The Use of Assortative Mating for Heritability Estimation.

EEVE (1953, 1955,1961) has suggested that assortative mating of parents be used to obtain data for estimation of heritability by parent-offspring regression. The advantage of this technique over the standard random mating procedure is that the variance of midparent values is increased, thereby providing greater accuracy in the estimation of the regression coefficient. Possible bias introduced as the result of phenotypic correlations between parents, given certain assumptions, was shown by REEVE to be negligible or to involve only second-order effects. Before this method is generally adopted, one further problem should be pointed out. Many quantitative characters for which heritability estimates are desired involve maternal effects. With a random mating design it is possible to obtain an estimate of heritability free from maternal effects by the use of regressions of offspring on sires. Thus, ignoring higher order (additive) epistatic terms, the sire estimate contains only variation due to the additive effects of genes ( b = 1/2 h 2 ) , whereas the regression of offspring on dams yields an estimate which includes maternal effects ( b = 1/2 h" + M). On the other hand, a completely assortative mating scheme does not allow the detection of maternal effects, since with such a procedure the regression of offspring on sire and on dam will be exactly the same. Thus the heritability of a character for which maternal effects are important will be overestimated by the use of assortative mating of parents. To demonstrate this point, data on heritability of developmental time collected in both ways in the flour beetle, Tribolium castaneum, may be of interest. From more detailed studies to be reported elsewhere maternal effects are known to be important for this character.