The Taxi Problem
In the 1970s, Daniel Kahneman and Amos Tversky presented subjects with a now-classic problem. A cab was involved in a hit-and-run accident at night. Two cab companies operate in the city: Green (85 percent of cabs) and Blue (15 percent of cabs). A witness identified the cab as Blue. Tested under the actual nighttime conditions, the witness correctly identifies the color 80 percent of the time and is wrong 20 percent of the time. What is the probability the cab involved was actually Blue?
Most people, including many with statistical training, answer somewhere around 80 percent — anchoring on the witness’s reported accuracy. The correct answer, using Bayes’ theorem, is closer to 41 percent. The reason is that the base rate — the fact that only 15 percent of cabs are Blue to begin with — matters enormously, and most people’s intuitive reasoning discounts it almost entirely in favor of the more vivid, specific-feeling information (the witness’s identification).
This phenomenon — systematically underweighting general statistical information (base rates) in favor of specific, individuating information, even when the specific information is less reliable than it feels — is called base rate neglect. It’s one of the most replicated findings in the heuristics-and-biases literature, and it shows up in medical diagnosis, legal judgment, financial forecasting, and — with particular force — in how people evaluate the accuracy of personal readings.
The Base Rate of a Personality Statement
Apply this to a divination reading. Suppose a reading says: “You often present a confident exterior, but privately you carry more self-doubt than people around you realize.”
What’s the base rate of this statement — that is, what fraction of people, regardless of their chart, their cards, or their numbers, would recognize this as broadly true of themselves?
Almost everyone manages some gap between their internal experience and their outward presentation; total transparency between inner state and outer behavior is rare enough to be notable when it occurs. A statement like this likely has a base rate well above 80 percent — it would ring true for the overwhelming majority of people who read it, independent of any information about who they specifically are.
But base rate neglect means this is not how the statement gets evaluated, experientially. When you read the statement and recognize yourself in it, the recognition itself feels like the relevant information — “yes, that’s me” — and the question of how many other people would have the same reaction doesn’t enter into the experience at all. The statement isn’t compared against its base rate. It’s evaluated in isolation, the way the witness’s identification was evaluated in isolation from the base rate of Blue cabs.
This is, in essence, the cognitive mechanism underlying the Barnum effect, named for the showman’s apocryphal claim about having something for everyone. Bertram Forer’s 1948 experiment gave students a personality description assembled from horoscope columns and had them rate its accuracy as a description of themselves specifically — ratings averaged around 85 percent, despite every student receiving the identical text. The text’s base rate of “feels true to readers” was close to 100 percent. No individual reader had access to that information. Each reader experienced only their own match.
Why “Specific-Feeling” Doesn’t Mean Low Base Rate
The taxi problem illustrates a particular trap: information that feels highly specific and individuating (a witness pointing at a color, a chart placement described in technical-sounding language) can have a much higher base rate — apply to a much larger fraction of cases — than its specificity suggests.
Divination readings are often constructed, whether deliberately or through the natural evolution of interpretive traditions, in ways that maximize this gap. A statement framed in the technical vocabulary of a system — “your Mercury in the eighth house suggests a tendency toward private, intensive modes of communication” — sounds like it’s describing something narrow, derived from a specific chart placement that not everyone shares. And in one sense it is: not everyone has Mercury in the eighth house. But the behavioral content of the statement — “private, intensive modes of communication” — may have a base rate, among the general population, of well over 50 percent, regardless of chart placement. The technical framing borrows specificity from the rarity of the astrological configuration while attaching it to a behavioral description whose actual rarity is much lower.
This combination — narrow-sounding technical framing wrapped around broad-based-rate content — is, whether by design or by the selective survival of effective interpretive language over centuries of practice, close to optimal for triggering base rate neglect. The reader experiences the specificity of “Mercury in the eighth house” (which feels rare, because it sort of is) as if it transfers to the specificity of the behavioral claim (which isn’t rare at all).
The Reverse Problem: When Specificity Is Real
It’s worth being fair to the other side of this. Not every statement in a divination reading has a high base rate, and base rate neglect cuts both ways — it can make a genuinely rare and specific statement feel less remarkable than it should, if the reader doesn’t appreciate how unusual the match actually is.
If a reading says something quite specific — “you have experienced a significant transition involving a geographic move within the last two years, connected to a change in a close relationship” — and this happens to be exactly true, the base rate of this compound statement (geographic move + relationship change + two-year window) is considerably lower than the “confident exterior, private self-doubt” example. A hit on a genuinely low-base-rate statement is more informative than a hit on a high-base-rate one — but base rate neglect means both hits tend to be experienced with similar intensity, because the feeling of a hit doesn’t carry information about the base rate that produced it.
This means that someone trying to evaluate a reading’s accuracy honestly faces a genuinely difficult task: distinguishing hits that are impressive because they’re improbable from hits that feel impressive but aren’t, because the feeling of “that’s accurate” doesn’t come tagged with the statement’s base rate. Without that tag, all hits tend to register similarly, and the overall impression of “this reading was accurate” ends up dominated by the number of apparent hits rather than their individual improbability — even though a single low-base-rate hit might be more meaningful than ten high-base-rate ones.
Why This Is Hard to Correct
Base rate neglect is among the more stubborn biases in the literature — explicit training in statistics reduces it somewhat but doesn’t eliminate it, and even people who can correctly solve the taxi problem on a test often fail to apply the same reasoning “in the moment” to a situation that feels personal rather than abstract.
Part of the reason is that base rates require a kind of information that isn’t naturally present in the situation generating the judgment. When you read a personality statement and ask “does this describe me?”, the information “what fraction of people would also say yes to this?” isn’t part of your experience — it’s a fact about a population you’re not looking at, while you’re looking at yourself. Retrieving and applying that information requires a deliberate, effortful step that the automatic “does this fit?” evaluation doesn’t include.
This is structurally similar to the hindsight bias problem discussed elsewhere: the corrective information exists, in principle, but isn’t part of the natural cognitive process that generates the felt experience, and has to be deliberately imported from outside that process — which most people, most of the time, don’t do, because the felt experience doesn’t signal that anything is missing.
What Knowing the Base Rate Changes
For someone evaluating a reading, the practically useful question — though an unnatural one to ask in the moment — is something like: if I showed this specific statement to a hundred random people, how many would say “yes, that’s me”?
For most personality-style statements in most readings, the honest answer is “a large majority” — which doesn’t mean the statement is false, exactly, but means that its truth isn’t telling you anything about whether the system that produced it has any validity. A system that produced only statements with this property would be functionally indistinguishable, in terms of felt accuracy, from a system with genuine predictive power and a system with none at all. The felt accuracy is consistent with all three.
What would actually distinguish a system with genuine signal is statements with low base rates that nonetheless land — specific, improbable, falsifiable claims that turn out to be true at a rate exceeding what their base rate would predict by chance. This is a much higher bar than “it felt accurate,” and it’s a bar that, as the research on astrological and similar systems discussed elsewhere in this series shows, has been difficult to clear under controlled conditions.
The statement fits you. It probably fits the person next to you too. The question that matters isn’t whether it fits — it’s how many other questions you’d have to ask before you found one that didn’t.