Justin’s Notes

The Same Data, Read Alone or Read Together

2026-07-23 · AI-generated insight

Here is the rare case where two notes stand in direct rebuttal, and reading them side by side reveals that the disagreement is not about the facts but about a single statistical move: whether to treat genetic loci one at a time or as a correlated whole.

Lewontin's claim is precise and, on its own terms, correct: at any randomly chosen locus, human variation is overwhelmingly within-group rather than between-group. From this he draws a taxonomic verdict — racial classification is "of virtually no genetic or taxonomic significance." The inference runs from a per-locus statistic to a global conclusion.

Winegard and Carl locate exactly where that inference breaks. The per-locus figure is real, but loci do not vary independently; their differences are correlated across populations. Their face analogy is the sharp instrument: nose size alone barely distinguishes the sexes, yet the joint pattern across all features classifies with 95 percent accuracy. Information invisible in any single dimension emerges when the dimensions are considered together.

What makes the pairing genuinely interesting is that it isolates a general epistemic hazard, one that outruns this particular controversy. A quantity can be small marginally and large jointly. Summing or averaging a set of "individually negligible" differences can systematically erase structure that lives entirely in the correlations between them. Lewontin's argument is, in this reading, a textbook instance of aggregating away a signal by refusing to look at covariance.

But notice what the rebuttal does not touch. Lewontin makes two claims: the taxonomic one and a social one ("positively destructive of social and human relations"). Winegard and Carl's correction bears only on the first. Even if the statistical structure is real, the social claim is a separate proposition requiring separate argument — and the fact that Lewontin fused them, letting the moral verdict ride on the taxonomic one, is itself the mirror image of the error he committed. He bundled two things that should be examined separately, just as he unbundled loci that should have been examined together. The methodological lesson cuts in both directions: know when your variables are secretly correlated, and know when your conclusions secretly are not.

Woven from these notes

In brief, Lewontin ignored the fact that differences among human populations are correlated. Although population variation at any one genetic locus tends to be small, global population structure becomes clear if one examines correlated differences across loci. For a simple analogy, consider men and women’s faces. If one takes any particular characteristic, say nose size, there is likely to be more variation within a sex than between the sexes (and it would be nearly impossible to classify faces by sex with any accuracy using only nose size). However, if one considers all facial characteristics together (which is, after all, how we actually experience human faces), then sex differences become sufficiently clear that an observer can guess the correct sex more than 95 percent of the time.

— Bo Winegard, Noah Carl Superior: The Return of Race Science—A Review
It is clear that our perception of relatively large differences between human races and subgroups, as compared to the variation within these groups, is indeed a biased perception and that, based on randomly chosen genetic differences, human races and populations are remarkably similar to each other, with the largest part by far of human variation being accounted for by the differences between individuals. Human racial classification is of no social value and is positively destructive of social and human relations. Since such racial classification is now seen to be of virtually no genetic or taxonomic significance either, no justification can be offered for its continuance.
— Lewontin