("this looks like the dangerous kind") was fast and usually good enough, long before anyone had abstract numbers like "70% lawyers, 30% engineers" to weigh against a written description. This isn't a broken brain. It's a fast one, built for a world that mostly didn't hand out base rates in numeral form.

Where it misfires

The shortcut breaks down specifically when a real, reliable statistic is available, alongside a vivid but weak or non-diagnostic description, and the description simply crowds the number out. The more concrete and story-like the individuating detail, the harder the underlying rate is to hold onto at the same time.

The live academic debate, briefly

This bias is the central case study in the live debate. Kahneman and Tversky's original "heuristics and biases" account treats base rate neglect as a genuine reasoning error, the representativeness heuristic (judging by how much something resembles a category) overriding correct probability math. Psychologist Gerd Gigerenzer's ecological-rationality research pushed back specifically here: when the identical problem is reworded from abstract probabilities into natural frequencies ("10 out of 1,000 people" instead of "1%"), both laypeople and trained professionals solve it correctly far more often, with no extra instruction. On Gigerenzer's reading, the brain was never bad at this kind of reasoning, it was being fed the information in a format it doesn't naturally track. This page doesn't take a side; both views are live in the research, and the practical lesson is the same either way: this isn't stupidity, it's a shortcut that responds enormously to how the numbers are presented.

How it activates, step by step

1. You're given, or already know, a base rate: how common something actually is across a whole population.
2. You're also given a specific, vivid description of one case within that population.
3. Your brain matches the description against a mental stereotype or pattern, rather than combining it mathematically with the base rate.
4. Your final judgment reflects mostly, sometimes almost entirely, the pattern match, not the actual math.

Base rate neglect isn't just something that happens to you passively, several propaganda techniques are built specifically to exploit the gap between a compelling story and the actual numbers.

False urgency / crisis framing

Presents a rare event or threat as an urgent, ongoing pattern. A handful of vivid, recent cases substitutes for the real, usually far lower, rate at which the thing actually happens, and the urgency itself discourages anyone from stopping to ask what the real base rate is.

Cherry picking

Hands you a small number of vivid, representative-feeling cases and lets base rate neglect finish the job. You never ask how common the pattern really is across the full population, because you were only ever shown a hand-picked slice of it.

Stereotyping

Relies on a description matching a group stereotype to stand in for the group's actual statistical makeup, the same representativeness shortcut this bias runs on, aimed deliberately at a category of people instead of a single ambiguous case.

Outside politics

Extended warranty and insurance sales routinely lean on one vivid, scary anecdote ("my neighbor's screen shattered the day after the warranty ran out") to override the far lower actual failure rate, which is usually available but boring. In its more predatory form, investment scams use a single dramatic success story to imply a probability of payoff nowhere close to the real, near-zero base rate of the scheme actually working.

What increases it

Strong: vivid, concrete individuating information. The more story-like and specific the description of a case, the harder the base rate is to hold onto alongside it, this is the core representativeness mechanism Kahneman and Tversky identified.

Moderate: how extreme the base rate is. Research on base rate problems generally finds neglect is often worse at the extremes (very rare or very common), and milder when the base rate sits closer to 50/50.

Moderate: low numeracy. Research using tools like the Berlin Numeracy Test finds people lower in numeracy are more strongly swayed by how numerically equivalent information is framed, which matters directly here since framing (probabilities vs. natural frequencies) is the single biggest lever known for this specific bias.

What does NOT reliably predict it

Professional expertise is not automatically protective. In physician David Eddy's often-cited 1982 study, the great majority of physicians badly overestimated the odds a patient with a positive mammogram actually had breast cancer, trained medical judgment didn't spare them from the same neglect a layperson shows on a simpler version of the same problem.

State-dependent factors

There isn't a bias-specific study directly measuring stress or time pressure here, so this is a reasoned extrapolation rather than a settled finding: because correctly weighing a base rate takes a slower, deliberate calculation while pattern-matching to a description is fast and automatic, time pressure and cognitive load plausibly push judgment further toward the fast route, consistent with general dual-process research, though not confirmed by a study built specifically around this bias.

Terrorism risk perception and probabilistic reasoning
A 2017 study measured probabilistic reasoning ability, including base rate tasks, alongside self-reported perception of terrorism risk and terrorism-related behavior change in a sample of 263 adults. Weaker probabilistic reasoning was significantly associated with higher perceived risk and more behavior change, even though terrorism's actual base rate is extremely low compared with the everyday risks most people don't restructure their lives around (Denovan, Dagnall, Drinkwater, Parker & Clough, Frontiers in Psychology, 2017).
Documented study Political/civic
The mammography problem (Eddy, 1982)
Physicians were given the real numbers for a breast cancer screening scenario: a 1% base rate, an 80% true positive rate, and a 9.6% false positive rate. Asked the probability that a woman with a positive result actually has cancer, the correct answer is about 7.8%, but most physicians estimated somewhere around 70 to 80%, roughly a tenfold overestimate, despite having every number they needed in front of them (Eddy, in Kahneman, Slovic & Tversky (Eds.), Judgment Under Uncertainty, 1982).
Documented study Workplace
Buying the extended warranty after hearing one bad story
Illustrative, not a specific cited study: someone hears a friend's dramatic story about a device breaking right after the warranty expired, and buys extended coverage for their own purchase, without ever checking (or being told) the manufacturer's actual failure rate, which is usually low enough that the warranty isn't worth its price for most buyers.
Illustrative Consumer

