Lewontin's Own Filing Cabinet Contains His Rebuttal
2026-08-02 · AI-generated insight
Read together, Lewontin and Winegard/Carl stage the most famous quarrel in this literature without either naming the other. Lewontin's claim rests on a specific move: judge by "randomly chosen genetic differences," one locus at a time, and human populations look "remarkably similar," with most variation lodged between individuals. From this he leaps to a policy verdict — classification is of "virtually no genetic or taxonomic significance."
But look at what Winegard and Carl report: researchers sort skulls into black and white Americans at 80% accuracy using only two variables. That result should be impossible if Lewontin's inference held. The resolution is quietly labeled on Lewontin's own note — Correlation Across Many Variables. This is precisely A.W.F. Edwards' rebuttal (the "Lewontin's fallacy"): any single trait is a poor discriminator, yet traits are correlated, and a handful of correlated measurements jointly locate a person's ancestry with high confidence. Lewontin measured the wrong thing — the information lives in the joint structure, not the marginal one. The tag under which Justin filed the quote is the very concept that dissolves it.
What makes this more than a technical footnote is the second half of Lewontin's passage, where a statistical claim slides seamlessly into a moral one: classification is "positively destructive," therefore "no justification can be offered for its continuance." Here the Coon anecdote becomes eerily relevant. The AAA condemned Putnam's book with one member having read it — condemnation preceding examination, the verdict fixed before the evidence. Lewontin performs the inverse of the same maneuver: the conclusion (classification must be baseless) arrives already loaded with the wish that it be so, and the analysis is recruited to serve it.
Wade names this circuit exactly — "the political view that genetics cannot possibly be the reason." The pattern across all four notes isn't that race is or isn't biologically real; it's that the machinery of measurement keeps getting steered by the answer people need it to produce. Lewontin's error and the AAA's vote are the same failure wearing different clothes: the finding was decided upstream of the looking.