Currently available software packages are not able to handle the joint gene
tic evaluation of pure and crossbred data correctly. Instead, the same trai
ts originating in the two purebreds and the crossbred are treated as three
separate traits and analysed in a three trait animal model. Using selection
index theory it can be demonstrated that this procedure contains model vio
lations even if only additive genetic effects are considered. The effective
ness of the correct model which allows different contributions of the paren
tal lines to the genetic variance of the cross including animal specific do
minance effects is assessed on the basis of different selection indices. Th
e sources of information chosen reflect those available in typical chicken
breeding programs. The genetic parameters required for index construction -
in particular the dominance variance and the genetic correlation between t
he pure and cross breds - were derived using a two locus model. If the pure
bred contribute extremely differently to the crossbred variance large losse
s in efficiency can be observed. With contributions of the sire line of 11,
19, 37 and 40% to the additive genetic variance of the cross corresponding
losses in efficiency of selection where 13.1, 6.5, 0.6 and 0.2% respective
ly. Not considering the dominance variance - depending on its proportion re
lative to the total genetic variance - lead to a moderate loss of efficienc
y from 0.1 to 2.6%.