The sample of one

My grandfather smoked and lived to ninety-four. This sentence has ended more arguments about health than any study, and it should end none of them.

The problem is not that the story is false. It is almost certainly true. The problem is that one case cannot tell you about a tendency, and health claims are claims about tendencies. Smoking raises the probability of several diseases. Raising a probability is not the same as guaranteeing an outcome, so a single person who smoked and lived long is not evidence against the claim. The claim never said it was impossible.

What makes personal stories so persuasive is that they are vivid and specific, while a statistic is neither. A number describing two hundred thousand people has no face. Your grandfather has a face, and a chair he used to sit in, and a particular way of laughing. The mind treats what it can picture as more real than what it cannot, and that preference is almost impossible to switch off by willpower.

There is one situation where a single case does settle a question: when a claim says something never happens. One counterexample destroys an absolute claim. But almost nothing interesting in health, education or economics is claimed absolutely, so the useful question when you hear a story is what it is being offered as evidence of.

The grandfather is real. He is simply evidence about one person, which is exactly as much as one person can be evidence about.