Novosibirsk, from 1959: Dmitri Belyaev selected silver foxes for one trait — tameness toward a human hand — and nothing else. Within ~10 generations the tame line grew floppy ears, curled tails, piebald coats and shorter breeding cycles: the “domestication syndrome,” riding along. This panel is runnable: the tame line climbs by the breeder’s equation R = h²·S. Select the tamest fraction, watch the mean move. Set heritability to zero — the Lysenkoist claim — and nothing moves. You cannot select what is not heritable.
source Belyaev & Lyudmila Trut, Institute of Cytology and Genetics, Novosibirsk (the farm-fox experiment, from 1959); Trut 1999 (American Scientist); the domesticated-fox genome, 2018. Real quantitative genetics, run in the country that had banned it. Room: THE WEIRD. Rendered, not quoted.
At Novosibirsk, Dmitri Belyaev ran real genetics quietly, in a country where the discipline had been declared bourgeois. His farm foxes were scored on a single axis: how they responded to an approaching human hand. Only the calmest fraction were allowed to breed.
Nothing was selected for coat, ear, or tail — only for tameness. Lyudmila Trut carried the line for decades. The experiment is still running; the tame foxes seek human contact like dogs.
Selecting for tameness alone dragged a suite of changes with it: floppy ears, curled tails, piebald (star-blaze) coats, shortened muzzles, and shorter, off-season breeding cycles. None was bred for directly.
This is the domestication syndrome — the same cluster seen in dogs, pigs, and cattle. One selected axis, many correlated responses. The engine models the selected axis; the syndrome is the correlated cargo it pulls.
h² is the fraction of the trait’s variance that is additive-genetic — the part parents pass on. It is the lever arm of selection: response scales with it, exactly.
R = h² · S
If h² = 0 the trait is real but not heritable: selection has no purchase, and R = 0. If h² = 1 the offspring mean shifts by the full selection differential. The table samples the identity across many h².
| h² | S | R = h²S |
|---|
Set the heritability, the phenotypic spread, the fraction bred, and the number of generations. Truncation selection breeds only the tamest fraction p; the engine derives the selection intensity, the differential, the per-generation response, and the cumulative climb.
Truncation selection on the tamest fraction p gives a standardized selection intensity i = φ(z)/p, where z is the truncation point. Then:
selection intensity i = — · differential S = iσ = —
per-generation response R = h²S = — σ/gen
Cumulative climb of the tame-line mean (constant params):
| gen | mean tameness (σ) |
|---|
| case | h² | R | claim: rises? | ok |
|---|
Lysenko taught that acquired characters are heritably transmitted — that habit passes to the genome. Belyaev’s foxes are the mechanical rebuttal: the tame line climbs only because tameness is heritable and he bred the tamest.
“Tameness was taught, not bred.” DEAD. If tameness were merely taught — acquired within a life, h² = 0 — the breeder’s equation gives R = 0 and the line would not have moved. It moved by ~0.49σ/generation. It was selection on heritable variation.
The red move is Lysenko’s: set h² = 0 (the trait is only acquired, not heritable) yet assert the tame line still rises across generations.
Under the Lysenkoist claim the engine forces h² = 0 but asserts a rise. The breeder’s equation returns R = 0 — the mean does not move — while the claim insists it does. The witness (5) compares the claim to the mechanism and turns red. You cannot select what is not heritable.