Inside a selective school, the math stars seem to write worse, and the admissions office is why.

Suppose a school admits students who are either very strong at math or very strong at writing. Look at the admitted class and a pattern appears: the best mathematicians tend to be weaker writers, and the best writers weaker at math. It is tempting to conclude that the two talents trade off, that a mind built for proofs is built against prose. But the pattern may say nothing about minds. The admission filter may have manufactured it.

That is Berkson’s paradox: conditioning on selection or success can make two traits look negatively correlated even when they are independent, or positively correlated, in the full population.

How the gate makes the pattern

The student who got in with extraordinary math ability did not need extraordinary writing ability, and vice versa. Students weak at both were rejected, so they never appear in the sample. With that corner of the population missing, what remains slopes the wrong way: among the admitted, more of one trait goes with less of the other. If either key opens the door, then among the people inside, having one key makes it less necessary to have the other.

The same thing happens whenever selection depends on several routes to one outcome: getting hired, showing up in a hospital sample, becoming notable enough to be observed at all. So “correlation becomes anti-correlation at the extremes” is often a warning about filters rather than about the world. At the top of a selected distribution, people may look like they trade one trait against another even when the traits are neutral or positively related in the full population.

Why the selected sample misleads most

The natural objection is that the selected sample is the one that matters. These are the admitted students, the hired engineers, the famous artists, so surely their patterns reveal the deepest truth. But selected samples are often the most misleading ones. The more aggressive the filter, the more likely you are seeing the logic of the gate instead of the logic of reality.

The published research literature is a particularly insidious example. Underpowered studies that happen to find significance pass through the publication gate, manufacturing a body of evidence that looks far more consistent than reality.

So before believing an anti-correlation, ask one question: what process decided who ended up in the sample? If it was a gate that either trait could satisfy, the trade-off you are looking at may belong to the admissions office, not to the students.