The pattern across all three: a specific, vivid story consistently out-competes a true, relevant statistic, in security-conscious citizens, in trained physicians with the numbers on the page, and in ordinary shoppers alike. Expertise and stakes don't reliably protect against it.

First named

Psychologists Daniel Kahneman and Amos Tversky documented the effect in "On the psychology of prediction," Psychological Review, 80, 237 to 251 (1973).

The landmark study

Kahneman and Tversky told participants a panel had interviewed 100 people, either 30 engineers and 70 lawyers, or 70 engineers and 30 lawyers, and gave them a short personality sketch drawn at random from the group. Participants were asked to estimate the probability the person described was an engineer. Estimates barely shifted between the two base-rate conditions: people judged almost entirely by how well the sketch matched an engineer stereotype, largely disregarding whether engineers made up 30% or 70% of the room.

Since then

A pivotal reframing came from psychologist Gerd Gigerenzer and researcher Ulrich Hoffrage, "How to improve Bayesian reasoning without instruction: Frequency formats," Psychological Review, 102, 684 to 704 (1995). Using David Eddy's mammography scenario and others like it, they showed that restating the identical problem in natural frequencies rather than percentages allowed far more people, including physicians in later replications, to reach the correct answer without any training at all. This didn't erase the original finding, but it reframed what was causing it: format, not just faulty reasoning.

Cross-tradition note

See Section 02 for how Gerd Gigerenzer's ecological-rationality program isn't just enrichment here, his natural-frequencies research is the direct origin of the field's best-known correction for this specific bias, not a side note to it.

Score: 2 out of 5 (see the breakdown in the card after Section 01).

The in-the-moment question

Before trusting a snap probability judgment: am I judging this by how common it actually is, or just by how well this one case matches a story already in my head?

Why this is hard to catch

The vivid, case-specific detail genuinely feels like better evidence than a population statistic, even when it's the opposite. Noticing the neglect doesn't feel like catching yourself skipping a step, it feels like you were simply paying attention to the more relevant, concrete thing in front of you. That's what keeps this low on the scale, even though, unlike some other biases, a known, testable fix (asking for the numbers as frequencies) exists and reliably works once applied.

Individual
Before reacting to a striking single case, explicitly ask how common this is, actually, and go looking for the real number instead of trusting how well the story fits a pattern.
Decision-process
Reframe any probability question as a natural frequency: "10 out of 1,000," not "1%." This is Gigerenzer's actual tested intervention, and it has real, repeated experimental support for improving accuracy without training.
Systemic
Risk-communication guidelines in medicine and science journalism increasingly recommend reporting frequencies with a clear denominator, rather than bare percentages or relative-risk figures, specifically because of this research line.

Debiasing research note

The natural-frequency reframing is one of the more robust, well-replicated debiasing interventions in the whole cognitive bias literature, genuinely stronger evidence than "consider the opposite"-style advice used for other biases. But it works only when someone deliberately reframes the numbers first; it doesn't spontaneously correct an intuitive snap judgment made under ordinary percentage or probability framing.

Commonly co-occurs with

Illusory correlation also compounds with this one but doesn't yet have its own Filter page to link to.

Often mistaken for

Gambler's fallacy

Not a Filter page yet, so no link here, but worth getting right: both involve misjudging probability, but the gambler's fallacy is about wrongly expecting independent random events to "even out" in the short run, while base rate neglect (this page) is about ignoring a known statistical prevalence in favor of a specific description. Different mechanism, different kind of number being misjudged. See Section 01.

Compounding effect

Base rate neglect and the availability heuristic feed each other directly. A tendency to weight vivid, easily recalled cases more heavily (availability) is often exactly what supplies the compelling individuating information that crowds the base rate out in the first place, the two rarely operate in isolation.

Primary sources

Eddy, D. M. (1982). Probabilistic reasoning in clinical medicine: Problems and opportunities. In D. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty: Heuristics and biases (pp. 249–267). Cambridge University Press.

Accessible reading

Kahneman, D., Thinking, Fast and Slow (2011). See the chapters on the "Tom W." problem and representativeness. Beginner.
Gigerenzer, G., Calculated Risks: How to Know When Numbers Deceive You (2002, published in the UK as Reckoning with Risk). Moderately technical but written for a general audience, built directly around this bias. Intermediate.