Quincunx at South Kensington

Dear Mr. Pearson,

The quincunx at South Kensington has been misbehaving, and the misbehaviour is instructive. On the 9th of December, during the fourth month of the laboratory’s operation, a lead pellet lodged in the left branch of the funnel. Every subsequent pellet that struck it was deflected into the right compartment, and by the afternoon the lower rows had assembled themselves into a distribution with its mode displaced a full division from centre. A visitor watching the board fill would have concluded that the population itself was skewed. The population was not skewed. One pellet was stuck.

Nine thousand three hundred and thirty-seven persons have now passed through the laboratory, and their heights, head-lengths, and breathing powers arrange themselves in curves of the familiar shape. But the temptation I wish to put before you is this: a single deviant observation, arriving early, has power over every observation that follows it, because it is mistaken for a channel rather than a sample. The confirmations that come after - and they come in their hundreds, obedient to the curve - are read as corroboration of the deviant case, when in truth they are merely the distribution reasserting itself around a blocked passage.

I have begun keeping a separate leaf in the record book for observations that contradict the prevailing arrangement. There are eleven entries so far. The rule I am testing is simple: an anomaly is not evidence of a different law until it can be shown to deflect the arrivals that come after it, and to keep them deflected once the passage is cleared. Most anomalies, when freed - as I freed the pellet with the point of a pencil - leave the curve exactly where the curve would have been without them. The exception is the anomaly that survives this test, and it is worth a thousand that do not, not because it is more striking, but because it alone changes the shape of everything downstream.

The lodged pellet was the forty-first to pass through the funnel that day.

Yours faithfully, Francis Galton