Base rate neglect is ignoring prior probabilities when case-specific information arrives — even when the case information is worthless. It is the flagship error of the representativeness heuristic: judgment by similarity to a stereotype, not by Bayesian updating.
Classic demonstrations
Steve the librarian: shy, tidy, detail-oriented — most rank librarian above farmer, ignoring that farmers vastly outnumber librarians.
Engineer-lawyer study: with 70% engineers in the sample, a bland "Dick" description still yields ~50% engineer judgments — identical to the 30%-engineer condition. Worthless evidence swamps base rates.
Tom W: personality sketch drives specialty rankings; base rate of graduate fields barely matters.
Linda: "bank teller and feminist" ranked more probable than "bank teller" — conjunction fallacy because the detailed story feels more representative.
Fix
Start with the base rate (reference class), then adjust modestly for diagnostic evidence — the outside-view discipline in inside-view-outside-view and planning-fallacy.