Experimental evidence on limitations of the heritability parameter.

Two-way selection was practiced at three selection intensities for 21 day pupa weight inTribolium castaneum. Each of the 19 single-generation replicates of the project consisted of three populations, one each being selected upward and downward at 10, 30 and 50% selection intensities. Realized heritability for divergence was significantly lower for the least intense (50%) level than for the 10% and 30% selection intensities. Both expected heritabilities, calculated from the regression of offspring on parents, and realized heritabilities were much larger for downward than for upward selection. Expected and realized estimates of heritability were consistent within direction of selection, but estimates of heritability for divergence were of little value in predicting directional weight change. The variation observed for heritability estimates was similar to that expected from published prediction equations. The asymmetry of heritability estimates and the differential effects of selection intensity on the magnitude of realized heritability are not consistent with the usual assumption of a linear relationship between genotype and phenotype.