Selection experiments its effect on inbreeding depression, particularly on
W .G. Hill Livestock Production Science 63 2000 99 –109
103
2 1 N , s , 0, and zero if more deleterious. The almost 100 years and generations, and in which
e
cumulative consequence of many mutations segregat- responses have continued in the lines selected for
ing in the population, however, some of which high oil content or for high protein content in the
eventually become fixed, could be a substantial drop kernel plants Dudley, 1977. The populations are
in mean performance, particularly in overall fitness bred with only 30 or so selected plants maternal
for which most mutations are likely to be deleterious. half-sib families. More male parents are likely to be
Thus considerable attention has been given to the involved, but the lines are so superior to the natural
size of the natural population necessary for the rate base for the selected traits that selection of offspring
of elimination of deleterious mutations to be suffi- got by cross pollination outside the selected lines
cient that mean fitness does not drop substantially; seems most unlikely to occur. Response in the low
or, in the jargon, that the population is large enough lines has attenuated, but they are approaching levels
to prevent it going into non-sustainable mutational of zero oil content in the kernels.
‘meltdown’. Conclusions depend on assumptions of Continued selection has been practised in Dum-
the rate of occurrence and size of effects of deleteri- merstorf for increased body weight in mice, with 160
ous mutations: for a given input of variance, the animals selected each generation, a relatively large
smaller their mean effect, the harder the mutants are size for a selection experiment in laboratory animals
¨ to eliminate and the greater their cumulative in-
Bunger et al., 1990. Selection has now been fluence on mean fitness. Some theoretical arguments
continued for over 90 generations. At 6 weeks the indicate that populations as small as 100 could be at
high line mice average |65 g and the unselected risk Lynch et al., 1995; whereas experimental
controls 30 g, a response of about 12 phenotypic checks suggest that smaller populations are expected
standard deviations. Response continues, albeit at a ¨
to be safe for very long periods Gilligan et al., reduced rate L. Bunger, personal communication.
1997. For populations under artificial selection, In a review of mouse selection experiments to that
fitness is also likely to fall as a consequence of date, Eisen 1980 showed that long-term responses
pleiotropic effects of genes affecting the trait under increased with population size and Jones et al.
selection that have increased in frequency or become 1968 showed similar relationships in Drosophila.
fixed by artificial selection. Laboratory or domestic Selection experiments in animals with the largest
livestock populations are, however, protected against populations have been carried out by Weber 1990,
extinction by careful and good management, for who got 30 higher responses over 55 generations
example by giving many opportunities to breed and of selection for wing-tip height in Drosophila with
by health control. 1000 than with 200 selected parents.
The effect of population size on maintaining variation and selection response is confounded with