Lewontin Builds the House Gould Told Him Not To
2026-07-29 · AI-generated insight
Read together, Gould and Lewontin are on the same political side but hold incompatible strategies — and Gould's is the one that survives contact with Winegard and Carl.
Lewontin makes a two-step move. Step one is empirical: variation within groups swamps variation between them, so races are "remarkably similar." Step two is normative: therefore "no justification can be offered for" racial classification's "continuance." He welds the moral conclusion directly onto the genetic premise. Anti-racism, in his telling, rests on a fact about allele frequencies.
Gould refuses exactly this weld. Equality, he insists, "is neither an ethical principle... nor a statement about norms." It "just worked out that way," and "a hundred different and plausible scenarios" could have produced deep biological gradients instead. The ethical commitment to equal treatment must therefore float free of any particular empirical result — precisely because the empirical result was contingent and could have gone otherwise.
Now bring in Winegard and Carl. Their skull-classification datum — 80% accuracy from two variables, near-perfect from many — is the classic rebuttal to Lewontin: within-group variance can dominate at single loci while correlations across many variables still cluster people by ancestry. (Note that Lewontin files under "randomly chosen genetic differences"; the counter-note files under "Correlation Across Many Variables." That is the whole argument in two filing tags.)
Here is the pollination. If, like Lewontin, you have staked human equality on the claim that populations are statistically indistinguishable, then every advance in multivariate classification becomes a threat to your ethics. You have handed the moral high ground to whoever wins a technical dispute about covariance structure. Gould saw the trap: an anti-racism mortgaged to a specific empirical finding is an anti-racism awaiting foreclosure.
And the Coon episode shows what foreclosure looks like — a discipline voting to condemn a book almost no one had read, defending a conclusion it could no longer defend on the merits. When the ethics and the empirics are fused, and the empirics wobble, the only tools left are the raised-hand vote and the statement passed regardless. Gould's decoupling isn't a concession to race science; it's the thing that makes anti-racism unfalsifiable-proof.