One of the assumptions required to estimate genetic parameters by use
of the diallel mating design is that the genes in the parents are inde
pendently distributed. The objective of this study was to test the val
idity of this assumption. Diallel experiments including a fixed sample
and a random sample of parents were conducted in maize (Zea mays L.).
For the fixed sample, an eight-parent diallel among selected inbred l
ines was produced (original diallel). Ninety-six unselected inbred lin
es derived from a population, developed by intermating for five genera
tions crosses of the original diallel, were used to produce 12 eight-p
arent diallels (random diallels). Diallels were evaluated in a replica
tion-within-sets randomized incomplete block design in six environment
s. Estimates of additive (sigma(A)(2)) and dominance (sigma(D)(2)) var
iances from the pooled random diallels were significantly different fr
om estimates of sigma(A)(2) and sigma(D)(2) in the original diallel fo
r about half the traits. For six traits, the average level of dominanc
e was overestimated in the original diallel relative to the pooled ran
dom diallels. The estimated average level of dominance for grain yield
was two times greater in the original diallel than in the pooled rand
om diallels. Estimates of sigma(A)(2) seemed to be affected more than
estimates of sigma(D)(2) by repulsion phase linkages. Non-independent
distribution of genes causes differences in sigma(A)(2) and sigma(D)(2
). The diallel mating design should only be used to estimate genetic p
arameters when the parents of the diallel have been randomly selected
from a population in linkage equilibrium.