Replicated selection for body weight in mice

SUMMARY The variation in the response to selection was studied by replication of selected lines. A random-bred strain of mice was divided into six replicates. Two-way selection for 6-week weight was applied in an identical manner to each replicate, and each had an unselected control. Each line (6 large, 6 control, 6 small) was maintained by minimal inbreeding with 8 single-pair matings. The overall mean responses, both up and down, were linear and very regular for ten generations, with realized heritabilities of 40% upwards, 33% downwards and 37% for the divergence. The separate replicates, however, differed greatly in their realized heritabilities, with upward selection ranging from 25 to 46%, and downward selection from 16 to 50%. The theoretical prediction that, because of genetic drift, the standard error of a realized heritability is underestimated by the standard error of the regression of response on cumulated selection differential was borne out in this experiment. The empirical standard error, calculated from the observed variance between replicates, was more than twice as great as that of the regression. The empirical standard errors showed that the asymmetry between upward and downward responses was not significant. The variation between the replicates was ascribed mainly to random drift, which may seriously influence the conclusions about the realized heritability and the asymmetry of response that would be drawn from a single experiment with the population size of one of these replicates. After 23 generations of selection the large lines were approaching limits, and the limit appeared to be at the same level in all. The small lines showed an undiminished realized heritability after 23 generations, but the selection differentials were then so small that little progress was made. There was evidence of counter-acting natural selection. All aspects of productivity – proportion of fertile matings, litter size and weaning rate – declined in the control lines. The overall productivity of the large lines was a little below the controls, and that of the small lines was reduced to about half the level of the controls. The separate replicates differed from each other significantly in all the components of productivity.