We studied the evolution of litter size in natural and experimentally
manipulated populations of Columbian ground squirrels (Spermophilus co
lumbianus) in eastern Washington state and southwest Alberta. Litter s
ize at weaning in a large natural population (mean = 3.51 pups/litter,
248 litters) was significantly lower than the litter size (6 pups) th
at produced the most offspring surviving to yearling age class. This e
vidence contradicted Lack's ''optimal litter size'' hypothesis, which
predicts that the most productive litter size should approximate the m
ean. Litter size had no significant effect on the subsequent survival
or reproduction of mothers, contrary to the negative effects predicted
by the ''cost of reproduction'' hypothesis. Litter success varied amo
ng years, and good and bad years for reproduction could be experimenta
lly simulated with food supplementations. However, small samples of th
e largest litters rendered the ''bad-years'' hypothesis inappropriate
for application to our data. Proportional survival of offspring was re
latively constant among different litter sizes, contrary to the necess
ary condition of the ''cliff edge'' hypothesis of a dramatic decrease
in survival of young from the largest litters. The data supported the
''individual optimization'' hypothesis most strongly. As predicted, th
e number of surviving offspring increased with litter size, although i
n one population the six largest litters (2.4% of 248 litters) suffere
d reduced success. In natural and food-supplemented populations, chang
es in maternal body mass were associated with changes in litter size (
r = 0.205 to 0.926). Because survival of young remained relatively con
stant as litter size increased, these correlations resulted in greater
offspring success for mothers in better body condition.