The assortative mating amplifier: a pre-registered test of first-order Bulmer theory in a synthetic genetic pool

Assortative mating is predicted by quantitative-genetics theory to inflate trait variance by generating positive gametic-phase disequilibrium, the mirror image of the Bulmer effect. This paper reports a pre-registered simulation test of the first-order identity V_LD(1) = rho x V_A(0) in a five-locus synthetic pool seeded from gnomAD v4 non-Finnish European frequencies, with 400 diploid founders, 10 generations, 30 replicates, and mate correlations of 0.0, 0.3, and 0.6. In the frozen tier, the control arm fails its criterion because drift generates a disequilibrium background of 0.0656 in the same statistic; the rho = 0.3 arm passes; the rho = 0.6 arm fails condition (i). A second tier under Amendment A-S3 subtracts the within-replicate control background; both assortment arms then fail condition (i), at approximately half the predicted magnitude. A third tier under Amendment A-S3b tests the resulting half-factor prediction, (rho / 2) x V_A(0), on fresh replicates; both arms pass. The qualitative prediction of the theory is confirmed at every level; the quantitative identity holds in its half-factor form. No verdict of an earlier tier is altered by a later tier.