Why Does Everybody Know the Phone Rings When You Are in the Bath? You Only Count the Hits
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
People perceive relationships between things that are statistically unrelated, and the mechanism producing that error also produces stereotypes about groups nobody knows well.
What the error is
The error is perceiving a relationship between two variables where none exists, or perceiving a much stronger one than the data supports. It is not a failure to do arithmetic, since people shown a full table of numbers still report a relationship that is not there. It is a failure of what gets noticed and remembered, because the cases that confirm an expected relationship are attended to and encoded more readily than the cases that do not, so the sample recalled is not the sample observed.
The four cases people forget
Assessing a relationship correctly requires attention to all four combinations:
- •Both things present, which is vivid and easily recalled
- •Both things absent, which is almost never noticed at all
- •The first present without the second, which is recalled weakly
- •The second present without the first, likewise
- •Judgement rests mainly on the first of these
- •So a relationship appears wherever the first case is memorable
The distinctiveness effect
A particularly consequential version arises when two uncommon things co-occur. An unusual event involving a member of a minority group is doubly distinctive, so it is encoded more strongly than the same event involving a member of a majority group, and people later overestimate how often the two go together. Experiments demonstrate this with entirely fictional groups and neutral behaviours, showing the effect with no prior prejudice involved at all, which means the mechanism can manufacture a stereotype from nothing more than differing group sizes and the memorability of rare combinations.
How it resists correction
Being shown the numbers frequently fails to shift the belief, and the reasons are worth separating. A person who already expects a relationship attends to confirming cases more closely and interprets ambiguous cases as confirming, so the evidence they gather continues to support them. Asking somebody to test a belief usually produces a search for confirming cases rather than for disconfirming ones, which is a separate and well documented habit. And where the belief concerns people, a confident prediction can change behaviour on both sides until the expected pattern genuinely appears.
Where it does damage
The error underlies a range of persistent false beliefs and some of them are expensive. Clinicians using projective tests reported diagnostic signs that systematic study showed were unrelated to the conditions they claimed to detect, and continued to report them after being shown the data. Belief that admissions rise at a full moon persists among hospital staff despite repeated studies finding nothing, because busy full moon nights are remembered and quiet ones are not. Arthritis pain and weather, sugar and children's behaviour, and a great many folk remedies rest on exactly the same asymmetry of attention.
The takeaway
People perceive relationships that are not there because confirming cases are noticed and remembered while the other three combinations are not, particularly the one where neither thing occurred. When two rare things coincide the pairing is doubly memorable, which manufactures stereotypes about small groups from nothing. The full moon, projective tests and many folk remedies rest on it.